regional climate modeling: progress, challenges, and...

Journal of the Meteorological Society of Japan, Vol. 82, No. 6, pp. 1599--1628, 2004 1599

Regional Climate Modeling: Progress, Challenges, and Prospects

Yuqing WANG

International Pacific Research Center, School of Ocean and Earth Science and Technology, University ofHawaii at Manoa, Honolulu, Hawaii, USA

L. Ruby LEUNG

Pacific Northwest National Laboratory, Richland, Washington, USA

John L. McGREGOR

CSIRO Atmospheric Research Division, PB1 Aspendale, Victoria, Australia

Dong-Kyou LEE

Atmospheric Sciences Program, School of Earth and Environmental Sciences, Seoul National University,Seoul, Korea

Wei-Chyung WANG

Atmospheric Sciences Research Center, State University of New York at Albany, Albany, New York, USA

Yihui DING

National Climate Center, China Meteorological Administration, Haidian, Beijing, China

and

Fujio KIMURA

Terrestrial Environmental Research Center, University of Tsukuba, Tsukuba, Ibaraki, Japan

Frontier Research System for Global Change, Yokohama, Japan

(Manuscript received 11 September 2003, in final form 15 June 2004)

Abstract

Regional climate modeling using regional climate models (RCMs) has matured over the past decadeto enable meaningful utilization in a broad spectrum of applications. In this paper, the latest progressin regional climate modeling studies is reviewed, including RCM development, applications of RCMs to

Corresponding author: Yuqing Wang, IPRC/SOEST, University of Hawaii at Manoa, 2525Correa Road, Honolulu, HI 96822, USA.E-mail: [email protected]( 2004, Meteorological Society of Japan

dynamical downscaling for climate change assessment and seasonal climate predictions, climate processstudies, and the study of regional climate predictability.

Challenges and potential directions of future research in this important area are discussed, withfocus on those that have received less attention previously, such as the importance of ensemble simu-lations, further development and improvement of the regional climate modeling approach, modelingextreme climate events and sub-daily variation of clouds and precipitation, model evaluation and diag-nostics, applications of RCMs to climate process studies and seasonal predictions, and development ofregional earth system models.

It is believed that with the demonstrated credibility of RCMs in reproducing not only monthly toseasonal mean climate and interannual variability, but also the extreme climate events when driven bygood quality reanalysis and continuous improvements in the skill of global general circulation models(GCMs) in simulating large-scale atmospheric circulation, regional climate modeling will remain an im-portant dynamical downscaling tool for providing the needed information for assessing climate changeimpacts, and seasonal climate predictions, and a powerful tool for improving our understanding ofregional climate processes. Internationally coordinated efforts can be developed to further advance re-gional climate modeling studies. It is also recognized that since the final quality of the results fromnested RCMs depends in part on the realism of the large-scale forcing provided by GCMs, the reductionof errors and improvement in physics parameterizations in both GCMs and RCMs remain a priority forthe climate modeling community.

1. Introduction

Motivated by the need of regional climateinformation to understand regional climatechange and its impacts, regional climate mod-els (RCMs) were first developed mainly as adynamical downscaling tool to address globalchange issues. Since the first successful dem-onstrations of regional climate modeling byDickinson et al. (1989), and Giorgi and Bates(1989), much effort has been devoted to thedevelopment, evaluation, and application ofRCMs. The principle behind regional climatemodeling is that given detailed representationsof physical processes, and high spatial resolu-tion that resolves complex topography, land-seacontrast, and land use, a limited area modelcan generate realistic regional climate infor-mation consistent with the driving large scalecirculation supplied by either global reanalysisdata, or a general circulation model (GCM).

RCMs have been used not only to dynami-cally downscale GCM climate change simula-tions, but also seasonal climate predictionswith similar goals of obtaining useful regionalclimate information. As such, RCMs have be-come a critical component in the end-to-endassessment, or prediction system, where RCMbridges the spatial gaps between the GCM andother modeling components such as hydro-logical models that require regional climate in-formation (e.g., Miller and Kim 1996; Leung

et al. 1996). Regional climate modeling hasbeen demonstrated to be able to improve simu-lation at the regional scales, especially in re-gions where forcing due to complex orography,or coastlines, regulates the spatial distributionof climate variables (e.g., Giorgi 1990; Joneset al. 1995; Walsh and McGregor 1995; Wanget al. 2000; Wang et al. 2003). The regional cli-mate modeling approach has also been shownto be useful for improving our understandingof climate processes, such as cloud-radiationforcing, cumulus convection, and land surfaceprocesses (e.g., Pan et al. 1995; Paegle et al.1996; Dudek et al. 1996; Bosilovich and Sun1999; Schar et al. 1999; Barros and Hwu 2002;Sen et al. 2004a,b; Wang et al. 2004a). Earlierreviews on regional climate modeling can befound in Giorgi and Mearns (1991), Giorgi(1995), McGregor (1997), and Giorgi andMearns (1999).

