Transient Future Climate over the Western United States ...msnyder/papers/snyder2005_ei.pdf · for...

Copyright © 2005, Paper 09-011; 7,247 words, 8 Figures, 0 Animations, 4 Tables.http://EarthInteractions.org

Transient Future Climate over theWestern United States Using aRegional Climate ModelMark A. Snyder* and Lisa C. SloanClimate Change and Impacts Laboratory, Department of Earth Sciences, University of

California, Santa Cruz, Santa Cruz, California

Received 13 October 2004; accepted 28 February 2005

ABSTRACT: Regional climate models (RCMs) have improved our under-standing of the effects of global climate change on specific regions. The needfor realistic forcing has led to the use of fully coupled global climate models(GCMs) to produce boundary conditions for RCMs. The advantages of usingfully coupled GCM output is that the global-scale interactions of all compo-nents of the climate system (ocean, sea ice, land surface, and atmosphere) areconsidered. This study uses an RCM, driven by a fully coupled GCM, toexamine the climate of a region centered over California for the time periods1980–99 and 2080–99. Statistically significant increases in mean monthly tem-peratures by up to 7°C are found for the entire state. Large changes in pre-cipitation occur in northern California in February (increase of up to 4 mmday−1 or 30%) and March (decrease of up to 3 mm day−1 or 25%). However,in most months, precipitation changes between the cases were not statisticallysignificant. Statistically significant decreases in snow accumulation of over 100mm (50%) occur in some months. Temperature increases lead to decreases insnow accumulation that impact the hydrologic budget by shifting spring andsummer runoff into the winter months, reinforcing results of other studies thatused different models and driving conditions.

KEYWORDS: Climate change; Regional modeling

* Corresponding author address: Mark A. Snyder, Department of Earth Sciences, Universityof California, Santa Cruz, 1156 High St., Santa Cruz, CA 95064.

E-mail address: [email protected]

Earth Interactions • Volume 9 (2005) • Paper No. 11 •

1. Introduction

The increase in atmospheric greenhouse gas concentrations (i.e., CO2, CH4, andNO2) derived from anthropogenic sources has produced measurable changes inglobal climate since preindustrial times (Houghton et al. 2001). Increases in globalmean annual temperature and decreasing Arctic sea ice are two examples ofchanges attributed to anthropogenic emissions (Johannessen et al. 1999; Jones etal. 2000). The Third Assessment Report (TAR) of the Intergovernmental Panel onClimate Change (IPCC; Houghton et al. 2001) presented the most compelling andcomplete climate change research to date. However, a deficiency in our under-standing is the nature of regional-scale climate changes within the broader contextof global climate change.

Coupled atmosphere–ocean general circulation models (AOGCMs) have beenthe primary tools for investigating future climate change through the end of thiscentury in the IPCC TAR. AOGCMs are run with relatively coarse horizontalresolution (∼280 km × 280 km), which is not adequate for accurate climate simu-lation in regions of complex topography such as western North America (Bell etal. 2004). However, regional climate models (RCMs) have been used to addressthe study of climate at higher resolutions (40 km × 40 km – 100 km × 100 km) fora number of topographically complex regions. RCMs have been shown to improvethe representation of climate, over global climate models, based on comparisons ofmodel output to observations of monthly average temperature and precipitation inregions such as western North America, the European Alps, the Aral Sea region,and Scandinavia (Bell et al. 2004; Christensen et al. 1998; Giorgi et al. 1997; Kimet al. 2002; Leung and Ghan 1999a; Leung and Ghan 1999b; Small et al. 1999;Snyder et al. 2002).

Simulations of future climate at the regional scale have been performed for theentire continental United States (Giorgi et al. 1994; Giorgi et al. 1998; Pan et al.2001). For the topographically complex western United States, some studies havefocused broadly on future climate change over the entire region (Leung and Ghan1999a; Leung and Ghan 1999b; Thompson et al. 1998), while others have focusedon a smaller region at higher spatial resolution centered over California (Bell et al.2004; Kim 2001; Kim et al. 2002; Leung et al. 2004; Snyder et al. 2004; Snyderet al. 2002).

California is of interest for a number of reasons; the complex topography andlatitudinal extent of the state create many different climate regions that support awide variety of ecosystems, ranging from desert in the south, to temperate rain-forest in the north, and to alpine tundra in the mountainous regions. The state ishighly populated, and as such, water and climate issues are constantly scrutinized.Water resources in California are highly tuned to the interannual and longer-termvariability of precipitation. The state’s water delivery system is designed to trans-port water from northern California, where the majority of precipitation occurs, tothe large population centers, particularly in southern California (California De-partment of Water Resources 1998). California relies on accumulated snow storedin the mountainous regions to provide a steady supply of runoff during the drysummer and fall months. Analysis of observational records shows that historicallythe year-to-year snow accumulation is highly variable (Cayan 1996; Dettinger andCayan 1995) because of the natural variability of precipitation in this region.


Changes to the mean climate and variability due to increasing greenhouse gasconcentrations could drastically alter the water resources of California and westernNorth America.

