1. Introduction
The western United States is marked by the diverse conditions of regional climate and geography such as terrain and land cover. The complexity of regional climate is not only induced by the heterogeneity of the underlying surface, but also by the large-scale circulation that exhibits strong seasonal, interannual, and interdecadal variability. The footprint of interactions between atmospheric circulation and orography is apparent in many mesoscale climate patterns. Each climate regime within the western United States has its unique features that challenge climate modeling. Furthermore, each region is vulnerable to different sets of climate and environmental perturbations. Systematic analyses, modeling, and diagnostic efforts are needed to improve the current capability of seasonal and long-term climate predictions for understanding climate sensitivity and managing the region's resources.
In the maritime climate regime of the Pacific Northwest and northern California, water resources are derived predominantly from cold season precipitation and storage in the snowpack along the narrow Cascades and Sierra Range. Spatially, precipitation and temperature, hence snowpack, are modulated by the diverse orographic features and land–sea differences across the region. Temporally, they vary with synoptic systems associated with the large-scale circulation. It has been shown that the latter is strongly affected by the El Niño–Southern Oscillation (ENSO) that has a timescale of 2–7 yr (e.g., Cayan et al. 1999), and the Pacific decadal oscillation (PDO) that changes sign roughly every 20–25 yr (Mantua and Hare 2002). As a result, surface temperature, precipitation, and snowpack all exhibit interannual and interdecadal variations (e.g., Cayan 1996; Dettinger et al. 1998). Their effects are amplified in historical variations of streamflow, fish habitat, and wildfires (Mantua et al. 1997; Mote et al. 1999).
The southwestern United States is not only affected by ENSO variability, but also by the North American monsoon (NAM) system, which exhibits strong interannual variability. Studies of the NAM using more extensive observational networks and numerical models have only begun in the past decade (e.g., Stensrud et al. 1995; Higgins et al. 1997; Berbery 2001). Predicting precipitation is a more serious challenge in the Southwest and the intermountain West because terrain features in those areas have much smaller spatial scales. Furthermore, warm season precipitation, which is a significant source of water supply in the regions, has been more difficult to predict than cold season precipitation because of its strong dependence on convective parameterizations and land–atmosphere interactions.
As part of the Department of Energy's Accelerated Climate Prediction Initiative (Department of Energy 1998) to put in place the computational resources required for climate simulations and to develop the scientific and other infrastructure needed to carry out a full assessment of potential anthropogenic threats, a pilot project has been initiated to study possible impacts of climate change on water resources in the western United States. We have extensively analyzed the hydroclimate conditions of the region using both observations and climate simulations, as well as evaluated climate simulations in reproducing the hydroclimate features. Model evaluation is important because climate models are the most common tools used in climate prediction especially at the decadal or longer timescales. Because of the relatively coarse spatial resolution of global climate models used today, regional climate models are often used to provide regional climate information needed for resource management and impact assessment (e.g., Houghton et al. 2001). As global climate models approach higher spatial resolution in the future, it will also be important to understand the impacts of spatial resolution on climate simulations. In a workshop cosponsored by the National Science Foundation and the Department of Energy, the validity of the different downscaling techniques and research issues for the future were discussed (Leung et al. 2003b). A succinct conclusion of the workshop is that “all downscaling methods have been shown to be valid and more testing should be performed in the end-to-end assessment framework for climate change and seasonal forecasts applications.” If regional climate models are to be useful downscaling tools, it is believed that the most positive impacts should be seen in applications to regions such as the western United States, where mesoscale topographic features play an important role in defining the climatology of the region.
In the past decade, many studies have used regional climate models to study various topics at different geographic regions. The studies by Giorgi et al. (1993), Leung and Ghan (1998, 1999), and Kim (1997), for example, share similar objectives of evaluating the regional climate simulations of the western United States. These studies have demonstrated success in driving limited-area models using lower and lateral boundary conditions from global analyses or global climate models (GCMs) for long-term simulations. The regional models realistically simulated many aspects of the regional climate of the western United States. Furthermore, Leung et al. (1996), Leung and Wigmosta (1999), and Miller and Kim (1996) showed that regional climate simulations can be used to drive hydrologic models to simulate snowpack and streamflow in watersheds in the Pacific Northwest and California. The successful coupling of regional climate and hydrologic models has prompted the use of physically based models for climate impact assessment and possible applications of climate forecasts in water management.
This study aims at providing more extensive analyses of the region's climate, and evaluating a regional climate model based on a longer simulation at spatial resolutions comparable to or higher than previous studies. Our focus is on analyzing mesoscale precipitation characteristics and examining the degree to which they can be represented by a regional climate model. These features include seasonal mean and extreme precipitation, and precipitation variability especially that associated with the ENSO. This paper focuses on means and extremes; Part II (Leung et al. 2003a) will focus on ENSO variability. The datasets and regional climate simulations used in this study are described in section 2. Section 3 discusses seasonal statistics such as seasonal means and extremes, frequency distribution, temperature–precipitation relationships, and topography–precipitation relationships, and seasonal phase and amplitude, based on observation and simulation. Conclusions are provided in section 4.
