National Academies Press: OpenBook

Risk Assessment Techniques for Transportation Asset Management: Conduct of Research (2023)

Chapter: Section 9 - Pilot Testing and Results

« Previous: Section 8 - Approach Followed for Study Selection and Testing
Page 35
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 35
Page 36
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 36
Page 37
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 37
Page 38
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 38
Page 39
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 39
Page 40
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 40
Page 41
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 41
Page 42
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 42
Page 43
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 43
Page 44
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 44
Page 45
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 45
Page 46
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 46
Page 47
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 47
Page 48
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 48
Page 49
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 49
Page 50
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 50
Page 51
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 51
Page 52
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 52
Page 53
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 53
Page 54
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 54
Page 55
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 55
Page 56
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 56
Page 57
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 57
Page 58
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 58
Page 59
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 59
Page 60
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 60
Page 61
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 61
Page 62
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 62
Page 63
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 63
Page 64
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 64
Page 65
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 65
Page 66
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 66
Page 67
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 67
Page 68
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 68
Page 69
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 69
Page 70
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 70
Page 71
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 71
Page 72
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 72
Page 73
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 73
Page 74
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 74
Page 75
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 75
Page 76
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 76
Page 77
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 77
Page 78
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 78
Page 79
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 79
Page 80
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 80
Page 81
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 81
Page 82
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 82
Page 83
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 83
Page 84
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 84
Page 85
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 85
Page 86
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 86
Page 87
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 87
Page 88
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 88
Page 89
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 89
Page 90
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 90
Page 91
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 91
Page 92
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 92
Page 93
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 93
Page 94
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 94
Page 95
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 95
Page 96
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 96
Page 97
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 97
Page 98
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 98
Page 99
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 99
Page 100
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 100
Page 101
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 101
Page 102
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 102
Page 103
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 103
Page 104
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 104
Page 105
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 105
Page 106
Suggested Citation:"Section 9 - Pilot Testing and Results." National Academies of Sciences, Engineering, and Medicine. 2023. Risk Assessment Techniques for Transportation Asset Management: Conduct of Research. Washington, DC: The National Academies Press. doi: 10.17226/27130.
×
Page 106

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

35   9.1 Study 1 and Study 2: Using Deterministic and Probabilistic Tools to Forecast Key Asset Management Variables at a DOT State transportation agencies are required to prepare financial plans for their 10-year TAMPs. These plans need to include forecasts of investments needed, the corresponding bridge and pave- ment conditions, and the revenues the agency expects to receive during the forecast period. Federal regulations also require TAMPs to be risk based and to include mitigation strategies for high-priority risks. Although agencies and legislatures may have control over factors that can influence revenues, such as taxes, allocations, and stimuli, they have no control over numerous other factors, such as inflation; construction costs; economic conditions like recessions; prices of essential inputs like steel, cement, and fuel; and as recently evidenced, pandemics. While preparing forecasts, agencies need to account for the associated uncertainties and communicate such uncertainties and the resulting implications to their stakeholders. Study 1 and Study 2 focused on testing general approaches to using deterministic and proba- bilistic forecasts, respectively, for key asset management variables to manage asset risks. For test- ing, forecasting revenues from state sources at VTrans were selected. The same data were used for both the deterministic and probabilistic methodologies. The discussion and findings for both approaches are presented in this section, so that the differences between the two can be conve- niently compared and assessed. The methodologies tested can easily be applied for forecasting costs or any other parameter of interest to an agency. As detailed in Section 9.1.3, in Step 1 of the methodology, several components of state revenue data were available to VTrans. The description in this report goes into detail for one component, gasoline tax revenues to the state transportation fund, to illustrate the methodology followed. The same methodology was applied to the other state revenue components and the summary output data are included in this report. The total aggregate projections for the state transporta- tion fund revenues over an 11-year forecast period from FY 2022 to FY 2032 are also shown. Summary charts of forecasts using the deterministic and probabilistic tools are provided along with a brief discussion to show how they compare. A separate example, that of using national data to illustrate construction cost risk, was also piloted. This example illustrates the limitations of deterministic forecasting. A description of this example is provided in Appendix E and is not included in the main report. S E C T I O N 9 Pilot Testing and Results

36 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 9.1.1 Pilot Objectives and How the Results Help Inform Asset Risk Decisions The objectives of Study 1 and Study 2 were twofold. First was to illustrate deterministic and stochastic methods that can be used to forecast state revenues. A selection of tools, from a simple deterministic model to more complex autoregressive models such as triple exponential smooth- ing (an error, trend, seasonal, or ETS, model available in Excel) and probabilistic analyses using Monte Carlo simulation were used for the pilot effort. Second was to demonstrate how the uncertainty surrounding forecasts of various components of state revenues can be illustrated using confidence intervals. An informed understanding of the band of likely values for the different components of state transportation revenues can help agencies plan for the investments they need to make (including trade-off decisions as applicable) to achieve bridge and pavement conditions commensurate with expected costs and available funding. Agencies can then communicate these uncertainties and the associated implications to stakeholders, help them manage their assets optimally, and set expectations of likely outcomes that are consistent with the funding available. 9.1.2 Description of the Strategy or Tool Studies 1 and 2 tested the use and limits of statistical analysis tools with increasing complexity to illustrate the risk posed by uncertainty surrounding revenue assumptions. Forecasting essentially involves projecting future values based on the analysis of historical time-series data, while accounting for any estimated changes based on known or anticipated market conditions. In deterministic models, the output of the model is fully determined by the parameter values and the initial conditions. Stochastic models possess some inherent random- ness. In these models, the same set of parameter values and initial conditions will lead to an ensemble of different outputs. Appendix E provides a brief description of deterministic and stochastic forecasting methods. Deterministic Model For Studies 1 and 2, the compound average growth rate (CAGR) of the various subcompo- nents of state transportation revenues, as determined from historical data, was used to illus- trate a deterministic model for projecting future revenues. Because deterministic forecasting techniques do not account for uncertainty, one sample approach to incorporate risk through confidence intervals is illustrated. The technique described to incorporate risk in the determin- istic forecast could be one of many and is included purely as an example. An agency will need to consult with SMEs to choose the approach that best represents their situation. Stochastic Models For the stochastic forecasting techniques used in Studies 1 and 2, two methodologies were selected, the ETS and Monte Carlo analysis methods: • ETS: The ETS method was chosen because it employs a smoothing process whereby it assigns exponentially decreasing weightage to historical data and accounts for trends and seasonality. ETS is conveniently available as a forecasting tool in Excel spreadsheets. The methodology estimates parameters that give weightage to recent historical data, trends, and seasonality and through a process of iteration and minimization of error between computed and actual historical values, arrives at the optimum values of these parameters. These parametric values are used to forecast future values. The assumption of a normal distribution for the data set is used to establish a 95 percent confidence interval.

Pilot Testing and Results 37   • Monte Carlo: The Monte Carlo method uses statistical parameters for the historical data (mean, standard deviation) and, using a random probability of occurrence per iteration, com- putes a value for the forecast variable using an inverse probability computation. This computa- tion is performed for a large number of iterations, and the mean value of the forecasts for all iterations is taken as the forecasted value. The upper and lower bounds of a 95 percent con- fidence interval are established by determining the 97.5 percentile and 2.5 percentile values, respectively, from all the iterations. Applicability The ETS and Monte Carlo techniques for revenue projection can be just as easily applied to forecasting future costs to arrive at investments needed to achieve certain condition targets or to other components of a TAMP financial plan. Studies 1 and 2 do not assert that there is one most appropriate means to prepare forecasts. Rather, the studies focus on illustrating the use of tools to help agencies understand and demonstrate uncertainty around forecasted state revenues. The ability to illustrate the degree of uncertainty around revenue forecasts can be important to • Include as a risk in the TAMP’s risk register. • Justify the monitoring of revenue changes as affected by market conditions (such as changing gasoline and diesel sales and the resultant impact on a material component of VTrans’ Trans- portation Fund revenue) and how they may require adjustments to a capital plan. • Explain to legislators or agency executives the uncertainty surrounding forecasts. Legislators and executives influence the allocation of funds in specific categories to achieve specific condi- tion targets. If unit costs rise more than expected over 10 years or if state revenues do not keep pace (or decline), then those targets may not be met unless funding increases. 9.1.3 Methodology Used in Conducting the Pilot State revenue data from FY 2000 to FY 2021 (22 years of data) were acquired from VTrans. (Fiscal years at VTrans run from July 01 to June 30.) The chosen deterministic and stochastic tools, CAGR, ETS, and Monte Carlo, were used to forecast the applicable parameters, such as the volume of gasoline sales and the annual rates of change for the various revenue sources. The forecasted rates of change were then used to arrive at the forecasted revenue values. The impacts of any potential new federal funding on state revenues have not been included in the forecasts. These will need to be separately accounted for as further details become available. Step-by-step instructions for the methodology are contained in Table 9-1. Before using the tools, the research team analyzed the input data to identify outliers, factors controlled by legisla- tive action, and factors outside the control of transportation officials and regulators. This was accomplished through data review followed by telephone and e-mail discussions with VTrans and Vermont state officials and SMEs. Based on the feedback, the research team made certain assumptions when using the data for the pilot forecasts. The assumptions related to the treatment of outliers or unusual changes observed in the input data. The approach used was to separate factors within the control of government decision makers (such as legislative actions that influence revenues) from those factors outside their control (such as recessions and pandemics). While these assumptions and the supporting rationale are dis- cussed in more detail in the following methodology section, it is important to recognize that state agencies are advised to involve SMEs before conducting forecasts to arrive at assumptions applicable to their circumstances. The approach described in Table 9-1 may not universally apply to each situation.

38 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Methodology Steps Step 1 Obtain input data. For the pilot, the input data were gasoline and diesel sales volumes, revenues to the Transportation and Transportation Infrastructure Bond (TIB) funds from gasoline and diesel taxes and assessments, and revenues to the Transportation Fund from the purchase and use tax, motor vehicle fees, and other revenues. Step 2 Evaluate the data to identify unusual patterns and outliers. Step 3 Obtain feedback from SMEs to understand the reasons for such patterns and outliers and to determine the data that would be pertinent to the forecasting effort. The feedback may include identifying factors such as legislative actions that create a controllable in- �luence over the revenues. Step 4 Establish a data set that can be used to forecast future values that are independent of controllable in�luences. (Decision makers can incorporate the impact of controllable in- �luences, such as increases in taxes and assessments, into the forecasts after the forecast values independent of such in�luences are computed.) The data set will incorporate cer- tain assumptions relating to outliers for forecasting purposes. For example, in the case of gasoline sales, the data indicated that the sudden reduction in gasoline sales in FY 2020 was a one-off event triggered by COVID-induced lockdowns, and the FY 2021 sales volume showed an initial rebound followed by an apparent tapering off. Certain as- sumptions relating to partial recovery of gasoline sales volumes were made to address this one-off observation. Step 5 Calculate deterministic forecasts. Step 5a For deterministic forecasts, compute key factors such as the CAGR for the key variable being forecast. Step 5b Prepare deterministic forecasts of the dependent variable. Step 5c Through discussions with SMEs, establish methodology and compute con�idence inter- vals for the dependent variable. Step 6 Calculate probabilistic forecasts. Step 6a For stochastic forecasts using triple exponential smoothing, enter the input data into an Excel spreadsheet, and using the Forecast functionality under the Data tab, create a fore- cast using the Forecast. ETS function. The function can produce the 95% confidence interval along with key forecast parameters. (Forecast.ETS inherently assumes a normal distribution for the data.) Step 6b For Monte Carlo forecasts using Excel, select the input data and determine the appro- priate probability distribution that best describes the data. As an alternative to Excel, an agency can use commercially available software for Monte Carlo forecasting to replace steps 6b(i)–6b(iv) by following instructions speci�ic to the use of that software. Step 6b(i) Compute key parameters, such as the mean and standard deviation for the input data, for use in the forecasts. Step 6b(ii) Using the inverse of the probability function for a randomly generated probability value between 0 and 1, compute the forecasted value of the dependent variable in an Excel spreadsheet. Repeat for the number of iterations desired. For this pilot, 500 iterations were used. Table 9-1. Methodology for conducting deterministic and probabilistic forecasts.

Pilot Testing and Results 39   Step 1: Obtain Input Data Step 1 involved collecting data relating to the state revenue available to VTrans. For context, it was important that the research team understand the various sources of revenue that contribute to the funding available to VTrans and the relative significance of each. State transportation revenues for VTrans currently come from two funds—the Transportation Fund and the TIB Fund. The TIB Fund is a subfund of the Transportation Fund whose revenue can be expended only on certain long-lived transportation structures. The Transportation Fund has the following six sources of revenue: 1. A fixed cent per gallon gasoline tax 2. A fixed cent per gallon diesel fuel tax 3. A gasoline percentage of price assessment with a minimum and maximum cent per gallon equivalent 4. A motor vehicle purchase and use tax of 6 percent split with 4 percent to the Transportation Fund and 2 percent to the Education Fund 5. Motor vehicle fees 6. Other revenue from small transportation-related taxes and fees The TIB Fund has two sources of revenue: 1. Assessments on gasoline sales 2. Assessments on diesel sales The gasoline and diesel levies (taxes and assessments) currently applicable in Vermont are summarized in Table 9-2. Methodology Steps Step 6b(iii) Compute the mean and standard deviation for the predicted values for the iterations performed. The mean value is used as the forecast value. Based on this data compute the lower and upper limits to establish the 95% con�idence interval. Step 6b(iv) Repeat steps 6b(ii) and 6b(iii) for each year of the forecast period. Step 7 Repeat Steps 2 through 6 for each parameter being forecast and combine to get a con- solidated forecast. For Studies 1 and 2, the steps were followed for all revenue sources and the forecast values for each year were summed to provide an estimate of the aggre- gate forecasted state revenues. Table 9-1. (Continued). Fuel Type Tax Assessment TIB Assessment Cleanup Fee % Min. Max. % Min. Fixed Gasoline 12.10 4% 13.40 18.00 2% 3.90 1.00 Transportation Fund 11.345 DUI Fund 0.380 Fish and Wildlife Fund 0.375 Cleanup Fund 1.00 Diesel 28.00 3.00 1.00 S�����: State of Vermont Table 9-2. Gasoline and diesel levies in Vermont (cents per gallon) as applicable to the Transportation Fund and the TIB Fund.

40 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research As seen in Table 9-2, the total gasoline tax is 12.10 cents per gallon on gasoline sold. Out of this, 11.345 cents per gallon is available to the Transportation Fund. The tax available to the Transportation Fund on diesel sold is 28.00 cents per gallon. There is a separate percentage of price assessment of 4 percent on the price of gasoline sales subject to a minimum of 13.40 cents per gallon and a maximum of 18.00 cents per gallon. For the TIB Fund, the gasoline assessment is 2 percent of the price of gasoline sales subject to a minimum of 3.90 cents per gallon and the diesel assessment is 3.00 cents per gallon. Additional details can be obtained by accessing the 2019 Revenue Details Fiscal Facts Booklet from the Vermont Legislative Joint Fiscal Office website.15 Table 9-3 provides the historical average (over 22 years) and FY 2021 contribution of each rev- enue source to the Transportation Fund. Table 9-4 provides the historical average and FY 2021 contribution to the aggregate of the Transportation Fund and the TIB Fund. As can be seen from Tables 9-3 and 9-4, taxes and assessments on gasoline and diesel sales in the state have historically been a major source of revenue available to VTrans (over 30 percent). Revenues from the purchase and use tax and motor vehicle fees are equally substantial. Figure 9-1 shows a plot of the trend of the relative contribution of the various sources over a 22-year period as a percentage of the aggregate of the Transportation and TIB Fund revenues (i.e., the total revenue available to VTrans). It shows that revenues from taxes and assessments on gasoline and diesel have been the dominant source of revenue to the Transportation and TIB Funds throughout the 22-year period. These revenues (and those from the purchase and use tax and motor vehicle fees) have fluctuated annually. The aggregate of the gasoline and diesel revenues has been declining since 2014. Another source of revenue called “other taxes” was S�����: Computed from data provided by VTrans. Revenue Source 2021 Average 2000–2021 Gasoline + Diesel 33% 38% Purchase and Use 30% 26% Motor Vehicle Fees 30% 28% Other Revenue 7% 8% Table 9-4. FY 2021 and historical (22-year) average contribution of various sources of revenue to the Vermont Transportation and TIB Funds (%). Revenue Source 2021 Average, 2000–2021 Gasoline 24% 28% Diesel 6% 7% Purchase and Use 32% 27% Motor Vehicle Fees 31% 30% Other Revenue 7% 8% Total of Gasoline and Diesel from Fiscal Facts Report 30% 35% S�����: Computed from data provided by VTrans. Table 9-3. FY 2021 and historical (22-year) average contribution of various sources of revenue to the Vermont Transportation Fund (%).

Pilot Testing and Results 41   applicable between FY 2000 and FY 2002 and is no longer in force. Therefore, this source was not included in Studies 1 and 2. Steps 2 and 3: Evaluate Data and Obtain SME Feedback To better understand the reasons for the annual fluctuations illustrated in Figure 9-1, the research team further evaluated the data (Step 2). The evaluation looked at the fluctuations in the annual revenues and the corresponding annual rates of change. Following the evaluation, discussions were held with VTrans and State of Vermont officials (Step 3). The objective of this evaluation and the follow-up discussions was to better understand the factors that drove the observed trends and to recognize outliers. Based on the feedback received, influences of “con- trollable” factors, such as legislative actions, were identified and separated from uncontrollable factors. Based on this understanding of controllable factors, suitable adjustments were made to optimally represent the data as reflecting uncertainty from uncontrollable factors as further described here and in Appendix E. The adjusted data were then used in preparing the forecasts. The logic followed is that once these forecasts are prepared, agencies can separately incorporate expectations of controllable factors into the forecasts. The approach used for evaluating the gasoline revenue data is as follows. Descriptions of data evaluation for all the other state revenue sources are also included in Appendix E. Figure 9-2 shows the historical revenue to the Transportation Fund from state taxes and assess- ments on gasoline sales. Revenue increased substantially in FY 2005 and FY 2014 (correspond- ing to tax increases implemented in those years). These increases were followed by periods of gradual but steady decline. To further understand these observations, the research team com- puted annual rates of change of gasoline revenues to the Transportation Fund. These are plotted in Figure 9-3. An evaluation of the annual rates of change shown in Figure 9-3 confirms the correlation between the revenue increases and increases in gasoline taxes and assessments. The evaluation also shows the declines observed in the years following the tax increases (illustrated by the trend line shown as a dotted line). To isolate the impact of the one-off tax increases, the black line in Figure 9-3 uses an inter- polated annual rate of change during the years of tax increase. The assumption is that if the tax increases did not take place in those years, the annual rates of change for those years would Figure 9-1. Trend in relative contribution of various sources of revenue to the Vermont Transportation and TIB Funds (%). Source: Computed from data provided by VTrans.

42 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 30% 25% 20% 15% 10% 5% 0% –5% –10% –15% Figure 9-3. Annual rate of change (%) of gasoline revenues to the Vermont Transportation Fund, FY 2001 to FY 2021. Source: Based on data available from Vermont Legislative Joint Fiscal Office. R2 = 0.6788 Gasoline Revenues ($, millions) to Transportation Fund Linear (Gasoline Revenues [$, millions] to Transportation Fund) Figure 9-2. Gasoline revenues to the Vermont Transportation Fund ($, millions). Source: Based on data available from Vermont Legislative Joint Fiscal Office.

