Performance Measures in Snow and Ice Control Operations (2019)

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.

P A R T I I Guide for Performance Measures in Snow and Ice Control Operations

85 The purpose of the Guide for Performance Measures in Snow and Ice Control Operations (referred to as the guide henceforth) is to provide insight on developing a performance frame- work for snow and ice operation management, helping decision makers to identify adjustments that can be made to manage resources effectively through the use of the proposed performance measures. Guide Focus Defining performance measures is a collaborative activity that requires a careful look at the agency’s mission, goals, and operational objectives. It is likely that no two agencies will have the same set of performance measures to assess program success or effectiveness. As agencies seek to create a core set of performance measures in these areas, it is important to note the following: • No individual performance measure is a perfect representation of the complexity of snow and ice response. A more useful approach to performance measurement and reporting would be based on multiple measures that together provide a balanced report card of agency performance. The guide identifies seven such measures that can provide an agency with a balanced report card. Not all seven measures need to be used, but assessing their suitability is suggested for all agencies. • Not all performance measures that are important to an agency can be fully controllable by the agency’s response activity. For performance measures like safety, the linkage from the snow and ice activity performed to the measure may be indistinct, but overall trends may still be valuable for the agency to support investments in snow and ice management. • Starting the process of performance measurement is the first and often the most important step. The search for the perfect performance measure or a sophisticated analytical approach can sometimes seem an undue burden on an agency. However, by focusing on what can be done immediately and in some priority segments of their jurisdiction, agencies often find that there is a path toward continual improvement of the performance measures once they begin the process of measurement. • The process of selecting a performance measure should include an assessment of the data avail- able to support the measure. Agencies should know what data are needed, where the data are available, and what attributes apply to the data (e.g., timeliness, frequency, accuracy, coverage). • For snow and ice control, some level of subjectivity in performance measurement cannot be avoided. Field personnel are often the only source of road condition updates in states with large sections of rural roadways. In other cases, only a small number of roadways may have the instrumentation required to provide data for performance measurement. With the technological revolution promised by connected vehicles still a few years away, accepting a certain level of subjectivity is necessary. Introduction

86 Performance Measures in Snow and Ice Control Operations • Performance measures identified by agencies need to be simple and easily understood not only by their stakeholders but also by their own staff. The following section defines how input, output, outcome, and impact categories are used in the guide. Input and output measures are important to agencies in informing day-to-day tactics and decision making about event response. However, for the guidance provided in this report, the focus is on the outcome and impact end of the spectrum, and the guidance is geared toward enabling a greater consistency in collecting, analyzing, and reporting outcomes and impacts associated with snow and ice control operations. To that end, measures provided in the guidance can be generally considered lag measures and primarily more suited for retrospective and strategic decision making. Defining Inputs, Outputs, Outcomes, and Impact • Input measures. Input measures represent the resources spent or used in snow and ice control operations. These resources include fuel, labor hours, machinery or equipment hours, and units of anti-icing materials or abrasives. The level of inputs is directly proportional to an agency’s winter maintenance costs and, therefore, the inputs are easily and commonly measured by transportation agencies. Despite this, inputs alone are not enough for management to assess the efficiency, quality, and effectiveness of winter maintenance. • Output measures. Output measures quantify the resulting physical accomplishment of work performed using resources in winter maintenance. Outputs include the lane miles plowed or sanded, the number of lane miles to which deicing materials were applied, lane miles of anti-icing brine applied, and other accomplishments of the maintenance process in units of work. Outputs and inputs together are generally more useful than inputs alone because together they can help to define the efficiency of winter maintenance operations by determining what level of input was or will be required to achieve a level of output. These measures may also be based on time and storm event. • Outcome measures. Outcome measures generally attempt to assess the effectiveness of the winter maintenance activity, often from the perspective of the user or customer. Outcomes are inherently difficult to measure. A desired outcome of winter maintenance could be the improvement of safety, mobility, or user satisfaction. However, the relationship between outcome measures and outputs and inputs is complex and not easily described. • Impact measures. Impact measures are the long-term (sometime indirect) effects of the outcomes (i.e., what is expected to happen or to be accomplished by the winter maintenance efforts). Impacts are typically difficult to measure and evaluate given their wide time span and the fact that their occurrence is not guaranteed. Guide Audience The target audience for this guide is staff, particularly those in management roles, at agencies responsible for snow and ice control, such as state and local government agencies. The guide is intended to be a resource to assist agencies in monitoring their level of performance with respect to a number of different aspects of winter maintenance and making appropriate adjustments to manage resources more effectively. The guide will also be of interest to contractors working with highway agencies and seeking to understand how to evaluate their performance and communicate with their clients about it. Policy makers seeking to understand how performance with respect to winter maintenance activities can be measured while taking into account the many external factors that influence performance would also benefit from the guide, which is illustrated through the examples from

Introduction 87 peer agencies. Finally, researchers and contractors for agencies responsible for winter maintenance can use the guide as a resource for information about the state of the practice with respect to performance measurement for winter maintenance. Ongoing Challenges with Snow and Ice Performance Measurement Measuring agency performance during snow and ice response is an ongoing challenge. This is mainly due to the differences in seasons, regions, roadway types, agency types, customer expectations, travel patterns, funding availability, and available strategies and tactics, all con- tributing to an inability to develop a consistent set of measures that are easily agreed on by all. Table 1 identifies the primary differences, challenges, and associated implications considered in the development of the material included in the guide. Guide Organization and Use In the development of the guide, the researchers attempted to create a rational approach that allows flexibility to its users. Since the audience for the guide is time-constrained, the guidance is designed to be succinct and directly actionable. Also, the guidance is intended to support agencies Issues Challenges for Performance Measurement Implication for Guidance Frequency of snow and ice events in locations Regions and agencies that rarely experience snow and ice events have limited capability and no sustained investments in winter maintenance. However, when the rare event occurs, it needs to be responded to with available resources. The relative unfamiliarity of the traveling public with driving on snow and ice also creates additional challenges. On the other hand, agencies/regions that have high frequency of snow and ice events have dedicated programs geared toward winter maintenance. While the public may be more comfortable in handling snow and ice, they may have higher expectations for snow and ice removal. Defining strategic missions, goals, and objectives needs to be accomplished prior to identification of performance measures. Variations in snow and ice events between seasons Regardless of the frequency of events, seasonal variability in a region can complicate performance measurement. Agencies may need to determine how much of any change in performance can simply be attributed to weather event severity between seasons. Measures need to factor in seasonal and event-level variations. These measures could range from moving averages to severity indices that drive performance curves for an agency. Differences between travel patterns and travel demand and between roadways/ locations during snow and ice events An event, while similar in precipitation amount to previous ones, may have a different impact based on its timing and location. A snowstorm around peak travel times on a highly congested Interstate has a significantly different set of issues than one on a rural Interstate during the night. Each performance measure needs to account for functional class and travel demand. Winter maintenance organization setup Winter maintenance programs at agencies are set up differently, with a mix of public, public-only, and contractor- only arrangements. Maintenance responsibilities in an area are often fragmented and overlapping. While similar performance measures can be used, the organization setup can drive target setting for contract operations. Differences in winter maintenance Agencies differ in their approaches to snow and ice response. Some agencies have a higher threshold for proactive warnings, road closures, and vehicle restrictions than others. Similarly, agencies may differ in the required standard for level of service (LOS) of their infrastructure. Comparing an agency that uses a particular practice extensively versus one that rarely or never uses that practice could be problematic (e.g., one that closes the road versus one that does not). Performance measures need to be based on the existing tactics and the associated response and practices recovery criteria used by an agency. Differences in practices should be considered when comparing agencies. Table 1. Issues for snow and ice performance measurement and implications for guidance.

88 Performance Measures in Snow and Ice Control Operations with different existing capabilities, from those that are beginning the process of collecting data to high-capability agencies that are able to adequately capture outcome measures. The guidance is structured into four chapters and 10 steps, as shown in Figure 1. • Chapter I provides guidance on defining performance measures. To adequately identify performance measures, agencies need to develop clear objectives that drive agency practices and subsequently what occurs as part of the response. • Chapter II provides guidance on implementing the performance measures, including identifying the data needs and the capabilities required. • Chapter III describes the use of the performance measures, including defining a baseline, setting targets, and reporting performance. • Chapter IV provides the linkage to broader decision making and strategic planning in an agency, with continual evaluation and updates to the performance measures. Chapter IV: Reinforcing Performance-Based Management Step 9 – Integrate into decision making Step 10 – Evaluate process and identify improvements Chapter III: Using Performance Information Step 7 – Set targets and establish baseline Step 8 – Report performance Chapter II: Implementing Performance Measures Step 5 – Inventory current practices and gaps Step 6 – Identify data sources and needs Chapter I: Defining Performance Measures Step 1 – Review mission and goals Step 2 – Refine operational objectives Step 3 – Identify performance measures Step 4 – Develop analytic approaches Figure 1. Organization of guidance.

89 Four steps are required for defining performance measures, as shown in Figure 2. C H A P T E R I Defining Performance Measures Defining Measures Figure 2. Steps for defining performance measures. Step 1: Review Mission and Goals Determining what constitutes effective snow and ice control performance by an agency is linked to the agency’s mission and goals. As illustrated in Figure 3, the mission and goals direct the operational objectives of an agency, which are used to set the performance standards that drive the identification and development of performance measures. Therefore, the first step for an agency is to review its stated mission and goals and how they relate to snow and ice control. Snow and ice control is one part of a larger agency mission of safe, reliable, and sustainable operations. How critical snow and ice control is to the overall mission and goals drives the invest- ment in this area and the corresponding emphasis on performance measurement. The stated mission and goals also determine the constraints on snow and ice operational practices and poli- cies. For example, an agency with either an implicit or explicit goal of never closing roads will have a different set of performance expectations than one where road closures are used widely. In large part, missions and goals are driven by the type of jurisdiction the agency manages and the public-sector role in snow and ice control. Table 2 identifies key elements in missions and goals and how they drive performance expectations and ultimately the selection of performance measures. Step 2: Refine Operational Objectives Once the mission and goals of an agency are reviewed, operational and maintenance objec- tives can be developed or refined to meet those goals. Operational objectives are important since they directly correlate to performance measures. If there is an operational objective, it is Mission and Goals Operational Objectives Performance Standards Performance Measurement Direct Set Drive Figure 3. Link between goals and performance measurement.

90 Performance Measures in Snow and Ice Control Operations Question Options Impact on Performance Expectations and Measurement How important is snow and ice control to the agency’s mission? High priority. Occurrence of snow and ice is routine and needs to be managed, which is a primary component of operations during winter. Agencies are likely to have a dedicated winter maintenance program. Existing protocols allow for ability to focus LOS during events. With snow and ice removal a major portion of the budget, there is stronger emphasis on efficiency. Travelers are used to snow and ice conditions and have high performance expectations during such events. Medium priority. Snow and ice is a factor but not the dominant concern. Agencies are likely to have a maintenance program that includes defined roles for winter maintenance, but staff may be drawn from different roles/functions. Likely to not use as many resources as high-priority agencies do. High-severity events cause a challenge for agencies to respond and maintain LOS. However, recovery from such events is important. Travelers may not be as used to snow and ice conditions. Expectations of driving conditions may be unfounded. Low priority. Snow and ice conditions are not frequent in the jurisdiction. Agencies typically have ad hoc responses to snow and ice events, which are rare. With little or no dedicated resources, agencies tend to focus on recovery from the event, knowing that such an event would result in higher costs than expected. Is there a handbook/policy for snow and ice control? Yes A written policy allows for more consistent use of winter maintenance tactics across an agency’s different districts/regions. Policies for closures, vehicle restrictions, and messaging directly influence operational objectives for the agency that in turn determine performance. No Agencies without a written policy may end up with ad hoc approaches across the different districts/regions. Greater effort is needed to develop a common, agreed-upon set of performance measures for agencies of this type. What is the nature of the jurisdiction that the agency manages? Mostly urban Snow and ice control in large urban areas presents opportunities and challenges for performance measurement. Measuring reliability of travel, especially around important commuter sheds and peak hours (a challenge in normal weather) is critical during snow and ice events. Distinguishing the impact of weather from the regular spatial and temporal extent of congestion is a challenge. Agencies typically have high levels of instrumentation, monitoring and detection systems, and availability of probe data to support more detailed performance measurement. Mix of urban, suburban, rural Agencies with a mix of urban, suburban, and rural facilities are likely to require a prioritized approach to response and recovery, with clear identification on levels of priorities on their different facilities. These different priorities can result in significantly different performance targets within a region. Largely rural Largely rural areas have limited detection but also low volumes. They are likely geographically large and require a lot of time to respond to conditions. Expectations of high levels of service during snow and ice events may be unrealistic, but recovery from these events may be more critical. Recovery of roadways that are lifelines for rural connectivity becomes a priority. What is the public-sector role in snow and ice response? Public sector only Performance and cost control are important, but the ability of a public agency to ramp up or ramp down resources may be limited. Contracted services (fully or in portions of the agency) When snow and ice control is handled largely through contracts, there is a greater need for clear performance specifications around response and recovery and monitoring/reporting of such specifications to enable contractual compliance. The public-sector role in oversight also requires a stronger reliance on performance data. What type of operations does an agency manage? Freeways only Agencies are likely to have high levels of service and capabilities to manage snow and ice response on these roads since they are likely to be the highest priority. Arterials only Other aspects of levels of service, such as considerations of sidewalks, bike infrastructure, and signal progression, may drive performance expectations. Mix of roadways Agencies are likely to have a varying mix of performance requirements for their roadway types. Transit only Snow and ice conditions can greatly affect surface transit use. While similar concepts of performance measures apply for both transit and roadways, the definitions and the implementation of the measures may vary. What type of weather- response traffic management strategies are in use by the agency? Low level of operational strategy use At a low level of operational strategy use, an agency might be still influencing demand on travel through messaging. High level of operational strategy use At high levels of operational strategy use, agencies seek to actively manage the roadways before, during, and after an event. From a full suite of advisory and control strategies, agencies can set restrictions on speed, lanes, and vehicles, can open/close lanes/roadways, and can actively support demand management (route, time, mode choices). In such cases, it is important to revisit the expectations of the performance measures in light of these operational strategies. For example, a static, speed-based LOS or recovery criterion is not useful in a context of variable speed limits. Table 2. Impact of agency mission and goals on performance measurement.

Defining Performance Measures 91 necessary that there be a way to measure the achievement of the objective. In other words, the operational objectives drive the performance standards. The achievement of the performance standards then becomes the primary performance measure for the agency. These objectives drive the nature and cost efficiency of the response. An agency that sets a high level-of-service (LOS) performance standard during an event could incur more cost than one that has a lower or no LOS performance standard. Similarly, recovery objectives (e.g., how quickly roads/systems will be cleared) drive the level of response activity. Where possible, operational objectives should be defined so that they are quantifiable through one or more of the measures listed in the following subsections. For snow and ice control, outcome-related operational objectives are ideally defined within the seven categories identified in Figure 4. It is important to note that environmental sustain- ability in the context of snow and ice control is defined from an agency’s perspective rather than a societal perspective. From an agency’s perspective, sustainable operations are defined by their environmental stewardship, efficiency of response, and the satisfaction of their traveling public. Each of the seven objectives sets a performance standard that can then be measured and reported. These objectives span pre-storm preparations, during-event conditions, and post-event recovery. These objectives are also not mutually exclusive. In fact, achieving all the objectives may require a balancing act that accounts for the increased resources needed to meet higher LOS and recovery objectives. Implied in the objective is a definition of an event. The agency must have the ability to clearly distinguish when a weather event started and ended and the time when conditions return to normal (see Box 1). As previously stated, while start times may be easy to determine, end times of weather events can be more difficult to define. In addition, start and end times may be difficult to define for some events (e.g., icing events, multiday events, start-and-stop events, and overlapping events). If such events are infrequent, they can be treated as exceptions; if they are common, an agency can apply start and end time criteria that are based on field personnel observations and are long enough to account for the impact of the event. LOS During Event Operational objectives in this area correspond to maintaining travel during the event at an acceptable level. The acceptability may vary based on the type of roadway and type of agency. Figure 4. Categories of operational objectives.