2. Datasets and regional climate simulations
Two sets of data have been used to analyze the hydroclimate conditions of the western United States. The first dataset consists of daily maximum and minimum surface temperature and precipitation gridded at 1/8° for 1949–2000. This dataset was developed by the Surface Water Modeling Group at the University of Washington for land surface modeling. Gridded data was produced following the methodology outlined by Maurer et al. (2001) based on daily observations made at the National Oceanic and Atmospheric Administration (NOAA) Cooperative Observer (Co-op) stations. The data were gridded and spatially interpolated to 1/8° resolution using the statistical topographic–precipitation relationship developed by Daly et al. (1994), which is important for capturing the mesoscale orographic precipitation pattern that is a prominent feature of the western United States.
The 1/8° dataset covers the conterminous United States and the Columbia River basin north of the U.S.–Canada border. Compared to the 1/2° gridded meteorological dataset of the Climate Research Unit (New et al. 1999) with no mesoscale orographic precipitation adjustment, the 1/8° dataset depicts a more pronounced pattern of orographic precipitation especially over data-sparse areas such as the Great Basin and the intermountain West. The 1/8° dataset has been used to provide meteorological inputs to a macroscale hydrologic model for major river basins in the continental United States to produce realistic simulations of surface fluxes and streamflow (Maurer et al. 2002).
The second dataset consists of temperature and precipitation based on the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalyses for comparison with observations and regional climate simulations. The 2.5° gridded reanalyses data were spatially interpolated to the regional climate model domain for our analyses.
A 20-yr regional climate simulation has been performed using the fifth-generation Pennsylvania State University–NCAR (Penn State–NCAR) Mesoscale Model (MM5; Grell et al. 1993). Various versions of MM5 (e.g., Giorgi et al. 1993; Leung and Ghan 1999) have been used in the past for modeling the regional climate. The model was found to be stable for long-term simulation and to produce results comparable to the large-scale features of the global analyses or climate models used to drive the regional model. The model further produced regional climate features that were comparable to observations at the regional scale (e.g., Leung and Ghan 1998). Our goal in this study is to evaluate the capability of regional climate model in simulating the hydroclimate conditions of the climatically and topographically diverse western United States. Initial condition and lateral and lower boundary conditions of the simulation were derived from the NCEP–NCAR reanalyses. The simulation was initialized on 1 July 1980 with lateral and lower boundary conditions updated every 6 h based on the reanalyses throughout the simulation. These boundary conditions include large-scale temperature, atmospheric moisture, winds, geopotential height, surface pressure, and sea surface temperature, all spatially interpolated horizontally and vertically from the reanalysis grids and vertical levels to that of the MM5.
Simulation was performed using a nested configuration with 23 vertical levels. The larger domain covers the continental United States at 120 km and the nested domain covers the western United States at 40-km resolution. This configuration allows the western United States to be modeled at a higher spatial resolution in a smaller domain for computational efficiency without bringing the lateral boundaries, hence large-scale effects, too close to the region of interest. Figure 1 shows the topography of the fine domain and the six small regions defined for analyses, which are discussed in the sections below. Analyses were performed based on the simulation at the fine domain with 60 by 48 grid cells, from 1 September 1980 to 30 June 2000 after sufficient model spinup time.
The current version (version 3) of MM5 has almost all the capability needed to perform the climate simulation. It includes different options of physics parameterizations such as longwave and shortwave radiation, cloud microphysics, cumulus and shallow convection, turbulence transport, and land surface processes. Lateral and lower boundary conditions can be updated regularly based on large-scale conditions. We performed several short sensitivity runs to guide the selection of physics parameterizations. The 20-yr simulation was performed using the Dudhia shortwave radiation scheme (Stephens et al. 1984), the Rapid Radiative Transfer Model (RRTM) for longwave radiation (Mlawer et al. 1997), the Kain–Fritsch convective parameterization (Kain and Fritsch 1993), the Reisner mixed phase cloud microphysics scheme (Reisner et al. 1998), the countergradient turbulence transport scheme (Hong and Pan 1996), a shallow convection scheme, and the Oregon State University land surface model (Chen and Dudhia 2001).
The simulation was found to be particularly sensitive to the use of the shallow convection scheme that helps to reduce cold season precipitation in the Pacific Northwest to more realistic values. Simulation of warm season precipitation in the southwestern United States is sensitive to convective parameterizations. Among the various schemes tested including the Kuo scheme and the Grell scheme, the Kain–Fritsch scheme consistently produced higher, and therefore more realistic, convective precipitation associated with the North American monsoon in the semiarid Southwest. Results using the version 2 of the Community Climate Model (CCM2) radiation (Briegleb 1992) versus the Dudhia shortwave and the RRTM longwave schemes are similar provided that the CCM2 radiation scheme is modified to use the cloud amounts simulated by the convection and cloud microphysics schemes rather than those prescribed based on temperature and relative humidity as is implemented in MM5 version 3.