Pilot Testing and Results 43   follow the trends consistent with those observed in the preceding and succeeding years. The annual rates of change in FY 2005 and FY 2014 were computed as the average of the rates of change in the corresponding preceding and succeeding years. Figures 9-2 and 9-3 also show a substantial decline in revenues and the annual rate of change in FY 2020. These declines corresponded with the lockdowns and travel restrictions implemented to control the pandemic spread, the related decrease in gasoline sales, and the resulting decline in tax revenues. Figure 9-3 indicates a partial recovery in the annual rate of change in FY 2021, however, this did not result in a corresponding increase in overall revenue from gasoline taxes (Figure 9-2). The revenues from gasoline and diesel taxes and assessments are dependent on the volume of gasoline and diesel sales. The State of Vermont tracks monthly gasoline and sales volumes, and these data were also available, along with the gasoline and diesel revenue data. Because the rev- enue increases observed in 2005 and 2014 were a result of legislative actions and thus “influenced,” to isolate the trends from such external influencing factors under the control of the state (Step 4), gasoline sales volume data (instead of revenues) were evaluated. Historical annual gasoline sales volume between FY 1994 and FY 2021 are shown in Figure 9-4. Figure 9-5 shows the rolling 12-month sales volumes for the same period. Gasoline sales data (Figure  9-4) do indeed show a steady but gradual decline beginning in FY 2006 followed by a sharp decline in 2020. The FY 2020 decline can be attributed to the impacts of COVID-19. The total volume of gasoline sales for FY 2021 was approximately 276 million gallons. Rolling 12-month sales data from February 2022 indicate an increase from 276 million gallons in June 2021 to approximately 289 million gallons in February 2022 (Figure 9-5). While data are insufficient to form conclusions about the recovery of gasoline sales volumes to pre-pandemic (December 2019) levels of approximately 315 million gallons, the data indicate that annual gasoline sales volumes have recovered partially. These observations make intuitive sense and likely indicate a steady decline in gasoline sales between FY 2005 and late 2019 (pre-pandemic), consistent with the increased fuel efficiency of automobiles and the increased proliferation of hybrid-electric and electric vehicles. The decline in the second half of FY 2020 and the first half of FY 2021 corresponds with the sharp decline Figure 9-4. Gasoline sales in Vermont (gal, millions), FY 1994 to FY 2021. Source: Based on data available from Vermont Legislative Joint Fiscal Office.

44 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research in motoring traffic caused by COVID-induced travel restrictions and increases in work from home practices. The partial recovery in the second half of FY 2021 through February 2022 could indicate that although travel by the public has recently increased after the severe drop from pre- pandemic levels, it has not fully recovered. This indicates that the extent of recovery of gasoline sales and any continuing impacts (such as from continued adoption of work from home prac- tices) remain uncertain. While the FY 2020 to FY 2021 drop in sales may be an outlier, practi- tioners would do well to keep these observations in mind while forecasting to decide on the appropriate treatment to consider. For forecasting gasoline revenues, it was assumed (Step 4) for the pilot testing that the reduction observed in FY 2020 is an outlier and that eventually, gasoline sales would recover to some degree. However, it was assumed that the long-term trend (gradual decline) in gasoline sales volumes that were observed pre-pandemic (FY 2005 to FY 2019) would continue. For the deterministic method, the historical CAGR was used to forecast the gasoline sales volumes. For the ETS and Monte Carlo methods, gasoline sales volume was forecasted from FY 2019 to FY 2032 (excluding the pandemic impact) and the corresponding forecasted annual rates of change were computed. The rates of change computed for FY 2022 to FY 2032 were then applied to the actual volume observed in FY 2021 (after accounting for a partial recovery) to obtain forecasted gasoline sales volumes through FY 2032. The extent of the partial recovery in gasoline sales volumes was estimated as follows: • Pre-pandemic volume: 315 million gallons (FY 2019) • Forecast value as of FY 2021 excluding pandemic impact: 312 million gallons (assuming the historical trend in gasoline sales continues) • Actual gasoline sales as of FY 2021: 276 million gallons • Estimated reduction attributable to the pandemic: 315 − 276 = 39 million gallons • Actual sales as of November 2021: 283 million gallons 240 260 280 300 320 340 360 380 400 Ju n- 94 Ju n- 95 Ju n- 96 Ju n- 97 Ju n- 98 Ju n- 99 Ju n- 00 Ju n- 01 Ju n- 02 Ju n- 03 Ju n- 04 Ju n- 05 Ju n- 06 Ju n- 07 Ju n- 08 Ju n- 09 Ju n- 10 Ju n- 11 Ju n- 12 Ju n- 13 Ju n- 14 Ju n- 15 Ju n- 16 Ju n- 17 Ju n- 18 Ju n- 19 Ju n- 20 Ju n- 21 G as ol in e Sa le s Figure 9-5. Rolling 12-month gasoline sales (gal, millions) in Vermont, FY 1994 to FY 2021. Source: Vermont Legislative Joint Fiscal Office, Transportation, https://ljfo.vermont.gov/subjects/transportation.

Pilot Testing and Results 45   • Estimated recovery as of November 2021: 283 − 276 = 7 million gallons • Percent recovery: 7 ÷ 39 = 18 percent • Assumed recovery as of FY 2022 for pilot testing: 25 percent (increased from 18 percent com- puted as of November 2021 to an estimated 25 percent as of June 2022 to account for potential continued recovery by fiscal year end) The forecasted gasoline sales volumes were computed for FY 2022 to FY 2032 after assuming that the FY 2022 sales volume would show a partial recovery of 25 percent from the FY 2021 volume. Corresponding revenues were computed assuming that current gasoline taxes and assessments would continue for the forecast period. This methodology was applied to the forecasts prepared using each tool selected. The evaluation and assumptions (Steps 2–4) used for Studies 1 and 2 are illustrative. Before agencies prepare forecasts, it would be important for them to evaluate their data and consult with SMEs to arrive at the assumptions and parameters best suited to their data. Similar evaluations were performed for data relating to diesel sales, purchase and use tax, motor vehicle fees, and other revenues. When no clear explanations for observed data trends were available, discussions were held, or feedback was obtained via e-mail from both VTrans and State of Vermont Fiscal Office personnel (Step 3) to obtain a better understanding of potential influencing factors. This information was used to establish a data set that was representative of unknown risk factors and excluded impacts such as those attributable to legislative actions. The approach adopted was to create forecasts that accounted for unforeseen and uncontrol- lable risk factors. If appropriate, decision makers could then incorporate additional impacts, such as those caused by anticipated legislative actions, based on inputs from SMEs, the legislature, and other experts for arriving at final revenue forecasts. Key observations from the data evaluation for the other revenue sources are included in Appendix E. Step 4: Establish Data Set for Forecasting Based on the description for gasoline sales and for other components of the state revenues (Appendix E), the data set consisting of annual rates of change used for forecasting was prepared. This is summarized in Table 9-5. 9.1.4 Outputs from the Pilot The quantitative outputs from the pilots of Studies 1 and 2 consist of forecasted values for each revenue source for each year of the forecast period and the upper and lower limits of confidence intervals to illustrate the uncertainty in the forecasted values. A confidence interval shows the likely range of values associated with a statistical parameter of the data, such as the population mean. The confidence interval is usually determined by a multiplier of the standard deviation in a normally distributed set of forecast values. The 95 percent confidence interval of the predicted mean in such a distribution is represented as Confidence interval = mean of predicted values ± 1.96 × standard deviation, where the multiplier 1.96 corresponds to a 95 percent confidence interval. In the case of a deterministic forecast, a single output value is estimated for each year of the fore- cast period. Because multiple forecast values are not computed for each year, the computations

46 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research do not provide the data to establish confidence intervals to account for uncertainty in forecasts. This is a key shortcoming of deterministic forecasts. In such cases, to account for uncertainty in forecasts, agencies may consider estimating confidence intervals in consultation with SMEs or based on best judgment. For the pilot, it was assumed that the standard deviation of the historical annual rate of change in gasoline sales would be applicable for the forecasted values. This value was used in the equation to determine the upper and lower bounds around the forecast value. Other approaches may include computing rolling multi-year rates of change to identify upper and lower limits to illustrate uncertainty. The confidence interval is established based on an assumed normal distribution of the data (for ETS forecasts using Excel) and the selected probability distribution (for the Monte Carlo forecasts). For simplicity, based on a preliminary analysis of the available data, normal distribu- tions were assumed for the Monte Carlo simulations performed during this pilot effort. In the probabilistic forecasts, the confidence intervals were based on a quantitative analysis of the data, in contrast to the subjective approach described for deterministic methods. Gasoline Sales Diesel Sales Purchase and Use Motor Vehicle Fees Other Revenue Date Values Date Values Date Values Date Values Date Values Jun-94 Jun-95 −2.6% Jun-96 2.4% Jun-97 −0.5% Jun-98 1.1% Jun-99 2.8% Jun-00 1.2% Jun-00 Jun-00 Jun-00 Jun-01 4.1% Jun-01 Jun-01 5.0% Jun-01 −5.1% Jun-01 0.8% Jun-02 −2.3% Jun-02 −9.9% Jun-02 8.7% Jun-02 1.6% Jun-02 16.5% Jun-03 3.9% Jun-03 −0.5% Jun-03 3.9% Jun-03 8.4% Jun-03 21.3% Jun-04 1.4% Jun-04 11.2% Jun-04 4.7% Jun-04 7.0% Jun-04 −11.1% Jun-05 −0.1% Jun-05 −12.2% Jun-05 0.5% Jun-05 4.6% Jun-05 4.6% Jun-06 −3.4% Jun-06 11.4% Jun-06 −3.8% Jun-06 2.3% Jun-06 7.5% Jun-07 0.0% Jun-07 −3.5% Jun-07 −0.4% Jun-07 2.8% Jun-07 18.1% Jun-08 −2.3% Jun-08 2.6% Jun-08 −1.9% Jun-08 3.2% Jun-08 17.3% Jun-09 −2.0% Jun-09 −15.8% Jun-09 −16.5% Jun-09 −3.0% Jun-09 −24.1% Jun-10 −0.1% Jun-10 3.3% Jun-10 5.7% Jun-10 10.7% Jun-10 1.1% Jun-11 −1.3% Jun-11 3.8% Jun-11 10.5% Jun-11 −0.3% Jun-11 −1.6% Jun-12 −0.7% Jun-12 −3.7% Jun-12 6.2% Jun-12 1.7% Jun-12 2.2% Jun-13 −3.0% Jun-13 6.7% Jun-13 2.0% Jun-13 6.0% Jun-13 4.4% Jun-14 0.1% Jun-14 −1.3% Jun-14 9.9% Jun-14 1.4% Jun-14 2.1% Jun-15 −0.3% Jun-15 0.9% Jun-15 5.9% Jun-15 1.4% Jun-15 1.0% Jun-16 −1.6% Jun-16 8.8% Jun-16 3.1% Jun-16 2.4% Jun-16 −0.5% Jun-17 1.3% Jun-17 −8.6% Jun-17 3.0% Jun-17 5.1% Jun-17 1.5% Jun-18 0.0% Jun-18 4.4% Jun-18 6.1% Jun-18 −0.2% Jun-18 15.6% Jun-19 −0.4% Jun-19 −2.0% Jun-19 2.1% Jun-19 −0.7% Jun-19 7.0% Jun-20 −9.8% Jun-20 −4.6% Jun-20 −5.6% Jun-20 −2.1% Jun-20 −13.4% Jun-21 −2.8% Jun-21 3.8% Jun-21 2.4% Jun-21 1.0% Jun-21 −3.8% N����: One-off impacts to gasoline sales observed in 2020 are ignored. One-off impacts observed in purchase and use tax in 2005, 2009, and 2021 are ignored. Impacts on motor vehicle fees attributable to legislative actions in 2002 and 2007 are ignored. Where data were ignored, annual rates of change were interpolated from previous and succeeding years or, where succeeding year data were unavailable, an average of several previous years. For diesel sales and other revenues, fluctuations are generally symmetric around the trend line and the overall impact is nonmaterial, hence, no adjustments are made. Table 9-5. Annual rates of change (%) in sources of revenue to the Vermont Transportation and TIB Funds, excluding one-off impacts and legislative actions, used in the pilot study based on the assumptions described.

Pilot Testing and Results 47   In the following discussion, study outputs are described for gasoline tax revenues in the fol- lowing order: • Deterministic • Triple exponential smoothing using Excel • Monte Carlo Study outputs for the remaining sources of revenue for the Transportation Fund and gasoline and diesel assessments for the TIB Fund are included in Appendix E. In each case, the forecast values are shown along with the confidence intervals, followed by a brief discussion highlighting the interpretive component of the study outputs. Finally, the total revenues forecasted for the Transportation Fund, the TIB Fund, and an aggre- gate of the two are presented along with corresponding confidence intervals. Steps 5a–5c: Calculate Deterministic Forecasts for Gasoline Tax Revenues Key computations and assumptions used for the forecast are as follows: • Gasoline sales volume CAGR = −0.36 percent (Step 5a) • Post-COVID gasoline sales recovery in FY 2022 = 25 percent of the loss from pre-COVID volumes (assumed in Step 4) The forecasted values are computed (Step 5b) based on the assumption that gasoline sales will change in each year of the forecast period at a rate of −0.36 percent, the computed value of the compound annual growth rate of the historical gasoline sales volumes. The plot showing histori- cal and forecast revenues is displayed in Figure 9-6. Figure 9-6 shows the deterministic forecasts of revenues from gasoline sales to the Transporta- tion Fund for the forecast period. It can be seen that the forecasts do not provide an estimate of the variability of future values and thus do not incorporate uncertainty into the forecast. If deterministic forecasts are used, agency decision makers would need to incorporate estimates of variability in forecasts to comply with the requirement that TAMPs be risk based. Because quantitative values of such variability are not inherently computed in deterministic forecasts, Assumes partial (25%) recovery of COVID impacts by FY 2022, based on continued lower gasoline sales due to sustained residual COVID impacts Figure 9-6. Deterministic forecast of gasoline sales tax revenues ($, millions) to Vermont Transportation Fund for forecast period.

48 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research the estimates of uncertainty will likely be based on a review of historical data and application of best judgment based on inputs from SMEs (Step 5c). For the pilot testing, the standard deviation computed for the historical rate of change in the volume of gasoline sales (second column in Table 9-5) was used to establish an upper and lower bound of the confidence interval. As shown in Table 9-6, the upper and lower bounds of the con- fidence interval are taken to be 3.30 percent above or below the forecasted value computed using the CAGR. This corresponds to a fixed change of 2.94 percent from the previous year’s value for the upper bound and a −3.66 percent change from the previous year’s value for the lower bound. Table 9-7 shows the computed lower and upper bounds of the 95 percent confidence intervals for gasoline sales volumes for each year of the forecast period. As detailed in the Steps 2 and 3 sec- tion, the recovery of gasoline sales to pre-pandemic levels by FY 2022 was assumed as 25 percent from the FY 2021 values. Based on this recovery assumption, the forecasted gasoline sales volume for 2022 is 284 million gallons, which is consistent with observations at the end of November 2022. A change in the recovery assumption from 25 percent would yield different values of fore- casts along with the confidence intervals. Mean Standard Deviation 1.96 × SD −0.36% 1.69% 3.30% Max. Mean + 1.96 × SD 2.94% Min. Mean − 1.96 × SD −3.66% N����: Data are normally distributed. Confidence interval is based on statistical parameters of historical data. Table 9-6. Computation of the upper and lower bound of 95% confidence interval based on statistical parameters of historical data. Fiscal Year Forecast Lower Confidence Value Upper Confidence Value 2021 276.05 276.05 276.05 2022 284.11 274.70 293.53 2023 283.09 273.71 292.48 2024 282.07 272.72 291.43 2025 281.06 271.74 290.38 2026 280.05 270.77 289.33 2027 279.04 269.79 288.29 2028 278.04 268.82 287.25 2029 277.04 267.85 286.22 2030 276.04 266.89 285.19 2031 275.05 265.93 284.17 2032 274.06 264.97 283.14 N���: Assumes partial (25%) recovery from COVID-19 impact by FY 2022, based on continued lower gasoline sales caused by sustained residual COVID-19 impacts. Table 9-7. Projected gasoline sales volumes (gal, millions) for Vermont, calculated using deterministic techniques for the forecast period, along with confidence interval.

Pilot Testing and Results 49   Agencies may consider the approach and the related assumptions after careful consideration and discussions with SMEs. Table 9-8 lists the forecasted gasoline tax revenue to Vermont’s Transportation Fund for FY 2022 to FY 2032, assuming a constant annual change in the gasoline sales volume of −0.36 percent and no change in the revenue basis. The forecasted revenues are shown as declining from approximately $70 million in FY 2022 to approximately $68 million in FY 2032. Based on the computation for the confidence interval (Table 9-6), the interval for FY 2032 is between approximately $66 million and $70 million, that is, 3.30 percent above and below the forecast value of approximately $68 million. The corresponding plot of the historical and forecast values is shown in Figure 9-7. Confidence intervals estimated for the pilot were based on certain subjective assumptions made about the data. The only purpose of doing so was to Table 9-8. Historic and projected gasoline sales tax revenues ($, millions) to Vermont Transportation Fund using deterministic techniques for the forecast period and assuming partial (25%) recovery to pre-COVID levels. Fiscal Year Historical Revenue Forecast Value Lower Confidence Value Upper Confidence Value 2000 $51.80 2001 $52.50 2002 $52.60 2003 $54.00 2004 $54.30 2005 $65.50 2006 $63.80 2007 $63.60 2008 $62.60 2009 $60.60 2010 $61.00 2011 $60.60 2012 $59.30 2013 $59.90 2014 $76.50 2015 $77.80 2016 $78.00 2017 $78.20 2018 $78.20 2019 $77.80 2020 $71.00 2021 $67.30 $67.30 $67.30 $67.30 2022 $70.30 $67.97 $72.63 2023 $70.05 $67.73 $72.37 2024 $69.80 $67.49 $72.11 2025 $69.55 $67.24 $71.85 2026 $69.30 $67.00 $71.60 2027 $69.05 $66.76 $71.34 2028 $68.80 $66.52 $71.08 2029 $68.55 $66.28 $70.83 2030 $68.31 $66.04 $70.57 2031 $68.06 $65.80 $70.32 2032 $67.82 $65.57 $70.06

50 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research illustrate an approach to incorporate uncertainty, which is not inherently present in determin- istic forecasting techniques. Step 6a: Calculate Probabilistic Forecasts for Gasoline Sales Tax Revenues Using Triple Exponential Smoothing ETS forecasting options are readily available within the Forecast module under the Data tab in Excel spreadsheets. As stated previously, historical gasoline sales data through FY 2019 were used as input data, and the forecast function in Excel was used to arrive at forecasted values for FY 2020 through FY 2032. The projected annual rates of change based on the ETS forecasting results were then computed for FY 2020 through FY 2032. The forecasted rate of change for FY 2022 was applied to the FY 2021 data to compute the forecasted value of gasoline sales volume for FY 2022 and further increased by 25 percent as described in Steps 2 and 3 to account for the estimated partial recovery. Forecasted rates of change were then applied to compute the fore- casted gasoline sales volumes through 2032. Using the current tax, the corresponding revenues were computed for the forecast period. The FORECAST.ETS function in Excel also produces a 95 percent confidence interval of the forecasted values, assuming a normal distribution for the forecast. Excel automatically detects and optimizes weightage factors for past values, trend, and seasonality. Tables 9-9 and 9-10 pro- vide the results of the analysis and Figure 9-8 presents a plot of the forecasted gasoline revenues along with the 95 percent confidence intervals estimated by the Excel FORECAST.ETS function. The forecast for FY 2022 is approximately 283 million gallons, which is consistent with recent observations as of November 2021. It is only slightly lower than the deterministic forecast of Assumes partial (25%) recovery of COVID impacts by FY 2022, assuming continued lower gasoline sales due to sustained residual COVID impacts Figure 9-7. Deterministic forecast of gasoline sales tax revenues ($, millions) to Vermont Transportation Fund for forecast period, along with estimated confidence interval.

Pilot Testing and Results 51   284 million gallons for FY 2022. However, the confidence interval delivered by the Excel func- tion diverges substantially from the forecast value in the outer years. Otherwise, the forecasted values and the lower and upper bounds are similar to those forecasted using a simple determin- istic technique. The probabilistic forecast value varies from 283 million gallons in FY 2022 to 267 million gallons in FY 2032, with the confidence interval in FY 2032 ranging from approxi- mately 205 million gallons to 329 million gallons. As shown in Table 9-10, the forecasted revenue decreases from about $70 million in FY 2022 to about $66 million in 2032, with the confidence interval ranging from $51 million to $81 mil- lion. Figure 9-8 illustrates these values. Consistent with previous trends, the forecasted gaso- line revenues continue to decline gradually over the forecast period. The 95 percent confidence interval also diverges over the forecast period, ranging from about 7 percent above and below the forecast value in FY 2022 to approximately 23 percent above and below the forecast value in FY 2032 (Table 9-11). Step 6b: Calculate Probabilistic Forecasts for Gasoline Sales Tax Revenues using Monte Carlo Simulation The historical annual growth rates for gasoline sales volumes for FY 1995 through FY 2019, as shown in Table 9-5, were used as input data for the Monte Carlo simulations. Because the dis- tribution of these data reasonably approximated a normal distribution, the mean and standard deviation for the historical data were accordingly computed (Step 6b(i)). These were the key input parameters for the simulation. A total of 500 iterations was performed using randomly generated probabilities for each sample run for each year of the forecast period (Step 6b(ii)). The mean, standard deviation, and a 95 percent confidence interval for the computed mean of each sample run for each year were computed (Steps 6b(iii) and 6b(iv)). The mean value of 500 iterations for a given year of the forecast period was used as the forecast value for that year. An example of the probability distribution for the 500 iterations for the rate of change of gasoline sales applicable to FY 2032 is shown in Figure 9-9. Fiscal Year Forecast Lower Confidence Value Upper Confidence Value 2021 276.05 276.05 276.05 2022 283.21 263.10 303.31 2023 281.59 257.29 305.90 2024 279.97 251.57 308.38 2025 278.36 245.88 310.84 2026 276.74 240.18 313.30 2027 275.13 234.45 315.81 2028 273.51 228.67 318.36 2029 271.90 222.82 320.97 2030 270.28 216.92 323.64 2031 268.66 210.94 326.39 2032 267.05 204.89 329.21 N���: Assumes partial (25%) recovery of COVID-19 impact by FY 2022, based on continued lower gasoline sales caused by sustained residual COVID-19 impacts. Table 9-9. Probabilistic forecast of gasoline sales volumes (gal, millions) in Vermont, calculated using FORECAST.ETS in Excel for the forecast period, along with confidence interval bounding values.