92 Performance Measures in Snow and Ice Control Operations In some cases, an agency might not want to specify an LOS during the event, which is likely when an agency is willing to close roads as part of the winter response or if transit service is suspended. The reason for defining an LOS during an event is twofold: • It establishes an expectation based on severity and type of roadway on what the agency is expected to provide, and • When contractors are used, having an LOS during an event allows for monitoring the response quality more effectively. Collecting LOS data during an event is a complex task, and data are likely to be available only for selected segments. Three different approaches are possible, based on agency capability, as shown in Table 3. While none of these are perfect, an agency can make progress by starting with a few priority segments to define this objective. In addition, defining the LOS depends on the nature of the operational strategies involved. For example, speed-based LOS definitions are likely not well suited for an agency that relies on variable speed limits extensively. In such a case, the spread of speed distribution might be a better LOS measure. Once an agency selects an approach, it needs to specify the LOS objective based on severity of the event and the priority of the roadway, since it cannot be credibly expected to provide the same LOS for all types of events and all roadways. Figure 5 illustrates the relationship between severity and priority of roadway types. Recovery from Event This objective focuses on how quickly an agency recovers from the event to a normal condi- tion determined based on historical averages. The sooner the recovery the better, but quicker recovery requires a higher level of response resources. Also, all parts of the agency’s system are Box 1. Define a Policy for Identifying Start and End Times for Snow and Ice Events Having a clear policy on what constitutes the start and end of a weather event is essential for monitoring performance because it underpins many of the core performance measures. While there are many possible options to define start and an end times, an agency needs to select one that works and allows it to clearly demarcate times that define before, during, and after an event. A potential approach could be based on estimated time that winter precipitation started and stopped for a given road segment; this may be logged from nearby Road Weather Information System (RWIS) snow accumulation data, observed weather radar, or agency personnel reports. For a region, start time could be estimated by the first recorded time stamp of precipitation within the group of environmental sensing station (ESS), weather radar, and agency personnel reports in the selected region. The end time would then be the last recorded time stamp of precipitation in the group. Alternatively, for a region, start and end times could be estimated by a determined percentage of sensors reporting precipitation or the lack thereof. (e.g., 6 of 10 sensors report precipitation at a specific time).

Defining Performance Measures 93 Ways to Define LOS During an Event Approach Comments Maximum accumulation during the course of the event (Measured). This LOS is calculated for sections where there is a Road Weather Information System (RWIS) station present to observe continuous accumulation only. This is limited by the amount of RWIS stations deployed and might be a good option for agencies with large-scale RWIS deployment. (Estimated/reported). This LOS is calculated for sections where field personnel reports are available during periodic intervals. Obtaining an estimate of accumulation from frontline personnel is difficult. However, such estimates could be collected from supervisory staff in the field. Maximum drop in roadway friction Roadway condition during the event (Measured). This LOS is calculated only for sections where there is an appropriate sensor present to report continuous grip factor or other friction readings. This is limited by the amount of deployed sensors and might be a good option for agencies with large- scale deployment. (Estimated). This LOS is calculated based on field personnel reports during the event. Estimates of road conditions can be qualitatively provided by tools like pavement snow and ice condition (PSIC) index. However, getting sufficient reports during an event may be problematic. Maximum allowable drop in speeds (Segment-based) This is based on point measurements of speed through agency-owned sensors and is possible where such sensors are deployed. (Probe-based) This is based on getting travel speeds on a significant portion of the network through probe data from third-party providers. Table 3. Approaches to collect LOS data during an event. Variation of LOS Objective by Severity and Roadway Type Low Medium High Priority Roads “A” Maximum LOS Objective Priority Roads “B” Priority Roads “C” Minimum LOS Objective Increasing Event Severity Decreasing Priority Figure 5. Variation of LOS objective by severity and roadway type.

94 Performance Measures in Snow and Ice Control Operations not likely to have the same recovery targets. Objectively defining what constitutes “recovery” or is a return to normal is the primary challenge in measuring this objective (see Box 2). Figure 6 illustrates the difference between the concept of LOS during an event identified in the previous section and this objective. Recovery from an event is typically specified in time to reach a specific condition after the event. The attainment of the condition can be measured in the following ways: • Return to normal or specific pavement conditions (reported by field personnel), • Return to normal or pre-event roadway friction/grip factors, or • Return to normal speeds measured by sensors or probes. Figure 6. Difference between LOS and recovery objectives. Box 2. Examples for Recovery from Event Alaska DOT defines the following priority levels for recovery, noting that winter conditions vary with response times depending on the severity and length of each winter storm (Alaska Department of Transportation and Public Facilities 2015). • Priority Level 1 – High-volume, high-speed highways, expressways, minor highways, all safety corridors, and other major urban and community routes. May take up to 24 hours to clear after a winter storm. • Priority Level 2 – Routes of lesser priority based on traffic volume, speeds, and uses. Typically, these are major highways and arterials connecting communities. May take up to 36 hours to clear after a winter storm. • Priority Level 3 – Major local roads or collector roads located in larger urban communities. May take up to 48 hours to clear after a winter storm. • Priority Level 4 – Minor local roads that provide residential or recreational access. May take up to 96 hours to clear after a winter storm. • Priority Level 5 – Roadways that are designated as “no winter maintenance” routes (e.g., Denali Highway or Taylor Highway). Generally cleared only in spring to open road for summer traffic. Other examples are Ohio DOT’s use of recovery of travel speeds on segments using data collected from a traffic speed data vendor and Wyoming DOT’s use of a road condition reporting system that allows trained citizen reporters to provide surface condition reports that are geo-tagged and time stamped at the transportation management center (TMC). In the case of Wyoming, these civilian reports are complemented by direct reports from snowplow operators on surface conditions through a tablet mounted in their snowplows. Outputs include a measure of time to recover surface conditions and a map of recovered and unrecovered segments.

Defining Performance Measures 95 Travel Reliability During Event This measure is closely related to LOS and attempts to determine the drop in travel reliability due to a snow and ice event. Travel reliability is called out as a separate objective because of its potential for providing more trip- or route-based assessment of impact (as opposed to more segment-level LOS). For example, reliability-based operational objectives may be used to focus on certain priority trips/corridors and periods during snow and ice events. This measure also takes advantage of new and emerging data sets that provide wide area coverage of travel speeds through the use of probe vehicles and third-party data providers. Travel time data can be obtained from the National Performance Management Research Data Set (NPMRDS). The NPMRDS is made available free of charge to state departments of transportation (DOTs) and metropolitan planning organizations (MPOs) by FHWA through acquisition of travel time data from a private vendor. The NPMRDS is obtained from anonymous GPS probe data from a wide array of commercial vehicle fleets, connected cars, and mobile apps. The NPMRDS uses travel times as a way to measure, monitor, and report the health of road networks. It is considered the baseline data set for meeting federal congestion and performance reporting regulations. As a first step for the analysis, base average travel times void of adverse weather for the entire winter season should be computed on the selected corridors. These base travel times will help in comparative analysis with the adverse weather events. Historical travel time data from multiple years should be analyzed to derive base average travel times for all periods of the winter season. The travel times can be aggregated to analysis periods of 5 to 15 min based on the desired precision for reporting the reliability measure. Extensive data mining of the NPMRDS is needed to derive the base travel times for the entire span of the winter season. The base conditions should be computed for each day and all times of day during the season. While computing the base travel times, care should be taken to exclude the historical data biased due to presence of extreme weather and incident conditions. After the base travel times are established, the average travel times over the selected corridors should be computed for the adverse weather events in the analysis year. For post-event perfor- mance measurement of the maintenance activity, for the selected corridors, the average travel times over the event horizon should be computed from the NPMRDS and aggregated to the same analysis period (5 to 15 min) as the base travel time data. The storm event duration can be a few hours during a single day or can span multiple days. The average base travel times are computed for the exact same duration (day and time) as the storm event from the historical data. Because NPMRDS probe data are sparsely available during nights and weekend off-peak hours, especially during storm events, the data can be supplemented with other reliable sources of travel time data to increase the coverage of travel time data during storm events. The planning time index (PTI) and buffer index (BI) are often used for travel time reliability evaluation. PTI signifies the total travel time the traveler would need for on-time arrival, while the BI illustrates the buffer time that a traveler should add to the average travel time to ensure on-time arrival. SHRP2 Report S2-L03-RR-1 (Margiotta et al. 2013) suggests using BI as a second- ary reliability metric because it can be erratic and unstable. These indices can be computed based on the following formulations: ( )=BI 95th percentile travel time – average travel time average travel time =PTI 95th percentile travel time free-flow travel time The PTI and BI should be computed for the event duration both for the exact same duration base travel times and travel times during adverse weather events.

96 Performance Measures in Snow and Ice Control Operations Figure 7 shows an example of travel times over a hypothetical corridor both during normal weather and adverse weather; Table 4 illustrates computation for BI and PTI. During adverse weather events, each of these indices over the selected corridors should be within an accept- able difference compared to the indices in the absence of adverse weather. In this case, if the acceptable differences were 30% for PTI and 10% for BI for the type of severity, this particular route would have failed the operational objective and performance standard. Safety-Related Objectives Important objectives of winter maintenance activities are to improve safety and reduce the risk of fatal crashes for all transportation system users in the midst of adverse winter weather. Crash data can provide valuable insight into snow and ice control operations as an indication of success of implemented maintenance activities. It is recognized that crashes, and especially fatal crashes, are an infrequent phenomenon. Further, travelers may choose to defer trips, leading to reduced demand on the system and potentially reducing crash rates during an event. However, the risk of crashes is higher for those that do travel. When adjusted for the reduced volume, crash rates during snow and ice are higher than on dry days. While many of the crashes that occur during winter weather are minor, objectives related to fatalities and serious injuries are consistent with the MAP-21 Safety Performance Management 0 5 10 15 20 25 30 6:00 6:10 6:20 6:30 6:40 6:50 7:00 7:10 7:20 7:30 7:40 7:50 8:00 8:10 8:20 8:30 8:40 8:50 9:00 9:10 9:20 9:30 9:40 9:50 A VE RA G E TR AV EL T IM ES (M IN .) ADVERSE WEATHER EVENT HORIZON Normal Weather (left) Adverse Weather (right) Figure 7. Travel times over a hypothetical corridor for normal and adverse weather. Performance Indicators Normal Weather Adverse Weather Difference Free-flow travel time (min) 13.82 13.82 – Average travel time (min) 24.7 32.02 7.32 95th-percentile travel time (min) 32.1 42.9 10.8 PTI 2.32 3.10 0.78 (34%) BI 0.30 0.34 0.04 (13%) Table 4. Performance indicators.

Defining Performance Measures 97 Measures Final Rule. The rule established five performance measures as the 5-year rolling averages for (1) number of fatalities, (2) rate of fatalities per 100 million vehicle miles traveled (VMT), (3) number of serious injuries, (4) rate of serious injuries per 100 million VMT, and (5) number of nonmotorized fatalities and nonmotorized serious injuries (FHWA 2013). As such, agencies are expected to develop methods to calculate the measures annually and also to set targets. Fatality measures also support overall highway safety improvement programs, including initiatives like Vision Zero, that are being considered in different parts of the country. Other safety objectives that may be included but are harder to set for snow and ice operations are secondary crashes during weather events and incident clearance times during weather events. Neither is as intuitive or clear-cut as fatalities and crashes. Defining secondary accidents continues to be a challenge, despite the significant role agencies have in incident management. Level of Customer Satisfaction The perception of roadway users is a valuable assessment measure since it allows agencies to better understand the needs of travelers in moments of distress. In this sense, the customer satisfaction objective can be used to measure the impact of current and future operational practices. Direct customer feedback, while subjective and expensive, can help ensure support for snow and ice control operations. With recent developments in apps and smartphone technology, new capabilities are starting to emerge for effective and cost-effective citizen engagement. A desired objective of customer satisfaction can influence how other operational objectives are set. Efficiency-Related Objectives Winter maintenance operations have a significant economic impact, both from an agency and a user perspective. Users’ economic losses due to winter events are mainly tied to safety and mobility issues, such as crashes and unexpected delays. For example, Sasha and Young (2014) estimated, through a conservative approach, that road closures on I-80 in Wyoming had an economic impact of around $11.7 million per 8-h closure for the freight industry. Road closures, however, can help agencies avoid elevated incident costs and the costs associated with rescuing injured and stranded motorists. Another important element is the lost economic activity associated with snowstorms. A study quantified the economic cost of 1-day shutdowns due to snowstorms in 16 states and two Canadian provinces and noted impacts in the $300 million to $700 million range for a 1-day shutdown (IHS Global Insight 2014). Although estimating the costs of winter maintenance is straightforward, estimating the economic benefits is challenging, and calculating the economic impact is even more difficult, with no consensus on the best approach. Cui and Shi (2015) suggested that the economic benefits of winter maintenance can be estimated by evaluat- ing energy savings in fuel costs, reduced wage loss from work absenteeism, reduced production losses, and reduced delays in the shipment of goods. In general, overall system efficiency objectives from a user perspective are captured by the previously described objectives. However, agencies are likely to have efficiency-related objectives that help ensure that the cost of maintenance operations is consistent with the severity of the event and the season. The more severe the season, the higher the expected cost of maintenance. Environmental Stewardship–Related Objectives Material resources have significant infrastructure and environmental impacts, such as in the case of chloride salts. Managing the environmental impacts of winter maintenance operations can be accomplished through appropriate salt management (Fay et al. 2013), the environ- mental impact of which has been studied in depth (Roth and Wall 1976; Hawkins 1971; Paschka

98 Performance Measures in Snow and Ice Control Operations et al. 1999; Ramakrishna and Viraraghavan 2005; Fay and Shi 2012). Research confirms that repeated applications of chloride-based deicers (salts) and abrasives, among other factors, adversely affect adjacent soil and water bodies, thereby affecting vegetation, aquatic biota, and wildlife (Buckler and Granato 1999; Levelton Consultants 2007; Venner Consulting and Parsons Brinckerhoff 2004). Abrasives such as sand (used to provide temporary traction improvement on icy surfaces) have been found to clog stormwater catch basins, harm aquatic animals, trigger respiratory problems, and remain in the environment even after cleanup (CTC and Associates and Wisconsin DOT 2009; EPA New England 2005; FHWA Office of Operations 2017). However, salt is an essential aspect of snow and ice control operations. Achieving other operational objectives requires the use of appropriate amounts of salt. However, if an agency is able to model and benchmark its salt usage based on historical trends and severity, an environ- mental stewardship objective could be tied to the ability of the agency to maintain the level of salt use as dictated by the model. The stronger the correlation between severity and salt use, the better the ability of an agency to monitor and report on this objective. Step 3: Identify Performance Measures Once the operational objectives are set, the identification of performance measures is fairly straightforward. Seven performance measures have been identified to support monitoring the seven objectives. However, each performance measure includes different variations for use depending on the agency’s current and desired capabilities. Table 5 shows how the operational objectives relate to the identification of performance measures. These performance measures were identified to support a wide variety of agency capabilities, types, and functions. Together, these measures provide a balanced view of outcomes and impacts associated with snow and ice control operations. Each of the performance measures is described in detail in the following sections. Table 6 highlights the efficacy of the measure in supporting agency and contractor decision making. Percent of Time Road Segments Meet Agency-Defined Level-of-Service Thresholds During Winter Storms This is an event-based measure that assesses whether service-level thresholds were maintained by an agency during an event; the measure thereby directly measures agency performance of Objective Identified Performance Measures Maintain level of service during event Percent of time road segments meet agency-defined level-of-service thresholds during winter storms Meet recovery criteria set by agency Percent of segments meeting time to regain or recover to acceptable criteria for agency-defined segments after the end of event Meet reliability targets for specific routes Percent of trips within accepted difference between measured travel time index and additional expected travel time index for snow and ice events for selected routes Support safe operations of the roadway Five-year rolling average of fatalities and injuries (number, rate) during a winter season Meet customer satisfaction ratings Customer satisfaction ratings for snow and ice response Support efficient use of resources to meet operational objectives Cost of snow and ice control to meet established performance criteria for a given winter severity Support environmental stewardship goals by optimizing material use Agency within acceptable difference between expected and actual use of salt and other materials in a season Table 5. Relationship between operational objectives and performance measures.