Three minor changes were implemented to facilitate long-term integration. First, fractional vegetation cover was updated rather than fixed at the initialized values by using mean monthly vegetation cover derived from the Advanced Very High Resolution Radiometer (AVHRR) data. Second, sea surface temperature in the Gulf of California was set equal to the sea surface temperature of the nearby grid cells in the eastern Pacific Ocean. This avoids forcing the model with reanalyzed surface temperature in the Gulf of California, which is simply part of the continent at 2.5° resolution. Third, different instantaneous or temporally averaged simulation data were archived at 3-hourly, 6-hourly, and daily intervals for two- and three-dimensional fields. These simulation data have been compared against observations and the NCEP–NCAR reanalyses as described below to evaluate the skill of the regional model.
3. Seasonal climate statistics
The major goals of this study are to develop an improved understanding and documentation of the hydroclimate conditions of the western United States, and to evaluate a regional climate simulation based on the 20-yr model integration. We examined the seasonal climate statistics in this paper. In particular, most analyses are focused on precipitation, which is arguably the single most dominant driver of the surface water budget of the region, and how regional climate is strongly affected by the mesoscale surface topography of the region.
a. Seasonal mean precipitation and surface temperature
Figure 2 shows the 20-yr mean warm [June–July–August (JJA)] and cold [December–January–February (DJF)] season precipitation in the observation, the NCEP–NCAR reanalyses, and the regional climate simulation. During the cold season, the observed precipitation at 1/8° resolution shows distinct spatial distributions that resemble the complex orography. The two precipitation bands along the West Coast correspond to orographic precipitation associated with the hilly coastal areas and the mountain ranges near the coast, namely, the Cascades and the Sierra Range. Further inland, precipitation decreases in the basins and the intermountain zone, and increases again as the air mass encounters the Rockies. At a much coarser spatial resolution of 2.5°, the reanalyses precipitation only shows a gradual reduction from the maritime coast to the continental west. The regional simulation follows similar spatial variations of the observation, except that the two precipitation bands along the coast were often merged into a single band, which is more reflective of the orographic signature associated with the Cascades and the Sierra Range. The simulated precipitation is typically larger than that observed in the basins and intermountain West. Precipitation in the northern Rockies was generally well simulated by the regional model.
During the warm season, precipitation is much less pronounced although orographic signatures are still evident. In the Northwest, stronger precipitation is found along the Olympic Mountains, the Cascades, and the northern Rockies, with the latter being the most dominant. In the Southwest, precipitation is found mainly in New Mexico and Arizona in association with the NAM. The spatial distribution of precipitation is accentuated by ridges of the Wasatch Range and the Colorado Plateau. The reanalysis precipitation shows much stronger and spatially extensive precipitation over the northern part of the domain and the Rockies. The major deficiency in both the reanalysis and the regional simulation is the lack of precipitation in the Southwest and the northern Rockies. Over 90% of the simulated precipitation in those areas was derived from convection processes; this suggests that model deficiency may be related to the convective parameterization and spatial resolution used in the simulation.
The seasonal mean surface temperatures based on observations, the NCEP–NCAR reanalyses, and the regional climate simulation are shown in Fig. 3. The spatial distributions of large-scale features are very similar among the three sets of data. Comparing the reanalyses with the regional simulation, the latter clearly resembles the observed temperature more in the Southwest and provides more realistic mesoscale features of both seasons. The simulated temperature during the cold season is very close to the observed, except it is warmer along the coast and mountain ranges and within the basins. The warm bias is generally within 3°C. During the warm season, warm biases are mainly associated with mountain ranges such as the Cascades, Sierra, and the Rockies. The simulation is also slightly too cool in the southwestern United States.
b. Extreme precipitation
A major concern of long-term climate change is a change in the frequency and severity of extreme events such as flood and drought that could result from variability of a more vigorous hydrologic cycle. Indeed a few studies that analyzed climate change simulations produced by GCMs suggest that extremes in daily precipitation increase either regionally or globally (e.g., Zwiers and Kharin 1998). However, there are issues associated with the skill of GCMs in reproducing even the observed mean precipitation, and interpretation of extreme precipitation corresponding to the coarse scales of GCM grid cells. By analyzing extreme precipitation based on the 1/8° observed precipitation and regional simulation, we hope to provide an improved documentation of the observed extreme precipitation and evaluation of how well it can be reproduced in regional models. This will lay a foundation for future analyses of extreme precipitation in an altered climate.
