Projected future climate

Overview

Figure 1. Annually averaged daily maximum temperature over the Colorado River Basin and interior West, as observed in 1960-1990 (left) and projected for the 2050s under RCP8.5 (average of 32 models; right). By the 2050s, the models project 4-8°F of warming above the 1960-1990 average (3-7°F above the 1991-2020 average) with climate zones moving uphill and northward. (Design: J. Lukas; Maps: The Climate Explorer, https://crt-climate-explorer.nemac.org/, based on data from Livneh et al. 2013, 2015 (observed) and CMIP5-LOCA; D. Pierce, Scripps Institution; http://loca.ucsd.edu; Pierce, Cayan, and Thrasher 2014 (projected)

As described in Recent climate change, the climate of the Colorado River Basin has become substantially warmer in the past 40 years, mainly or entirely due to human changes to the atmosphere and climate system. Meanwhile, although average precipitation in the basin has not obviously changed, as with temperature, the underlying atmospheric processes (e.g., circulation patterns, water vapor content) are already being influenced by the warming global climate.

Both basic physics and our most sophisticated tools (global climate models; GCMs) tell us that further warming will occur in the basin over at least the next several decades, at a similar or greater rate than what has been observed since 1980 (Figure 1). Neither the physics nor the GCMs present a clear picture of future precipitation change, however.

Relevance

Continued warming of the basin’s climate will further impact the surface water balance and hydrology, leading to chronic drought conditions by today's standards, unless a large increase in average annual precipitation occurs--which is unlikely. Long-term declines in annual precipitation are possible, and more likely in the Lower Basin. More warming will also further stress the basin’s ecosystems and species, potentially beyond thresholds for viability in some cases.

Methods

Global climate models

Global climate models (GCMs) are extraordinarily complex math-based software programs that simulate the Earth’s climate system. GCMs partition the Earth into thousands of 3-D gridboxes or cells, typically 30 to 80 miles (50 to 130 km) on a side horizontally, and use equations based on both observations and fundamental physical laws to represent the movement of energy, air, water, and other constituents between the gridboxes. GCMs are the main tools used to diagnose past and recent climate changes, and to generate physically plausible scenarios of the future climate. At the global scale, GCMs produce realistic simulations of key physical phenomena, broad-scale patterns, and statistical characteristics of the historical and current climate, though this realism weakens at finer scales. As biogeochemical processes (e.g., carbon cycling) have been added to many of the GCMs, they are also known as Earth system models, or ESMs.

Several dozen GCMs have been developed by over 20 modeling centers in 10 countries. The magnitude of projected future climate change differs among these GCMs, which reflects unresolved scientific uncertainty regarding some key climate processes and the resultant different ways that the modeling teams represent those processes in their models. A major limitation across all GCMs remains their horizontal resolution; the gridboxes are too large to reflect the complex terrain of mountainous areas such as in the Colorado River Basin, or to directly simulate processes like cloud formation or convective storms which occur at smaller, sub-grid scales. Clouds and convective storms are indirectly represented in GCMs through parameterizations: generalized values based on observations or additional modeling.

Simulations of future climate from GCMs are known as projections, as opposed to predictions or forecasts, because the projections are conditional on an assumed future trajectory for greenhouse gases and other human influences on climate. To represent the uncertainty in the future emissions of greenhouse gases, scenarios or storylines now called representative concentration pathways (RCPs) have been developed. These are labeled according to their impact on the Earth’s surface energy balance by 2100, in units of W/m2:

• RCP2.6 (low emissions)
• RCP4.5 (medium-low)
• RCP6.0 (medium)
• RCP8.5 (high emissions).

Most published analyses of climate change focus on RCP 4.5 and/or RCP8.5; RCP2.6 is considered by many experts to be implausibly optimistic in its assumed reduction of emissions.

Under the auspices of the Coupled Model Intercomparison Project (CMIP), the available GCMs are run under standardized protocols, including emissions scenarios as described above, to produce future climate projections to support the periodic Intergovernmental Panel on Climate Change (IPCC) reports. Most of the currently available GCM analyses of local impacts, e.g., for Colorado River Basin hydrology, are based on CMIP5, whose output was originally released in 2011-2012. The most recent set of projections, CMIP6, were released in 2019-2020 and are now being run through a similar chain of subsequent modeling and impact analyses as CMIP5.

The potential climate futures for the Colorado River Basin portrayed by the CMIP5 models are very similar to those seen in CMIP3 in terms of the change in temperature and precipitation (ensemble mean and spread) under a given emissions future. The analyses of CMIP6 thus far indicate that the range of potential precipitation changes for the basin is very similar to CMIP5, but the temperature change is greater in the CMIP6 models. However, there is increasing evidence that the CMIP6 models as a group are overestimating a key cloud feedback on global and regional temperature, and correcting for this bias brings them into line with CMIP5.