Defining Performance Measures 99 winter maintenance activities. Service-level thresholds vary by type of roadway and winter severity but are often set by an agency as a surrogate for crash risk and to some extent mobility needs. Service-level thresholds are also often used for monitoring contractor performance. Measure Definition This performance measure assesses whether one or more preestablished service-level thresholds were maintained by an agency during an event. A service-level threshold defines the acceptable road conditions during an event. The percentage of the event duration during which the service threshold was met is compared to the chosen threshold for the segment. Agencies must be capable of monitoring the service-level threshold for the duration of the event. The measure an agency selects to report may vary based on the availability of data and resources required to calculate the data, what is suitable for local conditions, and operating policies and constraints. Objective measures of service quality during an event, such as measured accumulation amounts or measured grip factors, are preferred. Identified Performance Measures and Applicability to Agencies Type of Jurisdiction Type of Operations Modes Covered Urban Rural Mix Public Contracted Interstates Arterials Mix Transit Percent of time road segments meet agency- defined level-of-service thresholds during winter storms High Low. Unlikely to have LOS thresholds during event Low High High High Medium. More difficult to establish and report service levels for arterials Medium High Percent of segments meeting time to regain or recover to acceptable criteria for agency-defined segments after the end of event High High High High High High High High High Percent of trips within accepted difference between measured travel time index and additional expected travel time index for snow and ice events for selected routes High Medium Medium High Low since the measure is trip-based and different segments have different contract operations High Medium Medium High Five-year rolling average of number of fatalities and injuries during a winter season — — — High. Trends provide valuable information about agency priorities Low. Not controllable but still might be a valuable indicator — — — — Customer satisfaction ratings for snow and ice response — — — High High — — — — Cost of snow and ice control operations to meet an established performance criteria for a given winter severity — — — High High — — — — Difference between expected and actual use of salt and other materials in a season — — — High High — — — — Table 6. Efficacy of each performance measure in supporting agency and contractor decision making.

100 Performance Measures in Snow and Ice Control Operations Service-level thresholds can be stated in one of the following forms: • Road conditions–related. Based on assessment by field personnel into LOS A to F (see example later in the section). • Accumulation-related. Based on measured snow and ice accumulation (e.g., hours where accumulation was below maximum allowable depth target during an event). This is measured either by field staff or by spot-specific Road Weather Information System (RWIS) stations. • Friction-related. Based on measured grip factor (e.g., hours of grip factor rating of roadway being greater than target during an event). Grip factor is measured at specific sites along the roadway. • Travel speed–related. Based on measured travel speeds (e.g., hours where percentage of average travel speeds were above a minimum expected speed during an event). Speeds are expected to drop during weather events, especially if the agency uses strategies such as variable speed limits. • When a road is closed, the service level is null. The selected service-level threshold is defined by the agency and will vary by one or more of the following: • Roadway functional class. Higher levels of service may be expected for higher priority roadways (e.g., Interstates). • Observed severity of the event. Lower levels of service may be acceptable for more severe events. • Day of week. Lower levels of service may be acceptable for certain days (e.g., Saturdays, Sundays, holidays). • Time of day. Lower levels of service may be acceptable during certain periods (e.g., midnight– 6:00 a.m.). Implied in the Definition • Event duration. Agencies should select a period that consistently encompasses most winter weather events. Different agencies define different durations for a winter event (e.g., from first to last inch of accumulation or a combination of weather indicators). As an example, an agency may establish the following criteria: • Interstate highways should have no more than 1 in. of winter precipitation accumulation 100% of the time during a winter event with 1 in. or less accumulation per hour. • Interstate highways should have no more than 1 in. of winter precipitation accumulation 75% of the time during a winter event with more than 1 in. of accumulation per hour. • Other highways should have no more than 2 in. of winter precipitation accumulation at least 80% of the time during a winter event with 1 in. or less accumulation per hour. • Other highways should have no more than 2 in. of winter precipitation accumulation at least 50% of the time during a winter event with more than 1 in. of accumulation per hour. • Secondary roadways are not subject to this performance measure. A real-world example is from the Idaho Transportation Department, which reports the per- cent of time highways are clear of snow/ice during winter storms, with a target of maintaining at least 73% unimpeded mobility for the winter season. This measure is reported on sections having RWIS stations installed as the length of time with no accumulation present during the event. Weaknesses and Limitations of the Measure • What is considered an acceptable condition may differ between urban and rural areas or by road type.

Defining Performance Measures 101 • The accuracy of this measure will depend on the ability to report service levels consistently during an event, which could differ significantly between urban and rural areas or by road type. • Subjectivity is a concern for service levels based on field reports since different reporters may view road conditions differently. Use of a structured process can mitigate some of the subjectivity but not all. In general, field personnel like snowplow drivers are required to report conditions. • This measure is not appropriate for agencies that routinely use closures as a part of their winter maintenance strategies, particularly in remote or mountainous areas. • The flow of events is a critical factor because anticipated conditions may differ from actual conditions, which could lead to an under- or overestimation of the measure. Variations in the Measure • Agencies could establish service-level thresholds in different ways: – Accumulation-related threshold. Based on accumulated snow and ice between passes by snowplow; could be estimated by plow-driver observations or measured by RWIS stations; blowing and drifting snow could challenge the accuracy of these measurements. – Friction-related threshold. Based on measured grip factor reported by RWIS stations or plow trucks. – Travel speed–related threshold. Based on average measured travel speeds, which could have wide variations given lower traffic volumes during major weather events, particularly overnight; measure could also encourage higher speeds when they are not warranted, particularly where variable speed limit systems are in place. – Access-related threshold. Based on the capacity of available infrastructure (e.g., per lane opened). This could be an alternative to speed-related thresholds since they do not encourage the use of higher speeds. • Measures that rely on observations may be subject to increased subjectivity and have fewer data points to accurately calculate the time that a roadway segment did not meet the service- level threshold. • Training for individuals who report service-level threshold observations increases consistency in reporting. A structured visual inspection process may be used to minimize subjectivity. • Expanding the pool of reporters to include citizens, law enforcement, and other personnel increases reporting frequency for certain service-level thresholds. • Equipping vehicles with automatic vehicle location (AVL) for enhanced reporting may increase frequency of reports for certain service-level thresholds. • Leveraging mobile and RWIS data will increase accuracy for certain service-level thresholds. • The service-level threshold criteria are likely to vary by agency by roadway functional class, severity of the event, day of week, and time of day. • Service-level thresholds can be reported at any scale (e.g., segment, district, state). Data Elements Required for the Measure • Performance target(s) for service-level threshold. May vary based on road segment pri- oritization, hour of day, day of week, storm severity, and the respective definition for the service-level threshold. • Time stamp for beginning and end of winter event. Estimated time that winter precipita- tion started and stopped for a given road segment; may be logged from nearby RWIS snow accumulation data, observed weather radar, or agency personnel reports in the area. • Time stamp for observed service-level threshold. Time that the pavement status was observed. (This may be different from the time stamp of the submitted report.) • Service-level threshold. Calculated for each segment based on available reports and time for observed service-level threshold. • Road segment information. Road segments identified as a priority for calculating this measure need to be identified; if service-level criteria vary within an agency by road type or

102 Performance Measures in Snow and Ice Control Operations area, these designations need to be known for each road segment. Average travel speeds will be needed to calculate speed-related threshold. • Weather information. Data from RWIS sites regarding rate or total precipitation or grip will be needed to determine storm severity, an accumulation-related threshold, or a friction- related threshold. • Posted speed limit information. Information regarding posted speed limits should be used if agency changes speed limits in response to weather. • Road closure information. Information regarding the mileage and duration of road closures will be needed to calculate a closure-related threshold. Analytical Approach The analytical approach for this measure is as follows. 1. Consider agency capabilities, policies, and available data sources To help inform the selection of what service-level threshold to measure, an agency should consider the availability of data and resources required to calculate it, what is more suitable for local conditions, and the operating policies and constraints. Agencies must be capable of monitoring the service-level threshold for the duration of the event. 2. Define service-level threshold and criteria Agencies must select the type of service-level threshold to measure (e.g., based on accumu- lation, friction, travel speeds, or road closures) and ensure they have the capabilities to log all the required data during an event. A definition for the threshold should also be created, which may vary for different roadways segments and different events based on the roadway functional class, observed severity of the event, day of week, or time of day. This measure may not be calculated for all road segments in the network. Agencies need to develop a table that includes at least three severity levels (based on precipitation rate and timing of the storm). – Severity level – high – Severity level – medium – Severity level – low high For each of the severity levels, agencies need to define the priority segments for which the service thresholds can be defined. Priority segments can be based on functional class or on traffic volumes or other criteria that are important to the region (e.g., access to schools, emergency routes). For each of these segments, target service-level thresholds should be identified based on one or more of the following: – Qualitative assessment of road condition, – Measured speeds, – Measured road condition factors, and – Measured accumulation. 3. Define performance target(s) Set a performance target or targets for the accepted percent of time road segments meet the agency-defined service-level threshold(s) during a winter event. As the threshold criteria in Step 2 can vary, this target may also vary for each roadway segment and event based on the roadway functional class, observed severity of the event, day of week, or time of day. 4. Obtain start and end times for winter events The start and end times may be approximated for each road segment based on a nearby or regional observation points, which may be derived from: – Reports from agency personnel in the field observing when precipitation began and ended, – Agency personnel viewing real-time or archived weather radar, and – Initial RWIS reports of precipitation for winter event start time and consecutive RWIS reports showing no additional precipitation accumulation for winter event end time.

Defining Performance Measures 103 5. Obtain all observations of the service-level threshold that occurred during the event for affected segments As noted in Step 2, data logged by agency personnel in the field, law enforcement, citizen reporters, and RWIS stations should indicate a measurement of the service-level threshold and a time stamp of the observation. 6. Determine whether each observation met the defined service-level threshold Based on the targets established in Step 2, compare the entire duration for that segment to the prior observation (or winter event start time) and determine if it meets the threshold. If the observation does not meet the threshold, an agency may choose to calculate an estimate of the portion of time that the threshold was exceeded since the prior observation (or winter event start time). 7. Calculate the number of hours in which the defined service-level threshold was not met and the duration of the winter event For each roadway segment, total the number of hours that did not meet the service-level threshold. Aggregate this value for roadway segments that share a common performance target as determined in Step 3. Use the values from Step 4 to identify the number of hours of the winter event duration. 8. Calculate percent of time road segments meet agency-defined service-level thresholds during winter storms Take the aggregated number of hours that did not meet the service-level threshold and divide by the number of hours of the winter event duration. Divide this value by the number of segments included in the aggregated number to calculate the percent of time road segments meet agency-defined service-level thresholds during winter storms. 9. Factor in winter season severity This measure is sensitive to storm severity. As such, severity needs to be factored into the value calculated in Step 8. If the criteria in Step 2 and Step 3 already include adjusted targets based on the severity of the storm, additional consideration of severity may not be required. 10. Compare to performance target The calculated measure may be compared to assigned performance targets established in Step 3 (e.g., Interstate highways should have no more than 1 in. of winter precipitation accumulation 100% of the time during a winter event with 1 in. or less accumulation per hour, while other highways should have no more than 2 in. of winter precipitation accumu- lation at least 80% of the time during the same winter event). Percent of Segments Meeting Time to Regain or Recover to Acceptable Criteria for Agency-Defined Segments After the End of Event A key objective of maintenance operations is to reach acceptable pavement conditions. This measure is used to assess and report on the performance of winter storm management and response for an event (i.e., meeting expected time to regain acceptable pavement conditions). This measure provides support to event response decisions and post-event analysis, regarding infor- mation such as labor and material usage. Translating time-stamped surface conditions to winter weather event status allows an agency to see how quickly the segment recovered from an event. Measure Definition This performance measure assesses the amount of time that passes from the end of a winter event until an acceptable surface condition once again exists (e.g., bare pavement). This measure need not apply to all segments with the same criteria. The following criteria for this measure are defined by the agency and may vary for different roadway segments: • Acceptable condition. Defined percentage of the affected segment that must be at an acceptable criterion to meet target.

104 Performance Measures in Snow and Ice Control Operations • Performance target. Desired time to reach a defined pavement condition, which may vary by roadway segment priority. Pavement conditions may be defined using agency guidelines (see the Alaska DOT example in the previous section). • Roadway segment prioritization. Roadway segments for which to track this measure, and relative importance of selected segments for reaching desired pavement condition. • Storm severity. Given high sensitivity to storm severity, agencies should factor severity into the measure (e.g., develop the measure for different intensity of storm events taking into account the nature of the storm). Included in the Definition • Event duration. Agencies should select a period that consistently encompasses most winter weather events. Different agencies will have different durations that are considered an event (e.g., from first to last inch of accumulation or a combination of weather indicators). For instance, Alaska DOT uses the qualitative thresholds in Table 7 for determining the criteria for recovery. The recovery times are at the bottom of the table. Performance Target Performance Target Description Illustration LOS A (good winter driving conditions) Bare pavement is the primary goal. Good winter driving conditions exist when snow and ice have been removed from the driving lanes and excessive loose snow has been removed from the shoulders and centerline of the highway. Short sections of ice and packed snow are acceptable and can be expected within the driving lanes between the wheel paths, as well as on the centerline. Bare pavement may not be possible in the northern and central regions during periods of extreme cold weather. Generally, loose snow has been cleared, and traction is good for most vehicles properly equipped for winter driving. If required for traction, 100% of roadway has sand present. LOS B (fair to good winter driving conditions) Roads are passable with varying conditions. Drivers may encounter some standing water, packed snow, and icy patches covering the surface. Generally, loose snow has been cleared from the travel way, and traction is adequate for most vehicles properly equipped for winter driving. If required for traction, sand has been applied to hills, curves, intersections, and bridge decks. LOS B represents a fair to good level of service, which ranges from targets of bare pavement as much as possible on higher-standard or highly traveled highways to snow-pack or icy conditions on northern region roads as well as on lower-standard or low- volume roads. Traffic moves at reduced speeds, with isolated slowdowns or delays. Table 7. Alaska DOT qualitative thresholds.

Defining Performance Measures 105 Performance Target Performance Target Description Illustration LOS D (poor winter driving conditions) Travel is challenging for most vehicles properly equipped for winter driving. Moderate snow accumulation on roads (up to 4 in.). LOS D represents a marginal level of service where traffic moves slowly with substantial delays. Traction is marginal even for vehicles properly equipped for winter driving. LOS F (hazardous winter driving conditions) Travel is not advised. Considerable snow accumulation on roads (4 in. or more). Drivers may encounter snow drifts, berms, freezing rain, and glare ice. Traction is extremely poor even for vehicles properly equipped for winter driving. Note that the response time to reach a given LOS depends on the severity and length of the winter storm. Alaska DOT assigns an expected time threshold to clear roads in one of five priority levels: • Priority 1: May take up to 24 h to clear after a winter storm. • Priority 2: May take up to 36 h to clear after a winter storm. • Priority 3: May take up to 48 h to clear after a winter storm. • Priority 4: May take up to 96 h to clear after a winter storm. • Priority 5: Generally cleared only in spring to open road for summer traffic. LOS C (fair to poor winter driving conditions) Roads are generally passable with varying conditions. Drivers may encounter some standing water, loose snow, snow drifts, packed snow, and icy patches covering the surface. Patches of snow or ice exist even on the highest-standard roads, and these conditions may degenerate to predominately snow-packed or icy conditions throughout, with accompanying slowdowns or delays. On lower-standard or low-volume roads, the surface is snow covered (up to 2 in.) with substantial traffic delays. Source: Alaska DOT (2017). Table 7. (Continued). Weaknesses and Limitations of the Measure • What is considered an acceptable condition may differ between urban and rural areas or by road type. • The accuracy of this measure will vary by the frequency of checks made, which could differ significantly between urban and rural areas or by road type. • Subjectivity is a concern since different reporters may view the road conditions differently. A structured approach such as using the pavement snow and ice condition (PSIC) index can

106 Performance Measures in Snow and Ice Control Operations mitigate some of the subjectivity but not all. In general, field personnel such as snowplow drivers are required to report conditions. • The flow of events is an important factor because anticipated conditions may differ from actual conditions, which could lead to an under- or overestimation of the measure and toward bias in the final result. Variations in the Measure • This measure is typically used for specific locations and roadway segments driven by plow operators. • Training for individuals who report pavement surface condition status increases consistency in reporting. • Expanding the pool of reporters to include citizens, law enforcement, and other personnel increases reporting frequency for pavement surface condition status. • Equipping vehicles with AVL for enhanced reporting may increase the frequency of reports on pavement surface condition status. • Leveraging mobile and RWIS data will increase accuracy in pavement surface condition status. • Pavement condition criteria are likely to vary by agency: – Bare pavement is typically designated as a percentage of the affected segments that is bare (e.g., 95%). – Pavement condition criteria may differ within an agency for urban and rural areas as well as road type. – A structured visual inspection approach such as the PSIC may be used to minimize subjec- tivity (Blackburn et al. 2004). The PSIC uses visual characterization of the road, combined with traffic flow, to classify the roadway condition into one of seven levels, ranging from Condition 1 (bare/wet pavement where all snow and ice are prevented from bonding to the road surface) to Condition 7 [road is closed due to weather or road conditions (e.g., low visibility, drifting, glare ice)]. • The measure can be reported at any scale (e.g., segment, district, state). • A variation of this measure uses a time to return to normal speed or free-flow speed instead of bare pavement. With the greater precision offered by speed data, this variation may benefit from less subjectivity and potentially more time stamps, but it is not recommended because it could encourage unsafe driving speeds instead of improved performance of snow maintenance activities. Data Elements Required for the Measure • Road segment information. Road segments identified as a priority for calculating this measure need to be identified; if pavement criteria vary within an agency by road type or area, these designations need to be known for each road segment. • Time stamp for end of winter event. This is the estimated time that winter precipitation stopped for a given road segment; it may be logged from nearby RWIS site snow accumulation data, observed weather radar, or agency personnel reports in the area. • Pavement surface condition status. This is the percentage of a road segment that is at the predefined acceptable criteria; it is reported by agency personnel in the field or citizen reporters. • Time stamp for observed pavement surface condition status. This is the time that the pave- ment status was observed. This is different from the time stamp of the submitted report because drivers may not submit a report until arriving at their destination. Analytical Approach The analytical approach for this measure is as follows. 1. Define acceptable criteria and targets The definition of the pavement condition criteria must be established (e.g., at least 95% of the affected segment is free of snow and ice). Defined targets or designations for pavement