Previous studies of future climate in this region have found increased tempera-ture and decreased snow accumulation but are inconsistent in terms of changes inprecipitation in response to future climate change (some find increases whileothers decreases; Kim 2001; Kim et al. 2002; Knowles and Cayan 2002; Leung etal. 2004; Snyder et al. 2002). Snyder et al. (Snyder et al. 2002) used boundaryconditions derived from a global climate model (GCM) with monthly varying seasurface temperatures (SSTs) to examine the sensitivity of California climate todoubled preindustrial CO2 concentrations (560 ppm). That study found that underfuture climate conditions, the water resources of the state were dramatically af-fected. The results showed that snow accumulation decreased by up to 50% andthat the snow season terminated approximately one month earlier in the futureclimate case. Mean monthly temperatures increased throughout the state, with thelargest increases in the higher-elevation regions. Mean monthly precipitation in-creased in the northern half of the state and decreased in the southern half duringthe wet season (November–May).

Leung et al. (Leung et al. 2004), Kim (Kim 2001), and Kim et al. (Kim et al.2002) used transient forcing to drive regional climate models over the westernUnited States. In both studies by Kim, the Second Hadley Centre Coupled Ocean–Atmosphere General Circulation Model (HadCM2) was used to force the MASregional model (36-km resolution) for a total of 10 yr (2040–49). Results fromthose studies show increased monthly temperatures of up to 3°C, a large increasein cold-season precipitation, and decreased snowfall. In Leung et al. (Leung et al.2004), the Parallel Climate Model (PCM) was used to force the fifth-generationPennsylvania State University (Penn State)–National Center for Atmospheric Re-search (NCAR) Mesoscale Model (MM5; 40-km resolution) for 20 yr of a tran-sient scenario (2040–60). Their study found increased temperature of 1° to 2.5°C,no overall change in precipitation except for some summer drying, and a reductionof snowpack by up to 70%. For Leung et al. (Leung et al. 2004), the temperaturechange of the GCM and RCM was about the same magnitude over the entiredomain. However, the spatial pattern of the RCM temperature change was moredetailed because of better representation of topography. For precipitation, themagnitude of decrease in the GCM was not as great as in the RCM, and the spatialpattern of the RCM reflects the more detailed topography. For both Kim (Kim2001) and Kim et al. (Kim et al. 2002), the change in the GCM results for thefuture case was not reported. A separate examination of future snow accumulationusing statistically downscaled output from an AOGCM found decreases of up to50% by 2090 (with 2.5 times the preindustrial CO2 concentrations; Knowles andCayan 2002).

In this study, we examine climate change over California using an RCM drivenby a transient climate change scenario from an AOGCM (both models are de-scribed below). While equilibrium and transient forcing of an RCM have beenpreviously explored for this region (Kim et al. 2002; Leung et al. 2004; Snyder etal. 2002), our study looks further into the future by examining climate at the endof this century (2080–99); we also use a different pair of global and regionalmodels than previous transient studies. The use of multiple global and regional


model pairs for the same region is important as it provides a range of results forfuture climate, something policy makers have repeatedly requested (Franco andSanstad 2004).

2. Methods and modelsOutput from the NCAR Climate System Model version 1.2 (CSM1.2; Boville andGent 1998) was used to drive the regional climate model (RegCM2.5) for twoexperiments. Two archived segments of a continuous run, each 20 yr in length andcorresponding to the years 1980–99 and 2080–99, were obtained from NCAR atthe appropriate time resolution for driving the regional climate model (6 hourly).This continuous run is the CSM IPCC A1 scenario, which is a branch of anotherCSM run that began with the year 1870 and used time-dependent atmospheric SO2and greenhouse gases (Dai et al. 2001). For the time period 1980–99, CSM usedobserved values of greenhouse gases and atmospheric SO2 that were updated ineach year of the run. A comparison of the 1980–99 time period with observationsshows that the CSM compares well for surface temperature and precipitation (Daiet al. 2001). For the 2080–99 period, values of greenhouse gases and atmosphericSO2 were updated yearly in the model using calculated values based on the IPCCA1 scenario.

2.1. The AOGCMThe CSM1.2 is a component model with dynamic atmosphere, ocean, sea ice, andland surface models (Boville and Gent 1998). The atmospheric component modelis the Community Climate Model (CCM3), the ocean model is the NCAR OceanModel (NCOM), the sea ice model is the NCAR Community Sea Ice Model(CSIM), and the land surface model is NCAR Land Surface Model (LSM). Thecomponents are coupled together through a flux coupling component and run intandem. For this study, the horizontal resolution of the CCSM was T42 (2.8° × 2.8°latitude by longitude) and the vertical resolution was 18 levels.