Figure 4 shows the 95th percentile daily precipitation based on observation and regional simulation. A threshold of 0.01 mm day−1 is used to select rain days for the estimation of the 95th percentile. The observed extreme precipitation follows a very similar spatial distribution compared to the seasonal mean precipitation shown in Fig. 2, except it is accentuated more along the Sierra and southern California coast especially during the cold season. This suggests that the frequency distribution of precipitation along the Sierra displays characteristics that are different from most other regions. Along that mountain range, light events occur very often, but the magnitude of heavy precipitation (upper 5%) can reach over 40 mm day−1 which is comparable to that along coastal hills. This feature has been captured very well in the regional simulation. Generally, the 95th percentile precipitation is about 4 times as high as the seasonal mean precipitation except along the Sierra where it is 6–7 times as high. During the warm season, extreme precipitation is strongest in the northern Rockies of Canada. Again the orographic signature is prominent throughout the region. The model was also able to capture areas of stronger extreme precipitation at the higher elevation in the Southwest, although they are lacking in spatial extent and intensity.
c. Precipitation characteristics at six small regions
To further investigate precipitation distribution, Fig. 5 shows the observed and simulated total precipitation distributed as a function of precipitation rate at the six regions shown in Fig. 1. Each region covers an area of 6400 km2 defined by four RCM grid cells. Regions 1 (Cascades) and 4 (Sierra) both exhibit broad frequency distributions of orographic precipitation during the cold season that peak between 10 and 30 mm day−1 but extend beyond 40 mm day−1 in region 4. Warm season precipitation is much smaller and is distributed more narrowly between 0 and 20 mm day−1. These features are well captured by the model, except for the larger wet bias in region 4, particularly for intense precipitation events (more than 20 mm day−1) Located on the leeside of mountains or farther inland, precipitation in regions 2 (Columbia Basin) and 3 (northern Rockies) is typically much lighter compared to regions 1 and 4 during the cold season. The simulated precipitation follows the observed frequency distribution very well during the warm season. However the model produces too much cold season precipitation for intensity higher than 5 mm day−1 in region 2 and between 5 and 10 mm day−1 in region 3. Region 5 (Great Basin) and 6 (Southwest) are marked by light precipitation events that are distributed more evenly between the warm and cold seasons. Most precipitation events produce less than 10 mm day−1 of rain. The model generally underpredicts precipitation in the summer and overpredicts in the winter compared to observations; annual distribution of precipitation is well simulated by the model.
Figure 6 shows a comparison of precipitation characteristics based on observations and simulation at the six small regions, including precipitation amount, intensity, frequency, and inconsistency, which is defined by how often the simulation incorrectly categorized days with and without rain. The simulated mean precipitation is in closer agreement with the observations in regions 1, 3, and 5 than other regions. Precipitation intensity is well simulated in regions 1 and 3, but too high in regions 2 and 4 especially in the cold season. Precipitation frequency, as measured by the number of rain days per month are well simulated in most regions, except the simulated frequency tends to be lower than the observed during summer. In region 6, dry bias in the mean warm season precipitation is related mainly to the much lower precipitation frequency rather than lack of intensity.
The chances of finding mismatch between simulation and observation (i.e., wet vs dry days in the simulation and observation or vice versa) are less than 25% for regions 1, 3, and 4. Results are not very sensitive to the threshold used to define rain days. Synthesizing these findings, the model correctly simulates most precipitation aspects in regions 1 and 3. In regions 2 and 4, although the model correctly simulates precipitation frequency, precipitation amount per event is too high compared to observation especially during winter. The simulated precipitation intensity is too high in region 5, while the problem is related more to precipitation frequency in region 6.
d. Temperature–precipitation relationships
The hydrology of river basins in the western United States strongly depends on both precipitation and surface temperature. This is especially true for snowmelt-dominated basins where precipitation fallen during the cold season is captured largely in the form of snow and water is released as runoff during spring and early summer. To correctly simulate the amount and timing of streamflow, the relationships between surface temperature and precipitation must be realistically represented in the regional simulation. A warm bias during the cold season will result in too much precipitation contributing toward runoff during winter and reduced snowmelt during spring and early summer. Such a shift in the timing of streamflow has serious implications for water management of river basins if results from climate and hydrologic modeling are used.
To assess the regional simulation of temperature–precipitation relationships, Fig. 7 shows comparisons of scatterplots based on observation and simulation of daily mean temperature and precipitation during the cold season. Also plotted in the figure are the seasonal mean surface temperature and precipitation and frequency distribution of surface temperature for the six regions. The simulated surface temperature of regions 1 and 2 are about 2°–3°C too warm compared to observations. Furthermore, the frequency distribution of simulated temperature is more skewed toward the warmer side than that of the observation. As a result, more precipitation events are associated with above-freezing temperature in the simulation than the observation.
The regional simulation is quite close to the observed in regions 3, 5, and 6. Generally the simulated surface temperature is less than 2°C warmer than the observed. The frequency distributions of simulated surface temperature also resemble that of the observations. Precipitation is distributed quite evenly on each side of the mean temperature in regions 5 and 6. In region 3, more precipitation events are associated with warmer than normal temperature in both observations and simulation. For region 4 (Sierra), while more precipitation events are associated with warmer than normal temperature in the observations, the simulation shows a more even distribution of precipitation events above and below normal temperature. As a result, unlike other regions where the simulated runoff peaks occur earlier than that of the observed, runoff simulation in the Sierra area is delayed compared to the observations.
e. Topography–precipitation relationships
To understand the influence of orography on cold season precipitation, we closely studied the relationships between topography and precipitation in the observations and regional simulation. Figure 8 shows the surface topography and precipitation along five east–west-oriented transects that cut across different mountains and valleys. Both the observed precipitation and terrain have been smoothed from 1/8° to a spatial resolution similar to that of the regional model. The transects are 40 km wide, which corresponds to the width of one regional model grid cell.