Defining Performance Measures 107 condition may be used by different agencies, or within an agency for more rural or lower functional class roadways. For instance, an agency may define as acceptable a duration of 3 h to achieve 95% bare pavement for Interstate highways and urban U.S. and state roadways, and 24 h to 80% bare pavement for rural U.S. and state highways. 2. Identify affected road segments First, prioritize road segments for which to report this measure. Then identify which of these road segments have reports indicating they were affected by the winter event. Note road type and urban or rural designations, as necessary based on criteria in Step 1. 3. Obtain end time for winter event It may not be realistic to log a precise time for each road segment. The end time may be approximated for each road segment based on nearby or regional observation points, which may be derived from: – Reports from agency personnel in the field observing when precipitation ended, – Agency personnel viewing real-time or archived weather radar, or – Consecutive RWIS reports showing no additional precipitation accumulation. 4. Obtain the time of the first observation indicating the predefined acceptable pavement con- dition for each segment Reports received from agency personnel in the field, law enforcement, or citizen reporters should indicate the percentage of pavement that met the criteria for a given segment. Identify the earliest observation for each affected road segment identified in Step 2 that meets the defined pavement criteria as defined in Step 1. Note the time the observation was made, not the time the report was submitted; time may have lapsed for the report to be made until the observer completed a trip. 5. Calculate the time elapsed from the end of the winter event until the acceptable pavement condition observation for each segment to calculate time to regain acceptable pavement condition Find the difference between the times identified in Step 3 and Step 4 for each segment. Per agency reporting criteria, this may be documented in minutes or hours and be rounded accordingly. 6. Factor in storm severity This measure is highly sensitive to storm severity. A storm severity index allows for this measure to be used in comparisons across regions. An example of a normalization approach is to divide the measure by the storm severity index to create a composite measure that takes into account the nature of the storm. Another example would be to factor in severity in setting acceptable criteria. 7. Compare to performance target These values may be compared to assigned performance targets (e.g., less than 3 h to bare pavement for Interstate and U.S. highways and less than 6 h for state roadways). Note for each observation whether the performance target was met. 8. Calculate aggregated time to regain acceptable pavement condition Agencies may aggregate time to acceptable pavement for reporting or comparisons of responses by: – Weather event: average the measure for all road segments for a given weather event in the region to compare with performance following other storms. – Segment: average the measures for all storms at a given segment for the winter seasons to compare with performance at other segments. – Season: average the measures for all segments and all storms in the region or state to compare with performance in other seasons. At this stage, an agency will have estimated the overall percentage of time where thresholds were met during one storm event or season. The remaining steps in this analysis are optional and pertain to aggregating results and calculating the frequency of meeting targets.

108 Performance Measures in Snow and Ice Control Operations 9. Calculate frequency of meeting regain time target for each roadway segment for the winter season For each segment, identify the number of winter events where the performance target was met and divide by the total number of winter events for which a value of regain time was calculated for that segment. 10. Calculate frequency of meeting regain time target for each roadway segment type using all observations for the winter season For all locations within each roadway segment type (e.g., Interstate highways), identify the number of times that the regain time performance target was met and divide by the total number of observations for which a value of regain time was calculated over the winter season. For example, a total of 12 observations from six winter events are available from three Interstate segments; the regain time performance target was met nine times, for a frequency performance measure of 65%. 11. Calculate regional frequency of meeting regain time target for all observations for the win- ter season. For a regional measure, from all observations made at every road segment and for every winter event, identify the total number of times that the regain time performance target was met and divide by the total number of regain time measures made in the winter season. Percent of Trips Within Accepted Difference Between Measured Travel Time Index and Additional Expected Travel Time Index for Snow and Ice Events Agencies seek to provide reliable services through dependable travel times, as measured from day to day or across different times of the day. Therefore, consistency in travel time is an important measure of service quality and mobility for travelers since there is real value in understanding how congestion and service behave throughout the operation of a transportation system. This measure focuses on trip-making and the use of travel times to establish reliability measures. Measure Definition The travel time index (TTI) is measured as the ratio of the peak-period travel time to the free- flow travel time, with averages across urban areas, road sections, and periods being weighted by VMT. This measure of performance looks at the difference between measured TTI during storms versus prespecified additional TTI for key trips defined by an agency (i.e., specific origin– destination pairs) and for specific storm severity (or a range of severity). For example, if under normal conditions the TTI for a trip is 1.5 (i.e., it takes 1.5 times free- flow time), then the agency can establish an additional 33% increase in the index due to snow and ice for a medium-severity event, making the TTI equal to 2.0 for the specified condition. Then, the agency can estimate the percentage of the event duration during which the TTI threshold was met (i.e., the TTI is less than 2.0). Additional TTI impacts for an event can be: • Estimated based on expert judgment, • Calculated based on historical data, or • Modeled. The additional TTI selected by the agency will vary by one or more of the following: • Roadway functional class. Lower TTI may be expected for higher-priority roadways (e.g., Interstates). • Observed severity of the event. High additional TTI may be acceptable for more severe events.

Defining Performance Measures 109 • Day of week. Higher TTI may be acceptable for certain days (e.g., Saturdays, Sundays, holidays). • Time of day. Higher TTI may be acceptable during certain periods (e.g., midnight–6:00 a.m.). Implied in the Definition • Event duration. Agencies should select a period that consistently encompasses most winter weather events. Different agencies will define winter event durations differently (e.g., from first to last inch of accumulation or a combination of weather indicators). • Trips. Not all roads and segments will be monitored under this measure. Agencies would have to identify key trips that are of importance to stakeholders, such as: – Priority origin–destination pairs along key routes, and – Emergency routes. Weaknesses and Limitations of the Measure • Agencies must be capable of monitoring and recording travel time for the duration of an event. Historical data can be used to establish the additional TTI for each defined storm severity, or travel time can be estimated based on expert judgment. • This measure will only be calculated for key trips and not for all segments. The agency will identify a subset of trips for which this measure will be calculated. • Different from the LOS measure, which looks at segment data, this measure looks at a trip, which may encompass many types of roadways, agencies, and jurisdictions. • The additional TTI criteria are likely to vary by agency and jurisdiction. • The accuracy of this measure will vary by the ability to monitor travel time consistently dur- ing the event, which could differ significantly between urban and rural areas, or by road type. • This measure is not suggested for agencies that routinely use closures as a part of their winter maintenance strategies, particularly in remote or mountainous areas. • The flow of events is an important factor because anticipated conditions may differ from actual conditions, which could lead to an under- or overestimation of the measure and toward bias in the final result. Variations in the Measure • Agencies could also measure other reliability indicators, such as: – BI. Buffer time is the extra time required to make a trip. The index is the size of the buffer as a percentage of the average (i.e., calculated as the 95th percentile minus the average, divided by the average). – PTI. Planning time is the total travel time and includes buffer time. The index is measured as the ratio of the 95th percentile travel time to the free-flow travel time. Data Elements Required for the Measure • Historical travel time information. Historical travel time data are needed to define the criteria for the additional TTI. This information needs to be correlated with respective storm severity. • Current travel information. Current travel information is needed to estimate the TTI of selected trips in (near) real-time or at a defined frequency within the duration of the storm (e.g., every 30 min). • Time stamp for beginning and end of winter event. This is the estimated time that winter precipitation started and stopped for a given road segment; it may be logged from nearby RWIS site snow accumulation data, observed weather radar, or agency personnel reports. • Road segment information. Road segments identified as a priority for calculating this measure need to be identified; if service-level criteria vary within an agency by road type or area, these

110 Performance Measures in Snow and Ice Control Operations designations need to be known for each road segment. Average travel speeds will be needed to calculate a travel speed–related threshold. • Weather information. Data from RWIS stations regarding rate or total precipitation or grip will be needed to determine storm severity, an accumulation-related threshold, or a friction- related threshold. Analytical Approach The analytical approach for this measure is as follows. 1. Define input values for TTI A clear definition of the input measures necessary to estimate TTI must be established— for example, free-flow speed is calculated as the 85th percentile of off-peak speeds, where off-peak is defined as Monday through Friday, 9 a.m. to 4 p.m. and 7 p.m. to 10 p.m., as well as Saturday and Sunday 6 a.m. to 10 p.m. 2. Estimate historical TTI under normal and winter storm conditions Use historical data to estimate TTI for given storm severities. TTI is measured as the ratio of the peak-period travel time to the free-flow travel time. This measure is computed for the morning peak period (6 a.m. to 9 a.m.) and afternoon peak period (4 p.m. to 7 p.m.) on weekdays. Averages across urban areas, road sections, and periods are weighted by VMT using volume estimates, which can be derived from FHWA’s Highway Performance Monitoring System. Use of freely available data and tools like the NPMRDS are suggested to perform such analyses. 3. Determine additional TTI needed for different storm severity levels Define the percentage increase to the TTI during a storm. For example, an agency may establish the following criteria: – 80% of selected trips should be within TTI + 20% for a low-severity snow and ice event. – 80% of selected trips should be within TTI + 40% for a medium-severity snow and ice event. – 80% of selected trips should be within TTI + 60% for a high-severity snow and ice event. Note that defined targets or designations of TTI for a given storm severity may vary by agencies or within an agency’s scope if it is responsible for different functional class roadways (e.g., Interstate highways or urban U.S. and state roadways), and may also vary by population. 4. Track current travel time and determine TTI Analyze current TTI to assess whether maintenance activities are affecting travel time as expected in the selected parts of the network. Note that estimating TTI in real time can require high data collection and analytic capability, and therefore agencies can opt to estimate this value at a more realistic frequency (e.g., every 20 min). 5. Compare estimated TTI with actual TTI Estimate the percentage of the event duration during which the TTI threshold was met. Five-Year Rolling Average of Fatalities and Injuries (Number, Rate) During a Winter Season This measure allows for seasonal evaluations and can be a key input to both maintenance and incident management planning. It may allow for the identification of priority locations and other locations in need of specific safety interventions, technologies, programs, practices, and enforcement. Measure Definition This performance measure indicates the number of fatal and serious injury crashes as related to the winter season or winter weather events. Serious injuries are defined by the FHWA follow- ing the Model Minimum Uniform Crash Criteria “Suspected Serious Injury (A)” attribute found

Defining Performance Measures 111 in the “injury status” data element (Governors Highway Safety Administration and U.S. DOT 2012). The measure is calculated every winter season and averaged to account for seasonal differ- ences. Multiple seasons should be used to calculate an average to account for expected variations in crash rates. The exact measure of fatal and serious injury crashes an agency is able to report may vary based on the availability of resources and data, including: • Detailed crash records. Reliably documented weather-related causal factors, as well as time of the crash, specific location, and the number of persons involved, which can be easily queried and examined, facilitate calculation of a more comprehensive measure. • Accurate traffic volume data. Quality volume counts and estimates during winter events or the season to generate a reliable value of VMT could facilitate calculation of fatality rates. • Weather event data. Winter weather event start and end times to identify crashes occurring during the event. An agency may select one or more fatality- and injury-related measures, depending on data availability, for the following: • Fatal crashes versus fatalities. Crashes involving a fatality versus total number of individuals killed in fatal crashes. • Serious injury crashes versus serious injuries. Crashes involving a serious injury at worst versus total number of individuals with a serious injury (including those with a serious injury in a crash that also had a fatality). • Season versus winter events. Include versus exclude crashes that do not occur during a winter weather event. • Rate versus number. Based on the availability of estimated VMT. Included in the Definition • Season duration. Agencies should select a period that consistently encompasses most winter weather events. Different agencies will define durations for what is considered winter differ- ently (e.g., from first to last storm, based on frequency of events, based on temperature or a combination of weather indicators, or simply by calendar dates). • Number of seasons. Given expected variations in crash rates, a rolling average of multiple seasons should be used to manage regression to the mean and better indicate long-term trends; a 5-year average is consistent with the Safety Performance Management Measures Final Rule (FHWA 2013). Weaknesses and Limitations of the Measure • Availability of crash data is often delayed due to processing. In part, this is because some crashes cannot be immediately categorized since generally any injury that results in a death within 30 days is listed as a fatality. • The accuracy of fatality and injury rate measures will vary by the availability of actual traffic volume data and reliable estimates of traffic volumes during winter weather events, which could be particularly challenging for secondary and rural roadways. • Identifying individuals sustaining serious injuries in fatal crashes may complicate the ability to identify the total number of serious injuries versus number of serious injury crashes. • Collecting detailed reliable information on the factors that led to an incident, including the condition of the road at the time of the incident, is challenging. • Attributing crashes to winter maintenance activities is difficult since there may be multiple causal factors not related to snow and ice control—hence the need for detailed crash reports. • Relationships with safety management groups within a DOT are important to obtain dis- aggregated crash data. • This measure does not take into account crashes involving less severe injuries or no injuries.

112 Performance Measures in Snow and Ice Control Operations Variations in the Measure This measure can be reported in ways that may be more helpful and relevant to winter maintenance operations given the availability of resources and quality data: • Winter season fatal or serious injury crashes. The number of all fatal or serious injury crashes over the designated winter season period. This level of aggregation works for agencies where snow and ice events are common. • Fatal or serious injury crashes during winter weather events. Number of fatal or serious injury crashes occurring during winter weather events, given crash and winter weather event start and end times. This excludes dates and times when no winter weather event was present. This could be particularly useful for agencies where snow and ice events are sporadic. • Winter season fatal or serious injury crash rate. Number of winter season fatal or serious injury crashes divided by the estimated VMT occurring during the winter season, given availability of reliable traffic volume estimates. • Fatal crash rate during winter weather events. Number of fatal crashes occurring during winter weather events divided by estimated VMT occurring during those winter weather events, given crash and winter weather event start and end times and ability to calculate reliable estimates of traffic volumes during weather events. • Measures using fatalities and serious injuries. All of the previous measures could be calcu- lated using the number of individuals killed or injured instead of the number of respective crashes, which may be preferred for consistency with the Safety Performance Management Measures Final Rule (FHWA 2017) but is potentially more difficult to calculate given the unlikely availability of data. • Measures using serious injuries. All of the previous measures could be calculated using the number of individuals seriously injured instead of the number of serious injury crashes, which may be preferred for consistency with the Safety Performance Management Measures Final Rule (FHWA 2017) but is potentially more difficult to calculate given the unlikely availability of data. • Measures by road type. All of the previous measures can be calculated at any scale by road type and area (e.g., segment, segment type, district, statewide). In addition, agencies need to consider: • The quality of available crash and traffic volume data, as well as accurately defined winter weather event start and end times, will determine the ability to isolate the fatalities or injuries occurring during winter weather events and estimate the VMT during events, thus affecting the accuracy of the calculated measure. • If available, reliably documented causal factors on detailed crash records may be examined to identify specific areas to improve winter maintenance operations. Inconsistencies and sub- jectivity associated with reporting causal factors make them less desirable for calculating a performance measure, making the measure not as usable for comparison with other fatality measures. • Given transportation management center (TMC) coverage of the entire area of interest, incident data may be easier to obtain than crash records; however, TMC incident logs will likely have less detail. • Rolling 5-year averages are suggested to manage regression to the mean, provide a better picture of long-term trends over time, and be consistent with the Safety Performance Manage- ment Measures Final Rule. • Some agencies use different reporting scales for injury crashes and may need to adjust the calculation of this measure. • Agencies may calculate other measures to include all types of injuries or injury crashes during weather events instead of exclusively serious injuries.