Downscaling

To resolve some of the issues with their coarse resolution, raw GCM output is often downscaled, or post-processed to represent more local scales--at most 50 km (31 mi.), and often 12 km (7 mi.) and smaller. Downscaling methods fall into two general categories: dynamical or statistical. In dynamical downscaling, a higher-resolution regional climate model (RCM) is run over the domain of interest, driven by boundary conditions taken from a GCM simulation. This produces localized projections that are better grounded in the physical processes at these scales, but with far higher computing costs than statistical downscaling, which is based purely on spatial statistical relationships among observed climate variables. But because statistical downscaling relies on historical relationships, it may not correctly represent localized future changes in those variables.

Many different downscaled CMIP5 datasets have been developed and publicly archived, serving as the basis for both interactive plotting and analysis tools (i.e., Climate change portals) and impact studies and assessments. Reflecting the disparity in the computational costs, the vast majority of these datasets were produced through statistical downscaling.

The general benefits of downscaling are that it facilitates the use of GCM data in local-impact modes (such as hydrologic models) that require higher-resolution inputs, and provides end users with projections that are at least superficially more relevant to their needs. But downscaling does necessarily not make the depiction of future change more accurate than in the underlying raw GCM projection. Different downscaling methods applied to the same GCM projection can result in quite different projected climate changes for a given region or location, especially for precipitation. The widely used BCSD downscaling procedure, for example, leads to wetter future projections for the Colorado River Basin than the underlying GCMs show, and this "wettening" is now believed to be an artifact of the downscaling method.

Future basin temperature

Figure 2. Projected future temperature change for the Upper Basin compared to a 1971–2000 baseline, from an ensemble of 32 CMIP5 model projections under the moderate RCP4.5 emissions scenario with LOCA downscaling. The light blue lines are the 30-year running averages, plotted on the middle (15th) year, of the individual model projections of future temperature change, with the median projection shown as the dark dashed line. The box-whiskers plot at rights show the distribution of the 30-year average values at 2055 (2041–2070). (Figure: J. Lukas, adapted from Lukas and Payton, 2020; Data: D. Pierce, Scripps Institution; http://loca.ucsd.edu; Pierce, Cayan, and Thrasher 2014)

All of the GCM projections run for CMIP5 under RCP4.5 and RCP8.5 indicate that the Colorado River Basin will continue to rapidly warm in all seasons through the during the 21st century (Figure 2). By mid-century (2041-2070), the basin is projected to warm by +3°F to +8°F over a 1971-2000 baseline, and by +2°F to +7°F over the most recent 30-year period (1991-2020). As can be seen in Figure 2, the range or uncertainty in the projected warming is due to both the emissions scenario (RCP) and the GCMs' different climate sensitivities to each increment of greenhouse gases in the atmosphere. After 2040, the range of projected warming under RCP 8.5 increasingly diverges from RCP 4.5, reflecting the higher emissions and higher concentrations of greenhouse gases in RCP 8.5.

Even under a lower-end warming outcome (+4°F above 1971-2000), the “normal” years around 2055 would be warmer than the very warmest years of the last century. Slightly greater warming is projected for the Upper Basin than the Lower Basin, since the Upper Basin is further from the moderating influence of the Pacific Ocean. Warming is expected to be slightly greater in summer than the other seasons due to a land surface feedback that is more prevalent and potent in summer; when soils dry out, the energy that had been evaporating soil moisture can instead warm the land surface and the air above it.

Future basin precipitation

Figure 3. Projected future annual precipitation change for the Upper Basin compared to a 1971–2000 baseline, from an ensemble of 32 CMIP5 model projections under the moderate RCP4.5 emissions scenario with LOCA downscaling. The light blue lines are the 30-year running averages, plotted on the middle (15th) year, of the individual model projections of future precipitation change, with the median projection shown as the dark dashed line. The box-whiskers plot at rights show the distribution of the 30-year average values at 2055 (2041–2070). (Figure: J. Lukas, adapted from Lukas and Payton, 2020; Data: CMIP5-LOCA; D. Pierce, Scripps Institution; http://loca.ucsd.edu; Pierce, Cayan, and Thrasher 2014)

Compared to temperature, precipitation reflects a much more complex set of climate drivers and processes, and changes in precipitation are more difficult to diagnose and predict than changes in temperature. In general, the GCMs agree that the prevailing westerly storm tracks across western North America will shift northward on average (potentially drying the Lower Basin and parts of the Upper Basin), but also that individual weather systems will carry more moisture, consistent with the effect of warming on the atmosphere's moisture-holding capacity. The models do not agree on future changes in ENSO (El Niño/La Niña) events, which influence storm tracks and precipitation, especially in the Lower Basin.

