III. Comments on Each Chapter of the Draft Climate Science Special Report

III.1 CHAPTER 1: OUR GLOBALLY CHANGING CLIMATE

Summary

Overall, Chapter 1 provides a solid introduction to the topic of climate trends and associated confidence that accurately reflects current understanding. The focus is appropriate, with most of the emphasis on observed trends, but some discussion of projections. The treatment is compact, but mostly at a sufficient level of detail to effectively communicate both the conclusions and the nature of the underlying evidence.

The emphasis on multiple lines of independent evidence, featured in Key Finding 3, is central to the chapter’s impact. Throughout the chapter, an increased emphasis on documenting the findings that are based on multiple lines of independent evidence would make the chapter more effective.

The Committee thinks that the chapter can be improved in three major ways. First, the topic of extreme event attribution, a major development over the last decade, should be discussed. The introduction to extremes in Section 1.2.4 provides an appropriate discussion of trends in extremes, but the lack of consideration of extreme event attribution is a missed opportunity. Second, the long section on the hiatus in Box 1.1 of the draft CSSR gives that event much more prominence than is warranted. The main point of Box 1.1 is that internal variability can distort short-term trends. This is an important point, appropriately emphasized in Key Finding 5. Box 1.1 could be made more useful and consistent with the broad sweep of climate knowledge if it were retitled to address the role and magnitude of internal variability and if shortened substantially to provide more focused support for Key Finding 5. Third, the chapter would be substantially easier to read with a renumbering that creates a series of top-level sections. The current numbering somewhat awkwardly places most of the chapter contents in several subsections of Section 1.2. Renumbering as 1.3, 1.4, etc. would be a straightforward way to improve readability.

In addition to those three major points, the Committee has some further recommendations for improvements. Throughout Chapter 1, greater use of quantitative language, even with findings presented in qualitative terms, would be beneficial. A good example is Key Finding 5, where it is very hard to interpret “important, but limited influences on global and regional climate over timescales ranging from months to decades.” In cases like this, where the goal is to indicate that something plays a small (or a large) role, the point would be clearer and more complete with more quantitative framing. For example, rewording as “influences that can have important impacts, especially regionally, over months to years but are limited to a small fraction of global climate trends over decades” would better convey the message of the key finding.

Chapter 1 includes a somewhat awkward mix of observations and projections, most of which are discussed in greater detail in later chapters. Specifically, Key Findings 2 and 4, and Figure 1.4 concern projections. Chapter organization such that the text flows smoothly from observations to projections is appropriate for the chapter, but the introductory paragraphs could better prepare the reader.

A challenge in any climate assessment is how to present observed (past) trends in important climate variables, like global mean surface temperature. There are three important considerations in constructing such a quantity: illustrating the possible role of human influence for scientific purposes, aligning with policymakers’ needs, and data availability. For illustrating human influence, one could either compute a mean rate of change over the period of anthropogenic

forcing or compare recent with baseline, or “pre-industrial” averages (subject to data limitations). The period 1850-1900 is widely accepted in both scientific (e.g., IPCC, 2013) and policy circles as “pre-industrial” and therefore is a good baseline to use for a “before-and-after” comparison, where the “after” would be the most recent 20-30 years. Furthermore, this baseline period minimizes the influence of anthropogenic GHG emissions on climate, but is recent enough that an adequate observational record exists.

In the draft CSSR, however, 1901-1960 is generally used, in effect, to define “before,” despite the fact that considerable growth in anthropogenic forcing occurred during this period (see e.g., draft CSSR Figure 2.6). In other words, this approach to characterizing change suffers from the weakness that it is both too recent and too long to characterize “before,” in addition to the statistical weaknesses of this metric discussed in Section II.2 of this report. The chapter makes the legitimate point that global mean temperature is better known after 1900 than before, but an earlier period can be safely used.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

Overall, the chapter is well balanced and reflects the relevant scientific literature. The chapter could be made considerably stronger with discussion of extreme event attribution and reduced emphasis on the hiatus. The general area of extreme event attribution is so important that it may warrant a separate key finding. Alternatively, a sentence or two on extreme event attribution could be added to Key Finding 2. This should be coordinated with the recommended increased emphasis on event attribution in Chapter 3 (see Section III.3). Discussion of the risks of multiple interacting impacts, and the large magnitude of past sea level excursions could also be considered. A brief discussion of changes in ocean heat content would also be beneficial in this chapter.

Are the key findings presented clearly, and documented in a consistent, transparent and credible way?

In general, the key findings are clear, appropriate, and well documented, however some attention is needed.

Of the list of indicators, changes in ocean heat content and SLR are only mentioned in Key Finding 1; some discussion elsewhere in the chapter would be appropriate. For Key Finding 1, other candidate indicators that could strengthen the list include decreasing Arctic sea ice, depth of seasonal permafrost thaw, earlier snowmelt in rivers, and start and end dates of growing seasons. Also, the phrase “rapidly compared to the pace of the natural changes in climate that have occurred throughout Earth’s history” could be improved, with an adequately detailed explanation in the traceable account. Specifically, what does “rapidly compared to...” mean? Is there enough information to quantify past rates of change and their uncertainties, and compare them with recent changes?

For Key Finding 2, neither the text nor the traceable account provides justification for the phrase, “…although models tend to underestimate the observed trends.” The way the key finding is worded, a reader cannot determine whether the mismatch between observations and simulations is a serious issue. This should be clarified.

The concept of “no alternative explanations” needs further discussion to be understood by the intended audience. There are lots of alternative explanations. It is just that, for a number of very solid reasons, they are not credible or cannot contribute more than marginally to the observed patterns. It may be that the authors have conflated attribution of global temperature changes since mid-20th century (for which it is true that there are no alternative explanations) and attribution of “observed climate changes.” The missing elements are the requirements that explanations be grounded in understood physical mechanisms, appropriate in scale, and consistent in timing and direction. Saying there are no alternative explanations invites a strong (even if incorrect) rejoinder. This recommendation also applies to the similar statement in the ES. Additionally, some identifiable natural cycles (e.g., ENSO, northern annular mode) may themselves be influenced by human activities. Rewording Key Finding 3 to address these recommendations would strengthen its impact.

In the major uncertainties provided in the traceable accounts for Key Finding 4, the text should emphasize the uncertainty in the magnitude of climate feedbacks. It would also be helpful to name the major feedbacks, including the ice-albedo and cloud cover feedbacks and refer to the feedbacks discussion in Chapter 2 of the CSSR.

Are graphics clear, and do they appropriately reflect the major points in the text?

Chapter graphics are generally informative and appropriate, although clarification or additional detail should be provided for a few.

The caption for Figure 1.2 indicates that the temperatures are plotted relative to the 1901-1960 average. However, this cannot be the case because almost all of the temperatures from 1901 to 1960 are blue (negative). Instead, it looks like the reference temperature for the zero line is probably the 20th century average. Inclusion of standard deviations for each decade and explanation in the caption would improve this figure.

Figures 1.3 and 1.7 would benefit from an indication of the location of statistical significance of trends, and Figure 1.6 should show the envelope of model results in the time series of temperature anomalies.

Are likelihood / confidence statements appropriate, and justified?

The Committee did not identify any issues with the chapter’s confidence statements.

Are statistical methods applied appropriately?

In general, the Committee encourages analysis of trends based on regression or related slope-based techniques (that minimize end effects, and/or quantify uncertainty in the slope), rather than on differences between average conditions during a reference period and a later period. In Chapter 1 and throughout the report, it would be helpful to standardize time windows as much as possible, recognizing the intrinsic importance of calculating total warming since pre-industrial periods. See Section II.2 of the report for more detailed recommendations on this topic.

Is the chapter balanced? Are there areas that should be expanded, or removed?

The Committee recommends reframing and shortening Box 1.1 on the hiatus. One option is to discuss the hiatus within the context of other aspects of internal variability. This could include discussion of the limitation of evaluating short periods of record when looking for GHG signatures because of the difficulty in attributing trends to short periods. Further, short-term trends are not particularly useful for model evaluation because in many cases, we do not entirely understand what drives the short-term trends. A recent paper by Yan et al. (2016) would also be a valuable citation to consider including in the discussion of this topic.

Recommended changes to structure

Section 1.2 is longer than many chapters and subsections are uneven in effectiveness of conveying the intended message. The strongest sections quantify the trends they describe, are clear about the time periods under consideration, and attempt to provide brief explanations for the phenomena observed or modeled. Sections that need strengthening include those on precipitation, extreme events, and land processes. Integration of ocean heat content into the discussion of SLR and quantifying the changes and trends described in these sections would also benefit the chapter. The Committee further recommends reorganizing to make subsections into sections, with associated content changes based on comments provided earlier in this section for the chapter.

III.2 CHAPTER 2: PHYSICAL DRIVERS OF CLIMATE CHANGE

Summary

Chapter 2 provides an essential overview of the mechanisms of climate change. Much of the text is sufficiently detailed so that a scientifically literate audience can begin to understand how increases in GHGs can lead to large perturbations in the earth-atmosphere-ocean system. Text on the importance of feedbacks to this system is helpful. For example, the chapter makes clear the importance of water vapor in amplifying the radiative effects of CO₂ and other GHGs.

The Committee has some suggestions for improvement of the chapter. First, the text should emphasize from the start the interconnectedness of the Earth-atmosphere-ocean system. As written, there is too much emphasis on atmospheric processes, at least initially. The role of changing land

cover is not mentioned until 11 pages into the chapter, and the role of the ocean is not described until 18 pages in. Second, there is little mention of Chapter 2 in the ES. The Committee suggests that Figure 2.6 (with suggested edits provided here) could be included in the ES, along with Key Finding 1. Together, this material provides a strong demonstration of the changes in the drivers of climate. Key Finding 1 should also mention that anthropogenic forcing accelerated rapidly in the 1960s.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

A clear statement about the interconnectedness of the Earth-atmosphere-ocean system is needed early in the chapter. Climate change can be considered a redistribution of heat, water, and carbon within this interconnected system. The long-term consequences of anthropogenic climate change should be emphasized in the beginning paragraphs, with up-to-date references (e.g., Clark et al., 2016).

The chapter should clarify that the scientific and policy communities have devised a set of metrics with which to compare the relative effects of different perturbations to climate. These metrics include radiative forcing (RF), effective radiative forcing (ERF), global warming potential, and global temperature potential. Brief descriptions of each metric are warranted. In addition, the definition of ERF is not in line with that in the IPCC AR5 (Myhre et al., 2013). While ERF can be calculated in several ways, Myhre et al. (2013) clearly favor the approach that allows many rapid adjustments to forcing to take place, including that of land surface temperatures. Box 8.1 of Myhre et al. (2013) illustrates this widely accepted definition of ERF. The definition of climate sensitivity should be more detailed and the range of estimates for this important metric should be given. Finally, the text could refer to the envelope of climate projections for particular scenarios in Chapter 4 as a measure of how climate sensitivity varies across models. A succinct discussion of how climate sensitivity differs from transient climate forcing would also be helpful. Mention of the sources of uncertainties illustrated in the relevant figures in the draft CSSR, such as climate sensitivity, future GHG emissions, and ocean heat uptake, could also be useful.

Regarding the effect of aerosols on climate, the scientific community has moved on from the complicated and overlapping definitions of “direct effect,” “first indirect effect,” “semi-direct effect,” and so on. The text should adhere more closely to the new (and simpler) classifications of these effects: aerosol-radiation interactions and aerosol-cloud interactions, as described in IPCC AR5 (Boucher et al., 2013). Old terms should be mentioned once at most.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

This finding affirms the scientific consensus that anthropogenic emissions of GHGs have perturbed the radiative balance of the Earth. The Committee recommends clarifying the text by

revising to state that “… human activities have been, and increasingly are, the dominant cause…” The evidence base should include up-to-date references to changes in heat storage and other properties of the ocean. The Committee also recommends that the finding emphasize the rapid acceleration in anthropogenic forcing since the 1960s, as indicated by Figure 2.6.