Defining Performance Measures 113 Data Elements Required for the Measure Required Data • Fatal crash records. At a minimum, the number of fatal crashes occurring during each winter season month is needed. More detailed information, such as causal factors, location, time, and number of individuals involved in the crash, can help improve the measure. TMC incident records may not log all crashes or their severity, and official crash records have a time lag associated with them. • Injury crash records. At a minimum, the number of serious injury crashes occurring during each winter season month is needed. More detailed information, such as causal factors, location, time, and number of individuals seriously injured in fatal and serious injury crashes, can help improve the measure. TMC incident records may not log all crashes or their severity. Optional Data • Traffic volume. To calculate a rate of fatalities or fatal crashes, actual volumes from traffic count stations are preferred, but a volume estimate, adjusted if necessary to account for lower than normal traffic during winter weather conditions, will also suffice. • Time stamp for beginning and end of winter event. To calculate the number or rate of fatalities or fatal crashes occurring during winter weather events, the estimated times that winter precipitation began and ended for a given road segment are needed; these may be logged from nearby RWIS station snow accumulation data, observed weather radar, or agency personnel reports. Analytical Approach The analytical approach for this measure is as follows. 1. Identify data sources to establish measure parameters The availability of detailed crash reports and reliable volume estimates will inform the selection of what measures an agency decides to calculate. Establish relationships with DOT traffic safety staff or law enforcement to learn what crash information is available for calculating these measures and how it is reported. Crash data details such as time and location may be inconsistently reported for different jurisdictions, as may the classification of serious injury crashes; different agencies use different scales for classifying injury crashes. Agencies may also choose a period during which most winter events occur to define as the winter season. 2. Define performance targets Set performance targets for the chosen fatal and serious injury–related measures. Targets may be lower for highway types that receive higher priority for winter maintenance activities (e.g., Interstate highways). Establish a period (e.g., 5 years) for which to calculate rolling averages to be consistent with agency practices for reporting similar crash measures. 3. Identify the number of winter season fatal crashes and winter season serious injury crashes Reports received from law enforcement or DOT databases should include an entry for the date of the crash. With an electronic or manual query, retain only those records designating a fatal crash during the predefined winter season. Repeat this process for serious injury crashes. Verify that the database contains all records for the selected period, given inherent delays associated with classifying crashes. 4. Identify the number of winter season fatalities and winter season serious injuries Using the crash reports from Step 3, identify the number of individuals killed in the winter season fatal crashes. Then identify the number of individuals seriously injured in both winter season fatal crashes and winter season serious injury crashes.

114 Performance Measures in Snow and Ice Control Operations 5. Obtain start and end times for winter events The start and end times may be approximated for each road segment based on a nearby or regional observation point(s), which may be derived from: – Reports from agency personnel in the field observing when precipitation began and ended; – Agency personnel viewing real-time or archived weather radar; and – Initial RWIS report of precipitation for the winter event start time, and consecutive RWIS reports showing no additional precipitation accumulation for the winter event end time. 6. Identify the number of fatal crashes, fatalities, serious injury crashes, or serious injuries within the season Discard crash records used for the tabulation of measures in Step 3 and Step 4 that do not occur within the winter event times obtained in Step 5. Recount the number of fatal crashes, fatalities, serious injury crashes, or serious injuries to obtain a respective measure for winter events within the season. 7. Calculate the VMT occurring during the winter season or winter weather events Obtain available count station data for the winter season defined in Step 1 or winter weather event period obtained in Step 5 to provide a direct estimation of VMT when multiplied by the length of the segment that contains it. Do not include roadway segments or road types for which reliable winter event VMT estimates are unavailable. Depending on availability and reliability of volume estimates for adjacent segments, particularly for winter weather events, an agency may: – Apply the volume from traffic count stations directly to adjacent segments. – Multiply a winter event factor to the VMT estimates that is derived from observed traffic volumes during the winter events versus typical traffic volumes at count stations. The units for typical traffic volumes for this factor can vary to be consistent with the format available for the VMT estimates (e.g., annual or daily traffic). – Develop a rough estimate of VMT for segments with poor-quality data based on available information. 8. Calculate rates by VMT of fatal and serious injury crashes by winter event or season As desired, divide the measures calculated in Step 4 by winter season VMT calculated in Step 5 for every winter season. As desired, divide the measures calculated in Step 6 by winter weather event VMT calculated in Step 5 for every winter season. One or more variations of this measure may be calculated for different category road types (e.g., Interstate highways or the entire transportation network) per criteria established in Step 1. 9. Take a rolling average for all calculated measures To account for expected fluctuations in the number of crashes from year to year, use the criteria established in Step 2 to calculate a rolling average of values derived in Step 3, Step 4, Step 6, and Step 8. 10. Compare to performance target Calculated measures may be compared to assigned performance targets established in Step 2 (e.g., less than one serious injury crash per 100 million winter event VMT). Customer Satisfaction Ratings for Snow and Ice Response Although the results can be subjective and expensive to obtain (through comprehensive and representative surveys), direct customer feedback can be vital in ensuring support for snow and ice control operations. This measure provides support to event response decisions and post- event analysis by allowing an agency to see how perception of performance changes among its customers.

Defining Performance Measures 115 Measure Definition This performance measure analyzes the satisfaction of transportation system users during winter events by tracking traveler feedback at a regional or statewide level. Traveler satisfac- tion and demographic data are gathered through periodic surveys, focus groups, or other approaches. It is likely that traveler satisfaction will decrease during more severe events; therefore, severity can be considered in evaluating performance. Collecting information about demographics enables agencies to consider whether satisfaction varies significantly based on socioeconomic status. Data for the measure can be collected in two ways: • Seasonal survey of customer satisfaction. Conduct a survey once a season to gather data on snow and ice response perceptions from travelers. While conducting a survey is simple, the inability of travelers to recall the entirety of their satisfaction with the snow and ice response season may skew responses toward more recent events. • Survey after specific events. Conduct a survey after specific events through agencies’ exist- ing public-facing interfaces. However, such surveys might suffer from bias since obtaining a statistically sound sample is unlikely. Included in the Definition • Season duration. Agencies should select a period that consistently encompasses most winter weather events. Different agencies will consider different durations as a winter weather event (e.g., from first to last storm, based on frequency of events, based on temperature or a com- bination of weather indicators, or simply by calendar dates). • Event duration. Agencies should select a period that consistently encompasses most winter weather event durations. Different agencies will consider different durations for a winter weather event (e.g., from first to last inch of accumulation or a combination of weather indicators). The Missouri Department of Transportation (MoDOT) uses road rallies [a gathering of citizens to drive around while accompanied by a moderator who tracks their comments as they assess ride (i.e., road) quality], customer surveys, and report cards to monitor the degree to which the public accepts the agency’s performance. MoDOT places significant value on customer feedback; it spends approximately $200,000 each year on its public phone survey and a survey of the media and other partners (e.g., public officials and organizations like the Association of General Contractors). Customer relations personnel generally design the survey mechanism with input from the department on the agency-wide focus areas (Yurek et al. 2012). Weaknesses and Limitations of the Measure • Ideally, surveys should be conducted soon after the event has ended and for all events. How- ever, this might represent a significant cost, depending on the desired number of responses (or response rate) and the applied survey methodology. Depending on the sampling method- ology, there could be different biases introduced into the responses. • The quality of this measure is highly dependent on the quality of the information collected and is thus subject to all the limitations present for analyzing and reporting survey data. • The quality of this measure improves with consistent and representative observations over time and a higher response rate. Initial assessments may be not usable, but consistency in survey questions and methodology will allow for longer-term use as a performance measure. • The flow of events is a critical factor because anticipated conditions may differ from actual conditions, which could lead to an under- or overestimation of the measure and toward bias in the final result.

116 Performance Measures in Snow and Ice Control Operations Data Elements Required for the Measure • Demographic information. Socioeconomic data that can be used to characterize households. • Storm/season information. Data and times of all storm/season parameters (e.g., accumulation) for the surveyed areas. Analytical Approach The analytical approach for this measure is as follows. 1. Define criteria and targets for selected service measures The definition for each service measure and its acceptable level must be established— for example, customer satisfaction is defined as the society’s approval rate of maintenance activities or their perceived effectiveness. Defined targets or designations may vary by agency or within an agency’s scope if it is responsible for different functional class roadways (e.g., Interstate highways or urban U.S. and state roadways), as well as by population. 2. Estimate required sample size and geographical distribution It may not be realistic to expect the number of respondents to meet the sample size require- ment for a truly statistical and representative survey. Nonetheless, it is important to estimate what would constitute such a sample size. An agency may want to undertake a less rigorous sampling approach and include surveys on agency performance using on existing traveler information systems such as 511 and traveler information websites. 3. Determine survey distribution approach Surveys should be distributed soon after the end of weather maintenance activities—actual distribution time should be evaluated based on distribution and data analysis capability. This would allow agencies to use this information in post-event assessments. If seasonal surveys are conducted, surveys are suggested to be conducted in the middle of winter rather than in the non-winter season to minimize the bias due to recall issues. 4. Perform data analysis and summarize results to indicate the level of satisfaction Perform statistical analysis to assess whether maintenance activities reached targeted levels of satisfaction for different parts of the network and across population groups. 5. Calculate aggregated percentage Agencies may aggregate the estimated percentages for reporting or comparisons of responses by: – Weather event: average the measures for all population groups for a given weather event in the region to compare with performance following other storms. – Segment/region: average the measures for all storms at a given segment or region for the winter seasons to compare with performance at other segments or regions. – Season: average the measures for all populations and all storms in the region or state to compare with performance in other seasons. 6. Compare to performance target Values from Steps 4 and 5 may be compared to assigned performance targets. Note for each observation whether the performance target was met. Cost of Snow and Ice Control to Meet Established Performance Criteria for a Given Winter Severity Since agencies operate under a specified and limited budget, they should correctly translate their usage of different resources into cost of winter maintenance operations and use this information to assess the efficiency of their spending. The word “cost” here refers to a standard- ized cost, which can be the total cost per number of lane miles under the agency’s jurisdiction, for instance.

Defining Performance Measures 117 Measure Definition This performance measure is a highly visible parameter of local and state expenditures. The main challenge is to develop a monetization approach that captures the complexity that exists in most state and local agencies as well as their limited ability to control the costs. For example, the cost of materials may be determined by factors outside the jurisdiction’s control. Similarly, fuel prices fluctuate for reasons unrelated to winter maintenance activities, and personnel have multiple duties and realistically only spend a portion of their jobs on snow and ice control, making assigning personnel costs a challenge. Winter maintenance cost information can be viewed as the output of combining usage indicators of labor, equipment, material, and other resources with respective cost infor- mation. In this sense, this measure tracks the true cost of winter operations per storm and season. In order to estimate the efficiency of spending, the overall cost needs to be standardized by one or more characteristics of the maintained/served area and the severity of the event or season. There are many factors that drive cost, and these should be considered when estimating cost and efficiency of spending. These factors include: • Geographic size, • Functional class or roadway and priority segments, • Density of roadways (i.e., length of roadways per area), • Rural versus urban, • Microclimates and hot spots, • Timing of events, • Number of events, and • Intensity of events. Weaknesses and Limitations of the Measure • Overall cost is mainly defined by labor, materials, and fuel. While these are commonly tracked variables, they tend to vary in capability and accuracy. • Cost efficiency may not always be transferable to other locations. For instance, two agencies/ regions may have the same LOS objective but significantly different costs due to the nature of the areas. However, this measure can be compared with itself over time. • Determining the cost of winter maintenance may be easier for contracted operations where there is a well-defined contract value. • Detailed information on (unit) cost and weather data by event is needed to estimate this measure, making the agencies’ records the main source of information. • The flow of events is an important factor since anticipated conditions may differ from actual conditions, which could lead to an under- or overestimation of the measure and toward bias in the final result. Variations in the Measure • Agencies could also measure their costs through more general or detailed factors, depending on their capabilities, such as through: – Miles covered, – More detailed geographical areas, such as census tracks, – Cost per type of activity (e.g., anti-icing and deicing), and – Traffic level [winter average daily traffic (WADT)] that is serviced.

118 Performance Measures in Snow and Ice Control Operations Data Elements Required for the Measure • Unit cost. Cost of each unit of every resource used in winter maintenance activities. • Resource usage. Total amount of resources used, detailed by category. At a minimum, agencies should detail the amount of material, staff, equipment, and fuel used during each storm. • Weather information. Data from the RWIS regarding rate or total precipitation or grip will be needed to determine storm severity, an accumulation-related threshold, or a friction- related threshold. • Times of beginning and end of maintenance activities. The beginning and end times of all winter maintenance activities are needed as a proxy for the usage of some resources or to estimate the efficiency of the activities. • Mileage and functional classes of roadway. Total length of the maintained roads and functional classes. • LOS and recovery threshold. Analytical Approach The analytical approach for this measure is as follows. 1. Identify all tracked resources and collect usage data A clear definition of the input measures necessary to estimate winter maintenance costs must be established. This should be as comprehensive and accurate as permitted by the capability of the agency to track all available and used resources. 2. Define the criteria to measure efficiency for season severity level Establish the criteria to measure efficiency of budget expenditures per severity indicator. 3. Estimate an expected winter maintenance cost based on seasonal severity Use historical data to estimate the individual and overall costs of winter maintenance activities, noting storm and season severity, by creating a regression model between regional characteristics, level of service expectations, and severity indicators. It is likely that each region’s regression model will be different. However, it is more important to show a reasonable correlation between cost and severity that allows an estimate of expected costs for a certain severity. 4. Compare estimated costs with actual expenditures Determine estimated efficiency in expenditures for winter maintenance by comparing actual costs in the season to expected costs. A certain percent difference (e.g., plus or minus 10%) should be factored into the difference as imperfections in the model. Agency Within Acceptable Difference Between Expected and Actual Use of Salt and Other Materials in a Season An important objective of agencies is to achieve a specific LOS, providing a safe and reliable road network while using only the materials necessary, thus maintaining sustainable winter maintenance operations. This performance measure is used to assess the amount of materials used for a given season to achieve this objective. Material usage, when consistent with the goals of the agency and the severity of the conditions, will provide winter maintenance managers an overall perspective of how the agency is performing. Measure Definition This performance measure assesses the amount of material an agency uses in a given winter season for highway maintenance. This amount of material is affected by multiple factors. To use material usage as a performance measure, these factors must be taken into consideration so that consistency from season to season can be achieved.