As a result of these mixed future tendencies, under both RCP4.5 and RCP8.5 the GCMs indicate slight overall tendencies toward higher annual precipitation in the Upper Basin (Figure 3) and toward lower annual precipitation in the Lower Basin. Those tendencies are enhanced for the northern half of the Upper Basin (wetter) and the southern half of the Lower Basin (drier), reflecting north-south gradient across the western U.S. For the Upper Basin, the “wetter” of the 32 projections call for about 5–10% more annual precipitation by mid-century, while the “drier” projections call for 5–10% less precipitation. For the Lower Basin, the "wetter" projections call for 0–5% more annual precipitation, while the drier projections call for 10–15% less precipitation.

The GCMs show more pronounced tendencies for change in seasonal precipitation over the basin than annual precipitation. In winter (DJF), most models show increased precipitation over the Upper Basin. In spring (MAM), most models show decreased precipitation for the Lower Basin. In summer, most models show decreased precipitation for both the Upper and Lower Basins; however, the North American Monsoon is not simulated well in the GCMs, and the confidence in the projected changes in summer precipitation is lower than for the other seasons.

Multi-decadal natural variability (which is simulated by the GCMs) strongly influences the precipitation projections, as seen in the frequent excursions in the 30-year averages in Figure 3, up and down by 5% or more. These excursions, both in the models and in the real world, make it hard to discern trends that might be attributable to forced climate changes, such as shifts in average storm tracks. The ensemble medians of RCP4.5 and RCP8.5 throughout the 21st century are very similar, though after 2050 the RCP8.5 ensemble does have greater spread (right side of Figure 3), suggesting that at least some of the models show larger responses in precipitation as a result of the higher greenhouse gases in RCP8.5.

Most GCMs also project that the variability in precipitation over the western U.S. will itself change, increasing at all time scales, including greater year-to-year variability. This would mean more frequent occurrence of both very dry and very wet years, and more frequent transitions from very dry to very wet conditions.

Future climate extremes and hazards

Hot extremes

Figure 4. Projected increase in the future number of days over 90°F in the Western U.S. for mid-century (2036-2065) compared to a 1976-2005 baseline, from the weighted average of 32 CMIP5 model projections run under the RCP8.5 emissions scenarios, with LOCA downscaling. (Figure: J. Lukas, adapted from Figure 6.9 in USGCRP, 2017: Climate Science Special Report: Fourth National Climate Assessment, Volume I; CMIP5-LOCA data: http://loca.ucsd.edu; Pierce, Cayan, and Thrasher 2014)

Similar to what has been observed over the past 40 years in the basin and the Southwest U.S., the projected shift in the center of the distribution of temperatures (i.e., annually averaged temperature) is expected to manifest in a large increase in extremes at the warm tail of the distribution: extremely hot days and multi-day heat waves. Figure 4 shows that the number of days per year over 90°F is projected under RCP8.5 to increase by 20-60 days by mid-century over the middle and lower elevations of the basin; many high-elevation areas that do not currently experience temperatures above 90°F are expected to see them by mid-century.

Cold extremes

Cold extremes in all seasons are expected to decrease in frequency as the overall climate warms, although cold waves that do occur could be nearly as severe as historical cold waves.

Heavy and extreme precipitation

As noted in Recent climate change, warmer air can hold more moisture, also known as precipitable water (PW). Observations show that globally-averaged PW has increased in recent decades at a rate consistent with physical expectations, given the observed global warming. The CMIP5 models consistently project future increases in average PW globally as warming progresses. Changes in average PW for the Western U.S. are more variable across the CMIP5 models because at the regional scale, PW is also a function of the frequency of wet weather patterns that transport and concentrate moisture--changes in which being less certain. The CMIP5 models do project that extreme PW values for this region (e.g., highest daily PW in a 10-year period) will increase by about 15% by the 2050s.

With these projected increases in PW on the wettest days, extreme precipitation events over the region that includes the Colorado River Basin are also projected to become more frequent and intense. Over the the Southwest US (CA, NV, UT, CO, AZ, NM), what is currently a 1-in-5-year 2-day event is projected to become, by the 2050s, a 1-in-3-year event under both RCP4.5 and RCP8.5. [Jannsen et al 2014] The 1-in-20-year 1-day event is projected to increase in magnitude (rainfall depth) by mid-century by 10% under RCP4.5 and by 12% under RCP8.5 [USGCRP 2017]. These and other studies indicate that heavy and extreme precipitation days (>95th percentile of all days with precipitation) will contribute an increasingly greater fraction of the annual precipitation.