52 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research The rates of change and the confidence intervals computed for the 500 iterations for each year were used to compute the projected gasoline sales volume for FY 2020 and FY 2021 to estimate the forecasted FY 2021 volume, had COVID not occurred. These data were compared with actual recorded gasoline sales in FY 2021 to compute the estimated reduction in sales attributable to COVID impacts. As previously described, the pilot test assumed that gasoline sales would recover by 25 percent by the end of FY 2022, and forecasted gasoline sales were computed. N���: Assumes partial (25%) recovery from COVID-19 impact by FY 2022, based on continued lower gasoline sales caused by sustained residual COVID-19 impacts. Fiscal Year Historical Revenue Forecast Revenue Lower Confidence Value Upper Confidence Value 2000 $51.80 2001 $52.50 2002 $52.60 2003 $54.00 2004 $54.30 2005 $65.50 2006 $63.80 2007 $63.60 2008 $62.60 2009 $60.60 2010 $61.00 2011 $60.60 2012 $59.30 2013 $59.90 2014 $76.50 2015 $77.80 2016 $78.00 2017 $78.20 2018 $78.20 2019 $77.80 2020 $71.00 2021 $67.30 $67.30 $67.30 $67.30 2022 $70.08 $65.10 $75.05 2023 $69.68 $63.67 $75.69 2024 $69.28 $62.25 $76.31 2025 $68.88 $60.84 $76.92 2026 $68.48 $59.43 $77.53 2027 $68.08 $58.01 $78.15 2028 $67.68 $56.58 $78.78 2029 $67.28 $55.14 $79.42 2030 $66.88 $53.68 $80.09 2031 $66.48 $52.20 $80.76 2032 $66.08 $50.70 $81.46 Table 9-10. Historic and projected gasoline sales tax revenues ($, millions) to Vermont Transportation Fund for forecast period, along with confidence interval, produced using FORECAST.ETS in Excel.

Pilot Testing and Results 53   Projected gasoline sales for the remainder of the forecast period were then computed using the forecasted annual rates of change. The results are shown in Table 9-12. The FY 2021 sales volume in Table 9-12 is the actual recorded sales volume. The remaining values were computed from the forecasted rates of change. The Monte Carlo method forecasts a slight increase in gasoline sales for FY 2022 (approximately 285 million gallons) over the previ- ous two methods. While the values do show a declining trend over the forecast period, the annual rate of change is lower than the deterministic and ETS methods. The forecast value for FY 2032 is approximately 283 million gallons. The corresponding projected revenues and the 95 percent confidence intervals are shown in Table 9-13 and plotted in Figure 9-10. Table 9-13 shows that, consistent with an anticipated (and observed) recovery in gasoline sales volume in FY 2022, the forecasted revenue goes up to approximately $71 million in FY 2022. Assumes partial (25%) recovery of COVID impacts by 2022, allowing for continued lower gasoline sales due to sustained residual COVID impacts Figure 9-8. Probabilistic forecast of gasoline sales tax revenues ($, millions) to Vermont Transportation Fund for forecast period, along with confidence interval, produced using FORECAST.ETS in Excel. Fiscal Year Lower Confidence Value Upper Confidence Value 2022 −7.1% 7.1% 2023 −8.6% 8.6% 2024 −10.1% 10.1% 2025 −11.7% 11.7% 2026 −13.2% 13.2% 2027 −14.8% 14.8% 2028 −16.4% 16.4% 2029 −18.0% 18.0% 2030 −19.7% 19.7% 2031 −21.5% 21.5% 2032 −23.3% 23.3% Table 9-11. Probabilistic forecast of upper and lower bounds of confidence interval band for Vermont gasoline sales tax revenues (% of forecasted revenue), by fiscal year.

54 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research –1 2% –1 1% –1 0% –9 % –8 % –7 % –6 % –5 % –4 % –3 % –2 % –1 % 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10 % 11 % 12 % Figure 9-9. Monte Carlo forecast of distribution of annual rate of change (%) of Vermont gasoline sales in FY 2032, computed over 500 iterations. N���: Assumes partial (25%) recovery of COVID-19 impact by FY 2022, based on continued lower gasoline sales caused by sustained residual COVID-19 impacts. Fiscal Year Forecast Lower Confidence Value Upper Confidence Value 2021 276.05 276.05 276.05 2022 285.21 274.54 296.94 2023 284.72 273.61 295.47 2024 284.68 274.18 295.99 2025 284.46 274.14 296.13 2026 284.22 272.91 294.84 2027 284.27 273.87 294.67 2028 284.16 273.33 294.60 2029 283.75 271.84 294.69 2030 283.61 272.11 294.87 2031 283.27 271.54 293.42 2032 282.59 270.89 294.14 Table 9-12. Probabilistic forecast of gasoline sales volumes (gal, millions) for Vermont, calculated using Monte Carlo simulation for the forecast period, along with confidence interval bounding values assuming partial recovery to pre-COVID levels.

Pilot Testing and Results 55   Over the forecast period, the Monte Carlo method shows little variability, and the 2032 revenue from gasoline sales to the Transportation Fund is forecasted at approximately $70 million. The 95 percent confidence interval ranges between approximately $67 million and $73 million. A comparison of the forecasted values using the three methods (shown in Figure 9-11) indi- cates a consistent but gradual reduction in gasoline sales revenues over the forecast period. The magnitude of the annual change in gasoline sales and, thereby, the forecasted revenues varied slightly between the three methods, with the Monte Carlo method forecasting an approximately 0.1 percent average annual decline, the deterministic method forecasting a 0.36 percent average N���: Assumes partial (25%) recovery from COVID-19 impact by FY 2022, based on continued lower gasoline sales caused by sustained residual COVID-19 impacts. Fiscal Year Historical Revenues Forecast Value Lower Confidence Value Upper Confidence Value 2000 $51.80 2001 $52.50 2002 $52.60 2003 $54.00 2004 $54.30 2005 $65.50 2006 $63.80 2007 $63.60 2008 $62.60 2009 $60.60 2010 $61.00 2011 $60.60 2012 $59.30 2013 $59.90 2014 $76.50 2015 $77.80 2016 $78.00 2017 $78.20 2018 $78.20 2019 $77.80 2020 $71.00 2021 $67.30 $67.30 $67.30 $67.30 2022 $70.58 $67.93 $73.48 2023 $70.45 $67.70 $73.11 2024 $70.44 $67.85 $73.24 2025 $70.39 $67.84 $73.28 2026 $70.33 $67.53 $72.96 2027 $70.34 $67.77 $72.92 2028 $70.31 $67.63 $72.90 2029 $70.21 $67.27 $72.92 2030 $70.18 $67.33 $72.96 2031 $70.09 $67.19 $72.61 2032 $69.93 $67.03 $72.78 Table 9-13. Historic and projected gasoline sales tax revenues ($, millions) to Transportation Fund for forecast period, along with confidence interval, produced using Monte Carlo simulation.

56 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Assumes partial (25%) recovery of COVID impacts by 2022, allowing for continued lower gasoline sales due to sustained residual COVID impacts Figure 9-10. Probabilistic forecast of gasoline sales tax revenues ($, millions) to Vermont Transportation Fund for forecast period, along with confidence interval, produced using Monte Carlo simulation. Assumes partial (25%) recovery of COVID impacts by FY 2022, based on continued lower gasoline sales due to sustained residual COVID impacts $90.00 $80.00 $70.00 $60.00 $50.00 $40.00 $30.00 $20.00 Monte Carlo Upper BoundMonte Carlo Lower Bound Monte Carlo ForecastETS Forecast ETS Upper BoundETS Lower BoundDeterministic Upper Bound Actual Values Deterministic Forecast Deterministic Lower Bound Figure 9-11. Comparison of forecasts of gasoline sales tax revenues ($, millions) with Vermont Transportation Fund for forecast period, along with confidence intervals, produced using deterministic and probabilistic methods.

Pilot Testing and Results 57   annual decline, and the ETS method forecasting an approximately 0.6 percent average annual decline. The confidence interval for the Monte Carlo method was narrow at approximately 3.8 percent around the forecast values, whereas for the ETS method, it varied between approxi- mately 7 percent and 23 percent. For illustrative purposes, an approximately 3.3 percent band around the forecast values was assumed as a confidence interval, though no confidence interval was computed for the deterministic method. The wider confidence interval computed by the ETS method reflects a higher degree of uncertainty in the outer years as compared with the Monte Carlo method. As can be seen from the description of the quantitative outputs from this study, the determin- istic forecast (as illustrated using the historical CAGR for gasoline sales) does not inherently compute potential variability in the forecasted data to account for this uncertainty. To incorpo- rate risk into the forecasts, a somewhat subjective approach that involves discussions with SMEs and application of best judgment becomes necessary. On the other hand, the probabilistic meth- odologies do estimate the inherent uncertainty in the forecast values based on the computations to provide an estimate of the confidence intervals. For the remaining components of the Transportation Fund and TIB Fund revenue forecasts, a similar approach to that for the gasoline revenues was used and estimates of forecasts and related uncertainties (confidence intervals) were prepared. The computations are not discussed in detail in this report, but plots of the output data are provided in Appendix E, along with a summary description. Finally, estimated aggregate forecasted revenues for both funds, computed using the different methodologies, are also shown in Appendix E. Step 7: Repeat Steps 2–6 for Each Parameter Being Forecast and Combine for Total VTrans Funding (Transportation Fund  TIB Fund) The results from the forecasting described here and in Appendix E for the two funds are aggregated to display the total forecasted revenues available to VTrans for the forecast period for each method tested. The results for the deterministic forecast are provided in Table 9-14 and Fig- ure 9-12. The results of the ETS forecast are provided in Table 9-15 and Figure 9-13. The results of the Monte Carlo forecast are provided in Table 9-16 and Figure 9-14. These results are shown to illustrate how the different methods can be used to forecast state revenues and estimate the associated uncertainty in those forecasts. Doing so allows state agencies to incorporate risk into forecasts and to estimate a potential magnitude in the shortfall (or excess) of funds available to meet their asset management needs. Understanding the uncertainty can be valuable in making trade-off decisions and in communicating with stakeholders to set performance expectations. Figure 9-15 displays a comparative plot of the forecasts of total revenues available to VTrans from state sources (the Transportation and TIB Funds) using the three methodologies tested. The chart shows that the forecasted values are generally consistent across the three methods. Although an illustrative example of a confidence interval for deterministic forecasts was pro- vided in the Step 5 discussion, the deterministic method does not provide any estimate of uncer- tainty. However, the projected 95 percent confidence interval is computed in the ETS and Monte Carlo methods. State agencies will need to establish an applicable means to forecast uncertainty through internal discussions and feedback from SMEs. Impacts from potential legislative actions will need to be separately accounted for. Table 9-17 summarizes the key findings from the forecasts. Both the deterministic and the Monte Carlo techniques forecast state revenue from all sources in FY 2022 at approximately $305 million, whereas the ETS technique forecasts approximately $310 million. For FY 2032, the Monte Carlo method forecasts $332 million, whereas the other two methods forecast $350 mil- lion. The estimate of uncertainty in the FY 2022 forecast is consistent at about plus or minus

58 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Fiscal Year Actual Value Lower Confidence Value Forecasted Value Upper Confidence Value 2000 $181.20 2001 $185.90 2002 $198.80 2003 $207.80 2004 $214.60 2005 $209.00 2006 $209.90 2007 $220.90 2008 $223.10 2009 $203.60 2010 $228.10 2011 $236.10 2012 $244.50 2013 $251.20 2014 $274.40 2015 $281.70 2016 $279.60 2017 $285.60 2018 $294.00 2019 $297.60 2020 $278.80 2021 $294.83 $294.83 $294.83 $294.83 2022 $278.23 $304.79 $331.35 2023 $281.82 $308.83 $335.83 2024 $285.51 $312.97 $340.43 2025 $289.29 $317.21 $345.13 2026 $293.16 $321.56 $349.95 2027 $297.13 $326.01 $354.89 2028 $301.19 $330.57 $359.95 2029 $305.36 $335.25 $365.14 2030 $309.62 $340.03 $370.45 2031 $313.99 $344.94 $375.88 2032 $318.47 $349.96 $381.45 Table 9-14. Deterministic forecast of historical and projected VTrans revenues ($, millions) from Transportation and TIB Funds.

Pilot Testing and Results 59   $27 million between the techniques. However, the uncertainty in the FY 2032 forecast using the Monte Carlo method is plus or minus $31 million, while that estimated by the ETS forecast is over or under $51 million. The forecasted confidence intervals indicate the higher degree of uncertainty reflected in the ETS forecasts, particularly for the later years in the forecast period. The estimates of uncertainty can allow agency decision makers to plan the implementation of their asset management programs and be prepared to make trade-off decisions if necessary. The forecasts further assume an estimated recovery in gasoline sales in FY 2022 based on observations during the first half of the fiscal year. This assumption was made for illustrative purposes and to incorporate the latest information available into the forecasts, however, state agencies would need to decide, based on their data-specific situations, whether such assumptions were warranted. The key takeaway from the pilot effort is that the forecasts are relatively consistent using each of the three methods. The ETS and Monte Carlo methods inherently provide estimates of uncer- tainty that allow agencies to incorporate risk into the forecasts. To incorporate risk into deter- ministic forecasts, agencies will need to involve SMEs and make judgment calls that they are comfortable with and can defend during interactions with stakeholders. 9.1.5 Participating DOT Organizational Unit(s) Studies 1 and 2 were conducted in conjunction with SMEs and leadership from VTrans who have knowledge about state financial data, as well as analysts from the Vermont Legislative Joint Fiscal Office who monitor annual state revenues. In general, for revenue projections, it is impor- tant to have representation from agency finance experts, and information and input from the state financial analyst or an agency representative with knowledge about state revenues and assumptions. Figure 9-12. Deterministic forecast of total VTrans revenues ($, millions) from Transportation and TIB Funds for forecast period, along with estimated confidence interval.

60 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Fiscal Year Actual Value Lower Confidence Value Forecasted Value Upper Confidence Value 2000 $181.20 2001 $185.90 2002 $198.80 2003 $207.80 2004 $214.60 2005 $209.00 2006 $209.90 2007 $220.90 2008 $223.10 2009 $203.60 2010 $228.10 2011 $236.10 2012 $244.50 2013 $251.20 2014 $274.40 2015 $281.70 2016 $279.60 2017 $285.60 2018 $294.00 2019 $297.60 2020 $278.80 2021 $294.83 $294.83 $294.83 $294.83 2022 $281.61 $310.22 $338.83 2023 $282.88 $314.09 $345.29 2024 $284.36 $317.96 $351.57 2025 $285.96 $321.86 $357.76 2026 $287.63 $325.77 $363.90 2027 $289.36 $329.68 $370.01 2028 $291.11 $333.61 $376.11 2029 $292.89 $337.55 $382.21 2030 $294.68 $341.49 $388.31 2031 $296.47 $345.44 $394.42 2032 $298.26 $349.40 $400.54 Table 9-15. Probabilistic ETS forecast of historical and projected VTrans revenues ($, millions) from Transportation and TIB Funds.

Pilot Testing and Results 61   Many DOT divisions can use the methodology detailed in this pilot to incorporate uncer- tainties associated with financial forecasting. Depending on the area of application, DOTs may engage various SMEs with the appropriate understanding of the financial data. 9.1.6 Who in a DOT Could Use the Results and How Experts from various DOT offices could use the pilot results. Different DOT offices involved in forecasting could benefit from understanding the techniques. DOT SMEs projecting funding allocations for the 10-year TAMP financial plan could use the methodology detailed in this study to make funding projections for assets included in the TAMP. DOTs following this methodology are advised to detail the assumptions made in the projec- tions. If circumstances change in the future, the amounts projected will change. For example, if a DOT receives higher or lower state or federal revenues, then the assumptions will change, and the forecasts will have to be revised to reflect the changing circumstances. Under normal circumstances, most DOTs have clarity on the next 2 years of revenue avail- ability. Then, based on the prevailing financial environment, various assumptions are made, and longer-term funding allocations are forecasted. These forecasts serve as input for allocating funds for preservation, maintenance, rehabilitation, and replacement of assets in the 10-year TAMP financial plans. The results of such analyses can be used to estimate whether a DOT will have the funds needed to achieve its asset targets or if projected funding allocations will be inadequate to manage the assets. The same methodology can be applied to forecast whether the needed funding will be available to achieve a state’s desired state of good repair. Actual Values Upper Bound Projected Lower Bound Projected Projected Values 2% Moving Average (Projected Values) Figure 9-13. Probabilistic ETS forecast of total VTrans revenues ($, millions) from Transportation and TIB Funds for forecast period, along with 95% confidence interval.

62 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Fiscal Year Actual Value Lower Confidence Value Forecasted Value Upper Confidence Value 2000 $138.25 2001 $185.90 2002 $198.80 2003 $207.80 2004 $214.60 2005 $209.00 2006 $209.90 2007 $220.90 2008 $223.10 2009 $203.60 2010 $228.10 2011 $236.10 2012 $244.50 2013 $251.20 2014 $274.40 2015 $281.70 2016 $279.60 2017 $285.60 2018 $294.00 2019 $297.60 2020 $278.80 2021 $294.83 $294.83 $294.83 $294.83 2022 $278.37 $305.06 $330.92 2023 $282.90 $309.32 $336.19 2024 $285.77 $313.70 $340.21 2025 $290.04 $318.03 $345.23 2026 $295.29 $322.45 $349.38 2027 $298.18 $327.24 $354.67 2028 $302.93 $332.30 $360.75 2029 $306.98 $337.21 $368.24 2030 $310.27 $342.39 $372.99 2031 $317.46 $347.41 $377.97 2032 $319.22 $352.04 $382.14 Table 9-16. Probabilistic Monte Carlo forecast of historical and projected VTrans revenues ($, millions) from Transportation and TIB Funds.

Pilot Testing and Results 63   Figure 9-14. Probabilistic Monte Carlo forecast of total VTrans revenues ($, millions) from Transportation and TIB Funds for forecast period, along with 95% confidence interval. $150.00 $200.00 $250.00 $300.00 $350.00 $400.00 $450.00 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 20 15 20 16 20 17 20 18 20 19 20 20 20 21 20 22 20 23 20 24 20 25 20 26 20 27 20 28 20 29 20 30 20 31 20 32 Actual Values Deterministic Forecast ETS Forecast Monte Carlo Forecast ETS Lower Bound ETS Upper Bound Monte Carlo Lower Bound Monte Carlo Upper Bound Figure 9-15. Comparison of deterministic and probabilistic forecasts of total VTrans revenue ($, millions) from Transportation and TIB Funds for forecast period, along with 95% confidence interval for probabilistic forecasts.