The first transect near 48°N cuts across the Olympic Mountains and the Cascades before reaching the upper Columbia Basin and the Rockies to the east. In the observation, precipitation increases rapidly on the west side of the Olympic Mountains and the Cascades. The peaks of the precipitation are shifted about 1/2° (30–40 km) to the west of the crests. The rain shadow effect is significant especially in the Cascades where precipitation at the same elevation differs by a factor of 5 or more on the windward and lee sides of the mountain. In the regional model, the elevation of the Olympic Mountains is only about half of the actual elevation because smoothing was applied to preserve numerical stability in the model. The Cascades, on the other hand, have an elevation similar to that shown in the observed. The model correctly captured the double peaks in precipitation associated with the Olympic Mountains and the Cascades. Because of the lower elevation in the model, precipitation simulated at the Olympic Mountains is about two-thirds of the observed value.
Precipitation on the Cascades, including the westward shift in the precipitation peak relative to the crest, is simulated very realistically, although the simulated precipitation is consistently higher than that observed east of the Cascades. The latter hinted at the model deficiency in representing the rain shadow effect in the Cascades. A possible reason for such a deficiency may be related to the use of the sigma coordinate, which can result in numerical errors in advection and diffusion of moisture and hydrometeors that could become serious in the complex terrain of the west. In the simulation, precipitation only differs by a factor of 2–3 on the windward and lee sides of the Cascades. Plotting the precipitation-to-elevation ratio reveals similar relationships in both observations and the model simulation. This suggests that improved skill may be possible with a more precise representation of surface elevation in the model.
The second transect near 46°N begins with a small hill near the Pacific coast and cuts across the narrow Cascades and the Columbia Basin to reach the Palouse and the Rockies in Idaho. Observation shows a significant amount of precipitation associated with the 300-m hill near the coast. Over the Cascades, precipitation again peaks on the west of the ridge and the shift is near 1°. Precipitation reaches a minimum of less than 1 mm day−1 in the Columbia Basin and increases as elevation rises again in the Palouse and the Rockies. The regional simulation captured the general trend except that the model depicts a more gradual ascent from the coast to the Cascades, thereby missing the low-lying hill in the coastal area. As a result, the simulated precipitation shows only one peak near the coast in association with the Cascades. Precipitation in the basin is about twice as high as the observation, but otherwise resembles the observed spatial variations east of the basin.
The third transect near 43°N cuts across southern Oregon through the Cascade Range to the Colorado Plateau and the intermountain West. There is a narrow band of precipitation near the coast in the observation that seems to be related to the large terrain gradient near the coast. The model failed to reproduce the narrow precipitation band because of the smaller topographic gradient represented in the model or simply lack of spatial resolution to resolve the narrowbanded structure. With elevation varying much more smoothly across the Colorado Plateau, small undulations in the observed precipitation forced by smaller ridges are not reproduced by the model.
The fourth transect near 39°N is located in northern California; it cuts across the coastal hill and the steep gradient on the west of the Sierra, the plateau of the Great Basin, and the Wasatch Range. Observation shows equally strong orographic precipitation associated with the coastal hill and the west side of the Sierra. Because the coastal hill peaks at a much lower elevation in the simulation, the precipitation peaks associated with the coastal hill and the Sierra are not as well separated as in the observation. The model, however, produced a precipitation peak with the correct magnitude and at the right location west of the Sierra as in the observed. Beyond the Sierra, the precipitation amount is very low and the simulated precipitation follows a very similar trend except it is somewhat higher.
The fifth transect near 35°N cuts across narrow hills and valleys in the semiarid Southwest. Small peaks of precipitation are found on the west side of the Coast Range, the Sierra, and the Colorado Plateau in the observation. Again, coastal features are not resolved very well in the model at 40-km resolution. However, the model generally captures the small variations in precipitation across the complex terrain. In the Southwest, we also show the warm season precipitation because annual precipitation is more dominated by warm season events. In both simulation and observation, precipitation follows the terrain features with increased precipitation near mountain ridges. The major deficiency in the simulation is the much reduced precipitation on the west side of the transect that may be related to the much smoother topography used in the model.
f. Seasonal amplitude and phase
As shown in the above analyses, surface topography exerts a major influence on the seasonal mean and extreme precipitation in the western United States. Here we examine the patterns of seasonality changes over the various climate regimes as affected by large-scale atmospheric conditions as well as the complex terrain. Fourier analysis was applied to both the observed and simulated daily precipitation to highlight the phase of the 12-month cycle that corresponds to a single precipitation peak during the year. Figure 9 shows the spatial distribution of the amplitude and timing (or phase) of the 12-month cycle precipitation during the year. The amplitude, which is dominated by cold season precipitation over most areas, shows a very similar spatial distribution as in the seasonal mean DJF precipitation.