Defining Performance Measures 119 Agencies often review material usage as an indicator rather than a performance measure. Material usage alone is not an effective performance measure. Varying factors or practices from storm to storm or season to season will result in inconsistencies. Factors that may influence material usage include storm/season severity; varying levels of service; proactive approaches like treated salt, anti-icing, and pre-wetting; computerized dispensing equipment; maintenance decision support systems; calibration of equipment; and yearly weather patterns. To the extent possible, these factors should be considered when normalizing the usage of salt and other materials. It is important to model an expected amount of salt and material usage based on winter sever- ity and the objectives of the response. This requires historical data to correlate actual material expenditures to measured severity, miles of roadway by functional class, and LOS parameters for particular regions. The performance measure then is an acceptable difference between the expected and actual material usage (e.g., plus or minus 10%). Included in the Definition • Season duration. Agencies should select a period that consistently encompasses most winter weather events. Different agencies will define winter durations differently (e.g., from first to last storm, based on frequency of events, based on temperature or a combination of weather indicators, or simply by calendar dates). • Expected amount of material use. A modeled estimate of material use based on winter severity, LOS, and historical data. Weaknesses and Limitations of the Measure • Direct assessment of the environmental impact of winter maintenance practices is hard to measure for agencies, making material use a proxy. • Different techniques and quality of materials have a direct impact on the amount of material used to achieve the stated operational objectives. Operational approaches such as pretreatment can influence the final amount of material used. • Advanced capabilities require in-vehicle monitoring systems that report, in near real-time, material usage and minimize the need for manual data collection. • This measure relies on development and acceptance of a model for the expected amount of material use. However, anticipated conditions may differ from actual conditions, which could lead to an under- or overestimation of the measure and toward bias in the final result. Variations in the Measure • Agencies could also define their use of materials through more general or detailed factors, depending on their capabilities, such as: – Lane miles covered, – More detailed geographical areas, such as census tracks, and – Traffic level (WADT) that is serviced. Data Elements Required for the Measure • LOS. An agency must determine an LOS for its road system. LOS may vary for different road segments based on agency priorities and public use. Regardless of the LOS requirements (bare pavement, wheel tracks, or snow covered), the same LOS must be targeted consistently in order for material usage to be used as a performance measure. LOS may be measured in order to ensure under- or overachievement is not taking place. • Material used. An agency must have a consistent and accurate way of collecting the data for the amount of materials used (solids and liquids). This may be achieved by capturing the

120 Performance Measures in Snow and Ice Control Operations amount or weighing the material as it is loaded, using onboard data to collect the amount of material dispensed, or by operators keeping good records of the material being loaded and used and the material returned at the end of an event. These data should be compiled per event and per season and should include materials used in pre-storm operations and in-storm operations in order to capture all the material being used. • Treatment recommendations and calibration. Accountability may be the biggest factor in material usage as a performance measure. Consistency in material being dispensed by trucks and operators following recommendations is important to ensuring that material usage is uniform in every condition and every event. Equipment must be calibrated prior to a season and every time material changes or something with the equipment changes (like a breakdown or broken hose). Material usage must be consistent from operator to operator given similar conditions. • Model techniques and results. This measure is dependent on the expected amount of salt and material usage based on winter severity and the objectives of the response. This requires historical data to correlate actual material expenditures to measured severity, miles of road- way by functional class, and LOS parameters for particular regions. Analytical Approach The analytical approach for this measure is as follows. 1. Define material usage criteria and targets Material usage may be defined as the average amount of material used in a normal winter season. This may be determined by using normal winter weather criteria (snowfall amounts, temperatures, and number of events) as compared to the amount of material used in a season. It is expected that more material will be used in a more severe winter and less in a milder winter. However, agencies should define and develop a model to estimate a value of material usage, and this should be complemented by a data collection and operational protocol that takes into account that consistency in levels of service, methods, and materials is needed to ensure that performance is being measured. Historical data on winter severity, LOS objectives, and traffic volumes can be used to set expected usage levels. 2. Determine materials and treatments Materials and treatment methods should be consistent. If new methods or new materials are introduced, then results compared to previous years may vary. Accounting for these changes can help in determining if the performance level has increased or decreased. Every type of material used in winter maintenance has a working threshold. Recommendations for treat- ments should be made based on the performance of the material and the conditions at the time of application (pavement temperature, precipitation type, and LOS being achieved). The amount of material being applied should vary based on each changing condition. Treat- ment recommendations can be provided in many ways such as through charts, value-added services, and the Maintenance Decision Support System (MDSS). 3. Calibrate the equipment being used All equipment dispensing material is capable of being calibrated to determine the amount of material being dispensed at any given speed over the length of a roadway. This is true of solid materials and liquid materials. In order to have accurate data, especially when data are taken from in-cab controllers that collect data as the materials are being dispensed, it is important that the amount of material coming out of the vehicle matches exactly what is being recorded. Scales and weighing equipment that measure material being placed into vehicles must also be calibrated. Any change in material (wet salt, treated salt, or a switch to sand) requires all equipment to be recalibrated to be accurate. Note that all equipment should be calibrated, regardless of whether the material is selected for analysis or not.

Defining Performance Measures 121 4. Collect the data Data should be collected following each event, and data should be compiled if a yearly total is desired. Collection methods should be consistent and as accurate as possible. If done electronically, then a secondary check should be done by the operator to confirm the amount used. (Not clearing the unit prior to a shift or a malfunction may give inaccurate readings.) Manual collections should be done with scaled or loaded amounts as well. Liquid use should coincide with the lane miles covered. Weather conditions, pavement temperatures, wind conditions, precipitation amounts, and length of the event should all be recorded. 5. Compare to performance target The performance measure is an acceptable difference (e.g., plus or minus 10%) between the expected/modeled and actual material usage. These values may be compared to assigned performance targets. If there is consistency in the severity from storm to storm and season to season, then the amount of material used should coincide with the performance an agency achieved in its winter operation. As agencies implement more proactive approaches to opera- tions in order to achieve a more sustainable winter program, performance is likely to increase and material use likely to decrease. 6. Aggregate the measure Agencies may aggregate this measure for reporting or comparisons of responses by region. For this, agencies can take an average of the normalized measures for all storms at a given region for the winter seasons to compare with performance at other regions. Step 4: Develop Analytic Approaches The following subsections detail the steps to estimate severity index (storm and winter season) and performance curves needed for the standardization of the performance measures and definition of targets. Developing a Severity Index The concept of severity is fundamental to performance measurement for winter maintenance. The inputs required for management of winter maintenance are directly correlated to storm severity and to seasonal severity. A winter event is defined as any weather occurrence (with defined start and end times) that requires resources for preventing, minimizing, or regaining the loss of bare lanes. Winter events can include freezing rain, drizzle/sleet, snow, drifting/blowing snow, frost, ice/black ice, refreeze, or any combination of these conditions (Minnesota DOT 2015). Agencies responsible for winter maintenance operations use severity to classify individual winter events and the entire winter season based on event and overall characteristics, such as (total) precipitation and wind speed. Following the approach of what gets measured, gets managed, these indices allow agencies to factor in severity in the performance of their maintenance activities, enabling comparison between agency regions and seasons. The following subsections provide background on such indices and a suggested approach to develop them. Storm Severity Index 1. General steps for calculating a storm severity index (SSI) are: a. Identify data sources/availability and select storm elements. The first step consists of identifying the data sources (e.g., agency-collected data, RWIS, National Weather Service) and describing the appropriate storm event elements to be analyzed and used for the development of the SSI. Table 8 provides a summary list of elements. Note that every location has unique conditions that might need to be included to obtain more robust/ realistic results; therefore, agencies should use this list as a starting point to which they can

122 Performance Measures in Snow and Ice Control Operations add elements if necessary. For this particular methodology, only storm-related elements are suggested for use. Indicators of resources usage (e.g., amount of salt used) and output/ outcomes of winter events (e.g., number of incidents) are not suggested for inclusion to avoid double counting because such indicators will be used in developing the core perfor- mance measures. b. Decide the element’s measurement unit and categories. The next step is to decide how the storm elements should be measured and to identify cut points to categorize the data (if necessary). Regardless of the analysis approach, clear levels (i.e., categories based on specific thresholds) are useful to simplify data that might be too complex to interpret (e.g., meteorological data) or have a wide range (e.g., wind speed). At a minimum, agen- cies should distinguish low, moderate, and heavy characteristics of a given element (e.g., how much snow needs to fall to classify precipitation as high). This step also includes the removal or correction of outlier data points. Outlier data points can be described as data Storm Element Description Units* Type Classification of the storm based on type and amount of precipitation. Examples: • Freezing rain, light snow, medium snow, and heavy snow; and • Snow and ice. – Temperature (air and road surface) Temperature of the road before, during, and after a storm. This can be measured either continuously or by categories (e.g., warm, mid-range, cold). Dew point or relative humidity could also be considered. °F, °C Precipitation Amount of snow, rain, or ice that has fallen. This can be measured either continuously or by categories (e.g., moderate, below 4 in. or 10 cm, or heavy, over 8 in. or 20 cm). in., cm Rate Intensity of the precipitation within a given time frame. in./h, cm/h Drift Duration of drifting snow. h, days Wind Wind speed before, during, and after a storm. This can be measured either continuously or by categories (e.g., mild, moderate, strong) and include gusts and direction. mph, km/h, m/s Visibility Road visibility during a storm. This element measures how far a driver can see during a storm. mi, km, m Duration The time a storm affects a particular area. This measure could be h, days disaggregated into the time the storm was occurring and the period the effect of the storm was prolonged. Duration Number of hours from a defined start time to an end time. Can be tabulated as “weather hours.” h, weather- hour Forward speed How fast a storm is moving. The slower the storm, the longer its duration in a particular area. mph, km/h, kn, m/s Area covered The area affected by a storm, estimated through a spatial measure or population density. people/mi2, ln-mi, mi2, km2 Storm behavior Qualitative or quantitative measure indicating pre- and post-storm behavior. This measure identifies how a storm begins (e.g., starts as rain or snow) and finishes (e.g., finishes as rain or snow). – Topography Indicator of how the road alignment could affect the impact of the storm. This could be as a proxy through qualitative values, such as levels (e.g., 1 to 5) based on a subjective assessment of potential impact of the topography, or through quantitative values, such as proxy by elevation and changes in or average horizontal or vertical alignment. – Timing Qualitative or quantitative measure of severity based on the time of day and day of week the storm takes place, considering its duration in its entirety or partially (e.g., only when it starts). – *Other units may be used if applicable. Table 8. Elements to consider when developing a storm severity index.

Defining Performance Measures 123 that are not representative of the entire population. These points can occur as a result of instrument error or abnormal conditions (i.e., unusual weather events for the time of year or location, such as snow in the state Florida in December). c. Identify a proper methodology to define the analysis period (i.e., duration) and spatial scope (i.e., areas covered/affected). This step helps determine how much data to collect and how to discern between continuous winter events. d. Develop merging criteria. Once the elements and their respective units have been selected, agencies must decide how to combine them, taking into account the different types (i.e., qualitative and quantitative) and units. Various techniques can be used to develop such criteria; two commonly used are: – Econometric models (e.g., multiple linear regressions) that use historical information to estimate the contribution of each component to the severity of a storm. The difficulty of this approach lies in identifying initial values for each element, which should be accom- panied by a thorough analysis of historical records (see example of an econometric model in Box 3). – Consensus through survey or expert opinion, which can be used to assign initial or final weights to each element. Note that this approach is heavily based on the subjective perspective of the respondents; therefore, the sample selection process should attempt to include a variety of experienced respondents, such as managers, practitioners, and users of the system. e. Define severity criteria. This step provides the threshold for categorizing severity. This can be done by identifying the various percentiles of the historical distribution or scaling the obtained values to a specified range (e.g., 0–1, 0–10, or 0–100) and dividing it into quantiles. Box 3. Example of Econometric Model Nixon and Qiu (2005) provided the following multiple regression model to estimate the SSI: ( )[ ]= + + + −SSI 1 0.5 b ST Ti Wi Bi Tp Wp a   where SSI = storm severity index, ST = storm type, Ti = in-storm road surface temperature, Wi = in-storm wind condition, Bi = early storm behavior, Tp = post-storm temperature, Wp = post-storm wind condition, and a and b are parameters to normalize storm severity index range. Using the estimates for the factors, the SSI was calculated for 252 different storms based on the initial algorithm and scores. Then the initial scores were modified (using the a and b constants) so that the computed SSI values have an approximately normal distribution.

124 Performance Measures in Snow and Ice Control Operations f. Steps c and d provide a model to obtain a final SSI when more than one element is selected to characterize the severity of a storm. However, not all elements are required. Sometimes a single element, such as precipitation, may be selected as a proxy for severity for a simpler but less robust SSI as long as the element is strongly correlated with key input, output, and outcome measures. Winter Severity Index Similar to an individual storm, a winter season can be defined by many of its features, such as temperature averages and extremes, snowfall totals, average and highest snow depth, the dura- tion of winter weather conditions, and the aggregated value of its impact (e.g., economic loss). Overall, winter severity tends to be measured by aggregating, averaging, or normalizing one or more features over the length of the season. Despite similarities in variables in many states, the complexity of the methods to evaluate a winter severity index (WSI) varies broadly by state. This guide provides a general approach to develop a WSI that builds on previous research, to the extent possible. a. Identify a proper methodology to define the analysis period (i.e., duration) and spatial scope (i.e., areas covered/affected). This step helps determine how much data to collect and how to discern winter events within a season. b. Define reporting unit(s) of elements. This step ensures that all subsequent efforts are designed to correlate to and reach the desired objective(s). The units of analysis are usually defined as aggregated, averaged, or normalized measures of storm elements or SSIs. As for an SSI, indicators of resource usage and output/outcomes of winter events should not be included when creating the WSI since these will be used in the development of the core performance measures. c. Develop merging criteria. Similar to the SSI, agencies must decide how to combine the selected elements, taking into account the different types (i.e., qualitative and quantitative) and units. Various techniques can be used to develop such criteria. – Econometric models (e.g., multiple linear regressions) that use historical information to estimate the contribution of each component to the severity of a winter season. The difficulty of this approach lies in identifying initial values for each element, which should be accompanied by a thorough analysis of existing historical records. – Consensus through survey, which can be used to assign initial or final weights to each element. Note that this approach is heavily based on the subjective perspective of the respondents; therefore, the sample selection process should include a variety of experi- enced managers, practitioners, users of the system, and any other important (or affected) stakeholder. – Aggregation of a single variable over the winter period is a simple yet effective way to estimate overall winter severity. This is a suggested approach when agencies already have a record of the SSI of all storms during the winter period. d. Define severity criteria. This step provides the threshold for categorizing severity. Most approaches found in the literature depend, ultimately, on defining breakpoints of accumu- lated severity—for example, answering the question “How many days with more than 10 in. of snow makes a winter season mild, moderate, heavy, or extreme?” This can be done by identify- ing the various percentiles of the distribution or scaling the obtained values to a specified range (e.g., 0–1, 0–10, or 0–100) and dividing it into quantiles. Another approach is to develop an algorithm by which points are assigned on a daily basis based on observed weather conditions; these daily values are then summed to create monthly or seasonal scores. Steps c and d provide a model to obtain a final WSI when more than one element is selected to characterize the severity of a storm. If only one element is selected, however, it should be

Defining Performance Measures 125 normalized to obtain a less biased WSI. If in the previous example for SSI, precipitation is selected as the sole element of analysis, one could standardize it by coverage, yielding a WSI measured as average inches of snow per lane mile per day for heavy winter areas or average inches of snow per lane mile per month for milder winter areas. Developing Winter Performance Curves Winter maintenance performance curves are graphical representations of how an agency operates under various winter conditions. They illustrate the correlation between selected per- formance measures and indicators of severity. This guidance summarizes the methodology to develop performance curves into three distinct processes: 1. Data collection and review. Through this process, data are compiled and quality checked for outliers. Outlier data points can be described as data that are not representative of the entire population and can therefore lead to misleading or incorrect findings. These points can occur as a result of instrument error or abnormal conditions (i.e., unusual weather events for the time of year or location, such as snow in the state Florida in December). It is always sug- gested to perform basic quality checks of the data before attempting to develop performance curves. This can be accomplished by plotting the data (e.g., X-Y plots, boxplots, and distribu- tion plots) and calculating basic distributional statistics for each variable. Values that are far outside the range of typical values for that variable should be investigated. 2. Data analysis and visualization. This process entails the development of the performance curve. For this, the data are summarized into key performance indicators of historical events (i.e., individual storms and seasons) that are relevant to the preparation of performance curves, such as estimates of snow accumulation, wind speed, and temperature. Based on this information, plots can be developed with the selected indicator of performance on one axis and a severity indictor on another. 3. Sensitivity analysis. Both performance and severity indicators can fluctuate for a correspond- ing value (e.g., an agency may take 2 h to remove 6 in. of snow during one storm and 3 h to remove the same amount during another storm). Therefore, agencies should perform a sensi- tivity analysis to evaluate the effect of varying parameters on the performance curve. Matthews et al. (2017) present a flexible approach for developing WSIs for winter maintenance. The overall approach assigns daily scores based on weather triggers (i.e., variables) and their associated thresholds (Figure 8). The authors highlight four questions that should be addressed when using daily weather scores as the building blocks for a WSI: 1. What overall approach should be taken in characterizing daily weather? 2. What winter weather conditions, either individually or in combination with other variables, are potential triggers of road maintenance activity? 3. Where continuous variables are involved, how should threshold values be determined (e.g., what amounts of daily snowfall translate into winter maintenance activities of different amounts)? 4. How should specific daily scores be assigned?

126 Performance Measures in Snow and Ice Control Operations Source: Matthews et al., 2017. Figure 8. Approach to develop a WSI.