64 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research The methodologies can be applied to forecast costs of essential commodities, such as steel, asphalt, and concrete, that impact infrastructure projects planned for the 10-year TAMP. Such estimates can then be used to forecast future costs. 9.1.7 Challenges in Pilot Study Setup This pilot required historical data and clarity on the current state environment to help with the assumptions. The setup was not difficult because VTrans and the state legislative fiscal office provided the necessary data, responded promptly to questions, and clarified assumptions. The main expertise needed is to understand the forecasting methodology, have some statistical background, understand the impact of outliers, and make assumptions accordingly. 9.1.8 Resources Needed by a DOT to Implement the Study The Excel spreadsheet tool was used in this study. However, other commercial software tools are available to perform similar analyses. Detailed and granular historical data on sources of funding and associated revenues are an important resource needed for the study. VTrans had high-quality, detailed data that were used in this study. 9.2 Study 3: Asset-Level Risk Index—Bridge Risk Utility Index The Study 3 pilot illustrates an example strategy for analyzing risk in an existing BMS. This pilot was completed in the Kansas Department of Transportation (KDOT) BMS. It shows how a transportation agency can use its BMS to develop and calculate a risk index, then use the risk index to forecast scenarios with different funding levels for all applicable structures. KDOT uses an off-the-shelf commercially available bridge management software. The strategy discussed in the KDOT pilot is applicable to any configurable BMS that can accommodate formulae for the development and further use of complex indices. The risks ana- lyzed and the expected implementation effort for the pilot are as follows: • Threats: Structural failure caused by fracture-critical design, failure caused by scour criticality or waterway adequacy during flood events, and issues related to posting and rail safety. • Consequences: Bridge closures caused by event or incident and possible safety issues. • Implementation Effort: Define risk index input values and calculation for bridges. Define how the risk index fits into objective function, in this case, the overall utility function. (In an Method Description Forecast Value FY 2022 FY 2032 Deterministic Forecast 305 350 Range ±27 ±32 ETS Forecast 310 350 Range ±28 ±51 Monte Carlo Forecast 305 332 Range ±27 ±31 Table 9-17. Forecast of total VTrans state revenues ($, millions), with range of variability at beginning and end of FY 2022 to FY 2032 forecast period.

Pilot Testing and Results 65   optimization analysis, this function is used in a management system to maximize a selected value—the objective—under a set of constraints.) Run analysis to identify an optimal struc- tures work program, including risk. Analyze and summarize resulting metric projections. 9.2.1 Pilot Objective and How the Results Help Inform Asset Risk Decisions The objective of the Study 3 pilot was to use an agency’s existing BMS to test the incorporation of bridge risk into a utility function, calculate and forecast the bridge risk for each structure in the network, and hence forecast the overall risk carried by the network bridge population under different funding levels. Once incorporated, the risk utility function (as defined in Step 3 of Appendix F) can be an integral part of the bridge management process. Notably, the risk utility function can be used to analyze and inform specific risk mitigation decisions using the BMS. For example, by being able to evaluate and project the risk for bridges, an agency can identify high- risk bridges or groups of bridges for mitigation actions. Depending on the maturity of the BMS within an agency, personnel within the bridge man- agement group should be able to define risk mitigation actions that decrease the risk to indi- vidual structures. If these mitigation actions were incorporated into the BMS, the system would be able to use these actions in its overall benefit–cost optimization to include risk mitigation projects in the overall bridge work plan. This will ultimately help agencies to better manage the risk of bridge closure caused by scour and other event-related closures. 9.2.2 Description of the Technique or Tool This pilot used KDOT’s existing BMS to calculate a risk utility index for each structure in the BMS. Because the risk index is a “utility” index, the higher the utility, the lower the risk. The risk index used here is based on utility theory and is not calculated strictly based on threat likelihood, vulnerability, and consequence. The risk utility index is therefore a surrogate for these more formal and quantifiable measures. Although not directly related to threat likelihood, vulnerability, or consequence, the index value can provide a valuable measure of risk per structure, as well as a network-wide risk value. Once the risk utility index is defined, actions can be identified that mitigate this risk. Given the additional benefit in terms of risk, projects that also include risk mitigation are expected to receive higher priority for selection based on the benefit–cost optimization process. This method of defining a risk utility index can be used for any number of individual criteria associated with risk by the bridge manager. 9.2.3 Methodology Used in Conducting the Pilot The steps summarized in Table 9-18 outline the process for completing the pilot analysis of risk utility for bridges. These steps are described in detail in Appendix F. A bridge manager in an agency with a configurable BMS in place will be able to follow, delegate, and complete the steps in this methodology. The steps detailed in Table 9-18 were completed in the pilot using the existing KDOT BMS. The existing data within the KDOT BMS were used as the source data, including the NBI inspec- tion data for each qualifying structure. Existing formulae in the KDOT BMS configuration docu- ment, defined to calculate scaled utility values from select NBI item subcomponents, were used for calculating the risk utility index. The detailed formulae are provided in Appendix F. Five NBI item subcomponents relating to risk were identified, along with relative weighting and scaling

66 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research values for use in calculations. Relative weighting gives each selected subcomponent a weight- ing that totals 100. Then, the KDOT BMS was configured to incorporate the risk utility into the overall utility calculation. This allowed the risk utility to be included in the overall optimization that was set up within the BMS. Finally, scenarios were analyzed to determine how projects affect the risk utility subcomponents and how different funding levels affect the total risk utility index calculated for the network. For the pilot analysis, two scenarios were set up and run in the KDOT BMS to illustrate the effects of the different funding levels on the risk utility index. One scenario used a budget con- straint of $125 million per year. The second used a budget constraint of $250 million per year. All other scenario inputs were kept consistent. The BMS then calculated and forecasted the bridge utility index (the record of total bridge utility for the network) and, by extension, the risk utility index for each structure in the network. The system forecasted the overall utility and risk carried by the network bridge population under the two scenarios. KDOT’s BMS does not cur- rently support reporting of individual risk components of the overall utility index. Therefore, the risk subcomponents associated with each structure were produced as an output from the BMS analysis, and additional analysis was completed outside the system to look at risk specifically. In addition to the typically required steps identified in Table 9-18, analysis was undertaken to extract the resulting projects from each scenario. The additional analysis ultimately produced forecasts of the risk index component of the overall utility maximized in the BMS under the two different scenarios. To accomplish this additional analysis, the risk subcomponents were exported into a spreadsheet for the two funding levels analyzed. The previously mentioned for- mulae were used to calculate and forecast the risk utility index for comparison between the two Methodology Steps Step 1 Identify available source data. Step 2 Ensure the BMS can access the bridge inspection data. Step 3 Identify the formulae for calculating the risk utility index within the BMS. Step 4 Identify the relevant NBI item subcomponents and identify or agree upon the relative weighting value within the risk index. Step 5 Identify scaling values for each NBI item subcomponent selected to compute the risk utility index. Step 6 Con�igure the calculation of the risk utility index in the BMS. Identify or con�irm consequence and enter structure weight.Step 7 Step 8 Step 9 Identify risk mitigation actions. Step 10 Step 11 Identify analysis parameters. Step 12 Run analyses. Step 13 Analyze results. Con�igure deterioration models. De�ine trigger rules. Table 9-18. Methodology for developing a bridge risk utility index.

Pilot Testing and Results 67   scenarios. The detailed use of these formulae is described under Step 13 in Appendix F. Appen- dix F also includes the detailed steps for the additional analysis and how this was accomplished in a spreadsheet. Input Data Used in the Pilot This pilot study used data for the full bridge population residing in the KDOT BMS. Because the existing BMS was used in this pilot, all inventory and condition data were already avail- able. Regardless of the transportation agency, this information would be available in the bridge inspection system or a similar database and thereby almost certainly available to the BMS. This pilot study focused on the following five NBI standard inspection items because these data are available in the NBI data set and the existing BMS data: 1. NBI item 113—Scour Critical Rating 2. NBI item 92A—Fracture Critical 3. NBI item 71—Waterway Adequacy 4. NBI item 36A—Rail Safety Feature 5. NBI item 70—Bridge Posting The five subcomponents constituted the fundamental input data for calculating the risk utility index. An agency may select additional or different NBI items defined as contributing to risk within that agency. For example, the age of the bridge or the condition of the deck, substructure, or superstructure could be considered risk factors. The details of this calculation of the risk index, including formulae and the additional coeffi cients and weighting values used in the calculation, may vary depending on a BMS’s configuration or expert opinion within an agency. Examples of these values are included in Steps 3–5 in Appendix F. The two different scenarios, a program with a $125 million budget constraint and another with a $250 million budget constraint, were set up for the pilot in the BMS. For more general implementation of this method of incorporating structure risk into the agency’s existing BMS, the following assumptions were made for input data: • All data from the inspection system would be available in the BMS and no additional data would need to be gathered. This was assumed because all item subcomponents would be col- lected and expected to be populated in the BMS as part of the standard NBI data collection. • In addition to this base data, an agency would need to define the risk utility formula and any coefficients (such as weighting and scaling factors) for calculation of the risk component within the full utility function. Appendix F lists the formulae and weighting and scaling factors. • The agency would need to define one or more possible risk mitigation actions and associated improvements to the risk utility to be used as part of the modeling. • Analysis scenarios can typically be set up in the more common BMSs. As in the pilot, scenarios can typically be defined by funding level in these systems. Action Taken in the BMS This pilot was conducted using a fully functioning BMS. Depending on the configuration and functionality of the BMS, agencies may have to take additional steps to configure their systems. The scenarios were set up to maximize the utility (including the risk utility as a component), compute the utility index, and generate a list of projects for a period of 10 years. The BMS was specifically configured to use five subcomponents to calculate and incorporate the risk com- ponent of the utility index. The output projects and forecasts of the risk index from these two optimization analysis runs are described in the next section.

68 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research The system looks at the consequences of closure and structure weight assigned and generates a change in the risk value based on the projects selected by the optimization. Different types of projects have different effects on the risk subcomponents. For example, generally, a full bridge replacement would have a more significant effect on the risk subcomponents than a replacement of only the deck or superstructure. When analyzing the two scenarios, the system selects different projects based on their benefit–cost ratio. The programs of projects and their associated effect on the risk utility is generated as the output. 9.2.4 Outputs from the Pilot The output from this pilot study shows bridge risk calculated from the threats that are accounted for in the five subcomponents throughout the bridge population. The BMS forecasted this risk for two different scenarios: a $125 million program and a $250 million program. Refer to Appendix F for a detailed review and discussion of the results of the pilot study. For each scenario, all deterioration and improvement modeling considered by the BMS was used to create an optimum simulated strategy of projects for each structure in the network. While this modeling happens automatically in the KDOT BMS, it may not necessarily be docu- mented at the level of detail needed to extract associated reports from all BMSs. After the two scenarios were optimized in the BMS, the resulting simulated bridge projects for the next 10 years were available for review. The set of projects generated for each scenario was exported to an Excel spreadsheet. Example projects produced by the optimization in the BMS are presented in Step 13 of Appendix F. Depending on the BMS used, specific reports for the risk index component of the overall objective function (i.e., the overall benefit metric being maximized) may need to be configured in the system or extracted for separate analysis, as was required in this pilot. The risk index was forecasted in the KDOT BMS as a component of the overall utility index. However, the system would not report the risk index separately from the overall utility index. There- fore, a spreadsheet was created to show the forecast for the risk utility index individually. In the spreadsheet used in the pilot, the forecasted list of projects per bridge per year from the BMS runs was used to recreate the forecasts specifically for the risk index. The summarized steps for this additional output analysis (completed after running the optimization) are as follows: 1. Current data for each bridge were downloaded from the BMS database. 2. Projects were exported from the program planning screen of the BMS for the $125 million budget for 10 years. 3. Projects were exported from the program planning screen of the BMS for the $250 million budget for 10 years. 4. Benefits to the risk index for each project were calculated for each bridge. 5. The risk utility index before the project was calculated for each bridge. 6. A $125 million risk utility improvement was calculated for each bridge. 7. A $250 million risk utility improvement was calculated for each bridge. 8. Risk charts were developed to show the forecast of the risk index for the two scenarios. Refer to Step 13 of Appendix F for examples from each step. The resulting forecasts of the risk utility index component of the overall bridge network utility are shown in Figure 9-16. Figure 9-16 shows the forecasted risk utility index component of the overall bridge utility under the two scenarios. As the risk attributable to the factors considered in the pilot decreases, the risk utility increases. Thus, over the analysis period, the forecasted risk improves (i.e., utility trends higher) more for the $250 million scenario than for the $125 million scenario, as expected.

Pilot Testing and Results 69   In the pilot test scenario with $125 million funding, the BMS suggested the following projects that affect risk: • 10 bridge replacements with widening • 4 bridge replacements without widening • 25 deck replacements with widening • 15 deck replacements without widening • 68 reinforced concrete box (RCB) replacements The scenario with $250 million funding suggested these projects that affect risk: • 10 bridge replacements with widening • 4 bridge replacements without widening • 34 deck replacements with widening • 22 deck replacements without widening • 135 RCB replacements Figure 9-16 shows the risk index improving (i.e., the risk utility increasing) over time in both scenarios. This indicates that the projects chosen in the BMS improve the risk caused by scour, fracture criticality, posting, water adequacy, and rail safety over time. Also, as expected, the slope for the $250 million program line is steeper than for the $125 million program, indicating that improvement was quicker for the $250 million program. However, it can also be seen that even for the $250 million program, risk is not eliminated. That is, the risk utility index still does not reach 100 in the 10-year analysis period. In both cases, Figure 9-16 shows the risk index improving (i.e., the risk utility getting larger). This was partially a result of the modeling in the pilot wherein no deterioration of risk factors over time was modeled. In reality, this consistent improvement would not be completely true. For instance, the NBI score for scour may be expected to deteriorate slowly over time without mitigation. Deterioration of risk factors was not included in this pilot, as the analysis period was a relatively short 10 years. However, agencies that may have models for the deterioration of risk factors could include them in their implementation of the study. $125 million Program $250 million Program Figure 9-16. Forecasts of risk utility index component of overall bridge utility. Source: KDOT BMS.

70 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Although seemingly simple, the trends shown are the result of considerable modeling and benefit–cost optimization automatically carried out in the BMS analysis. The output also shows how risk was fully incorporated into the main analysis already carried out in BMSs. Funding in each of the two scenarios was not dedicated specifically to risk. Therefore, the forecasted risk trends were the result of projects being selected within the analysis based on benefit–cost optimization, which includes improvement of risk as a component within the overall benefit calculation. This full integration of a risk index into bridge management, along with considerations of condition, is a powerful strategy for enhancing bridge risk management in an agency. 9.2.5 Participating DOT Organizational Unit(s) This pilot required significant coordination with KDOT’s bridge management group. This group was already familiar with how the BMS was configured and had proposed risk-associated NBI items along with individual weighting and scaling factors. As a result, little additional effort was needed on the part of the KDOT bridge group. Technical experts were required only for the normal configuration of the BMS. No technical expertise outside the bridge program was needed for this pilot. Discussion and input from the director of the Division of Engineering and Design were con- sidered throughout the implementation of the pilot as well. 9.2.6 Who in a DOT Could Use the Results and How The results of this study could be used by a bridge management group to incorporate risk into future bridge management decisions. The impacts of projects on risk-related NBI items could be useful for project planning. Data analysts and modelers could also use the pilot results to develop summary reports for their agency’s risk management efforts. The resulting forecasts of risk under different funding scenarios would be useful to bridge managers, risk managers, and asset managers for planning. 9.2.7 Challenges in Pilot Study Setup KDOT had previously developed its planned configuration of the BMS and had its BMS mostly implemented. Specifically, including risk did not require substantial added effort. Because the pilot used standard items from NBI bridge inspection, no additional data capture for bridges was required. In general, if an agency has already implemented a commonly used BMS, the addi- tional setup and configuration to include risk in the utility index will not be difficult. At the time of the pilot, KDOT’s BMS did not have the reporting features needed to generate the risk utility index forecasts for the two different scenarios. For the pilot, data were exported from the BMS and analyzed using a spreadsheet. If the BMS were enhanced to enable these results to be easily reported, this step could be eliminated. Other commonly used BMSs are likely to have features to generate reports of these specific risk index projections under different scenarios. 9.2.8 Resources Needed by a DOT to Implement the Study The pilot study was conducted using KDOT’s existing BMS. Only a relatively small number of hours of subject matter expertise were needed to establish the final weighting and scaling factors and to configure the benefits of certain actions to the risk-associated NBI item values.

Pilot Testing and Results 71   Based on the pilot, an agency would typically need to commit some resource hours from the in-house personnel who configure and run the BMS. The configuration requires access to the BMS and the ability to run analyses and generate the resulting reports. An agency could also work with the database to set up a specific test environment to conduct the study, which would require additional effort. For successful implementation of the pilot, in addition to providing the technical resources, the agency should assign an implementation champion to engage appropriate stakeholders. The agency’s champion could be responsible for guiding or leading the engineering configuration, including the identification of scaling and weighting factors, actions, costs, decision rules, and analysis parameters. Finally, if specific risk-related inspection items that are not included as standard NBI items were identified as necessary, these would need to be added to inspections. 9.3 Study 4: Asset-Level Risk Index—Pavement Section Flooding The Study 4 pilot illustrates the methodology for defining and assessing a pavement section risk index at the asset level. The pilot was conducted with DelDOT using its existing PMS. The risks and consequences addressed in this pilot and implementation effort were as follows: • Threats: Flood events, sea level rise, high tides, storm surge, and hurricanes. • Consequences: Roadway becomes impassable from the flooding. • Implementation Effort: Define risk index input values for pavement management sections. Define how risk index fits into objective function, in this case, a pavement section flood- ing. In an optimization analysis, the objective function is used in a management system to define the value that is maximized or minimized under a set of constraints. Run analysis to optimize pavement network, including risk, and analyze and summarize resulting metric projections. 9.3.1 Pilot Objective and How the Results Help Inform Asset Risk Decisions The objective of this pilot was to demonstrate the use of a pavement section–level risk index that incorporates a pavement section’s risk of being flooded during storm events, sea level rise, high tides, or storm surge using an existing PMS. This study was piloted using the PMS in place at DelDOT, which may vary from PMSs implemented in other states. Using the methodology presented here, pavement managers can evaluate the flood risk for pavements, incorporate these risks into decision making, and target high-risk pavements for mitigation actions. Depending on the maturity of the PMS in an agency, risk mitigation actions can be identified and incorpo- rated into a benefit–cost optimization. This study piloted the methodology in DelDOT’s PMS to include flood-related pavement risks in the overall pavement work plan for better management of pavement risks. 9.3.2 Description of the Technique or Tool This pilot incorporated risk into the objective function (i.e., the overall “benefit” being maxi- mized) in the optimization the PMS performed. This pilot used existing Federal Emergency Management Agency (FEMA) flood risk data16 to demonstrate the impact of flooding. These data were used to exhibit the incorporation of national risk assessments that have a bearing on roadway closures.

72 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 9.3.3 Methodology Used in Conducting the Pilot The methodology steps describe how the analysis was conducted using DelDOT’s PMS. The eight main steps of the methodology are shown in Table 9-19 and summarized in this section. These steps were also followed in Studies 5 and 6, which were not piloted. Further detail for the Study 4 methodology is included in the associated section of Appendix G. The eight steps from Table 9-19 were completed in the existing DelDOT PMS. This culminated in defining coastal flood risks to pavement management sections and analyzing flood risk calcu- lations in the optimization of the pavement network. The configuration of DelDOT’s system made the inclusion of risk in the optimization analysis quick. The effort of such inclusions depends on the configuration of an agency’s PMS. Input Data Used in the Pilot This pilot used the information available within DelDOT’s existing PMS as well as existing GIS layers of the pavement management sections. The additional data included flood depths deduced for different percentages of annual chance flood events. The assessment also included a quantification of consequences (including terms for delay and/or detour costs, safety costs, repair costs, and any other associated costs). Numerous national flood risk evaluations have been completed for coastal and inland flood- ing. Many applicable national and agency-level sources were reviewed before the research team determined which data would be used in this study. These flood data resources are listed in Table 9-20. Methodology Steps Step 1 Identify available sources within the organization of vulnerability studies that have been conducted for the identi�ied threat or hazard type (�looding). Identify whether quanti�ied threat probabilities, vulnerabilities, and consequences are available from an existing assessment. De�ine likelihood, vulnerability, and consequence values for each vulnerable location within the study scope. Step 2 Determine and con�igure in the management system the calculation of the quanti�ied risk index component as a function of the threats, vulnerabilities, and consequences. Step 3 Decide on any deterioration models for changes in the threat probability, vulnerability, and consequence over time, possibly as the result of climate change. (Note that assumed deterioration rates can be constant.) Step 4 Identify risk mitigation actions, as well as the associated reduction in the risk index and improvements in the risk factors. Step 5 De�ine trigger rules. These are rules that trigger actions (treatments) and are used in the analysis to identify candidate actions (projects) for inclusion in the work plan. Step 6 Incorporate the risk index into the pavement management bene�it calculations. Step 7 Identify scenarios to be analyzed (such as funding constraints), as well as the speci�ic set of assets (i.e., the scope) of the analysis. Step 8 Run analyses based on the scenarios de�ined in the PMS, compile results, and report projections. Table 9-19. Methodology for developing a pavement flooding risk utility index.

