The climate change effects determined to affect highway systems and discussed in this report include changes in temperature and heat events, changes in precipitation and storm activity, and sea-level rise. These variables are identified as potentially affecting highways by a number of reports (NRC 2008; USGCRP 2009; CSIRO 2006; CIG 2007). Changes in these climate variables may directly affect existing stressors or may introduce new stresses on the highway system. For example, an increased number of heavy precipitation events or the 1-in-100 year storm event (i.e., a storm event with a 1% annual likelihood of occurring) may lead to flooding that cannot be handled by existing culverts and other components of drainage systems. Temperature increases may affect regional highway operations, affecting costs associated with snow and ice removal, as well as the change in environmental impacts associated with salt and chemical use (NRC 2008). Changes in precipitation may disrupt highway travel, construction activities, and compromise bridge integrity (see Section 5 for detailed discussion of climate impacts associated with these climate change effects). Additional variables, such as relative humidity and changes in solar radiation, are also identified but not considered in this report. Literature providing regional information for these variables is currently lacking, but these variables should be included as future research allows.
The compilation of regional projections began with a literature search using relevant, carefully composed search terms across relevant databases of publications on the environment, energy, technology, and government. The search was refined to include articles, government reports, and peer-reviewed publications with a published date post-2003 and available by June 2009. Effort was made to include seminal reports that became available after this date. Since this report focuses on regional effects, studies that are broader than the region-scale have been excluded. Sea-level rise studies are an exception, as much of this cutting-edge work pertains to global-scale projections. Any reports that used a data set from the same study are included as a group so that each primary data set is only represented once in the matrix (i.e., not double-counted), thereby avoiding over-emphasis on any one set of projections.
The climate projections provided in the Climate Change Effects Typology Matrix (Appendix C) were culled from studies of varying spatial and temporal scales, modeling parameters, downscaling techniques, and modeling methodologies. This matrix organizes the collected literature by U.S. region, time horizon, climate effect, and spatial coverage with the following characteristics:
We find some results are closely clustered while others range widely, posing a challenge for decision makers attempting to apply them. Attempts to streamline findings are further complicated because many of the studies do not formally quantify likelihood.
The bulk of the climate projections used in the regional temperature and precipitation tables in Chapter 3 and used to develop the maps in Appendix B in this report are from the CMIP3 database of climate model integrations by Michael Wehner of the Lawrence Berkeley National Laboratory for use in the 2009 USGCRP report, Climate Change Impacts on the United States. The USGCRP data are based on a compilation of aggregated climate model results for two IPCC Special Report Emission Scenarios. These emission scenarios are based on ranges of projections of a number of societal changes, such as changes in population, energy use, technological development, and unpredictable societal behavior (IPCC 2007a; CCSP 2007). The lower emissions (B1) and the higher emissions (A2) emission scenarios in this data set encompass a broad range of possibilities; however, they do not span the full range of possible emissions scenarios. In fact, current emissions rates are exceeding the A2 scenario. On the other hand, emissions scenarios below B2 have been developed for reports of the Intergovernmental Panel on Climate Change, and are being considered in domestic legislation and international negotiations.
This data set is an impressive collection of regional multi-model ensemble results. These values are averaged for each region from the corresponding climate model grid cells in the CMIP3 database. Using GCM-scale resolution is appropriate for this regional projected information of mean conditions of temperature and precipitation.74 However, the USGCRP report does not provide this information consistently for all regions. We have attempted to be as consistent and quantitative as possible in our documentation of these results to help facilitate comparisons between regions.
This dataset uses the following global climate models for determining the results of the A2 scenario (moderate-high emissions): ccsm3.0, cgcm3.1, cnrm, csiro, gfdl2.1, hadcm3, hadgem1, inmcm3, ispl, miroc_medres, miub_echo, mpi_echam5, mri_cgcm2_3_2a, and pcm. In addition to these models, the B1 scenario (low emissions) also includes results from: bccr_bcm_2_0, gfdl2.0, giss_aom, iap_fgoals_0_g, miroc_hires. The one exception is for Alaska, where a subset of top-performing global climate models was used: gfdl2.1, mpi_echam5, cnrm, hadcm3, and miroc_medres.
In addition, this report also includes several national maps developed using downscaled data. Statistical downscaling of the CMIP3 database was applied to the larger-scale grid-sized climate models to provide finer-scale results of fine-scaled variability of means and extreme conditions such as heat events that are provided in the figures of this report (Hayhoe et al. 2008; Maurer et al. 2002; Maurer et al. 2008).75 This downscaling is accomplished by first determining a statistical relationship between surface observations and climate simulations of the past for each region. This statistical relationship is then applied against the results of future climate simulations to provide fine-spatial projections of mean and extreme thresholds of temperature and precipitation. See Wood et al. (2002) for detailed description of this technique. Downscaled data is a useful tool for discussing variability and extremes that are not well captured in climate models (USEPA 2009). It should be noted that using the statistical downscaled data of the CMIP3 database to provide regional results instead of the CMIP3 database itself would introduce roundoff errors and possibly additional sources of errors.