127 The implementation stage involves the steps shown in Figure 9. C H A P T E R I I Implementing Performance Measures Implementing Measures Figure 9. Steps for implementing performance measures. Step 5: Inventory Current Practices and Gaps Review of Agency Organizational Practices The ability of an agency to successfully implement a series of comprehensive snow and ice performance measures will largely depend on the existing institutional capabilities. An agency with a strong history of data collection and analysis, staff with the necessary analytic skills, and the ability to establish and generate snow and ice performance measures will be better positioned for success. Agencies that foster collaboration and a culture of improvement will likewise be better able to collect and implement these performance measures. An agency lacking these institutional capabilities for performance management may have to start small in developing snow and ice performance measures that are realistic, rather than taking on too much too fast. As staff gain experience, data collection processes become more routine and automated, relation- ships improve with various stakeholders like snow maintenance and plow operator staff, and sufficient data become available to establish realistic performance targets, snow and ice performance measures can become more comprehensive and refined to provide more meaningful feedback into decision making and operations. The way an agency chooses to implement the data collection and calculation of a given snow and ice performance measure for mobility will vary based on its institutional capability. As an example, the return to bare pavement measure may be reported on an ad hoc basis via radio communications by snowplow operators with little to no attention to the frequency of reports or consistent approach regarding location or terminology by different operators. This level of data collection may be an appropriate starting point for an agency with no history or technol- ogy to more systematically collect these data. However, these data would allow for only a rough estimate of this performance measure for roadways with an available report of clear conditions

128 Performance Measures in Snow and Ice Control Operations or bare pavement and the time it was reported relative to the end of the event. Agencies with more established performance measurement programs may have the capabilities and resources to support training programs and enhanced, electronic reporting methods for agency personnel and citizen reporters that allow for more consistent, time-stamped reports, which would allow for a more precise, segment-level performance measurement for return to bare pavement. Table 9 provides a checklist that can be used for a targeted assessment of readiness for snow and ice performance measurement. To have a successful performance measurement program for snow and ice control requires the six components discussed in the following six subsections. The absence of any one of these components restricts the ability of an agency to measure and attribute performance to mainte- nance actions adequately. Ability to Determine LOS Before, During, and After an Event The concept of LOS is used in operational and performance assessments of facilities across transportation systems to indicate infrastructure status through traffic flow, density values, and other operational metrics. LOS provides a scale that is necessary for setting realistic and easily understood performance targets. The ability to determine LOS is necessary for the use of several response and segment-level indicators. A survey of the use of LOS in winter maintenance noted that almost two-thirds of the agencies (63%) used management staff to monitor routes to determine whether LOS guidelines had been met (CTC and Associates and Wisconsin DOT 2009). However, friction measurements require costly and specialized equipment that can traverse the road with sufficient frequency and spatial distribution to gain an adequate sample. Another approach is to use visual inspection and to miti- gate some subjectivity with training. The PSIC is a visual method used to characterize roadway conditions (Blackburn et al. 2004). The PSIC can be used to assess during-storm and post-storm Capabilities Necessary for Snow and Ice Performance Measurement Season Event Yes No Partial Yes No Partial Have a defined snow and ice policy that articulates the goals and objectives of the agency Have a policy to define start and end times of snow and ice events or season Have a process to document LOS before, during, and after an event Have a roadway classification system for defining expectations and the desired LOS Have the ability to track material, labor, and fuel use Have a process to determine severity Have the ability to collect both weather and pavement conditions Have the ability to monitor and analyze traffic impacts Have a road condition reporting system that enables retrieval of time-stamped data about events Have access to travel speed data in 5- or 15-min bins for all roads within the agency’s jurisdiction Have access to reliable crash data that include severity and time Have historical data available to assist with establishing realistic targets Table 9. Checklist to assess readiness for snow and ice performance measurement.

Implementing Performance Measures 129 performance. The PSIC is easy to use and low in cost, but it is a subjective measure because it is based on the perspective of the individual collecting the information. Use of measured travel speeds as a proxy for LOS may be less subjective than visual or field personnel observations (see Box 4). When reliant on posted or free-flow speeds to determine the variation due to an event, LOS estimates may provide an incorrect view of performance due to low-volume conditions or existing congestion on the roadway. However, when speed data are available on a segment basis over several winters and events, a better assessment of performance can be obtained using the data archive to develop an effective performance standard that can be tracked before, during, and after events. Ability to Track Material, Labor, and Fuel Use Agencies responsible for snow and ice control track material usage. Having this historical information enables agencies to estimate more accurately how much material they can expect to use at an aggregated (e.g., in a given budget cycle) and disaggregated (e.g., in a single storm) level. Given the important implications of material usage, these robust expectations then can be used to optimize environmental and budgetary/effectiveness programs. Although most agencies have some type of material management system, highly capable agencies demonstrate the ability to track material, labor, and fuel use at highly disaggregated levels and in different units. An Existing Technique to Factor Severity into Performance For snow and ice performance measures to be used effectively in decision making, agencies need a procedure to factor event or season severity into performance. Without such a procedure, performance measures are subject to criticism by stakeholders as not reflecting conditions or the unique challenges responders face between regions and differences in weather patterns between seasons. Ability to Obtain Road Condition Reports As noted previously, road condition reports typically are human observations and reports of driving conditions at various stages of a winter storm and during after-storm cleanup. Although these reports are subjective, they are valuable to travelers as they plan trips or consider alternatives and to maintenance personnel as they manage snow and ice treatment because the reports are available sooner than other sources of condition data. Box 4. Speed-Based Snow and Ice Performance Standards for Indiana DOT and Regain Time in Minnesota DOT McCullough et al. (2013) identified two performance standards and an LOS grade based on travel speeds collected over several winters in Indiana. As an example, an LOS scale was based on speed values collected over 2 years. As an alternative, to reduce the subjectivity of the return to bare pavement measure, Minnesota DOT has researched a normal condition recovery time measure based on traffic data. This measure would provide an automated measure for urban freeways using loop detector data on traffic speed, flow, and density to determine when roadways have returned to normal conditions. It is expected that this approach could be used to derive a similar but more objective indication regarding mobility as the return to bare pavement measure.

130 Performance Measures in Snow and Ice Control Operations Typically, the measurement and recording of road condition reports are performed by plow drivers or other maintenance personnel while out on the roads. Reports are often submitted by radio or onboard device. In addition, several agencies have implemented citizen reporting programs. In these programs, citizens are trained to report road conditions using the same definitions as DOT staff, effectively increasing the number of reports received. Several states also use a software system to record road conditions received from the field (see Box 5). For example, the road condition reporting system (RCRS) is often a focal point; it is populated by manual and automated data and information feeds, and it supplies information to various information dissemination mechanisms for traveler information and winter mainte- nance. Deeter et al. (2014) documented best practices for the RCRS and noted the following as having direct implications on snow and ice performance measurement: • Regular, automated intake of weather data collected on a timed cycle by systems external to the RCRS. • Generating automated performance measure data using time stamps of manual and auto- mated actions. Deeter et al. (2014) highlighted the successes of state DOTs in effectively integrating road weather data into road condition reporting systems and identified the following cases: • In Maryland, the CHART RCRS collects data from the RWIS network to deliver high-level roadway weather information to travelers. In addition to collecting weather information, the RWIS data are also automatically added to any incident entered into CHART to maintain a record of road conditions at the time of the event. A unique aspect to the Maryland approach is that attaching weather reports to incidents allows the department to identify connections between incidents and weather conditions. Box 5 Wyoming DOT’s Road Condition Reporting App The Wyoming weather response traffic management (WRTM) road condition reporting application addresses maintenance staff activities to report road conditions. These images (courtesy of Wyoming DOT) show a tablet mounted on a snowplow and the app user interface, which uses predefined field codes that correspond to particular road conditions. The app allows all functions that were previously performed manually or by phone by Wyoming DOT staff to be automated, minimizing radio chatter between drivers and the transportation management center.

Implementing Performance Measures 131 • In Ohio, basic air and pavement data from the 174 RWIS sites throughout the state are inte- grated into the Buckeye RCRS every 5 min. This intake of RWIS data provides additional information to travelers and enables internal Ohio DOT staff to view the weather informa- tion through the RCRS, which is particularly helpful when updating the manual road weather reports. • North Dakota DOT integrates wind speed and radar data into its RCRS. The wind speeds are stored in the RCRS and passed to the department’s traveler information website for display to travelers. Both the wind speeds and the radar data are visible to RCRS users to allow them to view current conditions and make decisions about weather pattern changes. • Idaho operates a module in its RCRS that collects RWIS data and assigns a circle of influence to the weather reports based on rules that are primarily dictated by the terrain surrounding the RWIS site (e.g., for mountainous terrain, the circle of influence is smaller than for flat areas). The weather reports are presented on a 511 website together with the circle of influence, allowing website visitors to understand what geographic area is most likely experiencing the conditions. • Iowa DOT operates a module in its RCRS to collect National Weather Service watches, warnings, advisories, and alerts in the Common Alert Protocol. A unique benefit is that the interface that collects the data can be used for additional alerts published using the protocol. Ability to Collect Both Weather and Road Weather Observations from a Wide Area Collection of weather and road weather data includes static and mobile data collection devices capable of measuring current precipitation, temperature, wind, and other conditions from several feet below the ground surface to tens of feet in the air (depending on specific sensors deployed). Weather and road weather data are used in real time by maintenance personnel. Similarly, the data are often fed into maintenance decision support systems. For example, the Idaho Transportation Department uses a network of about 100 RWIS stations that provide weather and pavement condition data every 15 min to track trends and inform winter maintenance operations decisions, develop storm severity metrics, and calculate performance measures. Ability to Monitor Traffic Impacts Across the Region Effective snow and ice performance measurement is directly related to the outcomes of the maintenance response, which in most cases translates to traffic impacts. With growing Intelligent Transportation System (ITS) capabilities in each state and greater private-sector–collected speed data, collecting and using traffic data (speeds, delays, travel times, crashes) in winter performance measurement is essential. The NPMRDS is a free and widely available resource of archived data with 1-month latency that is available to agencies for the development of snow and ice perfor- mance measures. The NPMRDS allows for increased consistency among agencies as a data source for reporting snow and ice performance measures. Moreover, as agencies enter into agreements to partner with Waze or obtain access to third-party data like INRIX, opportunities to leverage and integrate this information in order to enhance existing snow and ice performance measures or develop new measures may be explored; however, use of these data sources is unlikely to be cost effective unless they are also used for other purposes. Assess Current Gaps The ability of any agency to support robust and comprehensive snow and ice performance measurement depends on many factors. The Road Weather Management Capability Maturity Framework (CMF; http://www.ops.fhwa.dot.gov/tsmoframeworktool/available_frameworks/

132 Performance Measures in Snow and Ice Control Operations road_weather.htm) is a valuable tool that can be used to perform a general assessment of an agency’s capability level in key areas. The tool helps an agency assess the capability of road weather management programs, identify specific actions, identify where it needs the most improvement, and provide recommended actions for improvement. Through the use of the framework, agencies can quickly assess their relative strengths and weaknesses in important process areas for road weather management. Step 6: Identify Data Sources and Needs For performance management, weather and road data are important for normalizing per- formance between storms (e.g., measuring time to bare pavement after 2 in. of accumulation versus after 12 in. of accumulation). Weather and road data collection devices are typically posi- tioned throughout the road network in a pattern based on the needs of maintenance providers, but placement can be somewhat constrained by available power, communications, and access. Therefore, data collection typically is not available everywhere it is desired. Mobile data collec- tion onboard DOT vehicles helps supplement the data. Recent private-sector sources of data and national data sets like NPMRDS can provide additional means to monitor mobility measures. While these data sources may not be particularly useful individually, together they can serve to provide an effective guide to performance. Note that accuracy sometimes is less important than consistency (i.e., it may be less important for an agency to devote a large amount of resources to get perfect data than it is to collect data in a consistent manner to help compare performance from one year to another). Chapter I details the data needed for each performance measure and can serve as a starting point for agencies that seek to collect data for performance measurement.

133 Two primary steps are identified in this chapter, as shown in Figure 10. C H A P T E R I I I Using Performance Information Using Measures Figure 10. Steps for using performance measures. Step 7: Set Targets and Establish Baseline Performance target setting is an important and complex process through which agencies analyze expected performance and account for factors that will affect performance, including levels of available funding and the relative emphasis placed on achieving different targets for a transportation system (see Box 6). Specifically, some of the items that must be considered are: • Financial resources. A realistic projection should be made of what could be accomplished with available funding levels. • Technical considerations. Targets should be achievable based on current and forecasted conditions and trends and accounting for external factors that could affect performance levels. • Policy considerations. Targets should reflect existing priorities and policies and be based on public involvement, customer feedback, or legislative or executive direction. • Economic factors. Target setting should take into consideration how to maximize benefits in relation to investments or achieve the highest return on investment. • Correlation between targets. Potential correlations between targets that may influence their achievement should be highlighted (i.e., improving time to bare pavement while reducing the amount of salt and fuel used). As agencies go through multiple cycles of monitoring and evaluating performance, their ability to develop realistic targets and measure the outcomes resulting from target achievement improves. Throughout the transportation field, target setting generally remains part science and part art, informed by both qualitative and quantitative data and information. Time frames are an important consideration in setting targets. Short-, mid-, and long-range targets are all useful, and having a combination of all three to measure progress toward achieving longer-term goals

134 Performance Measures in Snow and Ice Control Operations could also be useful. The following are some different types of target-setting methods; rarely does a target-setting process fit neatly into one of these categories: • Policy-driven methods (established by executive management or a legislative body, which might arise because of public discontent with transportation issues), • Analysis-driven methods (based on modeling or other tools that provide information about expected levels of performance), • Consensus-based methods (established through a collaborative planning process with input from a variety of stakeholders), • Customer feedback–based methods (direct feedback from customers through surveys and outreach is used to help define targets), and • Benchmark-based methods (through comparisons with peer agencies). For snow and ice control, establishing targets likely will involve some combination of these approaches, particularly policy priorities, analysis, consensus, and customer feedback. Adams et al. (2014) identified methodologies for selecting target levels of service for maintaining and operating highway assets to improve agencies’ performance with respect to managing highway assets. Cambridge Systematics et al. (2010) provided additional information about methods that managers of state DOTs and other agencies could use for setting performance targets to achieve multiple objectives, interact with multiple decision makers and stakeholder groups, and use data management systems to support performance-based decision making. The steps for setting targets (see Figure 11) build on those created in a National Highway Institute workshop (National Highway Institute 2016). Step 1 for Setting Targets – Define Purpose Agencies should be able to link their targets directly to their goals. As such, it is important that agencies clearly state their goals so they can be converted into measurable objectives. For this, agencies need to: Box 6. Types of Targets Three types of targets exist (directional, aspirational, and realistic), and in most cases, transportation agencies consider all three before setting a final performance target. • Directional targets identify the direction of impacts desired. For snow and ice control, the target could be to reduce the amount of time to bare pavement or to improve the customer experience (as measured through satisfaction surveys). For safety, this target could simply be monitoring the trend in the desired direction (e.g., reductions in fatalities and injuries). • Aspirational targets are selected to reflect a policy priority or broader societal target, even if it might not be realistic (e.g., “zero fatalities” as part of Vision Zero programs). • Realistic targets account for available resources, trends, risks, competing objectives, and other factors that might affect performance. These targets are designed to provide a basis for assessing and tracking progress compared to a standard of performance believed to be obtainable.

Using Performance Information 135 • Assess why they are setting the snow and ice control operation targets; • Link performance to the agency’s overall and winter maintenance specific objectives, resources, and requirements; and • Use the agency’s overall and winter maintenance purposes to inform target parameters and guide target establishment. Step 2 for Setting Targets – Set Target Parameters The next step is to define the target parameters. Three important elements need to be defined for each target: • Target portrayal. The method for portraying the change in performance [e.g., percent change, number, rolling average, return to x-year value/level, and directional (up/down)]. • Time frame. The duration from the baseline that will be the basis for reaching the target (e.g., 1/5/10/20 years). • Scope. Boundaries and filters applied to the performance area to set the extent of the target; determine what the target should focus on [e.g., National Highway System (NHS), non-NHS, urban/rural, mode, and a specific geography]. Step 3 for Setting Targets – Assemble Baseline Data and Analyze Trends Agencies need to define the point of reference that will be used to guide the analysis of winter storm/season performance trends. For this, agencies need to: • Assemble historical winter measure data from available data sources and fill any data gaps, and • Plot the data to establish a baseline through the development of statistical analysis that can estimate trend lines. Step 4 for Setting Targets – Identify and Assess Influencing Factors It is important that agencies recognize the factors that they have control over and the forces that can change the course of their target outcome, such as the decision-making process, funding and existing conditions, winter staff and equipment resources and priorities, and external factors. Examples of influencing factors are: • Historical performance trends, • Policy directives, • Business culture barriers, Step 1 – Define Purpose Step 2 – Set Target Parameters Step 3 – Assemble Baseline Data and Analyze Trends Step 4 – Identify and Assess Influencing Factors Step 5 – Establish a Target Figure 11. Steps for setting targets.

136 Performance Measures in Snow and Ice Control Operations • Capital project commitments, • Budget and resource constraints, • Senior management directives, • Agency jurisdiction, • Agency goals and priorities, and • Planned operational activities. Other important considerations that agencies need to consider are: • Fiscal limitations and trade-offs, • Constraints and existing commitments, both within an agency and to stakeholders, • How factors will change over the time span of a target, • Risk associated with each of the factors (including magnitude and likelihood of risk), and • Documented assumptions involved with building each factor into the target-setting process. Step 5 for Setting Targets – Establish a Target The sequence of steps provided guides an agency through the process of identifying winter- maintenance–related goals and associated metrics, defining a baseline and trends, and selecting targets for road weather maintenance performance. This final step builds on this sequence to develop a preliminary list of feasible targets from which a final one can be selected. Figure 12 illustrates an example of the end product of this process. Note that targets can be set through different methods, including policy-driven, analysis- driven, consensus-based, customer feedback–based, and benchmark-based methods. Therefore, agencies need to provide clear governance, documentation, and rationale supporting the chosen target. This entails: • Determining who makes the final decision to pick the target, • Defining who provides oversight to setting the target, and Modified from National Highway Institute (2016). Figure 12. Example of baseline and target setting.