Pilot Testing and Results 73   So ur ce Fl oo d Le ve ls St or m S ur ge Ra in fa ll O ve r� lo w in g Ri ve rs /S tr ea m s H ig h Ti de s Se a Le ve lR is e Cl im at e Ch an ge As so ci at ed P ro ba bi lit ie s H is to ri ca l I nf or m at io n Fu tu re P ro je ct io ns Ap pl ic ab le L oc at io ns D at a D ow nl oa d FloodFactor17 Y Y Y Y Y Y Y Y Y Y National Potentially NOAA Sea Level Rise Viewer18 Y Y Y N Y Y National Y Climate Central Water Level19 Y N N N N Y N N Y National For a price Climate Central Year20 N Y N N Y Y Y Y N Y National Climate Central Risk Zone Map21 Y Y N N Y Y Y Y N Y National Hazus22 Y Y N N N N N N N N National Y FEMA National Flood Hazard Layer Viewer23 Y N Y National Y National Storm Surge Hazard Maps24 Y Y Y N Y N N N N Y Coasts National Potentially Coastal Emergency Risk Assessment25 Y Y Y Y N Y Y National N FEMA Flood Map Service Center26 Y Y Y Y Y Y National Y DNREC Flood Plan- ning Tool27 Y Y N Delaware Y University of Dela- ware Water Re- sources Center28 Y Y N Y Delaware Y Delaware Flood Risk Adaptation Map29 Y Y N N N Y N Y N Y Delaware Y Coastal Inundation Maps for Dela- ware30 Y N N N N Y N N N N Delaware Y FirstMap31 N Delaware Y DelDOT Gateway32 N Delaware Y For a price For a price N���: Blank cells indicate uncertainty about whether the data type was included or considered. NOAA = National Oceanic and Atmospheric Administration. Table 9-20. Flood data resources considered for use in developing the pavement flooding risk utility index.

74 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research To ensure that the study methodology could be applied to other states, nationally available data were used for analysis. As such, the FEMA Flood Map Service Center listed in Table 9-20 was selected as the source for national flood data. The following specific data, available through FEMA as part of the Flood Risk Database, were selected for analysis: • Flood depth grids for 1% annual chance flood events:33 – Coastal—Sussex County – Coastal—Kent County Of the three total counties in Delaware, only New Castle County does not have an associ- ated coastal flooding geodatabase, as the county’s border is along the Delaware River and is not coastal. Additional flood data were available for inland flooding in all three Delaware counties at 0.2 percent, 1 percent, 2 percent, 4 percent, and 10 percent annual chances. However, for this pilot, it was determined that including coastal flooding in DelDOT’s PMS analysis would be an appropriate starting point, with the assumption that the other flood depth grids could be analyzed later as desired. The flood data were applied to a GIS layer of DelDOT’s pavement management sections using ArcMap. This enabled the definition of the maximum flood depth anticipated for a 1-in-100-year (1 percent annual chance) flood for each pavement management section. FEMA 1 percent annual chance flood depth grids are available nationally and can be down- loaded by county from the Flood Map Service Center. To assist other agencies interested in implementing this method, additional detail about downloading and analyzing these flood data is included in Appendix G. Action Taken in the PMS Table 9-21 outlines the data used from DelDOT’s PMS as well as the calculated values included in the configuration of the analysis. Rows 7 through 21 were included in the PMS configuration as columns in the network master, added data tables, or Groovy Scripts guiding the analysis. For more detail, see Appendix G. 9.3.4 Outputs from the Pilot The first outputs of the Study 4 pilot were the current risk index for each section of road, and, by extension, the current overall risk index value for the road network. After the predictive modeling and benefit–cost analysis were run for a specific scenario, the PMS generated opti- mized strategies for each section in the road network. An optimized strategy is a series of simu- lated or recommended projects over time. These strategies were accompanied by the associated long-term forecasts of the risk level. Overall, the pilot study illustrates the ability of an existing PMS to analyze and help manage risk caused by road flooding. This process provides a tool for assessing current risk for every road section, identifying optimum strategies for managing the risk, and forecasting the resulting section- and network-level risk over the long term. For this pilot study, system configuration was completed in the development environment of DelDOT’s PMS. The additional configuration calculated and included risk as a dollar amount in the optimization analysis of the pavement network. After configuration, the following three scenarios were analyzed: • Baseline Scenario: The baseline scenario used for this study was an existing scenario set up within the DelDOT PMS and identified for use by a DelDOT pavement management expert. The scenario was set to optimize with maximization of network condition as the objective,

Pilot Testing and Results 75   ID Spreadsheet Category Formula Sources and Assumptions 1 Name From PMS Network Master Source: DelDOT’s network master in their PMS. 2 County From PMS Network Master Source: DelDOT’s network master in their PMS. 3 Begin Section From PMS Network Master Source: DelDOT’s network master in their PMS. 4 End Section From PMS Network Master Source: DelDOT’s network master in their PMS. 5 Length of Road (Miles) From PMS Network Master Source: DelDOT’s network master in their PMS. 6 Average An- nual Daily Traf�ic (AADT) From PMS Network Master Source: DelDOT’s network master in their PMS. De- fault is 100 (assuming when not available in PMS, then must be low). 7 Travel Cost per Vehicle per Hour = ([Hourly Value of a Person’s Travel Time in a Vehicle] × [Aver- age Vehicle Occu- pancy]) × 1.03(2021– 2010) Source: FHWA Work Zone Road User Costs—Con- cepts and Applications.34 8 Detour Length = 2 × [Length of Road (Miles)] Agency will measure, or get from NBI Structure da- tabase, or assume 2 × section length. 9 Detour Speed Default 50 mph. 10 Delay Cost per Day = [Travel Cost per Vehicle per Hour] × [Detour Length] ÷ [Detour Speed] × [AADT] 11 Max Flood Depth From FEMA Geodata- base This was computed by intersecting the FEMA 1% �lood layer with the PMS road section layer using GIS tools to extract the maximum �lood depth for each PMS road section. 12 Vulnerability = IF ([Max Flood Depth] > 0, 100%) If depth of �looding is > 0, then vulnerability is 100% (i.e., if the roadway is �looded at all then it is impassable by normal traf�ic). 13 Threat Prob- ability Probability of Event (Chance of Flooding) Use probability of the �lood depth grid layer as ap- plicable (typically, 0.2%, 1%, etc.). 14 Likelihood = [Vulnerability] × [Threat Probability] 15 Days of Delay = [Max Flood Depth] ÷ 2 In the absence of more accurate data, assume that the approximate number of days closed to traf�ic equals max �lood depth in ft ÷ 2. Table 9-21. Input data, formulae, and assumptions used in the PMS analysis of pavement flooding risk. (continued on next page)

76 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research meaning projects were selected to maximize overall network condition over time. The sce- nario included a list of either committed projects or scheduled projects for which funding was committed, as well as funding constraints for districts and work types for the remaining project selection. This scenario was compared with the risk scenarios to determine the effects of the additional risk configuration on the network analysis. The same funding constraints and general scenario setup informed by the identified baseline were used for all three scenarios. • Risk Scenarios: The other two scenarios were created by copying the baseline scenario and including minimization of flood risk as an additional objective for the optimization. All other inputs for the risk scenarios, including the committed projects, were kept consistent with the baseline scenario. When a scenario is set up to consider multiple objectives in DelDOT’s PMS, the objectives are weighted to sum to 1 (e.g., 0.3 + 0.7 = 1.0). The weightings indicate the importance of each objective in the optimization analysis. In the pilot, each risk scenario had two objectives: (1) maximize the average overall pavement condition for the network and (2) minimize the overall flood risk for the network. This resulted in scenarios that considered projects that would both maximize condition and minimize flood risk at the assigned weight- ing when selecting projects. Additionally, in each risk scenario, decision trees with an added treatment for mitigating flood risk were used in place of the general decision trees typically used. These decision trees included an additional reconstruction treatment that improves risk significantly by decreasing the maximum flood depth on a road section by 5 ft. – 0.25 Risk Scenario: In this scenario, the minimized risk objective was weighted at 0.25, while the maximized condition objective was weighted at 0.75. The inclusion of risk in this ID Spreadsheet Category Formula Sources and Assumptions 16 User Delay Costs = [Delay Costs per Day] × [Days of De- lay] 17 Safety Costs In the absence of more accurate data, assume no increase in crash rates, however, if roadway �lood- ing would cut off a community from emergency services, such as hospitals, or create a large detour, consider including risk of additional lives lost dur- ing a �lood event. 18 Repair Costs Assume very low water �low speed and therefore no direct repair costs. However, if an erosion wash- out is expected, then assume a unit cost per sq yard of �looded roadway. 19 Other Costs $0.00, unless other information available from agency. 20 Total Conse- quence per Event per Section = [User Delay Costs] + [Safety Costs] + [Repair Costs] + [Other Costs] 21 Total Risk per Section = Sum of ([Threat Probability] × [Vul- nerability] × [Total Consequence]) for each Flood Event (Threat) NCHRP 20-07/Task 378 method for calculating risk.35 Table 9-21. (Continued).

Pilot Testing and Results 77   scenario resulted in the selection of one risk mitigation reconstruction project in the nal year of the 15-year analysis period. – 0.5 Risk Scenario: In this scenario, the risk objective and condition objective were both weighted at 0.5. e more signicant weighting of risk in the analysis of this scenario resulted in the selection of 16 risk mitigation reconstruction projects over the length of the 15-year analysis period. Figures 9-17 and 9-18 show the comparison of forecasted average overall pavement condition (OPC) for the pavement network as it changes for each scenario throughout the analysis period. OPC is a state metric DelDOT uses to track its average pavement network condition. With the inclusion of risk as an objective, the selection of a risk-mitigating project typically comes at a slight cost to the OPC. is trade-o between competing objectives is expected and conrmed that the additional risk conguration worked as planned. Figure 9-17 illustrates the effects of each risk mitigation treatment on the average OPC throughout the analysis period. For all three scenarios, the average OPC is similar for the rst 8 years of the analysis period. is is partially because many larger projects are already included in the analysis for all three scenarios through the committed projects list. The committed projects are “locked in” and would not have been selected to minimize ood risk. 64.2743 64.2493 63.7155 63.5 64 64.5 Av er ag e O PC Average OPC - Baseline Average OPC - 0.25 Risk Average OPC - 0.5 Risk 20 22 20 23 20 24 20 25 20 26 20 27 20 28 20 29 20 30 20 31 20 32 20 33 20 34 20 35 20 36 Figure 9-18. Average OPC for nal year of scenarios with varying inclusion of risk. 63 64 65 66 67 68 69 70 71 72 Av er ag e O PC Average OPC - Baseline Average OPC - 0.25 Risk Average OPC - 0.5 Risk Figure 9-17. Average OPC for scenarios with varying inclusion of risk.

78 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research There is a notable difference between the baseline and 0.25 risk scenarios and the 0.5 risk sce- nario starting in 2031, when the average OPC begins to drop more significantly through 2036 for the 0.5 risk scenario. This is attributable to the selection of 16 risk-mitigating projects that have a positive effect on the total risk but are not as beneficial to the overall condition as the projects in the original work plan. As noted, this trade-off between flood risk and overall network condition was expected and shows that the new configuration worked. By the end of the analysis period, the difference between the baseline and 0.25 risk scenarios is minimal, as only one risk mitigation project was selected in the latter. The effect of weighting risk higher at 0.5 is more significant on the average OPC, though the difference is still within 1 point. This implies that while there is some effect on the average OPC, including more expen- sive projects that involve mitigating risk would not have a drastic effect on the overall condition of the pavement network. The treatments selected by the PMS for each scenario are provided in Table 9-22. The projects for each scenario included the same list of committed projects. The treatment types selected show that the PMS selected 1 reconstruction—risk mitigation project for the 0.25 risk scenario and 16 for the 0.5 risk scenario. The mix of treatments changes slightly between the baseline and the 0.25 risk scenario, but the 0.5 risk scenario exhibits a decrease in preservation and rehabilita- tion treatments (both structural and functional) to accommodate the additional reconstruction projects. Figure 9-19 shows the projected total risk in terms of dollars calculated throughout the analy- sis period for the three different scenarios. Note that the total risk increases at a steady rate, as the maximum flood depth is predicted to increase over time. This increasing maximum flood depth was predicted by increasing flood risks to Delaware properties from the FEMA data. Fig- ure 9-19 shows that the total risk calculated for the 0.25 risk scenario matches the baseline sce- nario until the risk mitigation project was selected in 2036 and the total risk decreases. In the risk scenario with a higher weighting of 0.5, the total risk drops from the baseline of $3.94 million to $3.67 million for the pavement network, accounting for a “savings” of approximately $300,000 annually in expected risk costs over the 15-year period. Treatment Types Baseline 0.25 Risk 0.5 Risk Chip seal 972 972 972 Chip seal + patch 6 6 6 Crackseal 138 138 139 Bituminous patching, 5% 28 48 50 Portland cement concrete patching 2 2 2 Preservation 424 427 419 Reconstruction—risk mitigation 0 1 16 Rehab functional 1,402 1,405 1,387 Rehab structural 369 370 359 Table 9-22. Count of treatment types selected for each scenario.

Pilot Testing and Results 79   9.3.5 Participating DOT Organizational Unit(s) This pilot was conducted with the involvement of the DelDOT pavement management team, PMS technicians, and GIS specialists. The activities included locating and analyzing data, con- figuring the PMS, and running the analysis within the system. However, should an agency fully implement this flood risk monitoring and management methodology, additional groups such as a hydraulics unit that conducts vulnerability studies could be involved. Additional field staff could be involved in determining specific flood risk mitigation actions that may need to be included in the reconstruction projects, for example, adjusting road height or improving drainage. This pilot study did not include the safety costs associated with road closures. However, if included, agency safety- and insurance-related staff could become involved. There is an opportunity for expanding and improving this study with additional and more specific data, depending on an agency’s resources and available data. For instance, rather than assuming vulnerability, an agency can incorporate real vulnerability study data into the analysis. 9.3.6 Who in a DOT Could Use the Results and How The results of this study could be used by a pavement management group to incorporate flood risk into future pavement management decisions. At a minimum, even if risk-mitigating treat- ments were not included in the PMS, the total risk of flooding could be estimated and forecasted into the future using this method. This could be useful information for asset managers and risk managers to anticipate risks in documents such as the federally required TAMPs. Because the study illustrates how to model and forecast risk into the future for different funding levels, pave- ment managers could work with finance staff to define potential budget adjustments. Pavement $3.0 $3.2 $3.4 $3.6 $3.8 $4.0 $4.2 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 To ta l R isk ($ , m ill io ns ) Total Risk - Baseline Total Risk - 0.25 Risk Total Risk - 0.5 Risk Figure 9-19. Total risk for scenarios with varying inclusion of risk.

80 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research managers could then develop recommendations based on the results of the assessment of network- wide flooding risk levels and trends that can be expected for specific budgets. 9.3.7 Challenges in Pilot Study Setup The setup of the pilot study involved a significant amount of troubleshooting within the PMS as well as difficulty with access to the system caused by technical issues. However, given that configuring software to complete new analyses always involves testing and troubleshooting, the challenges encountered were specifically not attributable to the methodology. Once the system is configured and flood depth data have been uploaded for appropriate management sections, then including risk in any future analyses is straightforward. Risk factors can then be easily included as an output for any scenario and as an input when desired. Depending on available data, moderate effort may be required to identify and define initial threats, threat probabilities, vulnerabilities, and consequences. Initially, at a minimum, many of the values can be estimated and defaults assigned in order to gauge the general magnitude of the risk index. 9.3.8 Resources Needed by a DOT to Implement the Study In the case of the pilot, the existing PMS was used, and significant hours of subject matter expertise were needed for the analysis of GIS data and configuration of the PMS. The agency would typically need to commit resource hours from the personnel who run the PMS and resources to determine the maximum flood depth on roadway sections, most likely from the GIS group. The personnel would need to have access to their PMS and be able to per- form the necessary additional configuration, as well as be able to run analyses and report results. At their discretion, agencies may decide to work with their PMS database administrators to set up a specific test environment in which to conduct the study. Finally, data are integral for implementing this study. Agencies may have different data avail- able for use in this study and may replace assumptions with real data when possible. Four main data components are needed, at a minimum, to complete this study: 1. GIS pavement section layer 2. 100-year flood depth grid from FEMA 3. Estimated mitigation costs 4. Configurable PMS in place In addition to the technical resources, an agency is advised to plan on assigning an imple- mentation champion to engage the appropriate stakeholders. The champion may be responsible for guiding the configuration, including the identification and quantification of initial threats, threat probabilities, vulnerabilities, consequences, mitigation actions, costs, decision rules, and analysis parameters. 9.4 Study 7: Program-Level Risk—Pavement Network Analysis The pilot for Study 7 illustrates a method for assessing risks to a pavement network at the pro- gram level using an existing PMS. The pilot was completed in the Idaho Transportation Depart- ment (ITD) PMS. This analysis exhibits the identification of high, expected, and low values for multiple variables to create an “envelope” of predicted conditions. The pilot was developed in ITD’s existing PMS but could equally well be developed for any available PMS for which similar inputs can be defined when creating and analyzing network-level optimization scenarios. The following was addressed in this study:

Pilot Testing and Results 81   • Threats: Variation in inflation rate; unplanned changes to available funding levels • Consequences: Impacts to forecasted condition metrics • Implementation Effort: Defined high, expected, and low values of chosen input values rep- resenting threats; ran multiple analyses to identify optimum strategies and forecast resulting metrics for high, expected, and low scenarios; and analyzed and summarized resulting metric projections 9.4.1 Pilot Objective and How the Results Help Inform Asset Risk Decisions The objective of this pilot study was to demonstrate a real-world example of the analysis of pavement program-level risk by forecasting condition based on high, expected, and low extents of program-level variables using existing tools. An expected envelope of predicted average net- work condition can be created to better inform decision makers about the risks to the network. The results of Study 7 would provide an agency with additional resources for decision making at a high level. It was important to consider fluctuation in high-level parameters that cannot be accurately predicted so that trade-off decisions could be made. This study technique could be implemented for a variety of potential impacts on the network, including the following examples: • Unexpected weather could increase the need for funds to repair or mitigate impacts. The strategy exhibited in this study could be used to determine the impacts of moving funding from the pavement management program. Examples include flood events, tornados, and a longer-than-expected snow and ice season. • Increased construction costs could reduce the number of projects that can be contracted and reduce the spending power of the originally projected funding level. The approach used in this study helps account for this scenario by projecting the future funding levels required to maintain the current level of service with potential increases in construction costs. 9.4.2 Description of the Technique or Tool ITD’s existing PMS was used to forecast network-level pavement condition for three scenarios. Scenarios were defined by different levels of variables at the program level, including inflation rates and available funding levels. The variables were selected because of their associated uncer- tainty. ITD’s existing PMS was already configured to allow modifications to funding levels and inflation rates. An agency could define additional variables that are at risk of varying significantly or significantly impacting the network condition. A typical PMS will be configurable to account for change in defined variables. 9.4.3 Methodology Used in Conducting the Pilot The methodology used for this study followed the four steps shown in Table 9-23. A detailed description of the processes and data used to complete this pilot study is provided in Appendix J. The steps shown in Table 9-23 were completed in the pilot within the existing ITD PMS. The initial variables analyzed in the pilot were construction inflation and available funding. Variation in construction inflation was determined using historical pavement-related construc- tion costs. Variation in funding was estimated by the ITD pavement management lead. Three scenarios were set up in the ITD PMS, and results were analyzed to illustrate the effects of risk of variation in the two parameters on the average condition of the pavement network.