Key Finding 2 confirms the large uncertainties in quantifying the effects of aerosols on climate, but uses a mix of old and new terminology to describe the interactions of aerosols with the climate system, making it confusing and hard to follow. The description of evidence base should adhere to IPCC AR5 terminology (Boucher et al., 2013) and references should be updated. As written, some of the evidence base is listed in the “uncertainties” section of the traceable accounts instead of in the “description of evidence base” section. Revising the traceable accounts to clarify the evidence vs. uncertainties would strengthen the finding. This finding should also emphasize the large regional forcings of aerosols over polluted areas and the potentially large consequences of these forcings. While global aerosol concentration is decreasing over recent decades, there is also much evidence that aerosol is increasing in developing countries, with potentially large consequences for regional climate. The text should also clearly state that the net effect of aerosols is cooling. Finally, the albedo effect of light-absorbing aerosols deposited on snow and ice should be mentioned.

This finding emphasizes the importance of feedbacks to the climate system, and is important for the intended audience. Examples of climate feedbacks would also be helpful in conveying this finding. More attention should be paid to the earth-atmosphere-ocean as an interconnected system, with changes to the ocean likely persisting for millennia. The Committee discourages ranking of the uncertainty in feedbacks e.g., “Cloud feedbacks carry the largest uncertainty of all the feedbacks...” Relative magnitudes of these uncertainties are not known. A graphic that specifically illustrates Key Finding 3 would also be helpful to the reader.

Are graphics clear, and do they appropriately reflect the major points in the text?

The Committee recommends that the Figures be updated to include more recent years, if possible. Figure 2.2 is very difficult to interpret, and relies on a non-standard definition of ERF. All feedbacks also appear to follow from temperature when in fact, feedbacks can arise directly from aerosol-cloud interactions and land albedo change can follow directly from land use change. The Committee suggests the diagram be revised and simplified to look more like Figure 8.1 in Myhre et al. (2013) or Figure 2.1 in Forster et al. (2007).

Figure 2.4 is outdated now that atmospheric CO₂ concentration has passed 400ppm. Figure should either be updated or deleted.

Figure 2.6 is an interesting figure that would benefit from clarification of some of the legend text in the caption, e.g., “Aer-Rad Int.” and “BC on Snow + Contrails.”

Figure 2.7 is probably unnecessary, as it adds little to the central message of the chapter.

The Committee recommends including a graphic that specifically illustrates Key Finding 3. Examples of existing relevant graphics include Figures 9.43 and 9.45 in Flato et al. (2013).

Are likelihood / confidence statements appropriate, and justified?

Yes, these statements are appropriate and justified in Chapter 2.

Are statistical methods applied appropriately?

Most Figures and Table 2.1 show confidence intervals, although the error bars in Figure 2.4 are not defined. A few values in the text lack an indication of uncertainty, as noted in Appendix A, Line Comments.

Is the chapter balanced? Are there areas that should be expanded, or removed?

Section 2.1 should be expanded as described previously, to include more information on the interactions of the ocean and land cover with the atmosphere. Section 2.2 should focus on all metrics of climate change, not just RF and ERF.

Recommended changes to structure

None beyond those previously described.

III.3 CHAPTER 3: DETECTION AND ATTRIBUTION OF CLIMATE CHANGE

Summary

This chapter is intended to convey the message that the observed changes in global climate since the mid-20^th century are detectable and largely attributable to human influences, which is an important point that is referenced in other parts of the draft CSSR. There have been several advances in detection and attribution of climate change, particularly the capability to attribute regional-scale climate change, extreme weather and climate events (or classes of events) to human influences. The fact that this chapter has only one key finding, which is focused only on the change in global mean surface temperature, is indicative of a missed opportunity. Both the IPCC Fourth Assessment Report and AR5 contain chapters that collected detection and attribution findings across a wide range of subjects. This information is distributed across several chapters in the draft CSSR. By the logic of including detection and attribution results in the chapters that covers that topic, the key finding in Chapter 3 should appear in Chapter 1.

The Committee recommends the following substantive changes to Chapter 3:

• The chapter should contain a more comprehensive evaluation of detection and attribution, refer more to IPCC reports, and place greater emphasis on the latest detection and attribution advances in both methodology and results. Input from an expert in detection

• and attribution could be beneficial in ensuring the latest understanding and advancements in the field are appropriately captured in the draft CSSR. The chapter should also clearly identify and provide a substantially more in-depth discussion of the major scientific questions that have received attention since IPCC AR5 and NCA3, particularly with regard to attribution of extreme weather events.

• The introduction, which the Committee found extremely dense and rather unintelligible for the intended scientifically literate audience, does not serve the intended purpose of introducing the reader to the topic. The introduction should include a better explanation of the conceptual approach to detection and attribution, and the detailed description of the methodology should be encapsulated in an appendix on methods.
• The remainder of the chapter should better link examples of detection and attribution to the discussion of these topics in other chapters of the draft CSSR by referencing relevant sections. There is now a rich literature on detection and attribution of climate change that should also be cited in this chapter, and where appropriate in other chapters. Some recommended citations are provided in the next section.
• The chapter could also benefit from some emphasis on the importance of this detection and attribution science for determining whether human influence on climate variables (and on individual extreme events or classes of extreme events) can be distinguished from natural occurrences. This discussion could then inform decisions on climate policy, adaptation, legal liability, etc.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

Much of the material in Chapter 3 is drawn from the IPCC AR5 (Bindoff et al., 2013). There are several other specific topics and papers that could also be cited to strengthen the message and content of this chapter. For example, discussion of the optimal fingerprinting technique and recent updates and applications of this method (e.g., Zwiers et al., 2011), as well as studies that use data assimilation as an underlying technique (e.g., Hannart et al., 2016) should be included. Citation of other attribution papers could include Schurer et al. (2013), Stone et al. (2013), Stern et al. (2014), Zwiers et al. (2013), Andres and Peltier (2016), and Hulme (2014).

Greater emphasis on the most recent advancements in detection and attribution is also warranted. The Committee recommends reviewing the NASEM report, “Attribution of Extreme Weather Events in the Context of Climate Change” (2016a) and references therein.

The Committee strongly recommends including a discussion of the nature of, or challenges in, detection and attribution, e.g., detecting and attributing changes in means vs. trends or extremes. Other examples of how detection and attribution approaches have evolved in the recent literature are also warranted. This is similar to the text already in the draft CSSR indicating that changes in extreme temperature now can be detected with greater confidence (NASEM, 2016a). Finally, some discussion is needed of the extreme values associated with a given averaging period (e.g., daily, monthly, seasonal, or annual records).

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

This key finding includes three statements that describe, in different ways, the human influence on the global mean surface temperature and does not go much beyond what was already documented in IPCC AR5. The three statements are also, to some extent, redundant with each other and with findings in other chapters.

An additional key finding about extreme events, which is the topic of many of the more recent detection and attribution studies, would substantially improve the chapter.

Are graphics clear, and do they appropriately reflect the major points in the text?

Figure 3.1 (the only graphic in the chapter) is very clear and makes its point well. However, it could be better linked to the chapter text.

Are likelihood / confidence statements appropriate, and justified?

Yes, the likelihood and confidence statements are appropriate and justified.

Are statistical methods applied appropriately?

In general, detection and attribution methods are statistical in nature, and this is conveyed in the chapter. On the other hand, there is a basic question about the amount of data (length of record) necessary to detect a trend in a climate time series. A statement to that effect should be included in this chapter, which would also be relevant to Chapter 1. The statement should be clear about how much more data is needed to detect a change in a trend (e.g., the hiatus) vs. detecting a trend. In addition, the description of multi-step attribution and attribution-without-detection methods is vague and hard to follow. Even the example is too abstract and does little to help the reader understand the material. The description of the risk-based approach to attribution is likewise vague and overly general. The section describing this approach would benefit from a mathematical expression to quantify the discussion and make it more concrete.

Is the chapter balanced? Are there areas that should be expanded, or removed?

As noted previously, the introduction should include a better explanation of the conceptual approach to detection and attribution, and the detailed description of the methodology should be encapsulated in an appendix on methods. The bulk of the chapter should then be devoted to describing examples of detection and attribution that are relevant to the other chapters of the draft CSSR. This could include a timeline, table, or other way to indicate how much the field of detection and attribution has changed in recent years. The challenges associated with model dependence and difficulties with attribution of extreme events could also be articulated more fully. The Committee suggests that the chapter would be strengthened by adding a key finding that highlights advances in the detection and attribution of features of climate change that go beyond simple global mean surface temperature. For example, “The science of event attribution is rapidly advancing with the understanding of the mechanisms that produce extreme events and the development of methods that are used for event attribution” (paraphrased from NASEM, 2016a).

Recommended changes to structure

See Summary comments and response to previous question.

III.4 CHAPTER 4: CLIMATE MODELS, SCENARIOS, AND PROJECTIONS

Summary

Chapter 4 provides necessary background about the growth of CO₂ concentrations, both in the recent past and projected in the future. The chapter also describes how global climate models (GCMs) and regional downscaling, either using regional dynamical climate models (RCMs) or statistical methods, transform information about changes in forcing by GHGs and aerosols into information about the climate system, in the past, present, and future. It is important to characterize the nature of the changing concentrations of GHGs and aerosols and the implications these have for the physical climate system, so this chapter represents a valuable portion of the report.

As written though, the chapter is difficult to read. The three topics named in the title of the chapter are treated in quite different depth: emissions scenarios are much more prominent than models and projections. Moreover, the draft is not balanced in terms of the discussion of GCMs and RCMs—the regional performance of GCMs is given short attention, and RCMs are given much more prominence than is commensurate with the rest of the draft CSSR. In particular, there is insufficient discussion of the limitations of RCMs, which could result in inadequate support for Key Finding 4.

It is important that Chapter 4 carefully articulate the advancements in climate modeling over time, including the evolution from atmosphere-centric to Earth system models, and focus that discussion on recent advancements such as are represented in the step from Coupled Model Intercomparison Project (CMIP) 3 to CMIP5. The discussion of the difference between the IPCC Special Report on Emissions Scenarios (SRES) approach and the RCP approach to emissions scenario development should be clearer, and the choice in the draft CSSR to focus on RCPs 4.5 and 8.5 should likewise be clarified. For example, it is implied that RCP4.5 represents a low emissions future, but RCP2.6 defines a much lower emissions future and one roughly consistent with the Paris Agreement. There are hypothetical scenarios (such as constant concentration and zero emissions) that should also be more clearly defined and described. There are ample reports and published papers documenting the similarities and differences in the two generations of emissions scenarios and GCMs (i.e., SRES-CMIP3 and RCP-CMIP5) that can be cited. Finally, the chapter is overly dependent on a single report (Kotamarthi et al., 2016) for much of the assessment discussion. Citation of the research literature underpinning the state of assessment science should be substantially increased.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

An important omission from this chapter is a discussion of the advances in climate modeling, both in GCMs and RCMs, that have been made since IPCC AR5 and NCA3. In particular, the CMIP5 generation of coupled model experiments has been executed and published and whose results were not extensively used in NCA3. This chapter would benefit from a pointwise description of the differences between CMIP3 and CMIP5, including both modeling advances and scientific findings. Such comparisons have been made and are published. For example, two recent reports from NOAA are available comparing the two generations of models and their results for North America

(https://docs.lib.noaa.gov/noaa_documents/NESDIS/TR_NESDIS/TR_NESDIS_144.pdf and http://cpo.noaa.gov/ClimatePrograms/ClimateandSocietalInteractions/COCAProgram/COCAArchive/TabId/390/ArtMID/1263/ArticleID/358942/Comparing-Two-Generations-of-Climate-Model-Simulations-and-Projections-of-Regional-Climate-Processes-for-North-America.aspx), and the papers cited in these reports are useful resources for this chapter.

It is unclear what value there is in including discussion of the World Climate Research Programme COordinated Regional climate Downscaling Experiment (CORDEX) in the draft CSSR. Results from CORDEX are not available and the RCM simulations in that experiment are run at 50-km spatial resolution, which is no longer significantly higher than typical GCM resolution, and based on a very limited and older set of GCM runs with a single SRES scenario.