The mean climate of the west is affected by the Aleutian low and a high pressure cell centered offshore of southern California. These semipermanent pressure systems are responsible for bringing moisture to the western United States during wintertime. A northward shift and weakening of these pressure cells cause a rather abrupt reduction in precipitation during the transition to spring and summer. Therefore, in much of the west, precipitation peaks between November in California and the intermountain West, and December or January in the Northwest. Embedded within the general trend, however, are isolated warm season regimes typically located on the east side of the Cascades and the Sierra along 118°W where precipitation peaks around June and July. The large-scale transition to the warm climate regime occurs on the east side of the northern Rockies. Such a transition is affected by both terrain features, as evident from the spatial variations of the transition zone, and large-scale circulation as moisture is depleted inland following the dominant westerly and southwesterly flows, causing significant reduction in cold season precipitation.
In the Southwest, there is a transition from a cold season regime near the coast to a warm season regime in Utah, Arizona, and New Mexico where precipitation peaks in April–June. This transition is closely related to the NAM system that brings in upper-level moisture from the Gulf of Mexico and lower-level moisture from the Gulf of California and the Pacific Ocean (e.g., Adams and Comrie 1997; Stensrud et al. 1995) in concert with the monsoon wind reversal. However, features of topographic variations are also apparent in the transition region. The correct representations of large-scale circulation and mesoscale topographic influence are therefore both needed to realistically simulate the spatial distribution of seasonal variations in the western United States, which define climatic features important at the regional scale.
Since the transition boundary between cold and warm season regimes is related to large-scale circulation and topography, as evident from Fig. 9, we examine this phenomenon further by showing time–longitude plots of precipitation in Fig. 10 along the five transects described earlier. Along transect 1, precipitation maximizes between November and January on the western half or maritime regime. A localized transition to the warm season regime occurs in the observation near 115°W, which lies on the lee side of a crest (refer to Fig. 8). A large-scale transition from cold to warm season regimes occurs over the Rockies east of 113°W, again on the lee side of a major crest. On the windward slopes, orographic forcing of stratiform precipitation is strong. On the lee side, the rain shadow effect suppresses cold season precipitation. As a result, while precipitation is more uniformly distributed throughout the year on the windward side of the Rockies, the precipitation maximum shifts to the summer months on the lee side.
Similar relationships between the transition locations and mountain crests exist for all the transects shown. For example, localized transition from cold season to warm season regimes occurs near 118° and 116°W on the lee side of the ridges along transects 3 and 4, respectively. The regional simulation captures the large-scale transitions rather well, except it misses some localized transitions because of the coarser spatial resolution and wet biases in cold season precipitation in the basins and intermountain zone that tend to dominate the seasonal cycle.
4. Conclusions
This paper investigated a number of regional climate aspects of the western United States, together with an evaluation of a regional climate simulation at 40-km horizontal resolution using observations. The regional simulation reproduces many mesoscale climate features that are important in the western United States. The overall spatial distributions of temperature and precipitation are well simulated by the regional climate model. Obvious biases in the simulation include oversimulation of precipitation in the basins and intermountain west during the cold season, and undersimulation in the Southwest in the warm season. The regional simulation retains the large-scale features of the reanalyses while adding useful information at the regional scales. The spatial distributions of temperature and precipitation are significantly better represented in the regional simulation than in the reanalyses.
The frequency distributions of precipitation at six local regions are also reasonably well reproduced by the model. Unlike in some GCM simulations (e.g., Risbey and Stone 1996) where precipitation tends to come mainly in the form of drizzle, the regional simulation produces precipitation over a wide range of precipitation rates comparable to observations. Indeed, the 95th percentile precipitation of the simulation resembles that of the observations in both magnitude and spatial distribution, especially during the cold season. There is, however, a tendency for reduced precipitation frequency rather than intensity in the simulation during summer in the Northwest and Southwest, which explains the insufficient summer mean precipitation in those areas. With over 90% of simulated precipitation derived from convection, further investigations of model sensitivity to convective parameterizations and spatial resolutions are warranted in future studies. It is also worth investigating model sensitivity to sea surface temperature in the Gulf of California, which plays an important role in triggering convection. In the current model simulation, sea surface temperature in the Gulf of California is assigned values equal to that of the neighboring eastern Pacific Ocean because of the lack of information.
An important application of regional climate modeling is in coupling with hydrologic models to study the impacts of climate variability and change on water resources. Since surface temperature and precipitation are the two main drivers of hydrologic models, an examination of the two variables as a whole is important. Because of the general warm biases in the simulation, there is a tendency for more precipitation events to be associated with warmer temperatures. In the Northwest, this implies a tendency for more precipitation in the form of rain rather than snow during winter—a bias toward more runoff during the cold season, earlier snowmelt and runoff peak in spring or early summer. In the Sierra where more observed precipitation events tend to be associated with warmer temperatures than that in the simulation, the tendency of the runoff bias is reversed.