82 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research The following three scenarios were set up for analysis within the system: 1. The best-case scenario, in which the most optimistic values for construction inflation rate (3.55 percent) and funding level (20 percent increase) were used 2. The expected scenario, in which the most likely values for construction inflation rate (8.30 per- cent) and funding level (expected and already included in the PMS, detailed in Table J.2 in Appendix J) were used 3. The worst-case scenario, in which the most pessimistic values for construction inflation rate (13.00 percent) and funding level (30 percent decrease) were used The specific inputs and outputs are summarized as follows. Refer to Appendix J for more detailed information. Input Data Used in the Pilot The input baseline funding levels used in the scenarios were obtained from an existing sce- nario in the ITD PMS that was defined by the ITD pavement management lead. The construc- tion inflation rate values were determined by completing a Monte Carlo simulation on historical construction cost data provided by ITD. The funding levels from the existing scenario in the PMS were predetermined for the first 7 years of the analysis period to cover the committed projects. The funding constraint was set to $130 million for the annual budget for the remainder of the analysis period (2028–2040). The 20 percent increase and 30 percent decrease for the best-case and worst-case scenarios were applied to funding levels in 2028–2040 following the years with committed projects. ITD provided historical data that reflected the construction costs of ITD pavement projects. The historical construction cost data were used to estimate the future variation in construction costs. The data set for projecting changes in construction costs was specific to ITD pavement projects. If program- or agency-specific data are unavailable, historical national inflation data could be used by agencies interested in implementing this study. The processes for using both agency-specific and national data to estimate variation in inflation rates are detailed in Appendix J. The data provided by ITD are included in Step 2 of Appendix J. The three scenarios defined for optimization in the PMS are shown in Figure 9-20. The top left corner of the chart represents the overall best-case scenario in which the construction infla- tion rate is low and there is a 20 percent increase in funding. The bottom right box in the chart Methodology Steps Step 1 De�ine which scenario input variables to consider as risk factors in the sce- nario (i.e., which variables are likely to change and have a signi�icant impact on the pavement network condition). Step 2 Determine levels of variation for input variables (using historical data, esti- mates based on expert opinion, Monte Carlo analysis, etc.). Step 3 De�ine scenarios that will be analyzed to provide best, worst, and expected cases (each scenario de�ined as a speci�ic combination of input variables, such as in�lation rates and funding). Step 4 Run analyses in asset management system, analyze the results, and report them. Table 9-23. Methodology for analyzing pavement network–level risks.

Pilot Testing and Results 83   represents the overall worst-case scenario in which the construction inflation rate is high and there has been a 30 percent decrease in funding. The worst-case scenario is expected to pose the highest risk to the overall pavement network condition, as this would lead to a significant reduc- tion in treatments applied to pavements. Action Taken in the PMS Each of the three scenarios was set up and run within ITD’s PMS. A scenario identified by the ITD pavement manager was copied and edited for this analysis. For additional details, refer to Appendix J. 9.4.4 Outputs from the Pilot The results generated from running scenarios in the ITD PMS are outlined in this subsection. A typical PMS could be set up to return a selection of outputs for analysis and ITD’s PMS accom- plished this. The pilot assessed the effects of changes in network-level variables on the average pavement network condition. ITD determines average pavement condition using an internal index called the Overall Con- dition Index (OCI). The OCI is the weighted average of many different pavement performance factors and varies from 0 to 100.36 A score of 0 represents the poorest pavements and 100 rep- resents the best pavements. The ITD PMS calculated and forecasted the OCI for the pavement network for each of the three scenarios. The OCI rating was used in the study to exhibit the effects of changes in funding and inflation on the average condition of a pavement network. Table 9-24 shows the average OCI for the pavement network for the expected, best-case, and worst-case scenarios. For the final year of the analysis period (2040), Table 9-24 shows the values 60, 47, and 69 for the expected, worst-case, and best-case scenarios, respectively. As expected, the OCI value recorded in the final scenario year is the lowest for the scenario with the worst-case funding and highest construction inflation rates. The OCI value recorded in 2040 is the highest for the scenario with the best-case funding and lowest construction inflation rates. Figure 9-21 shows the average change in OCI over the analysis period for each of the three scenarios. The envelope created by the scenarios is shown in the figure, where the best-case scenario (dotted line) produces the highest projected condition ratings, the worst-case (dashed line) produces the lowest projected condition ratings, and the expected case (solid line) is in the center. Funding Best Expected Worst +20% −30% Co ns tr uc tio n In �la tio n Best 3.55% Expected 8.30% Worst 13.00% Figure 9-20. Scenarios analyzed in ITD’s PMS.

84 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 0 20 40 60 80 100 O CI Avg OCI - Worst Case Avg OCI - Expected Avg OCI - Best Case Figure 9-21. Projected OCI for best-case, expected, and worst-case scenarios for funding levels and ination rates, 2021–2040. Scenario Year Average OCI Expected Worst Case Best Case 8.30% I �lation Original Funding 13.00% I �lation 30% Decrease in Funding, 2028–2040 3.55% I �lation 20% Increase in Funding, 2028–2040 2021 92 92 92 2022 91 91 91 2023 89 89 89 2024 88 87 88 2025 85 85 86 2026 83 83 84 2027 81 80 82 2028 81 79 83 2029 80 77 83 2030 79 75 82 2031 78 72 82 2032 76 70 81 2033 75 67 80 2034 74 64 78 2035 71 62 76 2036 69 59 74 2037 67 56 72 2038 64 53 71 2039 62 50 70 2040 60 47 69 Table 9-24. OCI values calculated by the ITD PMS for each scenario. e condition of the network for the rst 7 years of the analysis period is the same or similar, which reects the work plan of committed projects that were xed for this time period. Although committed projects could still be impacted by ination if they have not been bid, changing the ination rate in the PMS does not impact the cost of committed projects. Depending on the PMS conguration, an agency could include impacts of ination on committed projects by changing the construction costs directly entered into the system for the planned construction year to account for ination on those specic committed projects. e initial OCI of the network for all three scenarios was 92. Figure 9-21 shows the condition deteriorating in the expected and worst-case scenarios starting in 2027, following inclusion of the committed projects. For the best-case scenario, the overall condition improves from 82 to 83 in 2028, then begins slowly declining starting in 2030. e decline in condition in all three

Pilot Testing and Results 85   scenarios after 2030 was partially explained by all three inflation rates being higher than ITD’s typically assumed 2 percent inflation rate. Because the changes in construction costs specific to ITD’s pavement projects were used to determine the inflation rates, they were more applicable for these scenarios than the 2 percent rate ITD usually uses in analysis. This may be true for many agencies that are not using agency-specific data in their assumptions. These results suggest that significantly increased investment beyond adjusting for the typical 2 percent inflation could be required to maintain the network conditions. The output from the study could be used to support a case for additional funding to decision makers. At the least, the results of the study could be used to inform stakeholders of potential impacts on the condition of the pavement network. This would create the foundation for build- ing a strategy to account for program-level risks. 9.4.5 Participating DOT Organizational Unit(s) This pilot study was conducted in conjunction with the pavement management group at ITD. Support was also received from the agency’s information technology department as technologi- cal issues interrupted access to the PMS. As this study did not require configuration of the system itself, minimal effort was required from the pavement group, aside from determining the input parameters to use in the study. The pavement management groups in most agencies will be able to obtain the information to complete this study. As needed, the groups could use input from the financial team on future inflation rates and estimated future funding for pavement for the scenarios tested in the study. 9.4.6 Who in a DOT Could Use the Study Results and How Pavement managers could use the results of this pilot study to analyze the impacts of vary- ing inflation rates or funding levels on the overall network condition. With analysis results, the pavement managers could determine and support recommended changes to funding levels or targets. For instance, using the best-case scenario in this pilot, a pavement manager may suggest increasing funding levels to minimize the deterioration of the overall condition of the pavement network. The pavement manager would also be able to provide supporting evidence for changing the state of good repair targets with charts and graphs produced from this pilot. 9.4.7 Challenges in Pilot Study Setup This study used an existing PMS that had been configured for inputting various parameters. Therefore, the setup and implementation of the strategy presented here did not involve a high level of effort from the agency. Agencies that have an operational PMS in place will not find it difficult to identify ranges of input values in order to define scenarios for analysis. The required effort is dependent on the availability of data and the involved staff ’s familiarity with forecasting values and determining the variable ranges. 9.4.8 Resources Needed by a DOT to Implement the Study In the pilot for Study 7, ITD supplied historical construction cost data for unit prices across various activities and materials. These were analyzed to identify the worst (95th percentile) and best (5th percentile) of the likely inflation rates. Historical data were not used to determine likely, best, and worst scenarios for budget levels, as variation in funding is typically best estimated by financial staff. Rather than performing a separate analysis, the pavement manager provided an estimated variation in future funding based on a discussion with financial staff. Depending on

86 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research the level of detail desired, an agency implementing this method would need resources and tools to determine the variation levels to be used to set up scenarios. When setting up and running scenarios, the agency will need to commit some resource hours from the technical personnel who typically manage the PMS. These are personnel who set up, run, and analyze these scenarios, then report the results. As detailed in other studies, an SME who could also serve as the champion could help guide the study by detailing the relevant scenarios, providing guidance on the input variables, and communicating the results to other stakeholders. In some cases, analysis of data is necessary to predict potential variation in variables. This may require additional resource hours dedicated to locating appropriate data and determining the appropriate input values for analysis. 9.5 Study 9: Crosswalk Between Climate Change Vulnerability and Risk Terminologies Study 9 includes detailed, user-friendly guidance on how to transform the information col- lected in a typical climate change vulnerability assessment output into risk information that would be comparable to other risks. This includes a focus on how to assess the “likelihood” of climate change risks and a customizable template for a likelihood rating rubric to serve as a start- ing point for assessing climate risks. With the template rubric, agencies could scale up or down or modify the time frame of the analysis as needed. The guidance included in this study is applicable to any climate change and extreme weather threats. This includes rising temperatures, changing precipitation, sea level rise, severe storms, flooding, and wildfires. Risk is a function of the likelihood and the consequences of an event occurring. In many cases, depending on the methodology used to evaluate exposure, sensitivity, and adaptive capacity in a vulnerability assessment, it may be possible to approximate risk from these components. For example, Risk  Exposure and Sensitivity (i.e., Likelihood of Impact) versus Adaptive Capacity or Criticality (i.e., Consequences of Impact) Table 9-25 defines key terminology used throughout Study 9. These terms are traditionally used for vulnerability assessments and risk assessments. The study provides guidance on how these two sets of terms are interrelated. 9.5.1 Pilot Objective and How the Results Help Inform Asset Risk Decisions The objective of this pilot was to develop guidance for agencies interested in assessing climate risks but not sure how to apply the FHWA Vulnerability Assessment and Adaptation Frame- work35 to create outputs that tie into other risk assessment and risk management processes at the agency. 9.5.2 Description of the Technique or Tool Draft guidance was prepared, including summary text, tables, and graphics that explained how the terminology often used in climate change vulnerability assessments (e.g., exposure, sen- sitivity, and adaptive capacity) compares with terminology used in other risk assessments (e.g., likelihood, consequences). It also explains how to translate the information collected in a typical climate change vulnerability assessment output into risk information that would be comparable

Pilot Testing and Results 87   to other risks. In addition, the guidance included an approach to evaluate the likelihood of cli- mate change risks so that they can be evaluated relative to other risks. For example, the following likelihood rating scale was provided for the pilot agencies to test for different climate-related risk events (Table 9-26). 9.5.3 Methodology Used in Conducting the Pilot This study was piloted by two agencies: Michigan DOT (MDOT) and Minnesota DOT (MnDOT). Each pilot agency was provided with the draft study guidance as well as a worksheet to help them complete each section. Term De�initions Exposure Linked to likelihood of an event occurring. Refers to whether an asset or system is located in an area experiencing direct effects of climate variability and extreme weather events. Exposure is a prerequisite for vulnerability. Sensitivity Linked to likelihood of damage or disruption occurring, or consequences of an event. Refers to how an asset or system responds to, or is affected by, exposure to a climate change stressor. A highly sensitive asset will experience a large degree of impact if the climate varies even a small amount, where as a less sensitive asset could withstand high levels of climate variation before exhibiting any response. Criticality Analogous to consequence. When discussing risks that are threats, criticality of an asset is quoted, typically as a measure of the consequence if the asset fails. It is independent of the likelihood of failure. Adaptive capacity Linked to consequences of damage occurring. In some assessments, used in place of criticality. Refers to the ability of a transportation asset or system to adjust, re- pair, or �lexibly respond to damage caused by climate variability or extreme weather. Vulnerability The potential for a negative consequence to occur under a speci�ic threat. Quanti- tatively, a vulnerability will be treated as the probability of a consequence occur- ring given a speci�ic threat. It is also the degree to which a system is susceptible to or unable to cope with adverse effects of climate change or extreme weather events. In the transportation context, climate change vulnerability is a function of a transportation system’s exposure to climate effects, sensitivity to climate effects, and adaptive capacity. Likelihood A measure of the probability of a risk event occurring. Quantitatively, it may be measured directly as a probability, or it could be a surrogate measure of probabil- ity and be estimated on a simple scale through expert elicitation. Consequence The implication of what actually happens or the possible outcomes resulting from a decision.37 It can be thought of as a measure of the effect of an event or sequence of events occurring. Quantitively, the term consequence can be thought of as the incurred cost. It may be measured in terms of a dollar value or as a utility (or dis- utility). Less rigorously, for example, the utility of the consequence may be given in terms of a simple 1–5 scale. Risk The positive or negative effects of uncertainty or variability on agency objectives. Risk includes uncertainty, variability, vulnerability, threats, and opportunities. Quantitatively, risk will be measured as the product of probability or likelihood of an event and its consequence. If the risk consists of multiple events, then it will be measured as a sum of the product of likelihood and consequence of each event. Typically, it will be the expected value of the overall consequence of an asset or system being in a particular state. Table 9-25. Key terminology for climate change vulnerability–risk terminology crosswalk.

88 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research The core teams from MDOT and MnDOT filled out one worksheet per agency with their col- lective feedback. MnDOT held an additional working session with its TAMP Advisory Group to vet the guidance, with a wider sampling of staff using the worksheet prompts to gather additional feedback. Steps of the Study 9 methodology are in Table 9-27. Step 1: Kickoff Call A kickoff call was held to orient each pilot agency to the draft guidance and worksheet and answer any questions about expectations or what they were being asked to do. Step 2: Review: Vulnerability Versus Risk Pilot teams reviewed the draft Vulnerability versus Risk section of the guidance and provided feedback via the following prompts in the worksheet: Table 9-26. Likelihood rating scale for climate change vulnerability assessment. Likelihood Rating Value Criteria for Discrete Climate-Related Risk Events Criteria for Gradual Climate-Related Risk Events Almost certain 5 Event is expected to happen about once every 2 years or more frequently (i.e., annual chance ≥ 50%) Event is almost certain to cross critical threshold Likely 4 Event is expected to happen about once every 3–10 years (i.e., 10% ≤ annual chance < 50%) Event is expected to cross criti- cal threshold. It would be sur- prising if this did not happen Possible 3 Event is expected to happen about once every 11–50 years (i.e., 2% ≤ annual chance < 10%) Event is just as likely to cross critical threshold as not Unlikely 2 Event is expected to happen about once every 51–100 years (i.e., 1% ≤ annual chance < 2%) Event is not anticipated to cross critical threshold Almost certain not to happen 1 Event is expected to happen less than about once every 100 years (i.e., annual chance < 1%) Event is almost certain not to cross critical threshold Methodology Steps Step 1 Kickoff call Step 2 Review: Vulnerability versus risk Step 3 Review: Assessing likelihood of climate change Step 4 Test: Assessing likelihood of climate change Step 5 Check-in and debriefing calls Step 6 Revising the guidance Table 9-27. Methodology for developing a climate change vulnerability–risk terminology crosswalk.

Pilot Testing and Results 89   • Are the definitions of vulnerability and risk assessment terminology clear? • Is the relationship between vulnerability and risk assessment terms clear? • Do these definitions and terms align with terminology your agency has used in the past? Step 3: Review: Assessing Likelihood of Climate Change Pilot testers reviewed the draft Assessing Likelihood of Climate Change section of the guid- ance and provided feedback via the following prompts in the worksheet: • Is the risk matrix overview of likelihood and consequence scores clear? If not, how could it be improved? • Is the guidance clear on how to define an appropriate climate risk event for evaluation? • Is the guidance clear on how to assess likelihood? What additional guidance is needed? • The draft guidance does not include much guidance on evaluating consequences because there is a more direct link between consequence severity and sensitivity and adaptive capacity data from a vulnerability assessment. Do you think additional guidance is needed to translate vulnerability assessment findings into consequences? What kind of guidance would be most useful? Step 4: Test: Assessing Likelihood of Climate Change Pilot testers were then asked to apply the draft guidance to assess the likelihood of a gradual and a discrete hazard of concern facing their agency as a way of testing the effectiveness and clar- ity of the guidance. The steps to assess likelihood at the time of testing were • Define the climate risk event. • Consider the applicable time period. • Consider a distinction between gradual changes and discrete events. • Develop and apply a consistent likelihood and consequence rating scale across all risk events. The worksheet supplied prompts to guide and document testing for how to assess likelihood: • How would you define a gradual climate risk event based on the guidance? • How would you define a discrete climate risk event your agency has studied before, based on your vulnerability assessment data and the guidance? • How would you rate the current or future likelihood of your climate risk event? Did you have any challenges determining the appropriate rating? • Is the example likelihood rating scale useful? Did you develop a likelihood scale tailored to your existing data or were you able to use the example as is? • Are you familiar with the resources provided? Did you use any of these to assess likelihood? • Is the Maryland example useful? Are there other details that would be useful to include? • On a scale of 1 (low) to 5 (high), how challenging was it to apply the guidance and why? The worksheet also supplied prompts to help the pilot agencies better understand how they might use the results: • How do you envision staff using the results of this exercise? • How useful is this guidance to staff developing TAMP risk assessments in translating existing climate change vulnerability assessment information? How could the guidance be improved to feed into that process? Step 5: Check-In and Debriefing Calls Each pilot agency had a check-in call with the research team to discuss any questions or challenges reviewing and testing the draft guidance. After completing the worksheet, each pilot agency also had a debriefing call to discuss their final takeaways and responses to the prompts.

90 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Step 6: Revising the Guidance The research team updated the guidance based on the feedback received from the two pilot DOTs to improve the accessibility and usability of the guidance. The final guidance can be found in Appendix L. Input Data Used in the Pilot For testing the guidance (Step 4 in Table 9-27), the pilot agencies used common sources of climate projections (e.g., National Oceanic and Atmospheric Administration [NOAA] 2017 pro- jections, U.S. Army Corps of Engineers Sea Level Change Curve Calculator, FHWA Coupled Model Intercomparison Project [CMIP] Climate Data Processing Tool, U.S. Climate Explorer) to understand the likelihood of the expected changes, associated uncertainty, and how to apply that understanding in an asset management risk assessment. Agencies that have conducted a vulnerability assessment for some of their assets are likely to have exposure data for relevant climate hazards, which can also be used to assess likelihood. MDOT and MnDOT both used climate data sources from past work in their regions. MDOT used climate data sources to assess the likelihood of fluctuating Great Lake water levels and flooding. MnDOT used the data to assess the likelihood of increasing temperature and flood- ing. MDOT noted that the guidance did not include a recommended data source for fluctuating Great Lake water levels. The revised guidance in Appendix L includes the NOAA Lake Level Viewer as an additional data source for assessing likelihood for other Great Lake states that may be interested in applying the guidance to the same hazard. An example of how MDOT applied the guidance to flooding is provided in the following. MDOT conceptually applied the guidance to flooding as part of this pilot to vet the guidance. A sample of their worksheet response follows: 1. Discrete climate hazard: Urban flooding. 2. Defined climate risk event: Urban flooding causing some economic and social disruption This example is defined by the consequences of the event. A flood event could also be defined by the magnitude of precipitation (e.g., 4 in. per hour) or the frequency of occurrence (e.g., 100-year flood). 3. Likelihood rating: Almost certain. The event is expected to happen about once every 2 years or more frequently (i.e., annual chance ≥ 50%). For the purposes of this conceptual exercise, the rating is based on observed frequency of disruptive flood events by staff. A more detailed analysis of historical flood data and climate projections or use of existing exposure data from a climate vulnerability assessment is recom- mended to better justify the rating. For an example on how to formally analyze climate data and justify a likelihood rating, see the Maryland DOT example provided in the final guidance in Appendix L. 9.5.4 Outputs from the Pilot The ultimate output of the pilot test is the improved final guidance in Appendix L, which incorporates feedback received from the two pilot agencies. Feedback that required revisions to the guidance is summarized in Table 9-28. Some of the specific challenges pilot agencies encountered while working through the draft guidance are described in greater detail in the section following the table. For future users of the final guidance, the output is a mapping of vulnerability to risk assess- ment terms and data to assess likelihood of climate risk and more easily integrate that informa- tion into asset management planning and decision making. Although MDOT and MnDOT both

Pilot Testing and Results 91   walked through this process conceptually, both agencies felt that, with their suggested improve- ments listed in Table 9-28, they would be able to produce useful climate risk information to then integrate into their asset management processes. 9.5.5 Participating DOT Organizational Unit(s) MDOT and MnDOT served as pilot agencies for Study 9. The organizational units involved in this study were the DOT’s asset management and planning staff. Both agencies had completed a vulnerability assessment and had started to include climate change in their TAMPs. This experi- ence made MDOT and MnDOT strong candidates for piloting how to translate the outputs of a vulnerability assessment to risk information as a more applicable framing for informing the TAMP. MnDOT also engaged its TAMP Advisory Group to complete the worksheet and discuss how climate change can be considered in the TAMP. For one agency, the provided likeli- hood rating scale differed from the scale they regularly use for risk-rating exercises Revised guidance encourages agencies to consider using or modifying an existing scale to incorporate climate hazards and ensure a consistent approach. Lack of suggested data sources for as- sessing the likelihood of �luctuating Great Lake water levels Revised guidance includes the NOAA Lake Level Viewer as an example resource for assessing Great Lake water- level �luctuation. It also suggests that agencies consider using available local resources, which MnDOT did when applying the guidance. Suggestion for additional guidance on how to use the output of the study to inform transportation planning deci- sions and the possible impact of cli- mate change on speci�ic assets and operations of the transportation net- work While guidance on these topics is outside the scope of this particular study, other resources are available on these topics. A reference to the FHWA climate resilience home page has been added to the �inal guidance. This web page includes a variety of resources on how to man- age climate risks. Feedback from pilot agencies requiring improvements to the guidance How feedback was addressed in the �inal guidance Misinterpretation of the use of the ex- ample risk matrix overview provided versus risk matrices already being used by an agency Revised guidance clari�ies that the provided risk matrix overview is an example from FHWA, to be used, if needed. Additional guidance related to the example can be found in the FHWA report. The revised guidance also clari�ies that DOTs adapt their existing risk matrices to include climate change hazards in order to ensure a con- sistent approach. Challenges de�ining climate risk events, particularly gradual climate hazards Revised guidance provides additional information on how to de�ine climate risk events. There is also a new section on how to de�ine gradual climate hazards in par- ticular, as these are uniquely challenging to assess. Fur- ther the guidance clari�ies that agencies may �ind it use- ful to coordinate with a state climatologist or other in- ternal or external experts on climate change risks. Discomfort from some staff assessing the likelihood of climate risk events, as they felt they lacked climate exper- tise Revised guidance clari�ies that agencies may �ind it use- ful to coordinate with a state climatologist or other in- ternal or external experts on climate change risks if they feel they lack the expertise to apply the scale. Table 9-28. Summary of Study 9 pilot feedback and actions taken in response.