One of the new advances heralded in Chapter 4 is the use of unequal weights in combining multiple climate models to arrive at consensus results. While it is true that previous studies have used equal weighting, it should be mentioned that this is not only due to expediency or to a desire not to offend certain modeling groups—there are studies indicating that equal weighting of climate model output is statistically unsurpassed by any unequal weighting scheme in terms of prediction skill, at least for some applications (e.g., Peng et al., 2002; Peña and van den Dool, 2008; DelSole et al., 2012). The model weighting discussion in Flato et al. (2013) may also be appropriate to reference in this chapter. Finally, the scientific and statistical advantage of the new method by Sanderson et al. should be highlighted.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

This key finding is not linked to the rest of the chapter or to Figure 4.1, where it could be illustrated. The key finding is presented with no uncertainties and only one citation (granted, an IPCC chapter), and yet is given “high confidence.” The supporting language in the rest of the chapter should be clear that this finding refers to a “constant concentration” scenario, not a “zero emissions” scenario—the latter would result in almost immediate, if gradual, decline in CO₂ concentration. The message for this key finding is better worded as it appears in the ES and the same language could be used here.

There are multiple statements in Key Finding 2. The first part about the current level of CO₂ concentration and its future growth should be given a separate confidence level (probably “high,” given the body of evidence cited).

The evidence base for this key finding is consistent with the confidence levels indicated. However, more evidence is needed in the traceable account for the statement that economic growth has begun to decouple from fossil fuel combustion.

This finding is more of a methodological decision than a finding and the evidence base provides inadequate support. It relies entirely on a single federal report in the gray literature (Kotamarthi et al., 2016), with a vague reference to a large body of literature—key examples from the latter should be cited. The portion of the key finding that “These techniques allow the scientific community to provide better guidance on the use of climate projections for quantifying regional-scale impacts” is given “medium to high confidence.” However, the science, as documented in the traceable accounts, does not support high confidence on this broad statement. Confidence depends on the specific guidance, and the specific impact, so the statement is overly vague and should be revised. The statement in the traceable accounts that downscaling is “broadly viewed” as robust should also be documented or deleted.

Are graphics clear, and do they appropriately reflect the major points in the text?

Figure 4.2 is confusing and could be deleted. The statement “calculated in 0.5°C increments” is not appropriate for the intended audience and the essential information is already conveyed much more effectively in Figure 4.1.

Figure 4.3 is an effective graphic, but would be better placed in Chapter 12 (Section III.12).

The Committee was divided about the value of Figure 4.5, with some asserting that it does not add to the report narrative. It depicts results with an RCM run at different resolutions, so it is not a good choice for demonstrating the difference between GCMs and RCMs. A replacement that specifically illustrates differences between GCMs and RCMs could be more useful.

Figure 4.6 adds little to the draft CSSR because it is stripped of the context provided in the original Hawkins and Sutton paper, where the regional uncertainties are visibly different from the global uncertainties, and where the total uncertainty grows with time. While it is important to show results for Alaska and Hawai’i when such results are relevant, the results for these regions in Figure 4.6 are not sufficiently different from the results for the contiguous United States (CONUS) to warrant inclusion. Moreover, even though the point made by the figure is important, it is not well linked with the relevant Chapters. This figure could be revised and included in Chapter 5, where it would make sense to complement Figure 5.4, or it could be moved to an appendix.

Are likelihood / confidence statements appropriate, and justified?

As stated in the discussion about key findings, this chapter would benefit from including uncertainties wherever possible, stronger traceable accounts, and greater balance in discussion of GCMs and RCMs.

Are statistical methods applied appropriately?

The discussion of the rate of change of CO₂ concentration in Section 4.2.5 suggests that finding an analogue in the paleoclimate record requires a match to the rate of change. The last sentence in Section 4.2.5 conflates magnitude of change and rate of change, without comment. As mentioned several times in the draft CSSR (e.g,. page 158, lines 18-19), the long-term impact of human activities on climate can be assessed in relation to the paleoclimate record only in equilibrium, so the rate of change of CO₂ concentration seems to be irrelevant. Some clarification of the relationship of these two seemingly different statements is needed.

Is the chapter balanced? Are there areas that should be expanded, or removed?

The treatment of GCMs and RCMs is uneven. For example, the list of features that are represented in a GCM on page 160 is rather odd in that it is neither comprehensive nor particularly representative of the important features that one expects a GCM to faithfully reproduce. Some clarification of the nature of this list is needed.

The description of RCMs and their advantages is much more coherent and comprehensive, but the list of shortcomings of RCMs is incomplete. In addition to what is mentioned, the chapter should discuss the mismatch between the way that GCMs and RCMs represent subgrid-scale physical processes and the fact that many RCMs lack two-way interaction, which results in an inevitable gradient in important quantities between the domains of the GCM and RCM. For example, unmatched boundary conditions on the downstream side of RCMs lead to unique biases; the grid spacing of RCMs, e.g., 50 km in the North American Regional Climate Change Assessment Program (NARCCAP) and CORDEX, is not very different from the grid spacing in GCMs being used in CMIP6, so the advantage of RCMs is not clear; the specification of GCM output at the lateral boundaries of RCMs introduces uncertainty and error; and considerable “hidden physics” is included at the lateral boundaries in the form of sponge conditions or other engineering accommodations for the mismatch in dynamic features at the interfaces.

Recommended changes to structure

The Committee recommends a number of revisions and reorganization of sections to better focus the chapter scope and improve the readability.

Section 4.2 should include an introductory paragraph specifically mentioning that there are different ways of addressing scenario uncertainty, depending on the objective. Sections 4.2.1 through 4.2.4 describe different ways of approaching the relationship between emissions, concentration, and temperature change, and this should be summarized in the introduction.

Section 4.2.1 second paragraph (page 154, lines 1-10) is difficult to follow and the purpose of the calculation is not described. The paragraph should be rewritten for clarity and motivation, and it should reference Swain and Hayhoe (2015).

Section 4.2.2 on Shared Socioeconomic Pathways seems out of place and adds little to the report. This section could be omitted.

Section 4.2.3 discusses the global mean temperature scenario approach and pattern-scaling, but it is unclear whether this technique is used in the rest of the report. Also, the approach

seems more related to impacts, in that it bypasses uncertainty in scenario evolution and deals more with specific impacts. It could be omitted as it pertains more to NCA4 than to the intended scope of the draft CSSR. Or, if kept, it should be revised.

Section 4.2.4 is back to cumulative C emissions, which again relates to mitigation policies. This fits better with Section 4.2.1, so omitting 4.2.2 and 4.2.3 would lead to a more logical order.

Section 4.2.5 does not fit well in its current location and would be more appropriate in Section 4.3.

Sections 4.3 and 4.4 contain materials that would fit better in a methods appendix (Appendix B of the draft CSSR is already a start).

Section 4.3.2 discusses CORDEX (page 161, lines 25-33) and could be omitted. See the earlier comment noting that the value of including CORDEX in the draft CSSR is not apparent.

Section 4.3.3 focuses on Empirical Statistical Downscaling Model (ESDM), but results do not figure prominently in the draft CSSR. The abbreviation is not used elsewhere, and outside of traceable accounts, “downscaling” appears only in Chapter 8. There is also no discussion of how ESDMs are evaluated, e.g., is there any dependent/independent data testing? If so, how well do these models perform in such tests? Finally, the section is overly reliant on Kotamarthi et al. (2016). This section should only be retained if considerably revised.

III.5 CHAPTER 5: LARGE SCALE CIRCULATION AND CLIMATE VARIABILITY

Summary

This chapter is well written and flows nicely. The chapter covers modes of climate variability in the tropics and mid-latitudes, and discusses recent advances in quantifying the role of internal variability on past and future climate trends. Some of these topics have seen advances in science and conceptual understanding since the NCA3 and IPCC AR5. The Committee has some suggestions for improving the chapter that are included here.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that the chapter accurately reflects the scientific literature, except in details of the discussion of the Atlantic Multidecadal Oscillation (AMO) and Pacific Decadal Oscillation (PDO). In particular, Newman et al. (2016) strongly caution against the interpretation that U.S. temperature and precipitation variations that occur concurrently with the PDO are indeed an impact of the PDO. Also, Newman et al. (2016) indicate that the PDO does not have a preferred time scale. The AMO has been defined different ways (average sea surface temperature over a region or leading pattern from Empirical Orthogonal Function analysis), and the instrumental record is too short to detect an oscillation with a putative 50-70 year period. It may be a statistical artifact, or it may result from interdecadal fluctuations in aerosol concentrations. Language should be changed as appropriate to reflect this literature, either removing references to these quasi-oscillations or including alternate judicious views for balance. See the Line Comments in Appendix A with additional suggestions for the AMO. The Committee did not think that any critical content areas were missing from the chapter.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

This key finding is generally presented clearly and well documented, but authors could consider adding that storm tracks are shifting poleward (e.g., Norris et al., 2016). Also, because this finding states only “medium to high confidence,” the inclusion of a likelihood statement could be confusing to interpret and may not be appropriate. Finally, it is not clear why the “Low” confidence box is checked in the traceable accounts.

Key Finding 3 is not well grounded in the text. The relationship between temperature and atmospheric specific humidity is not discussed in the chapter and should either be discussed, removed, or moved (and discussed) to Chapter 6.

Are graphics clear, and do they appropriately reflect the major points in the text?

Figure 5.1 is not referenced in the text and depicts an old and unreasonably over-simplified zonally averaged picture of the general circulation that is not realized in nature, other than to some degree in the Hadley cell. The Committee suggests removing the figure and instead explaining the processes briefly in the text (see detailed recommendation in Line Comments, Appendix A).

Are likelihood / confidence statements appropriate, and justified?

Likelihood and confidence statements are appropriate and justified, but see previous comment for Key Finding 1.

Are statistical methods applied appropriately?

A discussion of the statistical significance (even if qualitative) should be added where appropriate. In particular, the discussion of teleconnections to Central Pacific or Eastern Pacific El Niño-Southern Oscillation (ENSO) events is based on a very small number of events, and should be caveated.

Is the chapter balanced? Are there areas that should be expanded, or removed?

The chapter is relatively well balanced in its content. The Committee recommends expanding the discussion of model fidelity in simulating natural modes of variability, and as appropriate, the connection with temperature or precipitation over the United States. This is cited as a source of uncertainty for the Key Finding 2 justification and therefore needs to be supported by the text.

III.6 CHAPTER 6: TEMPERATURE CHANGES IN THE UNITED STATES

Summary

This chapter addresses changes in mean temperature and extreme temperature in the United States, which are of foundational importance in discussing climate change and informing the development of NCA4. Results are generally consistent with NCA3, though some differences have arisen because of changes in model weighting, variables considered, and averaging period. Chapter 6 is generally well written and flows nicely, but could be improved by expanded discussion of extreme heat, the influence of the Dust Bowl on the observed record, and other topics detailed here.

The Committee has the following concerns about the treatment of extreme events in this chapter:

• The extreme metrics were often difficult to understand, especially the definition of warm and cold “spells.” How brief are “brief periods”? And how much are temperatures above- or below-normal temperature? To clarify, a box or text should be added that explicitly defines each of the extreme metrics that are discussed (see Appendix A for additional extreme metrics that should be defined).
• The Committee strongly recommends additional discussion and justification of extreme heat changes. The data presented in Table 6.2 and Figures 6.3 and 6.4 seem inconsistent with Key Finding 2, apparently because of the extreme high temperatures during the Dust Bowl years. Some of this confusion with Key Finding 2 comes from the statement that “In recent decades … intense heat waves have become more common.” This could mean that the frequency of heat waves in the last couple of decades is greater than the frequency in the 1901-1960 period; or it could mean that there has been an upward trend in the last few decades. The latter is probably intended, but the language needs to be clarified and the issue needs to be addressed in more detail.
• The metric shown in the line plots in Figure 6.3 is confusing and needs further explanation. The Committee’s understanding was that each point represents the average (over all stations) of the highest temperature recorded during a particular year. This metric will be extremely sensitive to spatial distribution of stations and therefore to the approach of spatial averaging. The Committee recommends removing the line plots, using an area-based approach (for example, EPA metric at https://www.epa.gov/climateindicators/climate-change-indicators-high-and-low-temperatures or references therein) for depicting variations in U.S. temperature, or using a metric that is less susceptible to spatial inhomogeneity, such as days exceeding a local percentile threshold. Whichever approach is taken, the text or traceable account should include enough information for the reader to find or reproduce the plot.
• Other analyses have shown that even if the Dust Bowl is neglected, extreme high temperatures in the Midwest do not appear to have increased as they have in the western United States. This may be due to increased agricultural intensity (Mueller et al., 2016). It is important that the key findings accurately represent and explain this discrepancy.
• The Committee also suggests adding a paragraph that discusses maximum and minimum temperatures, or at least adding more language as to where this change is likely to be true (see page 224, lines 13-15). In the Midwest and Great Lakes region, the opposite may be true. Is this difference due to the model weighting, or is it spatially variable?