The importance of topographic control on regional climate conditions in the western United States is further illustrated through the examination of precipitation–topography relationships and mesoscale features in precipitation seasonality. Analyses of precipitation and topography along five east–west transects show significant impacts of surface terrain on the spatial distribution of precipitation. Rain shadow effects are strong along the Cascades and Sierra Range. The regional simulation was able to reproduce the rain shadow effects rather well. However, at 40-km resolution, major differences exist between the model topography and the actual topography at the same resolution along the coast. As a result, the model failed to generate sufficient precipitation along the coastal hills. Plotting the time evolution of precipitation along the transects shows that local transitions from cold to warm season precipitation regimes mainly occur on the lee side of ridges where cold season orographic precipitation is suppressed due to rain shadow effects. Again this indicates the importance of topographic control on the regional climate features of the western United States.
Acknowledgments
Funding for this study was provided by the Department of Energy (DOE) Accelerated Climate Prediction Initiative (ACPI). The ACPI pilot effort was supported largely by the DOE Office of Biological and Environmental Research through numerous contracts and subcontracts to the participants. All regional climate simulations reported in this study were performed on 64 processors of an IBM-SP3 at the Center for Computational Sciences (CCS), which is supported by the DOE Office of Science, at the Oak Ridge National Laboratory (ORNL). We thank John Drake and his colleagues at ORNL for the excellent computing support. The authors thank Andrew Wood and Dennis Lettenmaier at the University of Washington for providing the 1/8° gridded observation dataset used in this study. Last, but not least, we thank Tim Barnett of the Scripps Institution of Oceanography for providing the leadership needed to coordinate the multidisciplinary ACPI project. The Pacific Northwest National Laboratory is operated for the U.S. Department of Energy by Battelle Memorial Institute under Contract DE-AC06-76RLO 1830.
REFERENCES
Adams, D. K., and A. C. Comrie, 1997: The North American monsoon. Bull. Amer. Meteor. Soc., 78 , 2197–2213.
Berbery, E. H., 2001: Mesoscale moisture analysis of the North American monsoon. J. Climate, 14 , 121–137.
Briegleb, B. P., 1992: Delta–Eddington approximation for solar radiation in the NCAR Community Climate Model. J. Geophys. Res., 97 , 7603–7612.
Cayan, D. R., 1996: Interannual climate variability and snowpack in the western United States. J. Climate, 9 , 928–948.
Cayan, D. R., K. T. Redmond, and L. G. Riddle, 1999: ENSO and hydrologic extremes in the western United States. J. Climate, 12 , 2881–2893.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129 , 569–585.
Daly, C., R. P. Neilson, and D. L. Phillips, 1994: A statistical–topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33 , 140–158.
Department of Energy, 1998: ACPI: The accelerated climate prediction initiative. Pacific Northwest National Laboratory, Department of Energy, Germantown, MD, 30 pp.
Dettinger, M. D., D. R. Cayan, H. F. Diaz, and D. M. Meko, 1998: North–south precipitation patterns in western North America on interannual-to-decadal timescales. J. Climate, 11 , 3095–3111.
Giorgi, F., G. T. Bates, and S. J. Nieman, 1993: The multiyear surface climatology of a regional atmospheric model over the western United States. J. Climate, 6 , 75–95.
Grell, G., J. Dudhia, and D. R. Stauffer, 1993: A description of the fifth generation Penn State/NCAR mesoscale model (MM5). NCAR Tech. Note. NCAR/TN-398+IA, National Center for Atmospheric Research, Boulder, CO, 107 pp.
Higgins, R. W., Y. Yao, and X. L. Wang, 1997: Influence of the North American monsoon system on the U.S. summer precipitation regime. J. Climate, 10 , 2600–2622.
Hong, S-Y., and H-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124 , 2322–2339.
Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, and D. Xiaosu, Eds.,. 2001: Climate Change 2001: The Scientific Basis. Cambridge University Press, 881 pp.
Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization in mesoscale models: The Kain–Fritsch scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 165–170.
Kim, J., 1997: Precipitation and snow budget over the southwestern United States during the 1994–1995 winter season in a mesoscale simulation. Water Resour. Res., 33 , 2831–2839.
Leung, L. R., and S. J. Ghan, 1998: Parameterizing subgrid orographic precipitation and surface cover in climate models. Mon. Wea. Rev., 126 , 3271–3291.
Leung, L. R., and S. J. Ghan, 1999: Pacific Northwest climate sensitivity simulated by a regional climate model Driven by a GCM. Part I: Control simulations. J. Climate, 12 , 2010–2030.
Leung, L. R., and M. S. Wigmosta, 1999: Potential climate change impacts on mountain watersheds in the Pacific Northwest. J. Amer. Water Resour. Assoc., 35 , 1463–1471.
Leung, L. R., M. S. Wigmosta, S. J. Ghan, D. J. Epstein, and L. W. Vail, 1996: Application of a subgrid orographic precipitation/surface hydrology scheme to a mountain watershed. J. Geophys. Res., 101 (D8) 12803–12818.