2.2. The RCMRegCM2.5 was originally developed at NCAR for climate studies as a modifiedversion of the Penn State–NCAR MM4 weather model (Giorgi et al. 1993a; Giorgiet al. 1993b). RegCM2.5 includes the CCM3 radiation package (Giorgi andShields 1999), and is coupled to the Biosphere–Atmosphere Transfer Scheme(BATS) land surface model (Dickinson et al. 1993). The model domain for theexperiments in this paper is centered over California with 60 latitudinal grid pointsby 55 longitudinal grid points and a horizontal resolution of 40 km (Figure 1). Thevertical resolution is 14 levels and is represented by terrain-following sigma co-ordinates. The model has been validated against observational data for modern-dayclimate for this region and does a reasonable job of simulating the spatial andtemporal features (Bell et al. 2004).

2.3. ExperimentsTwo cases, “Modern” (1980–99) and “Future” (2080–99), were run usingRegCM2.5. The range of CO2 concentrations in the AOGCM runs was 338–369ppm for the period 1980–99 and 635–686 ppm for the period 2080–99. RegCM2.5


was run for 20 yr for each case, with CO2 values in the RCM set to 353 and 660ppm in the Modern and Future cases, respectively. These values are the midpointvalues of the AOGCM runs, chosen because the RCM does not have an option fortime-varying greenhouse gases.

3. Results3.1. Comparison of the Modern case RCM and GCM results to

observationsMonthly average temperature and precipitation output from the RCM for the entiremodel domain, excluding the boundary forcing zone, was compared with gridded

Figure 1. Topography in meters of the full RCM domain and map of Californiashowing the location and extent of the northern, mountain, and southernregions.


monthly observational datasets generated by the Climate Research Unit (CRU;New et al. 2000) in order to assess the model performance in our domain. TheRCM results for the Modern case (1980–99) show a negative difference (RCMminus observations) in surface temperature of between 2° and 4°C on a monthlybasis compared to CRU monthly average temperature data for 1980–99 (Figure 2).The driving GCM data (CSM1.2) are 1°–2°C cooler than the observations (1979–98) on an annual average basis for western North America (Dai et al. 2001). It is,

Figure 2. Monthly and domain average surface temperature for the Modern runcompared with CRU monthly temperature data (1980–99). Units are de-grees Celsius. The CRU mean is plotted as the red solid line, with thestandard deviation plotted as the red dashed lines. The RCM mean isplotted as the blue solid line, with the standard deviation plotted as theblue dashed lines. The difference (CRU minus RCM) is plotted as the blacklong dashed–short dashed line.


therefore, likely that some of the colder-than-observed temperatures in the RCMare inherited from the driving conditions.

A comparison of monthly average precipitation from the RCM for theModern case shows positive precipitation differences of up to 2.0 mm day−1 (RCMminus observations) from October through April (the wet season; Figure 3). TheRCM has a negative difference of up to 0.6 mm day−1 (drier than observations) forJuly through September. The differences between the GCM driving data (CSM1.2)and observations of precipitation on an annual average basis are up to +5 mmday−1 over western North America (Dai et al. 2001). CSM also overestimatescloud cover over land.

Figure 4 shows the average temperature for June–July–August (JJA) over thedomain from the RCM Modern case, GCM Modern case, and observations ofsurface temperature from CRU. The RCM is cooler than observations in each

Figure 3. Same as in Figure 2, but for precipitation.


Figure 4. Average temperature for JJA for the RCM Modern case, GCM Moderncase, and CRU observations (1980–99). Units are degrees Celsius.


month, especially over the Sierra Nevada and the Central Valley. The GCM iscooler than observations over some of California, but a few grid cells in southernCalifornia show temperatures warmer than the observations. The RCM pattern oftemperature for these months matches much better with the observations than theGCM (Figure 4). We find that the RCM is cooler than the observations in allmonths over much of California (not shown). The RCM pattern continues to agreewell with the observed patterns of temperature for the other months. The GCM isgenerally cooler than the observations over California for all other months, and thepattern is consistently a poor match to the observed pattern (not shown).

Precipitation over the RCM domain is shown in Figure 5 for December–January–February (DJF) from the RCM, GCM, and CRU observations of precipi-tation. The RCM tends to overestimate precipitation in all three months over theSierra Nevada and along the coast of northern California (Figure 5). There is goodagreement between the pattern of precipitation produced by the RCM and theobservations. The GCM tends to underestimate precipitation over California anddoes not capture the precipitation highs in the Sierra Nevada and along the coastof northern California (Figure 5). For all other months, the RCM continues tounderestimate the amount of precipitation over California, while reproducing thepattern of the observations quite well (not shown). The GCM underestimates theamount of precipitation in most months, except for the summer when it overesti-mates precipitation in central California (not shown). The GCM pattern of pre-cipitation is poorly matched to the observations in all months.

3.2. RegCM2.5 temperature results

We defined three regions within California, based on the similarity of climate andgeographic features in those regions, to aid in our analysis (Figure 1). The northernregion includes coastal and inland California north of the Los Angeles area. Themountain region includes the Sierra Nevada, Cascade Range, and higher-elevationregions in northern California. The southern region includes coastal, inland, andhigh desert regions of southern California.

The mean monthly temperature results for the entire domain show increases ofup to 6°C in the “Future” relative to the “Modern” case, with the greatest warmingin the Sierra Nevada in May (not shown). Temperatures along the coast increasedby 1°–2.5°C on a month-to-month basis. Inland temperatures increased up to 3°C,with the greatest increases at higher elevations (not shown).