SPECIFIC REVIEW COMMENTS RELATED TO THE STATEMENT OF TASK

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that Chapter 6 generally reflects the scientific literature with accuracy. An exception is the discussion of extreme heat, as detailed previously. Additional discussion of changes in minimum and maximum temperature (daily highs vs. lows, and / or trends in winter vs. summer), both for past variations and future projections (page 224, lines 13-15) is also suggested.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

The change in annual average temperature should be expressed as a range that reflects the uncertainty in the estimate. Also, the estimated increase between 1901 and 2015 is less than the low end of Key Message 3 in NCA3 that stated, “U.S. average temperature has increased by 1.3°F to 1.9°F since record keeping began in 1895.” This difference needs to be discussed in the text. It would be useful to note that for most of the United States, the observed warming is consistent with anthropogenic forcing (Figure 6.5).

The Committee thinks the portion of the key finding referencing the paleo record and recent warming is likely overstated. The IPCC AR5WG1 provided a similar finding and attributed only medium confidence. Further, uncertainties associated with proxy records and reconstructions make it challenging to assign such a high confidence.

The description of evidence base does not contain appropriate information to support Key Finding 1. While it is true that previous assessments demonstrate that the United States has warmed, the specific amount of warming—and more importantly, the actual data sources and their uncertainties—are not given, and extremes are covered in Key Finding 2, not 1. Sea surface temperatures are barely discussed in the chapter and are not mentioned in Key Finding 1, so it is odd the topic is included in the traceable account, and while the data sources are given, no details are provided to show how the main conclusions are reached.

Key Finding 2 requires clarification and consistency of extreme events with the figures and evidence described, as stated in the Summary comments for this chapter. The statement that “the temperature of extremely warm days has increased across much of the West,” and “intense heat waves have become more common” is in direct contradiction (in message) to Table 6.2, which shows decreases in the warmest day of the year, and decreases in the warmest 5-day 1-in-10 year

event. This discrepancy needs to be addressed. Also, the description of evidence base provided for this key finding should include a discussion of how extreme temperatures during the Dust Bowl years have impacted relative changes in extreme temperatures over the recent period. The role of this event needs to be discussed in key findings (perhaps given its own key finding, or a discussion box). Further, it is difficult to understand how a statement that includes increases in extreme warmth can be associated with a high confidence or extremely likely statement, given that most of the graphics in this chapter show a decrease in extreme warmth in the historical record.

The Committee recommends expressing the projected change in terms of a range, rather than “at least 2.5°F.” The range of 5.0°F-7.5°F is due to scenario uncertainty, and it would be appropriate to list the range of expected warming for each of the two emissions scenarios instead. Also, the description of evidence base is too general in citing broad assessments when it would be more appropriate to cite specific literature. Indication of what data set the projections are based on is needed, and how model weighting is applied (if it is). Finally, quantitative statements linked to “extremely likely” should include the appropriate ranges computed using multiple GCMs. The numbers used here do not match Table 6.4. There may be an undocumented mismatch in the area indicated, definition of “late century,” and which RCPs are considered, and this should be noted.

Similar to Key Finding 3, the description of evidence base should indicate what data set the projections are based on, and how model weighting is applied (if it is).

Are graphics clear, and do they appropriately reflect the major points in the text?

Figure 6.2 is not cited in the text and requires additional detail to provide support for chapter messages. Specific considerations are provided in Appendix A.

Figure 6.6 is not cited in the text and should either be removed or moved to Appendix B of the draft CSSR, where model weighting is discussed. The figure is also challenging to interpret and requires more explanation. The metric “distance from observations” would likely be confusing to the intended audience, and most scientists would require some knowledge of how that distance was calculated.

Figure 6.9 adds little to the chapter besides illustrating large geographic themes. It could be noted here that the empirical statistical downscaling improves on the coarse climate model output, by establishing a more geographically accurate baseline for number of days per year. Some of the changes are strongly tied to that baseline, which in turn is strongly tied to topography. That is, locations where minimum temperature is rarely < 32^oF (southern Arizona, gulf coast) see only very small changes.

Table 6.2, specifically the fact that nearly half of the extremes presented here have gotten cooler, not warmer, does not support the assertions in Key Finding 2. Context should be provided to explain this discrepancy.

Tables 6.4 and 6.5 should include uncertainty ranges.

Are likelihood / confidence statements appropriate, and justified?

All key findings contain both a likelihood and confidence statement. Only one should be listed—probably the likelihood statement.

Are statistical methods applied appropriately?

No statistical significance of historical trends is provided. The Committee strongly recommends reporting past trends and future projected changes with a range of values using commonly accepted methods. See Section II.2 of this report for more detailed recommendations about the treatment of trends and statistics. Figures and tables should show statistical significance of changes in temperature. Text describing projected temperature changes, including captions, should indicate the number of models or simulations used to calculate the average change.

Is the chapter balanced? Are there areas that should be expanded, or removed?

For the most part, the chapter is balanced in the topics covered, with noted exceptions. The Committee suggests additional discussion of changes in daytime high temperature vs. nighttime low temperatures, that are consistent with the recommendations in Chapter II about extreme events.

Recommended changes to structure

As part of the restructuring recommended for Chapter 3, some of the attribution information could be moved to Section 6.2.

III.7 CHAPTER 7: PRECIPITATION CHANGE IN THE UNITED STATES

Summary

This chapter is structured with a series of subsections that address historic changes (annual and seasonal, then snow, extremes, extratropical cyclones, and detection and attribution) and a second series of subsections that address projections (seasonal means, snow, extremes, and hurricanes). This structure is easy to follow, and addresses the main topics. Given the importance of precipitation to water resources and hazardous extremes and the reality that these will be among the costliest manifestations of climate change, it is appropriate that the CSSR authors have broken out snow, as well as extratropical cyclones and hurricanes, as separate sections. Noting that Chapter 8 discusses drought, since drought is mentioned numerous times in Chapter 7, would also be helpful.

The Committee identified multiple sections where the chapter would benefit from further clarification and discussion of the breadth of available literature. The use of different historical periods in Section 7.1 is confusing for the reader. Some of this may be unavoidable given that results are reported from many publications that have made their own decisions as to historic periods. Nonetheless, the Committee suggests trying to identify trends over the last century (more or less), and the period of greatest GHG emissions, roughly the last 40-50 years. In some cases, it may be possible to replot results of others for these periods, or at least provide an interpretation that maps to these periods (or others that are defensible). Regardless, the time period evaluated should be clearly stated for all analyses. Additionally, the text is inconsistent in use of “ramp” vs

“step” trends (see also Section II.2). For instance, Figure 7.1 uses a step, which implies trend magnitudes that are half what they would be using a ramp. In most cases, ramp is preferable since manifestations of climate change occur gradually over time, with an exception being when some event may have caused an abrupt shift.

Chapter 7 seems to overstate the evidence for changes in precipitation extremes. For instance, the cited Westra et al. (2013) paper, reports that the number of statistically significant upward trends is larger (by a factor of 4 or so over the CONUS) than downward trends, but less than 9% of trends are statistically significant and upward and 5% would be expected due to chance. This suggests there may only be “weak evidence of increases in extremes” and the Committee recommends revising the text to better reflect the findings of relevant literature.

The snowpack discussion in this chapter focuses primarily on snow cover extent and lacks adequate discussion of snow water equivalent (SWE). Particularly over the West, where much of the annual runoff originates as snowpack, SWE in the springtime is critically important for hydrology and water resources, with snow cover extent being a much less important factor. The chapter should include some information about long-term SWE trends and increase discussion of this topic in the context of projections. There is recent work based both on observations and historical model reconstructions that could also be cited (e.g., Mote et al., 2016; Mao et al., 2015; and Margulis et al., 2016). An expanded section on snowpack, particularly SWE, could either be retained in this chapter or moved to Chapter 8, but should appear with a more comprehensive discussion in one.

SPECIFIC REVIEW COMMENTS RELATED TO THE STATEMENT OF TASK

Does the report accurately reflect the scientific literature? Are there any critical content areas missing from the report?

The chapter reflects the scientific literature reasonably well. Addressing the gaps noted previously with respect to precipitation change that affect hydrology will improve the chapter balance. For precipitation extremes, the Committee suggests reviewing the report, “Climate Stabilization Targets: Emissions, Concentrations, and Impacts over Decades to Millenia” (NRC, 2011). Although not as recent as some available literature, this publication addresses the topic and may be appropriate to include. The Committee did not think that any critical content areas were missing from the report.

Are the key findings presented clearly, and documented in a consistent, transparent and credible way?

Generally, the traceable accounts require the inclusion of more details about the science supporting the key findings and references to the literature. As written, there is not enough information to follow the line of evidence that underpins the findings. For example, Key Finding 3 points vaguely to “climate model projections and our understanding” which, combined with section 7.2.2, is insufficient to document how the calculations for Figure 7.7 were done in support of this key finding.

The Committee suggests deleting the first sentence of this key finding. The core of the finding is stated in subsequent sentences, and the fact that precipitation has increased slightly over the last century is primarily attributable to large scale droughts in the 1930s and 1950s. There are, however, important regional differences. The finding should also state the nature of changes over the post-1970 period, as noted previously.

This finding would be strengthened by focusing more specifically on the observation that, over the last century, heavy precipitation has increased in intensity and duration at a small, but statistically significant, number of stations. For stations where changes have been observed, a substantial fraction (about 80%) have been increases. The word ‘across’ implies ubiquity and therefore may not be the appropriate word choice. The finding should also include a statement about post-1970 trends.

This key finding would be much more impactful if it focused primarily on CONUS (and perhaps Alaska), and on SWE rather than snow depth and extent. As written, it is not supported by any figure or table, although Figure 8.3 could be relevant but is not mentioned here. See also the Summary for this chapter.

Are graphics clear, and do they appropriately reflect the major points in the text?

Generally, yes, however, additional detail is needed for some figures. For instance, Figure 7.7 (and others) would benefit from a more informative title and labeling of the y-axis. As currently displayed, it is very difficult to interpret. Also, how many of the CMIP5 models are represented in Figure 7.7? The across-model variations seem low.

In Figure 7.8 the spatial variability in projected changes also seems low and, if correct, bears explanation. It should also be noted for this figure whether it makes a difference if the return period is different. For the extreme value distribution EV1, the quantiles are just a fixed multiple of the mean, so changes in the mean are proportionately reflected in changes at any given return period. While the same does not apply for other distributions, it may well be approximately true, so perhaps something could be said about how other return periods change.

Chapter 7 would benefit from tables equivalent to those in Chapter 6 (Tables 6.1, 6.2, 6.4, and 6.5). Chapter 6 also noted that there were differences in changes in extremes, depending on which extremes were considered. A table showing changes for three to four definitions of extreme precipitation would be helpful, or a strong justification for selecting only the 2-day 5-year event for the bar charts in Figure 7.7 and 1-day 20-year event for the map in Figure 7.8.

A figure illustrating changes in snow cover extent could also be useful to this chapter and Figure 7.5 could be moved to Appendix B in the draft CSSR.

Are likelihood / confidence statements appropriate, and justified?

For the most part, yes, they seem appropriate.

Are statistical methods applied appropriately?

As detailed earlier, trends should include statistical significance statements whenever possible throughout the chapter.

Is the chapter balanced? Are there areas that should be expanded, or removed?

The chapter is reasonably well balanced.