92 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 9.5.6 Who in a DOT Could Use the Pilot Results and How Study 9 is intended to inform asset management planning. MDOT would like to use the guidance to expand its next TAMP to include a broader range of hazards, including climate hazards and network-scale impacts. MDOT also felt the results of the exercise would be helpful for identifying high risks and prioritizing next steps to address high risks. MnDOT viewed this pilot study as an opportunity to address a gap in its ability to evaluate asset resilience across assets. MnDOT conducts a climate resilience analysis for specific assets but does not have a process for elevating that analysis to the network level. 9.5.7 Challenges in Pilot Study Setup MDOT found the guidance easy to digest and apply to its needs. MDOT appreciated that the guidance was not overly prescriptive, which allowed for flexibility and applicability to multiple climate hazards, including climate hazards unique to the Great Lakes region (e.g., fluctuating water levels). MDOT completed the exercise for a gradual hazard (fluctuating water levels) and a discrete hazard (urban flooding). MnDOT found the guidance challenging to introduce to its TAMP Advisory Group. MnDOT and the TAMP Advisory Group completed the exercise for a gradual hazard (increasing tem- perature) and ran out of time to fully complete the exercise for a discrete hazard (flooding). MnDOT has a risk matrix that it is comfortable using for TAMP development and had difficulty adapting to the example in the guidance. Based on this MnDOT feedback, the guidance was revised. Clarifying language was added to explain that the likelihood scale is an example, and agencies are encouraged to modify their existing scales as needed to accommodate the inclusion of climate hazards. Both agencies found the gradual hazards difficult to assess, suggesting additional guidance or outside expertise may be useful. Assessing the likelihood of gradual hazards requires an agency to define a minimum threshold that would result in significant or measurable impacts to an agency’s assets or system. MDOT struggled to determine an appropriate threshold because fluctuating water levels have both a high and a low extreme. After discussion, MDOT concluded it would be effective to assess the high and the low extreme of fluctuating water levels as two separate risks. MnDOT felt it needed climate expertise to determine an appropriate threshold for tempera- ture. The exercise of determining a critical threshold also led to the conclusion that a gradual increase in temperature may simply not be a major threat to their state, given its moderate cli- mate. MnDOT is concerned about heat wave events and feels comfortable with and capable of applying the likelihood scale to a discrete heat wave event. To address this challenge, the guidance was revised to include additional examples and guid- ance to the content on defining climate risks. See Appendix L for the final guidance. 9.5.8 Resources Needed by a DOT to Implement the Study Pilot agencies used the draft guidance and an accompanying worksheet to test the guidance. Any DOT can implement the study using the final guidance provided in Appendix L. It may be helpful to gather climate data from local or external sources.

Pilot Testing and Results 93   MDOT requested an additional climate data resource be included in the final guidance to provide data on fluctuating Great Lake water levels. Following the DOT’s feedback, an additional resource was added to the table of climate data sources in the guidance. 9.6 Study 11: Institutionalizing Risk Management into Asset Management Plans The pilot for Study 11 illustrates an approach that can be used by transportation agencies to analyze risks to high-priority critical assets and highlight the importance of managing them by incorporating risk management strategies into every section of their asset management plans. For risk management to be used consistently to improve the management of assets, its use needs to be institutionalized within transportation agencies. If it is not, the use of risk management could wane as individual advocates and SMEs retire or move into other positions. The study builds from NCHRP’s active implementation frameworks,39 which state that any innovation must be ingrained or institutionalized into the policies and processes of an agency for the innovation to be sustained. Otherwise, use of the innovation could be short lived and confined to a few advocates within the agency without achieving enterprise-wide application. This study used large bridges as an example to illustrate how to ingrain risk management into every section of a TAMP. The approach used in this study can be applied to other high-risk critical assets. It can be used by any DOT to address risks to bridges, pavements, and other assets that an agency includes in its TAMP. The information included enhances the TAMP and brings focus on the assets that create the greatest long-term performance risk. It provides an agency the opportunity to include details about its short and/or long-term plans to address these assets. It also creates a process whereby agency personnel working on the TAMP can collaborate and work with other SMEs to identify, plan, and implement processes, investment strategies, and funding to address these at-risk assets. 9.6.1 Pilot Objective and How the Results Help Inform Asset Risk Decisions The objective of Study 11 is to present a data-driven strategy that will help an agency to insti- tutionalize risk management into its asset management plan. The strategy for institutionalizing risk management into a TAMP is illustrated by selecting one of the asset classes of interest to the agency. In this study, a subset of NBI large structures (i.e., large bridges) was analyzed. The analysis shows that large structures have the potential to contribute disproportionately to the long-term risk to achieving and sustaining bridges in a state of good repair. This strategy can just as easily be implemented for a different asset class. 9.6.2 Description of the Strategy or Tool Study 11 focused on MnDOT bridges that are 10 times the size of the average Minnesota NBI structure. This study illustrates how large structures represent a risk to sustaining condition targets, supporting freight mobility, and financing bridge investment strategies. The strategy detailed in this study can be applied to other assets and asset classes. Instead of large bridges, an agency’s assets may face risks from sea level rise, slope failures, or failure to meet performance targets. By focusing on assets that create risks to performance, the study expands how risk to assets is addressed beyond the risk management section of the TAMP. In past TAMPs, DOTs have addressed risk in only one section that focuses on risk manage- ment. This study shows how an agency can bring additional focus by addressing risk in all eight sections of the TAMP required by the FHWA asset management rule, 23 CFR 515: 1. Asset management objectives 2. Asset management measures and state DOT targets for asset condition

94 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 3. A summary description of the condition of NHS pavements and bridges 4. Performance gap identification 5. Life cycle planning 6. Risk management analysis 7. Financial plan 8. Investment strategies Background to the Minnesota DOT Study 11 Pilot Test Like many state DOTs, MnDOT manages a wide array of risks to its state of good repair. This section examines one of the risks MnDOT could consider—the risk posed by the subgroup of very large structures. Large structures present several risks. Their large size increases the cost to maintain them in good condition. Often, they are on high-volume roadways and experience above-average truck loadings. Truck volumes on large bridges can make them disproportionately important for freight mobility. If they deteriorate, they contribute disproportionately to poor or declining conditions. Often, as is the case in Minnesota, some are owned by local agencies that may not have the resources to afford timely maintenance and repair. The condition of the large bridges cited in this study does not indicate the structures are at risk of failure. On the contrary, based on their condition, it could be many years before some require significant rehabilitation. However, because of the size and cost of these structures, they pose a risk to sustaining the network-level state of good repair within a fixed bridge budget. A DOT can use the approach discussed in Study 11 to address the risk to any asset group. The TAMP provides agencies an opportunity to focus on specific subsets of assets that may pose condition and performance risks and require focused attention, detailed planning, and funding. This study’s results help highlight the importance of improving the condition and performance of such assets and can serve as a vehicle for communication with both internal and external stakeholders. The approach could also bring agency personnel together to work collaboratively and help institutionalize the practice of analyzing and planning asset management strategies to address high-priority assets and bring them to a state of good repair. This study examined the inventory of large bridges in Minnesota and describes how the risks they pose can be addressed in each section of the TAMP. For this study, a subset of struc- tures that have a deck area of more than 10 times (10X) the average size of all structures was selected, however, the definition of large bridges can be changed to suit the needs of each DOT. The average MnDOT structure is 527 sq m. This study, to illustrate the approach, defines large bridges as 10 times larger, or at least 5,270 sq m. This example is intended to show how these high-priority assets can be addressed in all TAMP sections. Similar exercises can be conducted on pavements and other assets. Focus Area to Demonstrate in the TAMP The data in this analysis do not indicate a risk to public health and safety in Minnesota. The rate at which these bridges will deteriorate will be slow, over decades. Also, MnDOT and local governments are probably addressing many of these structures. Nonetheless, these structures pose a long-term risk to sustaining Minnesota’s good bridge conditions. 9.6.3 Methodology Used in Conducting the Pilot For this study, the 2020 NBI data for MnDOT were downloaded and analyzed. The steps involved are listed in Table 9-29. The methodology helps an agency highlight and bring to the forefront high-priority at-risk assets and any shortfalls in funding needed to address them, as well as the implications of not addressing these assets. This study identifies structures that create the greatest risks to achieving

Pilot Testing and Results 95   network-wide performance. These large structures present a disproportional risk to both condi- tion and performance targets. Over time, as an agency uses the strategy detailed in the study, it will become part of the routine asset management process, highlighting the importance of addressing such types of asset risks in the TAMP. Input Data Used in the Pilot The Study 11 pilot used the 2020 NBI data publicly available on the FHWA website.40 State DOTs upload the NBI data every year, so accessing these data is not difficult. DOTs also have the Highway Performance Monitoring System (HPMS) data for pavements, which can be used for similar analysis of at-risk pavements. Summary of Data Analysis A summary of the analysis of the 2020 NBI data downloaded from the FHWA website is included in this report to provide the necessary background and context for the information included in this study. Minnesota structures are in better-than-average condition. The FHWA performance dash- board on all state DOTs shows the national average percentage of NHS structures rated poor is 4.8 percent.41 According to the 2019 TAMP,42 MnDOT’s percentage of poor NHS structures by deck area (sq m) was 3.2 percent.43 Table 9-30 includes a summary of the 121 10X structures. These 10X structures are all bridges. The table shows how the 10X structures disproportionately contribute to the total percentage of structures rated in poor condition and structures that have a condition rating of 5. Bridges are rated on a 0–9 scale as defined in the NBI guide, with 9 being a new bridge. Bridges rated 5 may deteriorate to poor condition within one or two decades if not treated. Bridges rated 4 and below are in poor condition (Figure 9-22). Although the 121 10X bridges are less than 1 percent of all structures, they are 20.3 percent of all structures by deck area. They make up 38.3 percent of all poor structures by deck area. They are 24.7 percent of all structures that have a condition rating of 5 and therefore are of longer-term concern. Methodology Steps Step 1 Download the data and format as needed. Step 2 Do a quality review of the data and eliminate data elements not pertinent to the study. Step 3 Analyze the data and summarize the results of the analysis. Step 4 Compute the average of the inventory of structures, then calculate the av- erage deck area. Step 5 De�ine what will be considered a large structure for the study. Step 6 Sort and �ilter the inventory of all structures and identify structures that meet the de�inition of a large structure. Step 7 Conduct additional analysis of this subset of large assets and identify those that contribute most to the risk of not sustaining a state of good repair. Step 8 Summarize the results of the Step 7 analysis. Step 9 Develop content for each of the eight sections of the TAMP focusing on this subset of at-risk large assets. Table 9-29. Methodology for institutionalizing risk management into asset management.

96 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 1 "IMMINENT" FAILURE CONDITION—major deterioration or section loss present in critical structural components or obvious vertical or horizontal movement affecting structure stability. Bridge is closed to traffic but cor- rective action may put back in light service 0 FAILED CONDITION—Out of service—beyond corrective action. Code Description N NOT APPLICABLE 9 EXCELLENT CONDITION 8 VERY GOOD CONDITION—no problems noted. 7 GOOD CONDITION—some minor problems 6 SATISFACTORY CONDITION—structural elements show some minor dete- rioration. 5 FAIR CONDITION—all primary structural elements are sound but may have minor section loss, cracking, spalling, or scour. 4 POOR CONDITION—advanced section loss, deterioration, spalling, or scour. 3 SERIOUS CONDITION—loss of section, deterioration, spalling, or scour have seriously affected primary structural elements. Fatigue cracks in steel or shear cracks in concrete may be present. 2 CRITICAL CONDITION—advanced deterioration of primary structural ele- ments. Fatigue cracks in steel or shear cracks in concrete may be present or scour may have removed substructure support. Unless closely moni- tored it may be necessary to close the bridge until corrective action is taken. Figure 9-22. Recording and coding guide for the structures inventory and appraisal of the nation’s bridges. Source: FHWA. Summary of 10X Structures Number Deck Area (sq m) 10X Structures 121 1,443,789 Ratio of 10X to All Structures (%) 0.9% 20.3% 10X Poor 7 109,212 10X Structures Compared with All Structures (%) 0.05% 1.54% 10X Poor Compared with All Poor Structures (%) 1.06% 38.3% 10X Structures in Condition 5 27 358,932 All Structures in Condition 5 1,719 1,453,739 Ratio of 10X Structures to All Structures in Condition 5 (%) 1.6% 24.7% Summary of All Structures NHS Bridges 1,388 2,557,483 NHS Culvert Summary 357 72,212 Non-NHS Bridges 6,112 3,965,397 Non-NHS Culverts 5,614 504,377 Total Structures 13,471 7,099,469 N���: Shading highlights data referenced in the textual discussion of this table. Table 9-30. Summary of all MnDOT 10X bridges and NHS and non-NHS structures.

Pilot Testing and Results 97   Table 9-31 shows the summary of the 2020 Minnesota NBI data. The percentage of NHS bridges rated poor is 2.2 percent and the percentage of bridge deck area rated poor is 3.2 percent. Table 9-32 shows the percentage of 10X bridges that are in good, fair, and poor condition. As shown in the highlighted cells, 4.3 percent and 1.7 percent of the 10X bridges on the NHS and non-NHS, respectively, by deck area are in poor condition. This is larger than the percentage rated poor by deck area, of all NHS and non-NHS bridges. Table 9-32 also shows a comparison between all the Minnesota NHS and non-NHS NBI bridges and the 10X NHS and non-NHS bridges. It shows that 3 of the NHS 10X bridges con- tribute to 4.3 percent of the NHS bridge deck area rated poor, whereas 31 of all NHS bridges contribute to 3.2 percent of NHS bridge deck area rated poor. Table 9-32 shows that 37.1 percent, 59.6 percent, and 3.2 percent of all NHS bridges by deck area are in good, fair, and poor condi- tion, respectively, whereas 43.0 percent, 52.8 percent, and 4.2 percent of all bridges are in good, fair, and poor condition, respectively. The summary of all NHS and non-NHS 10X bridges shows that 30.5 percent, 62.0 percent, and 7.6 percent by deck area are in good, fair, and poor condition, respectively. Comparing all NHS bridges with all 10X bridges shows that the seven 10X bridges contribute to 7.6 percent of all poor 10X bridge area. Table 9-32 shows that these seven 10X bridges represent 109,213 sq m, contributing to 39.8 percent of total bridge deck area in poor condition. Prioritizing these seven bridges for treatment will make a significant impact on the overall percentage of sq m rated poor and on MnDOT’s condition targets. Table 9-33 shows the number of 10X bridges and their deck area in sq m by a 0–9 scale based on the FHWA NBI structures ratings shown in Figure 9-22. Although the data show that these bridges do not present any immediate safety threat, they represent a long-term risk to achieving and sustaining MnDOT’s bridge condition targets within finite budgets. If any of these structures deteriorate more quickly than the TAMP anticipates, the department’s financial plan and target achievement could be put at risk. These 10X bridges are the focus of the pilot for Study 11. MnDOT could either adopt this defi- nition of large bridges or change the definition and/or terminology used to reference such large bridges. This study shows how the analysis summary of such bridges could be included in the next MnDOT TAMP in almost every section. For example, based on this analysis, MnDOT could consider incorporating information from the following section in various sections of the TAMP. 9.6.4 Outputs from the Pilot TAMP Section 1: Risk-Based Asset Management Objectives Section 1 of the TAMP details the asset management objectives of a DOT. In its 2019 TAMP, MnDOT has several key transportation asset management objectives. MnDOT objectives could also specifically address these 10X bridges. MnDOT was presented the following language for consideration as objectives: • Plan and systematically invest in MnDOT’s 10X bridges that are in poor condition to improve and sustain them in a state of good repair. • Develop a long-term bridge management plan to improve or sustain the conditions of the 10X structures that have a general condition rating of 5 or below. TAMP Section 2: Risk-Based Asset Management Program Targets The following targets are in the 2019 MnDOT TAMP: • A target of no more than 5 percent of NHS bridges by deck area in poor condition • A target of no more than 8 percent of non-NHS bridges by deck area in poor condition

NHS Bridges NHS Culverts All NHS Structures Condition Bridges (No.) Bridges (Area) Bridge % (No.) Bridge % (Area) Culverts (No.) Culverts (Area) Culverts % (No.) Culverts % (Area) (Area) % Good, Fair, Poor (Area) Good 637 949,877 45.9% 37.1% 148 30,472 41.5% 42.2% 980,349 37.3% Fair 720 1,524,596 51.9% 59.6% 198 39,710 55.5% 55.0% 1,564,306 59.5% Poor 31 83,010 2.2% 3.2% 11 2,030 3.1% 2.8% 85,040 3.2% Total 1,388 2,557,483 100% 100% 357 72,212 100% 100% 2,629,695 100% N���: Shading highlights data referenced in the textual discussion of this table. Table 9-31. Summary of MnDOT NHS bridges, culverts, and structures by condition (area in sq m). Condi- tion All NHS Bridges All Non-NHS Bridges Total NHS and Non-NHS Bridges NHS 10X Bridges Non-NHS 10X Bridges Total NHS and Non-NHS 10X Bridges Bridges (No.) Bridges (Area) Bridge % (Area) Bridges (No.) Bridges (Area) Bridge % (Area) Bridges (No.) Bridges (Area) Bridge % (Area) NHS Bridges (No.) NHS Bridges (Area) NHS Bridge % (Area) Non-NHS Bridges (No.) Non-NHS Bridges (Area) Non-NHS Bridges (%Area) Bridges (No.) Bridges (Area) Bridge % (Area) Good 637 949,877 37.1% 2,959 1,853,401 46.7% 3,596 2,803,278 43.0% 23 289,811 31.0% 16 149,995 74.1% 39 439,806 30.5% Fair 720 1,524,596 59.6% 2,640 1,920,697 48.4% 3,360 3,445,293 52.8% 45 605,597 64.7% 30 289,174 24.2% 75 894,771 62.0% Poor 31 83,010 3.2% 513 191,299 4.8% 544 274,309 4.2% 3 40,341 4.3% 4 68,872 1.7% 7 109,213 7.6% Total 1,388 2,557,483 100% 6,112 3,965,397 100% 7,500 6,522,880 100.0% 71 935,749 100% 50 508,041 100% 121 1,443,790 100% N���: Shading highlights data referenced in the textual discussion of this table. Table 9-32. MnDOT 10X bridges compared with all NHS and non-NHS bridges by condition (area in sq m).