For the three regions that we defined, temperature increased in every month.The mean temperature in each month in the Future case is greater than in theModern case (Figure 6). The maximum and minimum monthly temperatures,calculated from all months for each month of each case and then spatially averagedfor each region (Figure 1) are also greater in the Future case relative to the Moderncase. In many months, for each region, the increase in mean temperature in theFuture case exceeds the maximum temperature in the Modern case. Additionally,there are months where the increase in minimum temperature in the Future case isgreater than the mean temperature in the Modern case. Changes in the overalltemperature range (from maximum to minimum) occur in some months as well. InApril, for all three regions, there is an increase in the temperature range in the


Figure 5. Same as in Figure 4, but for average precipitation for DJF.


Figure 6. Monthly mean, maximum, and minimum temperature by region for theFuture and Modern cases. Units are in kelvins.


Future case relative to the Modern. We see the opposite occur in May, where thetemperature range decreases in all regions.

We tested monthly temperature changes for statistical significance in the north-ern, mountain, and southern regions. We used the Mann–Whitney U test, as inSnyder et al. (Snyder et al. 2002), to compare the values of monthly temperature,spatially averaged, for the Modern and Future cases to determine if they are fromthe same distribution. This test is similar to the t test, but with one importantdifference: the data does not need to have a normal distribution (Gaussian). Thismakes the Mann–Whitney U test well suited to climate data such as precipitationand snow accumulation that may not have a normal (Gaussian) distribution. Forconsistency, we also used the Mann–Whitney U test for temperature. Specifically,we are applying the test to monthly average values to determine if the values fromthe Modern and Future cases could have come from the same distribution. Themethod used here is as follows: 1) Perform a spatial average for each month, foreach variable, for each of the regions through the time series. This produces 12months × 20 yr × 3 regions × 3 variables × 2 cases. 2) For each variable, performthe Mann–Whitney U test on the 20 yr of monthly averages, for each region. Theresult of the test is the median difference between the Modern and Future datasetsand the p level. If the p level is less than 0.05, then the two datasets (Modern andFuture) are significantly different at the 95% confidence interval. Table 1 showsthe differences in the median temperature (Future minus Modern). We found thatfor all three regions in all months, the changes in temperature from the Modern toFuture case are statistically significant at the 95% confidence interval.

3.3. RegCM2.5 precipitation results

Mean monthly precipitation results show increases in some months and decreasesin others for the Future case relative to the Modern case (not shown). Precipitationincreases in northern and central California in both January and February by up to3.5 mm day−1 (21% increase) in the Future case (not shown). In January, the

Table 1. Monthly median surface temperature difference (Future − Modern) andsignificance by region (°C). Values in bold indicate significance at the 95% con-fidence interval based on the Mann–Whitney U test. Values shown are the differ-ences of the medians of the cases for each region.

MonthNorthern

median differenceMountain

median differenceSouthern

median difference

Jan 2.622 4.194 2.680Feb 2.319 3.634 2.153Mar 2.134 3.290 2.758Apr 2.206 3.097 2.959May 3.187 3.695 3.274Jun 2.876 3.626 2.513Jul 2.944 3.974 2.995Aug 2.610 3.169 2.465Sep 2.685 3.069 3.455Oct 2.639 2.640 3.160Nov 1.940 2.115 2.580Dec 1.640 2.319 1.945


largest increases are 2.5 mm day−1 over the Sierra Nevada (19% increase; notshown). In February, the greatest increases are 3.5 mm day−1 along the north coastand northern Sierra Nevada (21% increase). Decreases in Future precipitation of 2mm day−1 occur in the Sierra Nevada in March and May (30% decrease; notshown). Decreased precipitation of 3 mm day−1 along the north coast occurs inDecember (25% decrease).

By region, we again find increases and decreases in precipitation on a monthlybasis (Figure 7). In general, the mean precipitation values change very little be-tween the Future and Modern cases; there are, however, exceptions. In the moun-tain and northern regions in May, the mean precipitation decreases in the Futurecase. There are also changes in the maximum monthly precipitation in March,April, November, and December; in the northern region, the maximum precipita-tion decreases in the Future case relative to the Modern case.

Table 2 shows the results of the Mann–Whitney U test as applied to monthlyprecipitation for the northern, mountain, and southern regions. We found that onlythe mountain and southern regions in May showed statistically significant differ-ences in precipitation between the two cases at the 95% confidence interval.

3.4. RegCM2.5 snow accumulation results

Monthly mean snow accumulation is dramatically decreased by up to 400-mm(70%) snow water equivalent in all months in the Future case (not shown). Thedecreases are concentrated in the Sierra Nevada and southern Cascade Range. Inthe Future case, the end of the snow season occurs about one month earlier thanin the Modern case, with almost all snow melted by May.

Figure 8 shows the large decreases in mean, maximum, and minimum monthlysnow accumulation in the mountain region. In most months, the maximummonthly value in the Future case is less than the mean in the Future case. We findthat the variability of the snow accumulation is also greatly diminished, as shownby the decreased range of monthly snow accumulation values.