Using Performance Information 137 • Documenting key decisions and providing transparent information to explain the rationale around target choice: – Establishing project selection choices that will affect performance and link to target to decisions, and – Using the tools available to show link between funded projects and their impact on the target (see Box 7 for an example of how to present performance results). Step 8: Report Performance Communicating performance measures to stakeholders is important for snow and ice pro- grams. Some DOTs post annual performance measurement reports online, and these reports may include a section on winter maintenance activities. Also, some DOTs produce fact sheets and other publications that provide information about winter storm maintenance activities. However, fewer state DOTs have a dedicated, publicly available winter maintenance performance report. Among state DOTs that include external reporting are Minnesota (Minnesota DOT 2016) and Wisconsin (Wisconsin DOT 2014). Several other agencies collect and report performance internally. This type of report gathers data on winter weather to assess severity (e.g., precipita- tion and wind speed) and performance indicators to assess how the agency responded to each event (e.g., material use and cost). These data usually contain temporal and spatial information (i.e., when and where) to help agencies identify good performance as well as areas that need improvement; thus, agencies can use their resources more efficiently and achieve greater benefits. Reports on performance measures allow regional and county staff to compare resource use with that of their peers. For example, Minnesota DOT produces an annual Winter Maintenance Report (Minnesota DOT 2016) that summarizes maintenance performance and activities. This report gives context for the road mileage that is maintained and the resources used to complete the maintenance. The report also explains the dynamic nature of winter weather and the winter severity index that is used to normalize it. The report provides summaries of winter severity, material use, and costs Box 7. Minnesota DOT’s Use of Targets Minnesota DOT has two primary measures to gauge winter maintenance performance: return to bare pavement and public satisfaction (Minnesota DOT 2016). Because Minnesota DOT has calculated the two primary winter mainte- nance performance measures for many years, the targets are well established. The customer satisfaction survey results can be used to adjust the targets for return to bare pavement. The targets for return to bare pavement vary by road classification, specifically: • Super commuter: 0–3 h, • Urban commuter: 2–5 h, • Rural commuter: 4–9 h, • Primary collector: 6–12 h, and • Secondary collector: 9–36 h. The target for public satisfaction is a customer response of 7.0 or greater (out of 10). These findings are analyzed with historical data to establish a trend and target level of satisfaction.

138 Performance Measures in Snow and Ice Control Operations from season to season by district and statewide, ending with the final results of how often Minnesota DOT met bare-lane targets. Some agencies, such as the Idaho DOT, have experimented with dashboards to convey information with a higher frequency (see Figure 13). These dashboards portray the most recent processed information available to the agency, which in some cases can be (near) real time. Tools are available that can assist with generating easy-to-understand dashboards. Agencies can also use more advanced, centralized, and integrated systems that provide more advanced analysis and visualization tools, such as the Regional Integrated Transportation Information System (RITIS) developed by the Center for Advanced Transportation Technology Laboratory at the University of Maryland. RITIS is an automated data sharing, dissemination, and archiving system that includes many performance measures, dashboards, and visual analytics tools that help agencies gain situational awareness, measure performance, and communicate information, both between agencies and to the public. Source: Idaho Transportation Department (https://apps.itd.idaho.gov/apps/Dashboard/). Figure 13. An Example of the Idaho Transportation Department’s dashboard showing a snow and ice performance measure.

Using Performance Information 139 Reporting of performance measures is not a trivial task. The following steps for developing a robust performance measurement reporting strategy are provided based on guidance from the FHWA Transportation Performance Management (TPM) Toolbox (FHWA 2018). 1. Clarify audience. External and internal audiences require different levels of detail and information. 2. Define roles and responsibilities. Especially for external audiences, agencies need to follow protocol in terms of providing information to the public through existing public information officers. 3. Develop reporting parameters. For snow and ice, agencies need to define what information they can report on an event basis and what information is seasonal. 4. Refine and automate. Where possible, agencies need to create an automated approach to reporting performance. Relying on human input to create performance reports can require additional staff time.

140 Two important steps are identified for reinforcing performance-based management, as shown in Figure 14. C H A P T E R I V Reinforcing Performance-Based Management Reinforcing Measures Figure 14. Steps for reinforcing performance-based management. Step 9: Integrate into Decision Making The value of performance measures stems from their ability to support effective and timely decision making at multiple levels within an agency. The specific performance measures used by agencies are largely determined by the timing of the decisions. When used to influence decision making during a storm, a DOT might use measures of on-the-ground conditions such as storm speed, solid material application rate, pavement temperature, and air temperature. For post- storm assessments, measures such as a winter storm index, a winter mobility index, the amount of salt or other materials used, or lane miles plowed might be more suitable. When conducting annual reviews, performance measures that summarize the entire season of activities can be beneficial, including the number of snow events, the number of freezing rain events, total snow amount, and total number of incidents. Table 10 provides a high-level list of such decisions related to snow and ice removal that agencies make and how safety, mobility, and sustainability metrics can help. Using performance measures to guide snow and ice planning, investment decisions, strate- gies, and tactics can provide a clear basis for action. Once the decisions are made, the same performance measures can provide an assessment of the decision and enable adjustments. Using performance measures for supporting decision making is dependent on the reliability and con- sistency of the measures. While reliability is self-evident, consistency of measures is necessary to prevent a whiplash effect in decisions where a rapidly varying measure can create uncertainty in the nature of the decision. Logical times to review the role of performance measures in decision making are during snow and ice strategic planning and budgeting, annual maintenance reviews, and meetings and after-action reviews (see Box 8).

Reinforcing Performance-Based Management 141 Types of Decisions That Should Be Informed by Performance Measures Level of Support (high, medium, low; blank = not applicable) Safety Mobility Sustain- ability Level 1: High-Level Decision Makers 1 Making the case for additional investment H H M 2 Reporting on the cost effectiveness of current responses M M H 3 Gaining public and other decision makers’ support for program H H M 4 Improving decision making with respect to budgeting and programming funds H Level 2: Statewide/Regional Operations/Maintenance Leads 5 Maintaining adequate winter mobility in the state H 6 Supporting better allocation of funds between regions/districts H H H 7 Managing performance of contracted services H 8 Optimizing material management (balancing available supply and demand) H 9 Minimizing environmental impacts H 10 Supporting workforce development H 11 Supporting traveler information and emergency declarations H 12 Supporting asset equipment maintenance decisions H Level 3: Field Maintenance Supervisors 13 Supporting strategic decision making for event preparation and response H H 14 Supporting tactical strategies on where and how to respond H H 15 Improving the ability to relate crew performance to LOS H H 16 Providing effective feedback to field personnel on their performance H H H Table 10. Typical decisions faced by agencies responsible for snow and ice control. Box 8. Potential Areas and Times for Integrating Performance Measures into Decision Making 1. After-action reviews. For significant events, most agencies conduct an after- action review. Performance measures can be helpful in the after-action review by providing data and insights into what worked and what did not. 2. Annual maintenance reviews and preparedness meetings. Most agencies have a preparedness meeting at the beginning of each winter to discuss the plan for the upcoming season. Performance measures, especially those that sum- marize previous seasons’ performance, are valuable to guide future program direction. 3. Snow and ice budgeting process. Having outcome measures available can help make the argument for more or sustained levels of funding among various other competing priorities.

142 Performance Measures in Snow and Ice Control Operations Step 10: Evaluate and Improve Performance measures are not static. As program missions, goals, and objectives change, performance measures may have to be modified as well. Likewise, as an agency’s institutional capability grows, established performance measures can be refined, and more comprehensive performance measures can be developed in order to be more useful for making better-informed decisions. Even with a consistent set of objectives, progress toward achieving objectives can require changes to targets as the agency tracks the measure. Performance targets should be reexamined annually or at least biennially to determine how realistic they were and then be adjusted, if necessary, based on agency performance and customer satisfaction. The frequency of this review is dependent on the agency’s capability to collect and process data, and the agency will need sufficient historical data to analyze performance and compare it to historical trends. In general, an agency can improve snow and ice performance reporting every season through continuous process improvements as institutional capabilities improve (e.g., staff become more skilled in data collection, analysis, and reporting); collaborative relationships with plow operators and other stakeholders become better established; and decision makers and leadership see value in performance measurement. New investments in data and technology improvements as well as staff training opportunities may also improve the capability to calculate existing or new measures. As an agency becomes more familiar with established snow and ice control performance measures, opportunities may exist to improve those measures by refining them in ways that were not possible when they were initially developed (e.g., more specific for different areas, differentiated by roadway classification, and inclusive of more areas or roadway classifications). For instance, successful collaboration efforts with plow operators to report data may allow for greater cooperation in the successive winter season for more frequent reporting or more road- ways. Agencies may also consider development of additional measures to expand snow and ice control performance measurement. In particular, implementation of a new performance measure provides an opportunity for an agency to evaluate the effectiveness of the established data collection, analysis, and reporting process at the end of the first winter season it is used and implement it more broadly or incorporate modifications for improvement based on lessons learned.

143 Adams, T., Wittwer, E., O’Doherty, J., Venner, M., and Schroeckenthaler, K. (2014). Guide for Selecting Level-of- Service Targets for Maintaining and Operating Highway Assets. Contractor’s Final Report, NCHRP Project 14-25. University of Wisconsin–Madison. http://www.trb.org/Main/Blurbs/173327.aspx. Alaska Department of Transportation and Public Facilities. (2015). How ADOT&PF Approaches Ice and Snow Removal. http://dot.alaska.gov/nreg/blog/blog23.shtml. Alaska DOT. (2017). Winter Highway Performance Targets. Blackburn, R., Bauer, K., Amsler, D., and Boselly, S. M. (2004). NCHRP Report 526: Snow and Ice Control: Guidelines for Materials and Methods. Transportation Research Board of the National Academies, Washington, D.C. Buckler, D. and Granato, G. (1999). Assessing Biological Effects from Highway Runoff Constituents - Report 99-240. Washington, D.C.: U.S. Department of Interior and U.S. Geological Survey. Cambridge Systematics, Boston Strategies International, Gordon Proctor and Associates, and Markow, M. J. (2010). NCHRP Report 666: Target-Setting Methods and Data Management to Support Performance-Based Resource Allocation by Transportation Agencies. Transportation Research Board of the National Academies, Washington, D.C. CTC and Associates and Wisconsin DOT. (2009). Levels of Service in Winter Maintenance Operations: A Survey of State Practice. Wisconsin DOT Research and Library Unit. Clear Roads Pooled Fund Study. Cui, N. and Shi, X. (2015). Life-Cycle Sustainability Assessment of Highway Winter Maintenance Operations (Phase I). Tier 1 UTC for Environmentally Sustainable Transportation in Cold Climates. Deeter, D., Crowson, G., Roelofs, T. S., and Gopalakrishna, D. (2014). Best Practices for Road Condition Reporting Systems: Synthesis Report. Prepared for FHWA in cooperation with the Traffic Management Center Pooled Fund Study, FHWA-HOP-14-023. EPA New England. (2005). EPA 901-F-05-020. Retrieved from What You Should Know About Safe Winter Roads and the Environment: https://www1.maine.gov/mdot/winterdriving/docs/EPAwinterfacts.pdf. Fay, F., Akin, M., Shi, X., and Veneziano, D. (2013). Revised Chapter 8, Winter Operations and Salt, Sand and Chemical Management, of the Final Report on NCHRP Project 25-25(04). Washington, D.C.: American Association of State Highway (AASHTO) Standing Committee on Highways. Fay, L. and Shi, X. (2012). Environmental Impacts of Chemicals for Snow and Ice Control: State of the Knowledge. Water, Air & Soil Pollution, 223, 2751–2770. FHWA. (2013). Title 23 CFR Part 490 – Final Rule. Retrieved from the Federal Register, Volume 82, Number 11 (Wednesday, January 18, 2017). https://www.govinfo.gov/content/pkg/FR-2017-01-18/html/2017-00681.htm. FHWA. (2017). Highway Safety Improvement Program and Safety Performance Management Measures Final Rules Overview. 12 20. https://safety.fhwa.dot.gov/hsip/spm/measures_final_rules.cfm. FHWA. (2018). TPM Toolbox. URL: https://www.tpmtools.org/. FHWA Office of Operations. (2017). Saving Money and the Environment. (Publication No.: FHWA-SA-96-045). Retrieved from Road Weather Management Program. https://ops.fhwa.dot.gov/weather/resources/publications/ tech_briefs/cs092.htm. Governors Highway Safety Administration and U.S. DOT. (2012). MMUCC Guideline: Model Minimum Uniform Crash Criteria, 4th Edition. https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/811631. Hawkins, R. H. (1971). Street Salting, Urban Water Quality Workshop. Syracuse: State University College of Forestry. IHS Global Insight. (2014). The Economic Cost of Disruption from a Snowstorm. American Highway Users Alliance. Levelton Consultants. (2007). NCHRP Report 577: Guidelines for the Selection of Snow and Ice Control Materials to Mitigate Environmental Impacts. Transportation Research Board of the National Academies, Washington, D.C. Guide References

144 Performance Measures in Snow and Ice Control Operations Margiotta, R., Lomax, T., Hallenbeck, M., Dowling, R., Skabardonis, A., and Turner, S. (2013). SHRP2 Report S2-L03-RR-1: Analytical Procedures for Determining the Impacts of Reliability Mitigation Strategies. Transportation Research Board of the National Academies, Washington, D.C. Matthews, L., Andrey, J., Hambly, D., and Minokhin, I. (2017). Development of a Flexible Winter Severity Index for Snow and Ice Control. Journal of Cold Regions Engineering, Volume 31, Issue 3. McCullough, B., Partridge, B., and Noureldin, S. (2013). Snow and Ice Performance Standards. Publication FHWA/IN/JTRP-2013/21. Joint Transportation Research Program, Indiana Department of Transportation and Purdue University, West Lafayette, Indiana. Minnesota DOT. (2015). 2014–2015 Winter Maintenance Report at a Glance. Minnesota DOT. Minnesota DOT. (2016). Annual Minnesota Transportation Performance Report 2015. http://www.dot.state. mn.us/measures/pdf/12-2%20publicationsmall.pdf. Nagurney, A. and Qiang, Q. (2012). Fragile Networks: Identifying Vulnerabilities and Synergies in an Uncertain Age. International Transactions in Operational Research, 19, 123–160. doi:10.1111/j.1475-3995.2010.00785.x. National Highway Institute. (2016). Workshop: Steps to Effective Target Setting for Transportation Performance Management. Washington, D.C.: Federal Highway Administration. Nixon, W. and Qiu, L. (2005). Developing a Storm Severity Index. Transportation Research Record: Journal of the Transportation Research Board, No. 1911, Transportation Research Board of the National Academies, Washington, D.C. http://dx.doi.org/10.3141/1911-14. Paschka, M. G., Ghosh, R. S., and Dzombak, D. A. (1999). Potential Water-Quality Effects from Iron Cyanide Anticaking Agents in Road Salt. Water Environment Research, 71(6), 1235–1239. Ramakrishna, D. M. and Viraraghavan, T. (2005). Environmental Impact of Chemical Deicers—a Review. Water, Air & Soil Pollution, 166, 49–63. Roth, D. and Wall, G. (1976). Environmental Effects of Highway Deicing Salts. Ground Water, 14(5), 286–289. Sasha, P. and Young, R. (2014). Safety and Road Closure Benefits of Rural Interstate Variable Speed Limit Systems. ITS World Congress. Venner Consulting and Parsons Brinckerhoff. (2004). NCHRP Project 25-25/Task 04, “Environmental Steward- ship Practices, Policies, and Procedures for Road Construction and Maintenance.” Contractor’s final report. Chapter 8: Winter Operations and Salt, Sand, and Chemical Management. Wisconsin DOT. (2014). Annual Winter Maintenance Report 2013-2014: Keeping Wisconsin Moving During the Polar Vortex. Division of Transportation System Development Bureau of Highway Maintenance Winter Operations Unit. Yurek, R., Albright, N., Brandenburg, J., Haubrich, M., Hendrix, M., Hillis, D., and Zimmerman, K. (2012). NCHRP Project 20-68A, Scan 10-03: Best Practices in Performance Measurement for Highway Main- tenance and Preservation. Transportation Research Board of the National Academies, Washington, D.C.