III.8 CHAPTER 8: DROUGHTS, FLOODS, AND HYDROLOGY

Summary

This chapter is organized differently from Chapter 7, from which it logically follows. Chapter 7 includes first a historical context (basically trends) in the different subtopics then projections for each. The Committee recommends this structure also be used for Chapter 8 to provide a clear picture of what has been happening over about the last century, and what is projected to happen in the future. Also, wildfire (Section 8.3) does not fit naturally with the subject of the chapter as represented by the title, and probably belongs elsewhere in the report, perhaps Chapter 10. Finally, the title implies a rigorous consideration of hydrology (i.e., full hydrological cycle, including for instance groundwater). While there are well-defined subsections for droughts and floods, there is not for hydrology, creating a structural mismatch with the title. Perhaps the chapter could include some brief narrative about what is meant by ‘hydrology’ in this context, point out what is not covered, or revise the title to better reflect the chapter content.

A number of substantial improvements are strongly recommended for Chapter 8 beyond these organizational suggestions. The Committee recommends that the chapter authors consider consulting with hydrologic experts to assist in revising this chapter. More extensive input from researchers with such expertise would help ensure that the final text is more authoritative and balanced.

Most of the primary recommendations for this chapter are framed through the content presented in the key findings. Revising the chapter text to reflect these recommendations given for key findings will help to strengthen the chapter.

Key Finding 1 does not fully reflect the science regarding trends in droughts. While some specific regions have experienced recent droughts of record intensity, analysis of global and continental-scale trends indicates that drought severity and other statistics have actually declined (e.g., Sheffield et al., 2012; Andreadis and Lettenmaier, 2006; and Mo and Lettenmaier, 2015). Recent research finds that over about the previous 100 years, slight increases in precipitation (which are noted in Chapter 7) have overcome increased evapotranspiration (ET), resulting in generally increased soil moisture (Andreadis and Lettenmaier, 2006). Also, low flows (another indicator of drought) have become less common across much of the country, as documented in

references such as Lins and Slack (1999 and 2005), as well as other U.S. Geological Survey publications which could be cited and discussed.

Key Finding 3 does not accurately reflect the current state of understanding about the link between soil moisture and temperature. Changes in soil moisture depend entirely on the balance between precipitation changes and ET changes (presumably increases). A common misconception, which is reflected in some of the work on drought, is that potential evapotranspiration is strongly related to temperature, and hence temperature increases result in strong increases in ET. However, ET over most parts of the United States is dominated by net radiation, which in turn is dominated by solar radiation, which is not temperature dependent. Other factors that influence ET could be affected by warming and other climate trends, in particular, vapor pressure deficit and longwave radiation, are temperature dependent and solar radiation depends on cloud cover. Terms related to vapor pressure deficit are also controlled by wind, and there are studies showing that near-surface wind speeds generally have been going down. The potential of changes in these factors to influence future ET are not well understood yet, making it difficult to make statements about future soil moisture with high or even medium confidence.

The magnitude of projected snowpack decreases in Key Finding 4 may be understated. The draft CSSR could reasonably use words like “substantial,” as virtually all projections show large decreases in snowpack by mid-century. This key finding could also be strengthened by framing this topic as a change in an annual pattern rather than an episodic change (which is how droughts are typically framed). Further, including a sentence about how runoff timing and volumes are expected to change and how this change is linked to natural storage in the snowpack would improve this key finding and could be stated with high confidence.

Findings concerning trends in flooding are highly complex and spatially variable, and this key finding could be improved by revising the text to specifically articulate this. Within the existing literature, few locations show statistically significant changes in flooding nor have they been clearly linked to precipitation or temperature. Generally, a mixture of downward trends and upward trends are observed (e.g., Lins and Slack, 1999, 2005), and when upward trends are observed, they have been shown for a relatively small proportion of measurement stations. Other factors, including land cover, have been found to contribute to observed patterns (Vogel et al., 2011). There is some evidence of upward trends in precipitation extremes, but essentially none in floods, and this remains an outstanding research issue.

Additional Chapter-Level Summary Recommendations

This chapter could include a finding focused on snowpack and associated seasonal runoff timing changes, especially across the West. While this is not new, it is well understood and has clear hydrologic consequences. As shown in Mote et al. (2016), the exceptionally low spring 2015 snowpacks were pervasive across the West. Such conditions may become the norm in future decades. This important contributor to water scarcity has not only been detected, but also attributed to human-caused climate change. Mention of this observation-based finding before discussion of related future projections would improve this chapter.

The discussion of the California drought and attribution would be more appropriately balanced by including additional literature and stronger recognition of the known complexities and outstanding research questions. Collectively, existing studies do not use a sufficiently consistent formulation to lay out a clear case for attribution and this should be stated (e.g., Swain et al., 2014; Wang and Schubert, 2014; and Funk et al., 2014). The California drought is also unusual, as observed in the exceptional warmth in the winters of 2013-2014 and 2014-2015, especially the latter. This raises the question, as yet unanswered, of whether droughts in the western United States are shifting from precipitation control (as shown by Mao et al., 2015) to temperature control. There is some evidence to support a relationship between mild winter and/or warm spring temperatures and drought occurrence (Mote et al., 2016). This is a topic that could be addressed more strongly, with a view to changes in the full hydrologic cycle, which receives little coverage in this chapter otherwise.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that this chapter needs to provide a more comprehensive overview of the state of understanding of hydrologic change as documented in the literature. Addressing the gaps detailed throughout this chapter review will considerably improve the impact of this chapter.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

Some of the key findings should be revised, as described in earlier comments on this chapter.

Are graphics clear, and do they appropriately reflect the major points in the text?

Concerns noted above and in the Chapter 7 review for figures also pertain here. More specifically, for Figure 8.1, the Committee recommends using a more accepted method of showing variations in soil moisture in multi-model settings. One such approach is to use soil moisture percentiles rather than the raw model output. This approach better recognizes that inter-model differences are large, which is difficult to capture in the current figure, where the range in change is small and generally within the range of variability among models.

For both Figures 8.1 and 8.2, why distinguish between “small compared to natural variations” and “inconclusive”? Recommend simplifying and using stippling only.

The Committee recommends replacing Figure 8.3 with an off-line land surface model run with bias corrected inputs, which will represent elevation effects much better and remove the

considerable GCM biases. Or, include other simulations, perhaps with hydrologic models, if available.

As part of the revisions recommended for this chapter, the Committee suggests identifying new figures that reflect the revised text.

Are likelihood / confidence statements appropriate, and justified?

See previous comments for recommendations to improve likelihood/confidence statements associated with the key findings.

Are statistical methods applied appropriately?

Throughout this chapter, greater statistical context, particularly that on historical trends and attribution, would strengthen the chapter.

Is the chapter balanced? Are there areas that should be expanded, or removed?

The chapter requires a more robust discussion of the hydrologic context in order to accurately represent the hydrology component named in the title.

Recommended changes to structure

As stated in the Summary comments, the Committee thinks that it would be more effective to use the Chapter 7 structure with historical trends first, then projections.

III.9 CHAPTER 9: SEVERE STORMS

Summary

The Committee commends the authors for producing a very strong draft chapter. Minor revisions are described in this section, but no major concerns about the chapter were raised.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that for the most part the chapter accurately reflects the scientific literature, with one important exception. References to “challenging the IPCC AR5 consensus” with regard to findings in changes in tropical cyclone (TC) intensity and frequency might be overly broad. It appears that only the findings on frequency are subject to a qualitative challenge, since the first such “challenge” (page 311, lines 1-3) seems to question only the magnitude but not the sign of the hypothesized relationship between warming and intensification of TCs.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

The Committee recommends preceding this with an appropriate statement describing observed trends in TC properties in the North Atlantic. Without this, the relatively low confidence in attribution might be confused as low confidence in detection. It is important to be clear about the difference. For example, IPCC AR5 2013 was very confident in the existence of a trend in TC activity on the North Atlantic.

The Committee suggests adding an appropriate statement about expected trends in overall number (frequency) of TCs. The chapter language on page 309, lines 20-22 is particularly effective, and could be included as part of this key finding: “Both theory and numerical modeling simulations (in general) indicate an increase in TC intensity in a warmer world, and the models generally show an increase in the number of very intense TCs.”

The Committee is concerned that confidence in observed tornado trends may be less than “high,” owing to, e.g., issues of shifting completeness of observational network. If high confidence is in fact warranted, the Committee suggests adding some supporting information in the traceable account. As written, it is unclear how the traceable account supports the key finding, as it appears to be internally inconsistent. Compare, for example, page 321, lines 24-25 (“virtually all studies”) with page 310, line 28 “medium confidence that [human factors] contributed….”

Are graphics clear, and do they appropriately reflect the major points in the text?

The figures are a weak point in an otherwise strong chapter and the Committee recommends significant revisions.

Figure 9.1 has limited relevance, as it pertains to the western north Pacific region. If an effective figure pertaining to the North Atlantic can be found, that might be more useful.

In Figure 9.2 the only results of any apparent statistical significance pertain to the western Pacific region and thus are of relatively limited interest for this U.S.-focused draft CSSR. The results for locations near the continental United States appear to show very small differences having no statistical significance. If this is wrong, the Committee recommends providing supporting information, for example 95% confidence limits, on the differences. It appears that those limits are

very broad, meaning that the range of possible trends is very large—probably so large as to not constrain things enough to be interesting.

In Figure 9.3 it is unclear whether the apparent trend seen in the red curve is statistically significant. If it is possible to provide information supporting its statistical significance, the Committee recommends doing so.

Figure 9.4 could be improved by removing some panels and enlarging others. The upper right panel is too small to read easily. This could be remedied to some extent by zooming in on the United States. The lower left panel could be deleted, as it appears to be simply a map of measured extreme precipitation events with an editorial comment about atmospheric rivers (ARs) and the same point is made more effectively in the bottom right panel.

Are likelihood / confidence statements appropriate, and justified?

Yes, the statements appear appropriate and justified.

Are statistical methods applied appropriately?

Yes, statistical methods appear to be applied appropriately.

Is the chapter balanced? Are there areas that should be expanded, or removed?

Yes, the chapter is balanced.

Recommended changes to structure

The authors should consider how and where different types of extreme precipitation and flooding are covered in the draft CSSR and ensure linkages across chapters. Chapter 9 mentions flood risk associated with ARs, but the chapter on flooding (Chapter 8) does not discuss risk associated with ARs. Chapter 9 also covers convective storms and ARs and it would be good to point out that Chapter 7 provides a complete discussion of variability in precipitation, irrespective of specific physical mechanism(s) driving that variability. Mechanisms of variability are important when it comes to improving understanding, but given the draft CSSR is intended to inform NCA4 description of impacts, it is important to quantify as well as possible variability on all time scales, irrespective of physical cause. Finally, it would also be beneficial to indicate that floods due to storm surge are covered in Chapter 12, Sea Level Rise.

Projections of ARs indicate greater frequency and intensity. Does this translate to increased precipitation in California? The text blurs some of the important differences between ARs in California, where they can increase snowpack, and ARs in the Northwest, where they almost invariably remove snowpack.

The box about the “hurricane drought” is good. Is there a corresponding discussion elsewhere about how this might affect preparedness?

III.10 CHAPTER 10: CHANGES IN LAND COVER AND TERRESTRIAL BIOGEOCHEMISTRY

Summary

This chapter covers a great deal of ground, and generally does a good job describing the state of science in many of its topics. Many of these areas have seen advances in science and

conceptual understanding since NCA3, and certainly since the IPCC AR5. However, the Committee found many parts of this chapter problematic, and provides a number of suggestions for improving it here. These overarching comments are ranked here in roughly descending order of importance.

• The Key Findings are often not supported by the description of evidence base provided. They also do not match well with the chapter text, and are even inconsistent with it at times.
• The chapter puts too much emphasis on growing season length and albedo, and consistently plays down the direct effects of temperature and precipitation in driving ecosystem responses to climate change. The Committee recommends significantly condensing Sections 10.2.4 and 10.3.1, while more prominently acknowledging temperature and precipitation effects throughout.
• Following the last point, drought and tree mortality should be given a more in-depth discussion, given the extensive recent research and findings in this area, and the fact that this is one of the chapter’s key findings.
• Throughout, the text is prone to vague and weak statements, sometimes with no clear connection to the information that the authors intend to convey; for example, page 337, lines 9-11, page 339, lines 13-14, page 344, lines 2-5, page 345 lines 1-3, page 346, lines 16-18. Text should be precise and clear. Structurally, paragraphs in this chapter frequently lack strong topic sentences and combine multiple topics, often in a confusing way.
• The chapter title does not match the chapter’s content, as land use/land cover change is really only mentioned in the introduction and on page 342.
• None of the chapter’s key findings appear in the ES.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that the chapter generally accurately reflects the scientific literature in specific areas, and that no critical content areas are missing from the draft report. However, the discussion of some topics in the report should be expanded, while the emphasis on others should be reduced (see Summary), and better linkages to the key findings are needed.