Since the other two regions lack measurable snow accumulation (Table 3), weapplied the Mann–Whitney U test to monthly snow accumulation for just themountain region. We found that for the months of December through June, thechanges in monthly snow accumulation between the Future and Modern cases aresignificant at the 95% confidence interval.

3.5. Trends through time

In addition to monthly average results, we examined the trends through the lengthof the runs to determine rates of change for the variables (Table 4). Annual averagetemperature trends of + 0.019°C yr−1 were found in the Modern case, and in theFuture case the trend was +0.011°C yr−1. We found annual average precipitationtrends in the Modern case to be 0.006 (mm day−1) yr−1 and in the Future case tobe 0.018 (mm day−1) yr−1. For snow accumulation in the Modern case, modelresults indicate a trend of + 0.13 mm yr−1 in annual average snow accumulation.In comparison, in the Future case, there is a trend of 0.037 mm yr−1 in annualaverage snow accumulation. The trends, for all three variables, were found to notbe statistically significant when compared to the annual variability.


Figure 7. Same as in Figure 6, but units are in mm day−1.


4. Discussion

This study provides an examination of modeled climate over California for twotime periods using transient forcing. It fills an important gap in transient climatestudies of California by modeling the climate at the end of the century (2080–99)using a different global and regional climate model than previous studies. Leunget al. (Leung et al. 2004) pointed out the need to use multiple global and regionalmodels to adequately explore the uncertainty of future climate change.

For our domain, RegCM2.5 does a reasonable job of reproducing monthlyaverage temperature and precipitation for the time period 1980–99. Differences intemperature and precipitation apparent in the RCM compared to observations canbe attributed largely to the boundary conditions. The seasonal cycle of temperatureis reproduced quite well, after removing the difference. Seasonal precipitation isunderestimated in the winter and spring but overestimated in the summer. Overall,the seasonal cycle of precipitation is not as well represented as temperature.

This study also provides a good comparison to other studies using transientforcing for the midcentury. Leung et al. (Leung et al. 2004) found increasedtemperatures of 1°–2.5°C for a midcentury scenario (2040–60), which is about halfthe increase that we find for the end of the century. They also found that coldseason (December–January–February) precipitation decreased by up to 0.9 mmday−1. We found decreases in December precipitation of up to 1 mm day−1 innorthern California and increases in both January and December of up to 4 mmday−1.

Our snow accumulation model results show statistically significant decreases ona monthly basis for the mountain region. These decreases are of the same mag-nitude as those reported by Snyder et al. (Snyder et al. 2002), Knowles and Cayan(Knowles and Cayan 2002), and Leung et al. (Leung et al. 2004). Since precipi-tation is changing very little on a monthly basis, the total quantity of water (rainand snow) input into the system is not changing. However, the increased tempera-tures cause less of the precipitation to fall as snow at higher elevations, and thisproduces increased winter runoff and decreased spring and summer runoff. Undersuch conditions, the anthropogenic water storage and delivery system in this region

Table 2. Same as in Table 1, but for precipitation.

MonthNorthern

median differenceMountain

median differenceSouthern

median difference

Jan 1.568 1.189 0.194Feb 0.860 0.957 −0.870Mar −0.909 −0.259 −0.171Apr 0.305 −0.151 0.116May −0.679 −1.226 −0.163Jun −0.040 0.060 −0.019Jul −0.020 −0.118 0.005Aug 0.015 0.026 0.005Sep −0.066 −0.199 0.009Oct −0.067 0.392 0.378Nov −0.429 −0.733 −0.311Dec −1.448 0.884 0.073


Figure 8. Monthly mean, maximum, and minimum snow accumulation by regionfor the Future and Modern cases. Units are in mm snow water equivalent.

Table 3. Monthly median snow accumulation difference (Future − Modern) andsignificance for the mountain region (mm snow water equivalent). Values in boldindicate significance at the 95% confidence interval based on the Mann–WhitneyU test. Values shown are the differences of the medians of the cases.

MonthMountain

Median difference

Jan −37.315Feb −73.44Mar −81.0285Apr −61.084May −14.7975Jun −0.0183Jul N/AAug N/ASep N/AOct N/ANov −2.22245Dec −21.1655


would need to adapt to the new conditions of increased winter runoff from higherelevations and decreased runoff in the spring and summer months.

The comparison of the RCM and the GCM results for the Modern case showedsome similarities and differences between the models. While both the RCM andGCM underestimated temperature compared to observations, the RCM reproducedthe spatial pattern better than the GCM. For precipitation, the RCM generallyoverestimated monthly values but reproduced the spatial pattern, while the GCMunderestimated them and was poorly matched spatially with the observational data.The more detailed topography of the RCM is clearly contributing to the betterspatial match with the observations.

In general, the choice of a particular GCM climate change scenario that is usedto drive the RCM plays the greatest role in differences between RCM simulations(i.e., rate of CO2 increase varies across scenarios). In addition, differences in thedriving GCM used, model physics (both RCM and GCM), domain size and loca-tion, and RCM horizontal resolution are most likely contributing to differencesbetween the simulations.