Drought

The projected higher temperatures for the basin will likely lead to greater frequency and severity of soil-moisture droughts in the basin as the increases in evapotranspiration that will result from warming either exacerbate long-term decreases in annual precipitation (if they occur) or cancel all but the largest future increases in precipitation. CMIP5 projections of near-surface soil moisture consistently indicate decreased summer (JJA) soil moisture over the Colorado River Basin, and most parts of the Western U.S. [USGCRP 2017]

Future projections of hydrologic drought and its key indicators (snowpack, runoff) are detailed in Projected future hydrology.

Wildfire

Wildfire occurrence and total areas burned per year are projected to increase substantially in the Rocky Mountain West by mid-century as warming leads to drying of the land surface and of fuels. [Lukas et al. 2014]. The length of the fire season is also projected to increase. While several other factors have led to increases in large fires and in area burned since 1980, and will continue to influence wildfire trends in the future (e.g., historical fire suppression, development of the wildland-urban interface, increase in human ignition sources), anthropogenic warming has been identified as a significant driver of increasing wildfire occurrence. [Williams and Abotzoglou 2018]

Data and tools

There are several web tools that can be used to explore downscaled CMIP5 model projections of future climate in graphical form, and also download the data being shown. A few caveats to these tools:

• The multi-model mean or median, which is usually highlighted in maps and plots, is not necessarily the most likely future--it's just the middle of the distribution (range) of the model projections.
• Even within the same emissions scenario (RCP), the projected future climate change depicted by one tool can differ from others because:
  • Different subsets of the >35 CMIP5 models are used in each downscaled dataset
  • Each downscaling method may shift the change signal shown by the underlying models
  • Different historical baseline periods and future periods may be used to analyze the data
• To repeat a point made above, the local resolution of downscaled data does not mean that it is more accurate; i.e., more likely to represent the actual future climate change

A User Guide to Climate Change Portals provides descriptions of many different web tools for visualizing and downloading CMIP5 model projections, including the three listed below.

The Climate Explorer

The Climate Explorer interactive tool (developed by NOAA and its partners) allows users to generate zoomable climate change maps for temperature and precipitation for the U.S. and northern Mexico, and time-series plots of projected future temperature and precipitation for U.S. counties. Other areas for analysis, such as river basin, cannot be selected. All projected climate data are from the LOCA-downscaled CMIP5 dataset (32 models).

Climate Toolbox

The Climate Toolbox is a diverse set of interactive tools developed by researchers at U. Cal-Merced and partners. All projected climate data are from the MACA-downscaled CMIP5 dataset (20 models).

Climate Toolbox - Climate Mapper

The Climate Mapper tool allows users to generate zoomable climate change maps for temperature as well as precipitation and a broad set of drought, agricultural, and fire-risk indicators. To access the projections, "Future: Projections" must be selected under Choose Data - Time Scale in the upper left. Projected climate data are from the MACA-downscaled CMIP5 dataset (20 models).

Climate Toolbox - Future Time Series

The Future Time Series tool allows users to generate time series for temperature as well as precipitation and a broad set of drought, agricultural, and fire-risk indicators. To access the projections, "Future: Projections" must be selected under Choose Data - Time Scale in the upper left. Projected climate data are from the MACA-downscaled CMIP5 dataset (20 models).

National Climate Change Viewer

The National Climate Change Viewer (NCCV) is a versatile interactive viewer developed by USGS. It allows users to generate zoomable change maps, time series, monthly climographs, and scatterplots for multiple climate variables. This viewer's analysis areas are river basins/watersheds (HUC2, HUC4, and HUC8); a separate viewer allows analyses by state or county. All projected climate data are from the MACA-downscaled CMIP5 dataset (20 models), same as for the Climate Toolbox.

Additional resources

State of the Science Report

Chapter 11 of the State of the Science report details the projected future climate for the basin, in section 11.6, as well the methods and data used in those projections: Global climate models (GCMs; 11.2); CMIP (11.3); Emissions scenarios (11.4); Downscaling (11.5).

NCA4 Climate Science Special Report

The 2018 Climate Science Special Report (CSSR), Volume 1 of the Fourth National Climate Assessment (NCA4), describes climate models, emissions scenarios, and projections in Chapter 4, and details the future projections of climate variables and extreme events for the U.S. in subsequent chapters: Temperature changes (Chapter 6); Precipitation changes (Chapter 7); Droughts, floods, wildfire (Chapter 8); Extreme storms (Chapter 9).