Are the key findings presented clearly, and documented in a consistent, transparent and credible way?

The description of evidence base provided for Key Finding 1 seems to be referring to albedo effects only. Since the finding is about both albedo and GHG effects, the evidence should also address both. Note that this type of concern is a recurring one throughout this chapter.

Also, Key Finding 1 and Figure 10.2 seem somewhat inconsistent with the information in Figures 2.3, 2.6, and 2.7. In particular, no reason is given for starting in 1850 instead of 1750, the reader’s attention is not directed to Chapter 2, and the method of partitioning each contribution into LULCC and non-LULCC is not stated. The partition shown for CO₂ is plausible given Figure 2.7, but only if the enhanced land carbon sink is ignored. Agriculture is a major source of N₂O but this is true even if land use were not changing to increased arable land, so it seems a stretch to ascribe all N₂O emissions to LULCC. The discussion of nitrogen on page 341 does not address N₂O emissions or their relationship to LULCC. In summary, basing Key Finding 1 on one study (page 342, line 19) should constitute low confidence.

There is a major mismatch between this key finding, which is about the past, and the description of the evidence base, which is about the future. For this reason, the description of the evidence base is incomplete and a more thorough description of the data, evidence, and relevant studies should be included. In addition, there is strong evidence for impacts of drought on plant community structure, but the evidence for “increased occurrence and severity of drought” is not presented and not clearly supportable. Note also that, as described in Section III.8, it is far from clear that there is really an “increased occurrence of drought.” Additionally, the tone of Key Finding 2 is essentially opposite that of Key Finding 1. Key Finding 1 says the land is a net carbon sink and Key Finding 2 says drought is having an impact. Both can be right, but the juxtaposition requires explanation. Finally, this key finding could be better linked to the more extensive treatment of drought in Chapter 8.

This key finding is mostly about climate variables (length of the frost-free season) while the evidence is about ecosystem responses. One cannot conclude that the evidence supports the finding.

This key finding includes a very thin description of the evidence base that does not really support the assertions made in the finding. It should be expanded and clarified, or the key finding should be deleted.

Are graphics clear, and do they appropriately reflect the major points in the text?

The graphics are generally clear, but Figures 10.1 and 10.2 are probably not both necessary, and one might be replaced by a table. Discussion of figures in the text, particularly Figure 10.2, is too brief and should be expanded on for the figure to provide value to the chapter. Regardless, figures require better explanation in their captions. For Figure 10.1, “LULCC” should be defined and captions in general need to be clearer and more informative. The Figure 10.2 caption should refer to Figure 2.3 since the Myhre et al. (2013) forcings are shown, and include more detail.

Are likelihood / confidence statements appropriate, and justified?

Yes, the statements are appropriate and justified.

Are statistical methods applied appropriately?

The Committee is concerned about ad hoc time period choices and unqualified assertions of trends. This is addressed in general comments about the entire draft report (see Section II.2).

Is the chapter balanced? Are there areas that should be expanded, or removed?

The chapter could be better balanced, as detailed earlier in Section III.10.

III.11 CHAPTER 11: ARCTIC CHANGES AND THEIR EFFECTS ON ALASKA AND THE REST OF THE UNITED STATES

Summary

Some of the global consequences of climate change in the Arctic are potentially catastrophic and irreversible. There may also be physical thresholds beyond which these consequences become inevitable (even if they might unfold over centuries). For these reasons, this topic has both importance and policy urgency, and a thorough treatment in the draft CSSR is important. The third-order draft of this chapter is a sound foundation, and the Committee encourages the authors to consider the following points as they revise the chapter.

Are the key findings presented clearly, and documented in a consistent, transparent and credible way?

This key finding needs to be supported by stronger evidence than is currently provided on page 371, lines 26-35. As written, it contains illogical and confusing reasoning and contradictory conclusions. Are the satellite observations of the middle troposphere or the surface temperature? If longer records indicate that decadal variability dominates, why base Key Finding 1 on a study of temperature change since 1981? A strong topic sentence that summarizes the main message would help, instead of starting with “Satellite observations”. Additional detail should also be provided in the traceable account.

This key finding discusses sea ice projections, but the description of evidence base mentions only observations. Presumably the projections are based in some way on CMIP5 simulations, but specific literature should be cited.

The confidence level associated with Key Finding 4 seems low and the Committee recommends evaluating whether a higher confidence level might be appropriate. In either case, a more transparent reasoning process should be laid out in the traceable account. Recent work by Kirchmeier-Young et al. (2016) may also be relevant to cite here.

There is universal recognition that Arctic influence on mid-latitude weather is an area of active research (as pointed out in the draft CSSR) and the Committee supports including some discussion of this topic in the chapter, with citation of the full breadth of current research perspectives on this linkage. However, because no scientific consensus on this topic has been reached, the Committee strongly recommends removing Key Finding 5, so as to not place disproportionally high emphasis on a topic where there is currently little confidence.

Introduction: The second to last paragraph of the introduction mentions unique challenges associated with improving understanding of the Arctic. This is appropriate, but the Committee is concerned that this might leave the reader with the impression that we do not know enough to usefully inform policy, which is not the case. The final paragraph in the Introduction (page 371, lines 14-15) would be strengthened by stating not only that our understanding is improving, but also that it is advanced enough at present to effectively inform policy. It may also be worthwhile to explicitly state that Alaska is in the Arctic, making the United States an Arctic nation (the first sentence implies that it is not).

Permafrost: GHG emissions from thawing permafrost are an important mechanism by which the Arctic affects the rest of the planet. With this in mind, the key finding on GHG emissions from thawing permafrost should be stronger. Saying only that “The overall magnitude of the permafrost-carbon feedback is uncertain” is, while strictly true, not helpful. While the Committee recognizes that these emissions are quite uncertain, it is clear that these emissions have the potential to complicate our ability to meet policy goals like limiting warming to 2^oC, as is a target in the Paris Agreement. This should be stated. Further, the discussion of permafrost should be separated from discussions of snow cover and methane hydrates, with the entire discussion of permafrost provided in one contiguous section. The report emphasizes methane release from permafrost, which may not be appropriate. While permafrost is a source of methane, the text should explicitly note that at present, research indicates that more carbon is released from permafrost as CO₂ than as methane. Finally, it is important to be sure that GHG emissions from thawing permafrost are considered consistently throughout the draft CSSR. In particular, the discussion of remaining allowable emissions consistent with meeting the 2^oC goal (ES, page 27, lines 17-24) appears not to consider these emissions. The Committee considered this to be an important oversight. Similarly, the discussion of permafrost in Chapters 1 and 15 should be revisited in light of the above comments to ensure consistency.

Greenland Ice Sheet: Discussion of Greenland Mass Balance here overlaps substantially with Chapter 12, which provides a more thorough overview. The Committee recommend trimming this passage and referring to the equivalent in Chapter 12. The discussion that is provided in Chapter 11 is overly focused on recent observed trends in ice sheet mass loss. While this is an important topic, there should also be a short discussion of future trends. For example, is there a threshold beyond which eventual complete melting becomes inevitable? Do we know where this threshold is? (e.g. see Robinson et al., 2012). If so, how long would complete disintegration take? The implications of Greenland ice sheet mass loss for SLR and potential impacts on ocean circulation should be mentioned and linked to more detailed discussion elsewhere in the report, including that of model sensitivity in Chapter 15 and as previously noted, Chapter 12.

Sea ice extent: The projection is made that the Arctic will become “ice free (in summer) by mid-century.” It is further stated that “natural variability…future emissions, and model uncertainties … all influence sea ice projections.” This last statement is indisputable, but it would be helpful if something more specific could be said about the importance of future emissions on the fate of summer sea ice. In other words, how much control do we have (in principle) over whether and when summer Arctic sea ice disappears?

Arctic connections to mid-latitude weather: This is characterized in Chapter 9 as low confidence and low to medium confidence in Chapter 11. Regardless of which of these is most appropriate, the draft CSSR should be internally consistent. See the more detailed recommendation on this topic provided with Key Finding 5 earlier in this section.

III.12 CHAPTER 12: SEA LEVEL RISE

Summary

This is a strong chapter. It is well written, uses graphics effectively, and provides an excellent, comprehensive overview of the individual factors contributing to SLR, with particular emphasis on its spatial heterogeneity. The chapter represents a substantial departure from previous assessments of SLR (including the NCA3), and represents a substantial advance relative to previous

U.S. sea level assessments. Another particular strength of the chapter is the outlook beyond the year 2100. The Committee thinks that the potential rates of global mean sea level (GMSL) rise in the next century should also be discussed, because they are in the > cm/yr range, which poses particular challenges for coastal infrastructure, etc.

Notable changes relative to previous work include the revision of future GMSL scenarios, in line with the recent findings of the U.S. Interagency Sea Level Task Force (Sweet et al., 2017). The new scenarios now consider six discrete GMSL trajectories, in comparison with four used previously. In a further departure from previous assessments, the new, individual sea level scenarios are placed in context with published probabilistic projections of future sea level following standard RCP emissions scenarios (e.g., Kopp et al., 2014, 2016). The chapter considers and contextualizes the latest results from ice-sheet modeling that includes physical processes not previously considered at the ice-sheet scale. The chapter also breaks new ground relative to previous reports by providing some regional guidance on the expected departure of future relative SLR around the North American coastline, relative to GMSL estimates. This regionalized analysis also includes guidance on evolving recurrence probabilities of high water (flood events), which is particularly useful.

No fundamental deficiencies were found, however the Committee did raise several issues that should be addressed to improve the presentation of the material and overall clarity of the chapter, or highlighted and given even more emphasis.

• The Committee recommends considering the advantages of reducing the dates/time intervals in use (perhaps focusing on 1900 to 2000, and 1993 to present), if possible, for greater consistency. This would serve to simplify comparisons of past SLR with the tabulated future sea-level estimates expressed relative to the year 2000. The need for consistent, simple time intervals applies to the entire chapter, including Section 12.4.2, which mixes discussions of post-1970s and post-1900 eras.
• The elevated sea level (6m to 9.3m) during the Last Interglacial provides a powerful message that the polar ice sheets are sensitive to modest warming. Adding some discussion that sea level was likely even higher during previous interglacials, including MIS-11 (~400ka), when GMSL might have been 6-13m higher than today (Raymo and Mitrovica, 2012), and likely even higher still during the Pliocene (~3 million years ago; Rovere et al., 2014) could be considered. The Committee strongly recommends moving Figure 4.3 to Chapter 12, where it would be more effective and illustrative of GMSL sensitivity to past warming. Removing the CO₂ values from Figure 4.3, to avoid complications associated with the influence of orbital versus GHG radiative forcing during these past time periods is also recommended.
• While the accelerating rate of GMSL rise since the late 20^th century is described in the chapter, it is an important statement that could be emphasized further. This also applies to the notion that loss of land ice is overtaking thermosteric effects as the primary contributor.
• The chapter does a nice job of illustrating the radically different regional responses (fingerprints) to Greenland vs Antarctic ice-sheet retreat (Figure 12.1). However, the simple notion that North America faces greater risk from ice loss in Antarctica than from ice loss in Greenland is not as simply and clearly stated as it could be. This point should be emphasized, because it relates directly to the subsequent discussion on the potential for drastic Antarctic ice loss.
• The Committee noted that the impacts of changes in land-water storage (past and projected) are not sufficiently covered, although the Committee acknowledges that the land-water storage component is relatively modest and is considered in the likely ranges of SLR based on Kopp et al. (2014). Some additional discussion on this topic could be helpful.