By examining the monthly distributions of temperature, precipitation, and snowaccumulation (Figures 6–8) by region, we can learn something about the prob-abilities of future climate. For example, with temperature in every month, we findthat the maximum, minimum, and mean increased in the Future case relative to theModern. We can view the increase as a shift in the probability distribution oftemperature for the Modern climate toward warmer temperatures overall. TheFuture distribution for temperature indicates a very small probability that monthlymean temperature in the Future will be the same as the Modern. For monthlyprecipitation, the Modern and Future distributions are very similar, meaning it isprobable that precipitation in the Future climate will be indistinguishable from theModern climate. Future snow accumulation values are much less than those in theModern. The probability of large decreases in snow accumulation in the Futureclimate is therefore very high.

The monthly precipitation values from both cases were found to be statisticallysimilar. Therefore, any changes due to CO2 forcing are indistinguishable from thenatural variability of precipitation. Snyder et al. (Snyder et al. 2002) found in-creased precipitation in northern California and decreases in southern Californiafor a future climate scenario, and that the differences in precipitation between a560-ppm CO2 case and a 280-ppm CO2 case were not statistically significant.Based on the results in this study, the year-to-year variability of precipitation isgreater than any changes that occur because of increased CO2 forcing over the20-yr time period that we examined. Given that precipitation in this region ismodulated by oscillations in the climate system over long time scales [i.e., the El

Table 4. Trend in annual average values of temperature, precipitation, and snowaccumulation.

Modern Future

Temperature (°C yr−1) 0.019 0.011Precipitation (mm day−1) yr−1 0.006 0.018Snow accumulation (mm yr−1) 0.13 0.037


Niño–Southern Oscillation (ENSO) and the Pacific decadal oscillation (PDO)],climate model runs of greater duration may be needed to determine the actualeffect of climate change on precipitation. Several periods of these oscillations arelikely needed to get a complete result. Additionally, longer runs would be neededto assess whether changes in the frequency of these oscillations are occurring as aresult of greenhouse gas forcing and its feedbacks. However, it is also possible thatthe amount of increase in CO2 concentration in this study may not produce sta-tistically significant changes in precipitation in this region for any length of ex-periment. The frequency, intensity, and spatial patterns of natural variability arenot adequately represented in both global and regional climate models to the pointwhere we can say that there is no substantial difference between the models andobservations. The issue of natural variability in these models needs further con-sideration given the importance of variability to the climate of many regions,including western North America.

Based on the temperature results, we find a greater sensitivity of climate athigher elevations to increased atmospheric CO2 concentrations. Previous modelingwork using an RCM with doubled modern-day CO2 found a strong relation be-tween elevation and temperature in the European Alps (Giorgi et al. 1997). In thatregion, they found that the greatest increases in temperature occurred at higherelevations. They concluded that the increases are due primarily to decreased snowaccumulation resulting in more absorbed solar radiation at the surface. The moun-tain region, the region with the highest elevations in our domain, had the largesttemperature increases on a monthly basis (Table 1). We also found that the de-crease in the amount of snow on the ground in March through May and also inDecember, in the Future case relative to the Modern, results in an increase of about20 W m−2 (10 W m−2 in December) in mean absorbed solar radiation in each ofthose months (not shown). It is likely that the increased absorbed solar flux iscontributing to the greater increase in temperature in the mountain region relativeto the other regions.

The trends calculated for temperature, precipitation, and snow accumulation area unique result that we can examine only in studies using transient forcing. Theyear-to-year increase in greenhouse gases in the driving data allows us to assessthe response of the climate system to a more realistic forcing. The positive tem-perature trend from the Modern case is consistent with published results fromobservational data (Jones et al. 2000). Other modeling studies have shown theexistence of these upward trends as well. The small positive trends in precipitationmay suggest that the atmosphere is indeed storing more moisture and therebyincreasing the potential for more precipitation, as has been observed in the twen-tieth century (New et al. 2000). The positive trends in snow accumulation could bedue to increased snow accumulation at elevations that are still above the freezingline in both the Modern and Future cases. This has been observed in the westernUnited States in the present day, where, despite warming temperatures, snowaccumulation has actually increased (Mote 2003).

In forcing the RCM with output from an AOGCM, we are including importantfeedbacks from the global climate system. The interactions between the dynamicocean, sea ice, land surface, and atmospheric models can generate nonlinear feed-backs that may not be captured by using output from a GCM lacking these othermodels (ocean and sea ice models). The RCM does not include a dynamic ocean


model, and therefore dynamic feedbacks between the ocean and the atmosphere atthe regional scale are missing.

5. ConclusionsModel results from an RCM indicate statistically significant increases in monthlyaverage temperature and decreases in snow accumulation due to increased atmo-spheric CO2 concentrations for California. Precipitation differences between thecases were determined to be statistically indistinguishable. Given no net increasein precipitation, decreases in snow accumulation due to increasing temperaturewill likely lead to a significant alteration of the current spatial and temporal patternof runoff in California. This research lends further support to other studies (de-scribed above) that have reached the same conclusions.