There is always some degree of uncertainty associated with model projections. The Climate Change Effects Typology Matrix includes this information, when possible, in a column labeled "certainty." Modeling the climate system poses a number of challenges, including understanding and representing the climate system's processes and natural variability, and estimating future emissions and uptake of greenhouse gases (IPCC 2007a).
There are various techniques used to address uncertainty, including probabilistic approaches to quantify uncertainty, modeling various emission scenarios to produce a wide range of future possibilities, comparing present-day model results with observations, and engaging expert judgment to express uncertainty based on level of agreement and amount of evidence (IPCC 2007a).
The IPCC assessments (e.g., Fourth Assessment Report (AR4)) and the U.S. Climate Change Science Program (CCSP) Synthesis and Assessment Product (SAP) reports provide some guidance regarding likelihood and confidence and how this information can be used to filter and comprehend projected climate changes. Likelihood represents the assessed probability that the outcome will occur, and confidence characterizes the consensus across modeling groups or experts that the projections are correct.
Table A-1 outlines the likelihood and confidence for climate variables most relevant to the highway system: temperature rise, changes in precipitation, changes in frequency and intensity of storm events, and sea-level rise.
These likely and very likely indicators provide measures of a portion of the uncertainty and can act as a general guide in assessing the overall findings included in the Climate Change Effects Typology Matrix. However, they do not account for uncertainty associated with future emissions, uncertainty with downscaling techniques, uncertainty associated with the uptake of greenhouse gases, or any systematic errors in the climate models. Hence, the individual studies included in the Climate Change Effects Typology matrix may not reflect the same level of confidence or likelihood as described in Table A-1.
|Temperature Rise||annual mean||Very likely||High confidence|
|Seasonal mean||Very likely||High confidence|
|Extreme Heat Events||Very likely||High confidence|
|Changes in Precipitation||annual mean||Very likely||Not found|
|Seasonal mean||Very likely||Medium confidence|
|Change in frequency and intensity||Very likely||Not found|
|Intensification of storm events||Likely||High confidence (extratropical)|
|Sea-level rise||Cannot assess likelihood||Not confident in upper bound of SLR|
Table A-1: Likelihood and Confidence for Climate Variables Identified to Affect the Highway System. aCCSP (2007); bIPCC (2007a); Very Likely refers to a greater than 90% probability; Likely refers to a greater than 66% probability; High confidence represents an 8 out of 10 chance; Medium confidence represents a 5 out of 10 chance.
This report quantifies key aspects of the temperature and precipitation projections in the USGCRP dataset. The tables for each region describe the "mean," the "likely" range, and the "very likely" range for each of the three time frames addressed in this study. Those terms are defined as follows:
Since the studies in the Climate Change Effects Typology Matrix do not necessarily use the same terminology for defining uncertainty, we have limited our use of the terms "likely" and "very likely" in the report's narratives to the aforementioned definitions.
In general, the model projections are more uncertain the further they are into the future. A small range tends to be applicable for the near-term time horizon while a larger range of plausible values is appropriate for the long-term. Transportation planners that are less risk-averse may be more comfortable using the "likely" range, whereas planners that are more risk-averse may prefer to use the "very likely" range.
Although the projections described in this report are available at the regional scale, some care should be taken when applying them at regional or local scales. Given modeling efforts currently underway, finer spatial-scale information is likely to become available within the next few years.
The climate projections provided in the Climate Change Effects Typology Matrix were culled from studies of varying spatial and temporal scales, modeling parameters, downscaling techniques, and modeling methodologies. As a result, some projections are closely clustered while others range widely, posing a challenge for decision makers attempting to apply them. Use of the findings is also complicated because many of the studies do not formally quantify likelihood.
The climate projections contained in the Climate Change Effects Typology Matrix are at a regional-scale or finer spatial resolution. Fine spatial resolution is not directly available from global climate models with a typical grid cell distance varying in size from 50 to 250 miles. Downscaling techniques have been developed that transform the projected climate effect at large-grid cell resolution into a fine-scale resolution of the order of 20 miles or less (NECIA 2006). Such techniques fall into two categories: statistical downscaling and dynamical downscaling. Statistical downscaling determines a relationship between the climate model output of a climate effect for a past 30- to 40-year time period and the observed climate effect for the same time period. This relationship is then used to downscale the projected climate model output for that particular climate effect. This approach is best when the determined relationship is robust over time; that is, the processes governing the climate effect remain fixed with time. This may not always be an appropriate method for downscaling precipitation projections (NECIA 2006). Statistical downscaling is relatively quick and inexpensive. Dynamical downscaling uses a regional model equipped with small-scale processes and local topography. The climate model data are used as inputs around the boundaries of the regional model. Though this process tends to be expensive and time-consuming, it does include dynamical changes in response to large-scale forcing. Given the investment for this approach, dynamical downscaled studies generally provide projections based on the results from a single climate model.