Pilot Testing and Results 99   In addition, the pilot suggested that MnDOT consider selecting one or more of the following targets to address the 10X at-risk bridges and refine them for inclusion in the next TAMP: • MnDOT will repair, rehabilitate, and/or reconstruct the 10X bridges and bring their substruc- tures, superstructures, and decks to good or fair condition within the next 30 years. • MnDOT will continue to inspect 10X bridges that are in poor condition more frequently until they are brought to a state of good repair. • MnDOT will develop within 10 years individual bridge management plans using life cycle planning for 10X bridges that are in poor condition. TAMP Section 3: Summary Description of 10X Structures Section 3 of the TAMP addressed the Minnesota asset inventories. In addition, a description of these 10X bridges and their significant impact on statewide conditions was discussed to be included. If included, later TAMP sections could then describe the impact of 10X bridges on bridge budgets. The information included in the study is based on the analysis of published data. The analysis and explanation presented in this section provide an example of what an agency could include in this section of the TAMP for any asset that they consider to be at risk. Most TAMPs include in Section 3 the summary description of asset statistics about the inven- tory, such as in Tables 9-34, 9-35, and 9-36. Table 9-31 shows that MnDOT structures comprise 1,388 bridges and 357 culverts on the NHS. It shows the bridge conditions on NHS bridges by deck area to be 37.1 percent, 59.6 percent, and 3.2 percent good, fair, and poor, respectively. In addition, the table shows that 42.2 percent, 55.0 percent, and 2.8 percent of culverts on the NHS by area are in good, fair, and poor condition, respectively. Combining both NHS bridges and culverts, the analysis shows that 37.3 percent, Rating Number Area Rating of 10X (%) Ratio of 10X Compared with the Total Area of All Structures (7,099,469) 0 0 0 0% 0% 1 0 0 0% 0% 2 0 0 0% 0% 3 1 12,310 0.9% 0.2% 4 6 96,903 6.7% 1.4% 5 27 358,932 24.9% 5.1% 6 48 535,839 37.1% 7.5% 7 32 378,787 26.2% 5.3% 8 6 51,599 3.6% 0.7% 9 1 9,419 0.7% 0.1% 121 1,443,789 100% Table 9-33. 10X bridge condition compared with all structures on a scale of 0–9 based on FHWA NBI bridge rating (area in sq m). Owner Number Deck Area Deck Area as % of All Structures MnDOT 3,676 4,507,614 63.5% Counties 5,520 1,661,884 23.4% Towns 3,370 433,384 6.1% Cities 693 433,936 6.1% Other 212 62,651 0.9% Total 13,471 7,099,469 100% Table 9-34. MnDOT structures by ownership (area in sq m).

All Bridges All Culverts All Bridges and Culverts Condition Bridges (No.) Bridges (Area) Bridge % (No.) Bridge % (Area) Cul- verts (No.) Culverts (Area) Cul- verts % (No.) Cul- verts % (Area) (No.) % (No.) (Area) % (Area) Good 3,596 2,803,278 47.9% 43.0% 4,348 404,204 73% 70% 7,944 59.0% 3,207,482 45.2% Fair 3,360 3,445,293 44.8% 52.8% 1,506 161,666 25% 28% 4,866 36.1% 3,606,959 50.8% Poor 544 274,309 7.3% 4.2% 117 10,720 2% 2% 661 4.9% 285,029 4.0% Total 7,500 6,522,880 100% 100% 5,971 576,589 100% 100% 13,471 100% 7,099,469 100.0% N����: Shading highlights data referenced in the textual discussion of this table. Numbers may not total due to rounding. Table 9-36. Summary of all MnDOT NHS and non-NHS structures by condition (area in sq m). Non-NHS Bridges Non-NHS Culverts Non-NHS Bridges and Culverts Condition Bridges (No.) Bridges (Area) Bridge % (No.) Bridge % (Area) Culverts (No.) Culverts (Area) Culverts % (No.) Culverts % (Area) Non-NHS (No.) Area (% Good, Fair, Poor) Good 2,959 1,853,401 48.4% 46.7% 4,200 373,732 74.8% 74.1% 7,159 49.8% Fair 2,640 1,920,697 43.2% 48.4% 1,308 121,956 23.3% 24.2% 3,948 45.7% Poor 513 191,299 8.4% 4.8% 106 8,689.58 1.9% 1.7% 619 4.5% Total 6,112 3,965,397 100% 100% 5,614 504,377 100% 100% 11,726 100% N���: Numbers may not total due to rounding. Table 9-35. All MnDOT non-NHS structures by condition (area in sq m).

Pilot Testing and Results 101   59.5 percent, and 3.2 percent of all NHS structures are good, fair, and poor, respectively. It also shows that the 3.2 percent poor deck area on NHS structures is attributable to 31 bridges and 11 culverts. Table 9-35 shows the summary of all non-NHS structures. The summary of the 11,726 non- NHS structures shows that 49.8 percent, 45.7 percent, and 4.5 percent by deck area are in good, fair, and poor condition, respectively. It also shows that only 1.7 percent of the culverts by area are in poor condition, and that 619 non-NHS structures contribute to 4.5 percent of all non-NHS structures (bridges and culverts) being rated poor. Table 9-36 shows that a total of 661 NHS and non-NHS structures totaling an area of 285,029 sq m contribute to the 4 percent of area rated in poor condition. Based on how an agency defines large bridges, the subset of structures analyzed to identify assets considered for such a study will vary. This definition will also influence the number of structures an agency considers as presenting risk to its performance objectives for focus and needing to be prioritized for treatment in the next 10 years of the TAMP. The study also illustrated that further insight can be gained by analyzing which components of the 10X bridges rated poor contribute to their below-average condition. Table 9-37 shows that for both the bridges rated poor and those with a general condition rating of 5, substructures were the most common component that caused the poor rating. No poor 10X bridges had decks with a general condition rating of 4 or lower. Two decks were rated 5, but the rest were rated 6. A condition rating of 6 could be interpreted as a satisfactory condition. Superstructures were the next most common component that contributed to a poor rating. Table 9-38 shows the condition of the 27 10X bridges that have one or more components with a condition rating of 5 for fair. The table also shows that 20 of the bridges have substructures with condition rated 5, 11 have decks rated 5, and 11 have superstructures rated 5. Nine of the 27 are rated 5 only because of their substructures. MnDOT is planning on moving to an element-level rating that it believes more accurately reflects the condition of each bridge. By doing element- level analysis more insight can be gained on these nine bridges, which represent 32.5 percent of all MnDOT large bridges. Addressing these would significantly improve the condition of the inventory of the 10X bridges. Bringing substructures to a state of good repair could require replacement of the entire struc- ture. This will mean that the deck and superstructures are removed to fix the substructure. In those cases, costly bridge replacement may be the most economical long-term solution. In other cases, one pier or abutment with a deteriorating condition could be the cause for the bridge rating. In those cases, the one element may be repairable. Because of the variability around SFN Owner F.Class Year Deck Super Sub Culvert NHS Condition Low Value Area 69802C State Other Princ. Art. 1972 5 4 6 No NHS P 4 5,432 9217W State Interstate 1981 6 6 4 No NHS P 4 24,240 9217E State Interstate 1981 6 6 4 No NHS P 4 28,585 2440 State Min. Art. 1917 5 4 4 No Non-NHS P 4 14,327 62080 City Min. Art. 1982 6 6 3 No Non-NHS P 3 12,310 9036 State Min. Art. 1926 6 5 4 No Non-NHS P 4 10,670 2796 City Min. Art. 1929 6 4 5 No Non-NHS P 4 13,649 Total Area 109,212 N����: SFN = structure file number; F.Class = functional classification. Shading highlights data referenced in the textual discussion of this table. Numbers may not total due to rounding. Table 9-37. MnDOT 10X bridges rated in poor (4) condition (area in sq m).

102 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research substructures, each structure would need engineering analysis to determine how to prevent the substructure from deteriorating to a poor condition. The need for such analysis could justify developing long-term bridge management plans that incorporate life cycle planning strategies for each structure as a TAMP objective. TAMP Section 4: Performance Gaps The research team presented MnDOT with the performance gap detailed in this pilot for consideration to be included in the next MnDOT TAMP. The pilot included the data used in the study to illustrate the type of data that MnDOT could consider including in this section of the next TAMP. These data indicate how one subgroup of assets is contributing disproportionately to the risk of condition gaps. MnDOT will be establishing at least 2-year and 4-year targets for the next 10-year TAMP, addressing the FHWA requirements for performance management. In addition, MnDOT will be developing state targets that are aspirational and used for the 10-year TAMP planning horizon and MnDOT’s 20-year Minnesota State Highway Investment Plan (MnSHIP). MnDOT discussed having targets specifically for the 10X bridges or other bridges it defines as large bridges. SFN Prefix Year Deck Super Sub Culvert NHS Condition Low Value Area 9030 Interstate 1961 6 5 6 N NHS F 5 47,187 27624B U.S. Highway 1993 7 7 5 N NHS F 5 30,161 27624A U.S. Highway 1993 7 7 5 N NHS F 5 30,124 9600S State Highway 1978 6 5 6 N NHS F 5 28,761 9600N State Highway 1978 6 5 6 N NHS F 5 28,620 27831 Interstate 1967 6 5 5 N NHS F 5 24,949 27727 Interstate 1978 5 6 5 N NHS F 5 15,430 27816S U.S. Highway 1982 6 6 5 N Non-NHS F 5 13,598 27816N U.S. Highway 1982 5 5 5 N Non-NHS F 5 13,302 27586 U.S. Highway 1978 6 7 5 N NHS F 5 9,640 6805 State Highway 1960 6 5 5 N NHS F 5 9,622 07042 State Highway 1985 5 6 5 N Non-NHS F 5 8,358 72013 U.S. Highway 2005 7 7 5 N NHS F 5 8,008 69816 City Street 1982 5 6 7 N Non-NHS F 5 7,579 27792 Interstate 1968 5 6 5 N NHS F 5 7,494 62838 Interstate 1973 6 6 5 N NHS F 5 7,382 69802 U.S. Highway 1972 5 5 5 N NHS F 5 7,237 9067 Interstate 1958 5 6 5 N NHS F 5 7,052 9066 Interstate 1958 5 6 5 N NHS F 5 7,049 27855 Interstate 1967 6 5 5 N NHS F 5 6,648 27V63 City Street 2009 7 7 5 N Non-NHS F 5 6,251 27R23 U.S. Highway 2008 7 8 5 N Non-NHS F 5 6,072 27728 Interstate 1978 5 5 6 N Non-NHS F 5 6,025 27241 County Highway 1984 5 5 7 N Non-NHS F 5 5,977 19825 Interstate 1973 6 6 5 N NHS F 5 5,586 62026 U.S. Highway 1965 7 5 6 N NHS F 5 5,484 27799 Interstate 1968 5 6 5 N NHS F 5 5,335 11 11 20 358,932 N����: SFN = structure file number. Shading highlights data referenced in the textual discussion of this table. Numbers may not total due to rounding. Table 9-38. MnDOT 10X bridges with general condition or components rated fair (5) (area in sq m).

Pilot Testing and Results 103   As defined by 23 CFR 515.5, “Performance gap means the gaps between the current asset con- ditions and State DOT targets for asset condition, and the gaps in system performance effective- ness that are best addressed by improving the physical assets.” The performance gap will depend on the targets established for the assets for the next 10 years of the TAMP. As more 10X bridges in poor or worse condition are treated, the overall percentage of poor bridges will decrease. The pilot team discussed that once MnDOT finalizes the targets for bridges, the targets can be included in this section of the next TAMP. In addition, based on the targets, this section of the TAMP can show the reduction in the NHS and non-NHS bridges in poor condition based on the year of treatment. Also, based on the type and year of treatment, the deck area in good and/or fair condition will also change. This section of the TAMP also provides an opportunity to show the proportional increases in the assets in good and/or fair condition to reflect the improvement achieved by each year of treating the 10X bridges. TAMP Section 5: Life Cycle Planning MnDOT is considering developing bridge management plans for many of its major bridges that will incorporate life cycle planning strategies. Based on the mitigation strategies developed in the risk management section, each of the 10X bridges may well justify its own life cycle plan. The life cycle planning section of the 2019 MnDOT TAMP details two objectives: 1. Establish a long-term focus for improving and preserving the system. 2. Determine the funding needed to achieve the desired state of good repair. The approach to the 10X bridges discussed in this study was aligned to both objectives. The following additional objective was suggested for consideration in TAMP Section 1 on asset man- agement objectives: 3. Plan and systematically invest in MnDOT’s 10X bridges that are in poor condition to improve and sustain them in a state of good repair. This suggested additional objective further aligns MnDOT’s life cycle planning to support the objective of investing in the 10X bridges rated poor (4) or fair (5) and improving and sustaining them. The research team suggested that MnDOT could describe in this section of the next TAMP how it plans to improve and sustain the condition of these bridges with life cycle strategies, given the bridges’ disproportionate importance to statewide and NHS conditions. As discussed in this study, 10X bridges create unique risks to MnDOT. They pose risks to sustaining condition targets, staying within planned budgets, and supporting freight mobility if any structures need to be load limited; thus they merit special attention. For example, ITD identified long-term cost and performance risks caused by large structures in its 2018 and 2019 TAMPs. Individual asset management plans were developed for most of these large structures to mitigate the risks. TAMP Section 6: Risk Management The risk management section could summarize the risks posed by the 10X structures and put them into the context of all other identified risks. The risk management process involves at least the following steps: 1. Identify risks. 2. Assess them by likelihood and impact.

104 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 3. Prioritize them. 4. Adopt mitigation plans for the top-priority risks. 5. Monitor the top-priority risks. Section 3 of the TAMP identified the 10X bridges as being a risk to the long-term state of good repair. Section 6 provides an opportunity to assess the risks to the long-term state of good repair posed by the 10X bridges. Assessing the risks involved identifying the likelihood and impact of their influence on the overall state of good repair. The likelihood is a measure of the probabil- ity that risks will influence the state of good repair. The risk management section of this study provided an opportunity to discuss how the likelihood of each risk could change over time and influence achieving the state of good repair. Also, the pilot illustrates that MnDOT could deter- mine that the likelihood of risks impacting the state of good repair is low because of projects and programs that are in place. The assessment also could indicate how much of an impact the bridges’ condition will have on the state of good repair if the bridges deteriorate. The impact could vary over time and pose a long-term impact but not one that is important in the 10 years of this TAMP. Alternatively, MnDOT may determine the bridges are likely to deteriorate and create significant threats to bridge conditions and budgets. Assessment and prioritization of the risks involve putting them in rank order with all other risks. It will be up to the MnDOT staff to determine if these 10X bridges pose a greater or lesser threat to the state of good repair than other risks do. Because of the slow and long-term rate of change in bridge conditions, the 10X bridges could be assessed as a long-term but not a short- term risk. In that case, little or no action may be taken. On the other hand, the assessment could indicate that the already poor structures represent a significant risk to overall conditions and performance. How the 10X bridges are assessed depends on the context of the other risks MnDOT assesses. The mitigation section could discuss the mitigation plans for at-risk bridges. As noted earlier, as MnDOT develops its bridge management plans it could include individual life cycle plans for each structure or those in the worst condition. MnDOT has two different programs to manage bridges. One is a program for major bridge projects and the other is a priority preservation list. The 10X bridges could be prioritized within these two programs or addressed by developing special funding programs that recognize the high costs these bridges present. An alternative strategy could be to only monitor the bridges’ condition and defer mitigation for a future TAMP. The nature of the mitigation plan will depend on whether the assessment process ranks them as high or critical risks or relatively low risks. MnDOT could take several steps to monitor the 10X bridges. Summary bridge inspection reports could highlight changes in these 10X structures to keep the focus on them. The results of already programmed projects that improve these structures could be monitored for their impact on the 10X structures. Or, if mitigation strategies are enacted, the monitoring could ensure that the strategies have the desired effect. TAMP Sections 7 and 8: Financial Planning and Investment Strategies The financial planning and investment strategies of an agency are often influenced by its larger assets. Larger assets are disproportionately more expensive to fix and maintain. Main- taining and managing them requires funding that can pose financial risks to an agency’s asset management plan. These sections of the TAMP address the revenue sources and the funding needs of various DOT programs and include the investment strategies and the funding needed to achieve the

Pilot Testing and Results 105   federal requirements to meet NHS bridge and Interstate pavement targets. In addition, DOTs include the investment strategies and the funding needed to meet all state performance objec- tives and targets. Based on this pilot, TAMP Sections 7 and 8 also provide the opportunity to detail the fund- ing needed to address the 10X bridges and other assets that may be considered risks to achiev- ing various MnDOT objectives and bring them to a state of good repair. This would align with MnDOT’s 10-year TAMP and with the 20-year MnSHIP. There may be a need for a long-term and, in some cases, a short-term financial plan to fund the investment strategies. Short-term activities may include various types of maintenance and preservation treatments to ensure that the structures do not deteriorate. This may be a stop-gap strategy to provide time to develop and implement a long-term plan. Depending on the investment strategies developed to address the 10X bridges and how they are addressed in programs for the major bridge projects and the priority preservation list, funds may have to be “shaved off the top” to finance them. For example, both the Missouri and the Ohio departments of transportation set aside funds solely for large bridges. Another option could be a long-term bonding plan. A third option may be to only acknowledge the long-term need and leave a solution for future TAMPs. The detailed investment plan and the companion financial plan can be included in this part of the TAMP. 9.6.5 Participating DOT Organizational Unit(s) The asset manager from MnDOT was the lead person from the DOT engaged in this pilot. Developing TAMPs and working on consistency reviews are becoming more of mainstream activities in most DOTs. Though the agency personnel involved in such a study will vary depending on the assets that are the focus of the exercise, the same personnel involved in contributing to and/or developing the DOT’s TAMP are generally engaged in such analysis and reporting. 9.6.6 Who in a DOT Could Use the Results and How The results of this study can guide the DOT in its implementation of the TAMP. For example, this pilot study was on bridges, so bridge SMEs, project planners, and programmers would ben- efit and gain a better understanding of the types of projects that need to be implemented. As with all bridge and pavement infrastructure projects, several areas of an agency collaborate to plan, design, bid, and monitor the construction to ensure successful delivery of the project. Depending on the mitigation treatment decided on by the agency, appropriate lead time may be needed to design and bid the projects. Decision makers can use the results of this study to communicate the DOT’s plan to internal and external stakeholders. In addition, the study can help decision makers and financial planners in the DOT understand the assets that create the greatest performance risk and, as appropriate, develop a plan to mitigate the risks. The information in this study can be used by the DOT in its next TAMP. 9.6.7 Challenges in Pilot Study Setup Because NBI data are easily available, the study was relatively easy to complete. If a DOT is doing a similar study for an asset class for which data are not available, then it will need to plan ahead.

106 Risk Assessment Techniques for Transportation Asset Management: Conduct of Research 9.6.8 Resources Needed by a DOT to Implement the Study This study is timely and not difficult for DOTs to include in their TAMPs. The pilot was con- ducted using MnDOT’s published NBI data, so no additional DOT resources were needed. The data are already available to the DOT and the methodology is relatively easy to follow and refine to the specific needs of a DOT. Microsoft Excel, a spreadsheet tool, was used to conduct the analysis and is a tool used and available in all state DOTs. DOTs also have a process to develop their TAMPs, and the agency personnel working on the TAMP can incorporate the study results into the TAMP.

Next: Section 10 - Protocols Developed for Studies That Were Not Piloted »
Risk Assessment Techniques for Transportation Asset Management: Conduct of Research Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

The assessment of climatic and extreme weather risks is increasingly becoming important to the operation of transportation agencies. There are easy-to-use tools and techniques that can be implemented by agencies.

NCHRP Research Report 1066: Risk Assessment Techniques for Transportation Asset Management: Conduct of Research, from TRB's National Cooperative Highway Research Program, discusses how to assess risks and summarizes 12 studies that demonstrate how to enhance the measurement of risks, quantify risks, and better link risk management processes with the appropriate tools.

Supplemental to the report are a presentation and NCHRP Web-Only Document 366: Risk Assessment Techniques for Transportation Asset Management: Appendices.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!