We also found evidence of an elevation signal in the temperature response thatis likely due to snow–albedo feedbacks. Reduction of snowpack, due to CO2-induced warming, leads to increased absorption of solar radiation by the landsurface, leading to further increases in temperature in high elevation areas. Thetransition from snow-covered ground to bare ground in the winter months leads toincreases in the land surface temperature of up to 7°C.

Positive trends in temperature, precipitation, and snow accumulation in theModern case support observations made of these variables in the California region.A comparison of the trends in the Modern and Future cases shows that, while thetrends remain positive in the Future case, the rate of change in the trends decreasesfor temperature and snow accumulation and increases for precipitation. The de-creased rate of change for temperature could be an indication that the contributionof CO2 to warming is reaching a saturation point, where increasing the CO2 furtherwill not lead to the same amount of warming per unit increase in CO2 (Kothavalaet al. 1999). The decreased rate in the positive trend of snow accumulation is acombination of increased moisture in the atmosphere and the warming of high-elevation regions. This leads to increased snow accumulation above the snow line,while at the same time the snow line is at a higher elevation in the Future case,which leads to less area being covered by snow. The increased rate of change forprecipitation may indicate that the atmosphere is holding more moisture and that,as temperature increases, the increase in moisture is nonlinear.

While variability plays a key role in the climate of this region, it is difficult todetermine the effects that future global climate change will have on climate vari-ability. With experiments of only 20 yr in length in this study, we can only hopeto resolve half of one period of the PDO and possibly two or three periods ofENSO. Our runs are not long enough to adequately address the issue for a regionwith such large interannual variability. Future studies will need to employ longerruns, preferably of the same length as fully coupled global climate model runs (100yr or greater) in order to adequately represent the natural variability of this region.In addition, ENSO in the RCM will depend in part on the degree to which ENSOevents are captured in GCMs.

These results provide one estimate of the potential impact of climate change onCalifornia. More work is needed to determine the influence of using differentAOGCMs as driving conditions for the RCM. Further work in examining theeffects of increased spatial resolution in the RCM should also be done. Even at


40-km horizontal resolution, the topography is still underrepresented and somesmaller-scale processes may be neglected as a result.

This study shows that using time-varying climate forcing, and given continuallyincreasing greenhouse gases, monthly temperature will increase by substantialamounts by the end of this century with a high degree of certainty. In California,this temperature increase will affect the hydrologic budget of the state by decreas-ing snow accumulation in the winter months, which will lead to a deficit of runoffin the late spring and summer.

Acknowledgments. The authors wish to thank J. Bell, S. Bryant, L. Kueppers, and J.Sewall for helpful comments. We also thank the two anonymous reviewers. The Davidand Lucile Packard Foundation provided funding for this research.

REFERENCESBell, J. L., L. C. Sloan, and M. A. Snyder, 2004: Regional changes in extreme climatic events: A

future climate scenario. J. Climate, 17, 81–87.Boville, B. A., and P. R. Gent, 1998: The NCAR Climate System Model, version one. J. Climate,

11, 1115–1130.California Department of Water Resources, 1998: California water plan update. Bulletin 160-98.Cayan, D. R., 1996: Interannual climate variability and snowpack in the western United States. J.

Climate, 9, 928–948.Christensen, O. B., J. H. Christensen, B. Machenhauer, and M. Botzet, 1998: Very high-resolution

regional climate simulations over Scandinavia—Present climate. J. Climate, 11, 3204–3229.Dai, A., T. M. L. Wigley, B. A. Boville, J. T. Kiehl, and L. E. Buja, 2001: Climates of the twentieth

and twenty-first centuries simulated by the NCAR climate system model. J. Climate, 14,485–519.

Dettinger, M. D., and D. R. Cayan, 1995: Large-scale atmospheric forcing of recent trends towardearly snowmelt runoff in California. J. Climate, 8, 606–623.

Dickinson, R. E., A. Henderson-Sellers, and P. J. Kennedy, 1993: Biosphere–Atmosphere TransferScheme (BATS) Version 1e as coupled to the NCAR Community Climate Model. NCARTech. Note TN-387+STR.

Franco, G., and A. Sanstad, 2004: California Energy Commission Public Interest Energy Research(PIER) Program’s climate change research strategy. Proc. First Annual Climate ChangeConf.: From Climate to Economics: Anticipating Impacts of Climate Change in California,Sacramento, CA, California Energy Commission.

Giorgi, F., and C. Shields, 1999: Tests of precipitation parameterizations available in latest versionof NCAR regional climate model (RegCM) over continental United States. J. Geophys. Res.,104D, 6353–6375.

——, M. R. Marinucci, and G. T. Bates, 1993a: Development of a second-generation regionalclimate model (RegCM2). Part I: Boundary-layer and radiative transfer processes. Mon. Wea.Rev., 121, 2794–2813.

——, ——, ——, and G. Decanio, 1993b: Development of a second-generation regional climatemodel (RegCM2). Part II: Convective processes and assimilation of lateral boundary con-ditions. Mon. Wea. Rev., 121, 2814–2832.