• Some of the “emerging science” described in the chapter (DeConto and Pollard, 2016, Golledge et al., 2015) shows that the loss of marine-based ice (in West Antarctic for example) is a long term (millennial timescale) commitment, due to the slow thermal response (cooling of the ocean). The effective “permanent” loss of marine-based ice would obviously have lasting/irreversible impacts on U.S. coastlines and should be mentioned.
• The spatial pattern of recent and ongoing thermosteric SLR (indirectly illustrated in Figure 12.2) is somewhat marginalized. While the potential for future impacts caused by thermal expansion is smaller than from ice sheet loss, the thermosteric effects are already impacting locations in the western Pacific with U.S. economic, strategic, and humanitarian interests; and will continue to do so regardless of ice sheet loss. A similar point can be made for the ocean dynamical effects on regional sea level, which seems underemphasized, relative to the potential impacts they could have along the U.S. East Coast.
• The Committee appreciates the cautious treatment of new ice sheet modeling that implies the potential for much higher SLR in coming decades and centuries than previously reported (e.g., DeConto and Pollard, 2016). While it is important for the draft CSSR to consider the full range of physically plausible SLR, this discussion could be balanced by also mentioning alternative modeling (e.g., Ritz et al., 2015) that implies more modest future SLR. While Ritz et al. (2015) do not directly account for the glaciological mechanisms considered by DeConto and Pollard (Marine Ice Sheet and Marine Ice Cliff Instabilities), their work does provide an alternative view of Antarctica’s potential contribution to future SLR that should also be mentioned for completeness.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that the chapter accurately reflects the current scientific literature on this topic, although the discussion of drastic Antarctic ice-sheet retreat could be broadened by comparing the recent results of DeConto and Pollard (2016) with Ritz et al. (2015) as already noted. Discussion of ocean heat content and influence on SLR should also be provided and appropriate, up-to-date references added.

In addition to the comments made previously, the Committee recommends an expanded discussion regarding the onset of anthropogenic influences on SLR, and further recommends that the authors consider enhancing their graphics to illustrate the anthropogenic contributions to past (and future) GMSL rise, and perhaps a breakdown of the relative contributions to GMSL from the individual processes and sources described in the report. This would provide an important update to Figure 13.1 in IPCC AR5 (Church et al., 2013).

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

The Committee compliments the overall clarity of the chapter, and the well-written background content on what causes global and relative SLR. The chapter could be improved by consistent treatment of time scales wherever possible.

The Committee recommends the use of consistent time intervals in their discussion of past SLR, particularly avoiding mixing discussion of post 1880 and post 1900 GMSL in the same paragraph, if possible. In that case, the first sentence might read something like “Global mean sea level (GMSL) has risen by about 7.5 inches (about 19 cm) since 1900….” Some further discussion/clarification would be helpful, as to when in the 20^th century the anthropogenic influence on GMSL began. The traceable accounts reflect the current state-of-the science, and confidence levels are appropriate.

The Committee thinks it is important to state that very high (> 2.4m) SLR by 2100 is physically possible, but the language “cannot be ruled out” is vague, open to interpretation, and does not provide useful guidance. The traceable accounts reflect the current state-of-the art, and confidence levels are appropriate.

This key finding lists several locally important processes in a way that diverts attention from the fact that if future ice loss is dominated by Antarctica (vs Greenland), much of the U.S. coastline will experience considerably more relative SLR than the global average. The traceable accounts reflect the current state-of-the art, and confidence levels are appropriate.

Given the importance of long-duration winter storms on East Coast flooding in particular, the Committee recommends considering whether this key finding should be extended to include a comment on extratropical cyclones, in addition to Hurricanes. The traceable accounts reflect the current state of science, and confidence levels are appropriate.

Are graphics clear, and do they appropriately reflect the major points in the text?

The figures are generally clear and appropriately reflect the key points, although some specific recommendations are noted here.

As noted previously, the Committee recommends that Figure 4.3 (with the CO₂ values removed) be moved to Chapter 12, where it would be more effective at illustrating the potential sensitivity of the polar ice sheets to warming.

Figure ES.8 does not appear in Chapter 12 even though it shows SLR data. The Committee suggests that the figure be moved to Chapter 12, perhaps with a single representative city left as a figure in the ES, and discussed appropriately. Removal of the U.S. basemap would allow the individual time series to be expanded. At present, the axes on the individual panels are so small they are almost illegible. Furthermore, the y-axes should stop at 365, to reinforce that they are ‘days per year’ which is why the annual occurrences of daily flooding saturate near the end of the time-series, and a note added to the caption that this limit results in many of the curves having an inflection point. The choice of colors (blue and teal) could also be reconsidered for added clarity.

For Figure 12.2, panel labels ‘a,’ ‘b,’ and ‘c’ are missing, although they are mentioned in the caption. The Discussion of Figure 12.2c in the text could also better match the time period shown in the figure.

Figure 12.3 mixes meters and feet and should be edited to be consistent in the use of units.

Are likelihood / confidence statements appropriate, and justified?

Likelihood and confidence statements are generally appropriate and justified. The Committee also notes the importance of considering new science hinting at the potential for much higher future sea level than previously reported, but agrees that the confidence in this finding is still low and requires ongoing research. That said, the Committee questions the wording that GMSL > 2.4m by 2100 “cannot be ruled out” as this is too open to interpretation and could be misconstrued as ‘barely’ possible, for example. Given the importance of this issue, this wording should be reconsidered.

Are statistical methods applied appropriately?

The Committee commends the blending of discrete sea level scenarios with a probabilistic approach and has no recommendations regarding the statistical methods used.

Recommended changes to structure

The chapter is well balanced, although the introduction is thinner than other chapters and some modest rewrite might be considered. No specific edits are recommended.

III.13 CHAPTER 13: OCEAN CHANGES: WARMING, STRATIFICATION, CIRCULATION, ACIDIFICATION, AND DEOXYGENATION

Summary

The ocean has received increasing attention in climate assessment reports. Following the lead of IPCC AR5WG1, the draft CSSR treats SLR and other ocean changes in separate chapters. As the title of this chapter suggests, there are many aspects of ocean changes that are important, both for their impacts on the ocean and its ecosystems, and also for impacts beyond the ocean. The Committee thinks that more effort could be devoted to linking this chapter to broader climate system changes. In particular, the role of the oceans in storing heat, and the link between changes in ocean heat content and changes in sea surface temperature could be discussed. In addition, the

importance of ocean/atmosphere coupling in ENSO, mid-latitude storm tracks, and the thermohaline circulation, could also be better reflected in the text, and in turn the consequences of changes in ENSO for the United States and its territories (augmenting Chapter 5) could be emphasized, as is done in the Chapter 11 for the Arctic.

The chapter as a whole, including the key findings, was awkward to read even for those with knowledge of oceanography. Too many discipline-specific words or phrases are used with insufficient explanation. For example, the expression “ocean acidity” is used on a number of occasions, without any explanation of what it means. A number of words or phrases are explicitly noted in the Line Comments (Appendix A). Furthermore, it is rarely made explicit that any numerical value ascribed to a change in this parameter almost always refers to the surface ocean. The Committee recommends that this chapter be revised so as to improve consistency across the draft CSSR.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee felt that the chapter generally accurately reflects the scientific literature with two exceptions. First, Key Finding 1 represents an incomplete view of the evidence about changes in the Atlantic Meridional Overturning Circulation (AMOC). Second, the Committee suggests also noting the importance of changes in ocean properties (such as warming) for Antarctic ice sheet instability and SLR.

Are the key findings presented clearly, and documented in a consistent, transparent, and credible way?

The evidence for changes in the AMOC mentioned briefly in the traceable accounts, appears to rely on a single study, and contains no quantitative statements to put the changes into context. Other studies reach different conclusions (e.g., Rhein et al., 2013), assessing the then-available literature, stated that “there is no evidence for a long-term trend.” A fuller treatment of the issue is warranted, especially since it appears in the ES. This should include reference to more of the literature on this topic, including studies that emphasize the variability and challenges in assigning causes of AMOC trends. Moreover, if the 2 Sverdrup number is to be mentioned, it should be put into context with the total AMOC (e.g.,”may have slowed on the order of 10%”).

Key Finding 1 contains many topics and should be split into multiple findings. The last statement that upwelling along the U.S. West Coast has intensified is difficult to reconcile with other statements on page 454 (lines 2 and 10), indicating a more mixed picture both in the past and for the future, especially given the apparent attribution statement. Hence, the level of confidence assigned to this finding seems too high.

This key finding is not well grounded in the text. No specific details of the projected AMOC decline could be found in the chapter. The key finding also generalizes the projected average ocean acidity while the associated text focuses largely on the regional variability and the percent changes in acidity are not clearly traceable to the text. Little discussion is provided for the projections of the AMOC decline, with the text referring in places to the uncertainty of projections using earth system models. There appears to be an error in estimating the change in global ocean acidity between preindustrial time and 2100 that would result from scenario RCP8.5. Also, the unit “Sverdrup” appears nowhere in the document outside of this key finding and may not be appropriate for the intended reader and should be omitted. Finally, the traceable accounts should provide a more detailed summary of the information contained in the references provided.

Are graphics clear, and do they appropriately reflect the major points in the text?

The graphics are clear, but all three figures relate to changes in ocean chemistry, which constitute only two of the five topics named in the title. The legends generally lack some information needed to interpret the figures.

In Figure 13.1, there seems an excess of detail, although the legend still does not unambiguously describe the plots (i.e., it is unclear whether the green values refer to carbonate ion concentrations, or whether the x(CO₂) values are “wet” or “dry”). A citation for CO2SYS v2.1 should also be included.

The Figure 13.2 caption should be clear in stating that this is a change in surface ocean pH that has been estimated.

In the Figure 13.3 caption, the modeled density surface depicted should be included, as it is a key piece of information. This caption should also state that the data is on a particular density surface (26.5), as it was presented in the Long et al. (2016) source.

Are likelihood / confidence statements appropriate, and justified?

Yes, statements are appropriate and justified.

Are statistical methods applied appropriately?

Yes, statistical methods are appropriate.

Is the chapter balanced? Are there areas that should be expanded, or removed?

Mostly, however as noted above, the discussion of the AMOC is limited in scope and ocean heat content should be discussed. Also, the discussion of ocean acidification is somewhat confusing to those not in the field and would be improved by clarifying the various terms used. In particular, clarification that “acidity” is being used (apparently) as a synonym for hydrogen ion concentration, and “acidification” as an increase in that concentration is needed. Presumably the

use of the term “corrosive” is not (as most might think) referring to a chemical damage to a metal, but rather implies the potential for dissolution of aragonite (the more soluble form of biogenic calcium carbonate)? Reference is also made to concepts such as “sensitivity to ocean acidification” or “buffering capacity” without explicitly stating what these terms mean. Additionally, as the various chemical mechanisms for such changes are not clearly described, the distinction between open ocean and coastal acidification is hard to follow. Perhaps a box describing these mechanisms could be added to help with this.

III.14 CHAPTER 14: PERSPECTIVES ON CLIMATE CHANGE MITIGATION

Summary

This chapter provides a concise overview of the key concepts that frame the challenge of limiting damage from climate change through a combination of mitigation and adaptation, and is a readable account of the implications of the Paris Agreement. The framing is mostly based on a “reluctant participant” model, where progress with mitigation stops when pre-determined commitments are reached. In contrast, the presentations from the United Nations Framework Convention on Climate Change (UNFCCC) more typically present decarbonization as a process with emissions reductions that become increasingly ambitious through time, as technologies improve and nations work through the experience of institutionalizing low-carbon societies. For the purposes of understanding mitigation pathways, a notable omission (e.g., in Figure 14.1) is a Paris-compliant scenario, i.e., one that has a > 50% chance of stabilizing warming at less than 2°C.

The chapter’s key findings largely miss the opportunity to make what could be the chapter’s central point: a consequence of the essentially permanent nature of warming from CO₂ is that stabilization of CO₂ at any given concentration can only be achieved if CO₂ emissions fall to zero or become negative, to compensate for the remaining emissions of other GHGs and land-use change. Stabilizing warming at 1.5°C or 2°C requires emissions to fall to zero within a few decades, and even stabilizing warming at 3°C or 4°C requires zero emissions a few decades after that.