The variation in attributes between studies creates challenges in developing a universal methodology for drafting a regional narrative from the collected projections. Factors that may vary between studies and that can be important when comparing their results include the following:76
A.6 Criteria for Data Selection in this ReportFigure A-1
It is challenging to synthesize the array of projections provided by the Climate Change Effects Typology Matrix into a narrative discussion that can assist in informing future analysis of climate impacts on the highway system. The criteria used in this report for determining whether a study would be included in the regional narratives of section 4 were guided by discussions with climate experts and are illustrated in Figure A-1. There was a strong consensus for providing a plausible range of projections as opposed to a point value. This range can be provided by studies using multi-model ensembles and relevant emission scenarios (the B2 and A2 scenarios were suggested as the "lower" and "higher" emission scenarios).78 Further, while a projection associated with a single emission scenario for a multi-model ensemble study is acceptable for near-term discussions, as there is not much variation of greenhouse gas emissions between emission scenarios over this time frame, in this study we chose to use two emissions scenarios for each of the three time frames.79 The projections are, however, affected by the choice of emission scenarios for mid- to long-term projections, as the greenhouse gas emissions increasingly diverge between emission scenarios with time and should be provided across the range of low to high emission scenarios. Providing these ranges can arm transportation planners with a set of plausible scenarios to apply, as warranted, in determining the impacts on the highway system.
The USGCRP (2009) data meet the criteria established for incorporating a study into the narrative: the study was conducted recently, it is a multi-ensemble study, it provides a range of plausible scenarios based on low to moderately high emission scenarios, and the dataset has been aggregated for each region using the CMIP3 database (see section A.3 for further description). In addition, this dataset is uniformly available for each region and time horizon. Therefore, the USGCRP dataset is the primary source for annual temperature, seasonal temperature, and seasonal precipitation. Annual precipitation is not discussed in the narratives, as it averages out the important seasonal variations. One exception of the application of the USGCRP dataset is Alaska. The Alaska projections are based on the five climate models that are determined to be the top performers in simulating temperature and precipitation (Walsh et al. 2008).
Any additional regional temperature and precipitation projections that meet the model inclusion criteria are compared against the USGCRP projections. If the projections are significantly different, the second set of information is provided. It should be noted that this criterion for inclusion into the narrative tends to favor statistical downscaling results over the intensive dynamical downscaling results.
Extreme events such as heat events, and changes in precipitation intensity, duration, and frequency are provided regardless of how well a study meets the criteria discussed above. These studies may use dynamical downscaling followed by multiple regional model runs or intensive statistical downscaling.
Sea-level rise and storm surge projections are not provided by the USGCRP data set. Global sea-level rise studies are discussed at the beginning of Section 4 and are applicable across all regions. Studies that account for regional-scale or local-scale changes in sea level are discussed within the respective region.
74It should be noted that using the statistical downscaled data of CMIP3 database to provide regional results of mean temperature and precipitation instead of the CMIP3 model results directly would introduce roundoff errors and possibly additional sources of error. Statistically downscaled data are particularly informative when describing fine-scale variability or extremes. back
75Downscaled data is a useful tool for discussing variability and extremes that are not captured well in climate models (USEPA 2009). For access to these downscaled CMIP3 data, see Maurer 2007. back
76Not all modeling assumptions are provided in the "Climate Change Effects Typology Matrix" and therefore can be problematic in efficiently comparing these studies. For example: studies may or may not account for greater degree of ice melting when estimating sea-level rise or studies may or may not allow for evaporation when estimating precipitation amounts. back
77The IPCC developed four emission scenarios (A1, A2, B1, and B2) described in the IPCC Special Report on Emissions Scenarios (SRES) associated with four plausible storylines representing varying degrees of economic, regional, and environmental projected change as well as allowing for global integration. These studies tend to draw from the following 4 IPCC SRES: A1Fi: very rapid economic growth based on per capita, global population peaking in 2050, rapid introduction of new and more efficient technology being fossil-intensive; A1B: very rapid economic growth based on per capita, global population peaking in 2050, rapid introduction of new and more efficient technology being evenly distributed between fossil and non-fossil technology; B1: rapid changes in economy though slower growth than the A1 scenarios, same global population pattern as in A1, with new technology becoming clean and resource-efficient; A2: slowest economic development of all the scenarios based on per-capita growth, has the highest global population allowing for a continuous increase, with the slowest and most fragmented development. back
78It should be noted that this assumption assumes a somewhat linear relationship between greenhouse gas emissions and changes in the magnitude of the climate variables; hence, it does not account for complexities such as tipping points. back
79The issue of teasing out climate variability is particularly problematic in the near-term. For example, in the 2020s, the regional precipitation projection for the Pacific Northwest was found to lie within the range of natural climate variability, while regional projected temperature was outside of this natural range. The far-term projected temperature variable was found to be far outside the natural range (Mote et al. 2005). This highlights the difficulty in distinguishing between natural climate variability and the projected change. back