——, C. S. Brodeur, and G. T. Bates, 1994: Regional climate change scenarios over the UnitedStates produced with a nested regional climate model. J. Climate, 7, 375–399.

——, J. W. Hurrell, M. R. Marinucci, and M. Beniston, 1997: Elevation dependency of the surfaceclimate change signal: A model study. J. Climate, 10, 288–296.

——, L. O. Mearns, C. Shields, and L. McDaniel, 1998: Regional nested model simulations of


present day and 2 × CO2 climate over the central plains of the US. Climate Change, 40,457–493.

Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, andC. A. Johnson, Eds., 2001: Climate Change 2001: The Scientific Basis. Cambridge Univer-sity Press, 881 pp.

Johannessen, O. M., E. V. Shalina, and M. W. Miles, 1999: Satellite evidence for an Arctic sea icecover in transformation. Science, 286, 1937–1939.

Jones, P. D., D. E. Parker, T. J. Osborn, and K. R. Briffa, 2000: Global and hemispheric tempera-ture anomalies—Land and marine instrumental records. Trends: A Compendium of Data onGlobal Change, Carbon Dioxide Information Analysis Center, Oak Ridge National Labora-tory, U.S. Department of Energy.

Kim, J., 2001: A nested modeling study of elevation-dependent climate change signals in Californiainduced by increased atmospheric CO2. Geophys. Res. Lett., 28, 2951–2954.

——, T. K. Kim, R. W. Arritt, and N. L. Miller, 2002: Impacts of increased atmospheric CO2 onthe hydroclimate of the western United States. J. Climate, 15, 1926–1942.

Knowles, N., and D. R. Cayan, 2002: Potential effects of global warming on the Sacramento/SanJoaquin watershed and the San Francisco estuary. Geophys. Res. Lett., 29, 1891, doi:10.1029/2001GL014339.

Kothavala, Z., R. J. Oglesby, and B. Saltzman, 1999: Sensitivity of equilibrium surface temperatureof CCM3 to systematic changes in atmospheric CO2. Geophys. Res. Lett., 26, 209–212.

Leung, L. R., and S. J. Ghan, 1999a: Pacific Northwest climate sensitivity simulated by a regionalclimate model driven by a GCM. Part I: Control simulations. J. Climate, 12, 2010–2030.

——, and ——, 1999b: Pacific Northwest climate sensitivity simulated by a regional climatemodel driven by a GCM. Part II: 2 × CO2 simulations. J. Climate, 12, 2031–2053.

——, Y. Qian, X. D. Bian, W. M. Washington, J. G. Han, and J. O. Roads, 2004: Mid-centuryensemble regional climate change scenarios for the western United States. Climate Change,62, 75–113.

Mote, P. W., 2003: Trends in snow water equivalent in the Pacific Northwest and their climaticcauses. Geophys. Res. Lett., 30, 1601, doi:10.1029/2003GL017258.

New, M., M. Hulme, and P. Jones, 2000: Representing twentieth-century space–time climatevariability. Part II: Development of 1901–96 monthly grids of terrestrial surface climate. J.Climate, 13, 2217–2238.

Pan, Z., R. W. Arritt, W. J. Gutowski, and E. S. Takle, 2001: Soil moisture in a regional climatemodel: Simulation and projection. Geophys. Res. Lett., 28, 2947.

Small, E. E., F. Giorgi, and L. C. Sloan, 1999: Regional climate model simulation of precipitationin central Asia: Mean and interannual variability. J. Geophys. Res., 104D, 6563–6582.

Snyder, M. A., J. L. Bell, L. C. Sloan, P. B. Duffy, and B. Govindasamy, 2002: Climate responsesto a doubling of atmospheric carbon dioxide for a climatically vulnerable region. Geophys.Res. Lett., 29, 1514, doi:10.1029/2001GL014431.

——, L. C. Sloan, and J. L. Bell, 2004: Modeled regional climate change in the hydrologic regionsof California: A CO2 sensitivity study. J. Amer. Water Resour. Assoc., 40, 591–601.

Thompson, R. S., S. W. Hostetler, P. J. Bartlein, and K. H. Anderson, 1998: A strategy forassessing potential future changes in climate, hydrology, and vegetation in the westernUnited States, 20 pp.

Earth Interactions is published jointly by the American Meteorological Society, the American GeophysicalUnion, and the Association of American Geographers. Permission to use figures, tables, and brief excerptsfrom this journal in scientific and educational works is hereby granted provided that the source is acknowl-edged. Any use of material in this journal that is determined to be “fair use” under Section 107 or thatsatisfies the conditions specified in Section 108 of the U.S. Copyright Law (17 USC, as revised by P.IL.94-553) does not require the publishers’ permission. For permission for any other form of copying, contactone of the copublishing societies.


Transient Future Climate over the Western United States ...msnyder/papers/snyder2005_ei.pdf · for...

Documents

Transcript of Transient Future Climate over the Western United States ...msnyder/papers/snyder2005_ei.pdf · for...