The absence of a focus on the need to drive CO₂ emissions to zero means that the chapter is not as clear as it might be on the range of emissions trajectories consistent with any temperature goal, illustrated in Figure 14.3. Specifically, it is important to emphasize the point that, for any mitigation goal, slower action in the near term requires more aggressive reductions or larger negative emissions later in the century. The linear relationship between cumulative emissions and warming creates the clearest entry point for understanding possible futures and especially for appreciating the motivation for reducing emissions to zero.

The chapter also misses an opportunity to add value by framing the mitigation challenge as one of managing risk, which has two dimensions. One is the risk of impacts at any level of warming. Here, links to Chapters 6-9 and 15 would be helpful. The other is the probability that a given emissions trajectory holds warming below a given goal. For the first dimension, the opportunity is largely in laying out the issues. This chapter, indeed this draft report, is appropriately focused on setting the stage for a thoughtful presentation of impacts. Still, the discussion can be more informative with a deeply grounded discussion of risk. The second dimension is central to the theme of Chapter 14. Without a clear presentation of the probabilities of reaching climate goals, the presentation of the emissions numbers has limited value. While it is not the responsibility of this draft report to define a “right” probability of meeting a goal, it is important to frame the discussion in a balanced way.

One of the biggest challenges in framing discussions of mitigation is striking a useful balance between discussion of CO₂ and other climate altering substances. The overall sense of the

Committee is that the chapter puts less emphasis on short-lived climate pollutants and other long and short-lived GHGs than the topic deserves. The discussion of climate intervention is valuable, though it would be more useful with a careful discussion of the limited knowledge base concerning climate intervention, especially solar radiation management.

Another challenge is that natural scientists tend to use carbon “C” but economists and policy experts tend to use “CO₂”. From a natural science perspective, “C” is the more natural quantity to discuss, for several reasons. But the solutions, from discussions of carbon pricing to allowable budgets, are almost universally discussed in units of “CO₂”. The Committee thinks that this chapter (and the whole report) would be clearer and more useful with all of the quantities presented in units of “CO₂” where it is appropriate to do so.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

Most of the specific recommendations for this chapter are framed through discussion of the key findings. The key findings of Chapter 14 are all fundamentally consistent with the scientific literature, but they could be structured to more accurately capture the relative importance of several key concepts. In particular, none of the key findings emphasizes the point that stabilizing warming, independent of the target, requires that emissions of CO₂ and other long-lived GHGs fall eventually to zero. Further, none makes the point that the difference in the emissions trajectories that lead to stabilization at levels ranging from 1.5°C to 4°C turns out to be only several decades in the future for reaching zero CO₂ emissions.

Key Finding 1 presents the relationship between CO₂ and warming in a confusing way. A casual reading of the finding would be that decreases in CO₂ concentration resulting from natural partitioning into land and ocean sinks might lead to cooling and that there are important time lags between emissions and impacts on warming (or emissions reductions and impacts on cooling). Both parts of this are misleading. Many papers (see especially Matthews and Caldeira, 2008 and Solomon et al., 2009) show that warming from CO₂ is essentially permanent due in part to the long lifetime of CO₂ in the atmosphere and in part to the decreasing heat transfer to the oceans as they gradually warm. Matthews and Solomon (2013) make the important point that, if emissions stop, additional warming stops shortly thereafter. It is not really useful to discuss the lag between emissions reductions and concentration reductions because the CO₂ problem is essentially one of cumulative emissions, such that delaying action in the near term makes it more difficult to solve the problem in the longer term.

This finding is, in some sense, based on a logical inconsistency. RCPs 4.5 and 8.5 are constructed around the idea that there is not a goal of limiting warming to 2°C, which makes them intrinsically incompatible and challenging to discuss in a single context. Also, the stated “cumulative emissions would likely have to stay below 1,000 gigatons carbon (GtC)” is given without a citation and is inconsistent with the 790Gt C cited in IPCC AR5 2013. According to IPCC, cumulative CO₂ emissions through 2016 are about 555 GtC, leaving a remaining allowance of 235 (not 400) GtC.

Additionally, it is important to include the probability of reaching the target and to be clear on the assumptions about other GHGs and aerosols, and on the implications of those assumptions.

This finding would be clearer with an explicit acknowledgement of the link between climate stabilization and zero CO₂ emissions. Presenting the concepts in terms of emissions reductions after 2030 misses that key point. Key Finding 3 (and Figure 14.1) are both grounded in a specific conceptual model of what it means to comply with the Paris Agreement. In particular, the idea that “Continued ambition” should be read as emissions staying at 2030 levels is only one of many different possibilities. It is also possible (and more consistent with the way the Agreement has been framed by leaders in the UNFCCC) to interpret “continued ambition” as sustaining rates of decarbonization, rather than emissions levels. With this framing, “continued ambition” leads to decreasing global emissions, and “increased ambition” leads to more rapid emissions decreases. Additionally, this finding misses a central element in the UNFCCC narrative about the Paris Agreement, notably its role in building a “culture” of emissions reductions. Almost all of the analysis makes strongly value-laden assumptions about the way that initial emissions reductions influence prospects for future emissions. Without weighing in on which assumptions might be correct, it is important to note their influence on the assessment of the challenges associated with reaching any goal.

This key finding is currently written as a prediction about future policy emphases and the statement about “may gain attention” feels like a commentary on potential political dynamics. It would be clearer and more useful if presented as saying something about the state of knowledge about climate intervention. In particular, the statement could make it clear that at present, there is not sufficient knowledge to support a mature judgment about benefits and risks of possible use of intervention approaches, and some of these approaches could have unintended consequences and would not address all negative impacts of climate change (e.g., solar radiation management

does not lessen ocean acidification).With this in mind, the key finding could be reshaped to state that geoengineering solutions require additional research and there are preliminary indications that geoengineering could limit some, but not all, aspects of climate change. Finally, a National Academies committee tasked with evaluating climate intervention techniques developed separate reports on CO₂ removal/sequestration and albedo modification (NRC 2015a and 2015b), noting that the large differences in research needs and social risks warranted independent treatment. A similar distinction between climate intervention approaches could also be considered here.

Are graphics clear, and do they appropriately reflect the major points in the text?

Figure 14.1 presents several possible trajectories for future emissions, but it does not present any with a greater than 50% chance of stabilizing warming at no more than 2°C. Given the chapter’s emphasis on ambitious mitigation, there would be real value showing at least one trajectory with a greater than 50% chance of stabilizing below 2°C and one with a greater than 50% chance of stabilizing below 1.5°C. Relevant scenarios are shown in Figure 14.3.

Are likelihood / confidence statements appropriate, and justified?

The confidence statement on Key Finding 4 is difficult to interpret, based on the wording of the finding. As written, the draft report appears to assess confidence in the prediction that climate intervention will get increased attention and on the value for policy makers of increased attention. Presumably, the confidence should be associated with an assessment of the potential for climate intervention to contribute solutions or to the maturity of current knowledge.

Are statistical methods applied appropriately?

Yes, statistical methods applied are appropriate.

Is the chapter balanced? Are there areas that should be expanded, or removed?

There is no simple way to provide a comprehensive overview of the prospects for and challenges of mitigation in a few brief pages. Still, this chapter could set the stage more effectively with a clearer focus on the full range of possible future trajectories and on the critical issue of the probability of meeting any climate goal.

The treatment of aerosols and GHGs other than CO₂ could be stronger. The treatment of climate intervention would be clearer with an increased emphasis on the fact that climate intervention strategies are much less well known than climate change and that a reasonable foundation for decisions will require a big expansion of technology development as well as knowledge, especially in the area of governance and political dimensions.

III.15 CHAPTER 15: POTENTIAL SURPRISES: COMPOUND EXTREMES AND TIPPING ELEMENTS

Summary

The Committee found this chapter to be a welcome addition to the discussion of climate science and recommends it be expanded. It is the first time in a synthesis document of climate science that this topic has been addressed in a stand-alone chapter. The importance of recognizing compound extremes and tipping points (or thresholds) is fundamentally based in the inherent properties of complex systems and in the science of extremes in risk characterization. The chapter

covers the limits of risk quantification and two broad categories of low probability-high impact events (compound extremes and tipping points). The Committee has some suggestions for improvement of the chapter.

A more thorough introduction for this topic is warranted. One suggestion is to better frame the chapter in the context of climate change as a complex system of interacting components. Prediction is difficult based on knowledge of the components of the system alone, the history of the system matters, emergent features appear that are not necessarily observed in the individual pieces, and feedbacks make simple cause and effect rare.

• The chapter could be strengthened if revised to move in the direction of more emphasis on lower probability but high consequence outcomes emphasizing compound extremes and tipping points, e.g., methane hydrates influenced by ocean warming and pressure.
• Because surprises are unknown unknowns, it is suggested that “Potential Surprises” be removed from the title, or changed to “Potential for Larger Changes.”
• There is no mention of negative feedbacks that could potentially offset positive feedbacks. The Committee recommends including this for balance.
• It would be valuable to mention a few examples of some past surprises (e.g., ozone hole, rate of Arctic sea ice loss) and discuss when scientists have been surprised and the factors that contributed to that surprise.
• The chapter could be strengthened by illustrating how gradual climate change can lead to tipping points in built as well as natural ecosystems (see NRC, 2013).
• The chapter could include a more thorough discussion of characterizing risk (see NASEM, 2016b).
• Chapter 15 would benefit from the inclusion of known unknowns in the science, such as changing natural variability in a warming world, ocean-ice dynamics including potential impact of ice-sheet melt on ocean circulation, changing ocean ecosystems, and their interaction with the physical ocean environment, stratosphere-troposphere exchange.

Specific Review Comments Related to the Statement of Task

Does the chapter accurately reflect the scientific literature? Are there any critical content areas missing from the chapter?

The Committee thinks that the chapter could be updated with some more recent references, e.g., Clark et al., 2016; Liu et al., 2017; Drijfhout et al., 2012; and Koenig et al., 2014.

There is also the soil C “bomb” hypothesis, whereby metabolic/microbial activity adds heat to thawing soils resulting in a runaway carbon release. (e.g., Hollesen et al., 2015).

Are the key findings presented clearly, and documented in a consistent, transparent and credible way?

Key findings are generally presented clearly and appropriately.

Without including negative feedbacks, the confidence of Key Finding 1 may be overstated. The Committee recommends acknowledging this and considering whether the confidence level is appropriate.

Key Finding 3 also includes only positive feedbacks and certainly the models do not incorporate all processes that contribute to both positive and negative feedbacks. There is no obvious summary of “what is and is not included in the latest generation of CMIP5 models” in Chapter 9, as the description of evidence base suggests. Moreover, it might be helpful to mention the failure of climate models to simulate importantly different past climates like the Paleocene-Eocene Thermal Maximum.

Are graphics clear, and do they appropriately reflect the major points in the text?

Table 15.1 lists some potential tipping elements. Some of the terminology in the table is vague and could be made more explicit. For example, there is frequent use of the term “collapse” to describe a state shift (AMOC, ice-sheet retreat, sea ice retreat) when it would be more valuable to define the state shifts more explicitly. North Atlantic Convection could refer to the ocean, atmosphere, or both and should be clarified. Also, consider adding “freshwater forcing on ocean circulation” as a main impact pathway for Greenland and Antarctic ice-sheet retreat. Finally, ecosystem services are listed as a main impact pathway: the subject might better be left for NCA4, but if retained, “ecosystem services” is a rather broad term that could be made more explicit.

Figure 15.1 (left panel) should include potential climate tipping points for the entire globe, not just the Americas. In particular, the instability of marine-based ice in deep East Antarctic basins represents a large and scary unknown (Pollard et al., 2015; DeConto and Pollard, 2016; Aitken et al., 2016; Mengel and Levermann, 2014; etc.). The bubble in the figure should be re-labeled to read “instability of marine-based Antarctic ice,” rather than implying that just West Antarctica is vulnerable. Figure 15.1 (right panel) seems to be rather obvious (high-impact wildfire and drought events occur under hot, dry conditions) and could be deleted. This figure is also not referenced in the text and should be.

Are likelihood / confidence statements appropriate, and justified?

See comments for key findings.