Leung, L. R., Y. Qian, X. Bian, and A. Hunt, 2003a: Hydroclimate of the western United States based on observations and regional climate simulation of 1981–2000. Part II: Mesoscale ENSO anomalies. J. Climate, 16 , 1912–1928.
Leung, L. R., L. O. Mearns, F. Giorgi, and R. Wilby, 2003b: Regional climate research: Needs and opportunities. Bull. Amer. Meteor. Soc., 84 , 89–95.
Mantua, N. J., and S. R. Hare, 2002: The Pacific Decadal Oscillation. J. Oceanogr., 58 , 35–44.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific Interdecadal Climate Oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78 , 1069–1079.
Maurer, E. P., G. M. O'Donnell, D. P. Lettenmaier, and J. O. Roads, 2001: Evaluation of the land surface water budget in NCEP/NCAR and NCEP/DOE Reanalyses using an off-line hydrologic model. J. Geophys. Res., 106 (D16) 17841–17862.
Maurer, E. P., A. W. Wood, J. C. Adam, D. P. Lettenmaier, and B. Nijssen, 2002: A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States. J. Climate, 15 , 3237–3251.
Miller, N. L., and J. Kim, 1996: Numerical prediction of precipitation and river flow over the Russian River watershed during the January 1995 California storms. Bull. Amer. Meteor., Soc., 77 , 101–106.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102 (D14) 16663–16682.
Mote, P., and Coauthors. 1999: Impacts of climate variability and change, Pacific Northwest. A Regional Report for the USGCRP National Assessment, 109 pp.
New, M., M. Hulme, and P. Jones, 1999: Representing twentieth-century space–time climate variability. Part I: Development of a 1961–90 mean monthly terrestrial climatology. J. Climate, 12 , 829–856.
Reisner, J., R. J. Rasmussen, and R. T. Bruintjes, 1998: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124B , 1071–1107.
Risbey, J. S., and P. H. Stone, 1996: A case study of the adequacy of GCM simulations for input to regional climate change assessment. J. Climate, 9 , 1441–1467.
Stensrud, D. J., R. L. Gall, S. L. Mullen, and K. W. Howard, 1995: Model climatology of the Mexican monsoon. J. Climate, 8 , 1775–1794.
Stephens, G. L., S. Ackerman, and E. A. Smith, 1984: A shortwave parameterization revised to improve cloud absorption. J. Atmos. Sci., 41 , 687–690.
Zwiers, F. W., and V. V. Kharin, 1998: Changes in the extremes of the climate simulated by CCC GCM2 under CO2 doubling. J. Climate, 11 , 2200–2222.
Fine domain of the nested configuration used in the regional climate simulation. Shown are contours of surface elevation (200-m interval) at 40-km resolution and six small regions used in the analyses of hydroclimate conditions of the western United States
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Seasonal mean precipitation during (left column) warm JJA and (right column) cold DJF seasons based on (a), (b) NCEP–NCAR reanalyses, (c), (d) regional simulation, and (e), (f) observations. Shading interval is 1 mm day−1.
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Similar to Fig. 2 but for surface temperature. Shading interval is 3°C
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
The 95th percentile daily precipitation during (left column) warm and (right column) cold seasons based on (upper) regional simulation and (lower) observation. Shading interval is 4 mm day−1.
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Observed and simulated total amount of precipitation distributed as functions of precipitation rate at regions 1–6 shown in Fig. 1. Distributions for warm and cold seasons and annual totals are shown separately. Daily precipitation is binned at an interval of 2 mm day−1.
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Monthly mean precipitation amount (mm day−1), intensity (mm day−1), frequency (day month−1), and percentage inconsistency based on observation and regional simulation. Inconsistency is defined as the percentage of days when the model incorrectly classifies rain vs no-rain days compared to the observation
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
(Continued)
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Scatterplots of daily surface temperature and precipitation based on (left column) observation and (right column) simulation for the cold season for the six regions shown in Fig. 1. Also shown are the mean surface temperature and precipitation (horizontal and vertical lines) and the frequency distribution of daily surface temperature
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
(Continued)
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Precipitation (mm day−1), surface elevation (×102 m), and ratio of precipitation to surface elevation [mm day−1 (100 m)−1] along five east–west transects at various latitudes across mountains and basins of the western United States during the cold season based on (right column) regional simulation and (left column) the 1/8° dataset smoothed to resolution comparable to the regional simulation. Summer precipitation is also shown for the transect along 37.1875°N.
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
(Continued)
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Spatial distribution of the amplitude and phase of the 12-month cycle precipitation indicating the magnitude and timing of maximum precipitation during the annual cycle based on (right column) simulation and (left column) observation. Shading intervals are between 0.5 and 1 mm day−1 for amplitude and 10–30 days for phase. Note that each unit in the phase corresponds to 10 days before (negative) and after (positive) 1 Jan.
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
Time–longitude plots along five transects showing variations in precipitation throughout the year based on (right column) simulation and (left column) observations. Contour intervals are 1–2 mm day−1
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
(Continued)
Citation: Journal of Climate 16, 12; 10.1175/1520-0442(2003)016<1892:HOTWUS>2.0.CO;2