The objective of this paper is to provide atimely comprehensive review of the latestprogress in regional climate modeling. The fol-lowing areas are covered in this review: RCMdevelopment (section 2), applications of RCMsto dynamical downscaling for climate changeassessment and seasonal climate prediction(section 3), application of RCMs to climate pro-cess studies (section 4), and the study of RCMpredictability (section 5). Challenges and po-tential directions of future research in this im-portant area are discussed (section 6), with

Journal of the Meteorological Society of Japan1600 Vol. 82, No. 6

focus on areas where less attention has beengiven previously, such as the importance of en-semble simulations in detecting signals of cli-mate change or climate sensitivity, the appli-cation of RCMs to climate process studies andseasonal climate predictions, modeling extremeclimate events and sub-daily variation of cloudsand precipitation, process-oriented evaluationand diagnostics of model simulations, and thedevelopment of regional earth system models.

2. RCM development

2.1 Dynamical issuesThere are a large number of RCMs currently

in use that incorporate a variety of dynamicaltreatments. Almost all limited-area models aregrid point models, employing a variety of hori-zontal staggerings for the wind components. Anexception is the regional spectral model (RSM),such as the one developed at the National Cen-ter for Environmental Prediction (NCEP) byJuang and Kanamitsu (1994), and the one in-dependently developed at the MeteorologicalResearch Institute (MRI) by Kida et al. (1991).RCMs also employ a variety of time integrationschemes, including the split explicit schemeused in NCAR RegCM2 (Giorgi et al. 1993a)and semi-implicit schemes. A couple of RCMssuch as DARLAM (McGregor and Walsh1994) and CRCM (Caya and Laprise 1999)permit longer time steps by employing semi-Lagrangian integration.

Most RCMs are formulated using the hydro-static primitive equations. A few RCMs, suchas MM5, CRCM, and RAMS (Liston and Pielke2000), include nonhydrostatic terms, whichallow more accurate representation of phenom-ena, such as deep convection and mountainwaves that may produce large vertical motionwhen fine grids are used. However, in the con-text of regional climate modeling, improvementin the simulated climatology from the use ofnonhydrostatic formulations has yet to be dem-onstrated, because most RCMs have been ap-plied at relatively coarse spatial resolutionbetween 30 and 100 km. Some groups haveperformed multi-year simulations at 10–20 kmresolution (e.g., Christensen et al. 1998; Leunget al. 2003a,b). Higher resolution modeling canalso be achieved through model nesting to tele-scopically zoom into finer spatial resolutionwith one or more levels of nesting, and with the

use of nonhydrostatic formulations (Grell et al.2000).

Similarly, vertical resolution has also gradu-ally increased. Most RCMs use some type ofterrain-following pressure coordinates. RAMSand CRCM use terrain-following height coor-dinates, and the ETA model uses a step moun-tain vertical pressure coordinate. So far, in thecontext of regional climate modeling, no studyhas systematically examined the impacts ofdifferent choices of vertical discretization andvertical resolution.

2.2 Limited-area modelsLimited-area models require information at

their lateral boundaries for all meteorologicalvariables, which can be derived from global re-analyses or GCM simulations. This informationis usually incorporated via ‘‘one-way nesting’’,in conjunction with a boundary relaxation zone.Most RCMs follow a procedure described byDavies (1976), with exponentially decreasingweights and larger buffer zone, as advocatedby Giorgi et al. (1993b), to provide a smoothertransition between the prescribed lateralboundary conditions and the regional climatesimulations.

With ‘‘one-way nesting’’ the RCM circula-tion could differ from that of the host GCM.This is possible especially when large domainsare used, or in the tropical regions where theboundary forcing is relatively weak. Severalviewpoints exist as to whether or not this di-vergence between the RCM and GCM climatol-ogy reduces the credibility of regional climatesimulations. On one hand, Jones et al. (1995)believe that the synoptic circulation of the RCMshould not depart far from that of the drivingGCM, and hence conclude that the RCM do-main should be fairly small, such that theimposed large-scale circulation at the lateralboundaries can exert strong control over theregional simulation. On the other hand, someregional climate modeling studies, such asMcGregor (1997), use large domains with thephilosophy that the RCM simulation should beallowed to produce realistic intensities andfrequencies for each type of major synoptic sys-tem, without necessarily reproducing the dailysequence of the host GCM. With this view, onemay argue that RCM is useful not only forproducing regional climate information, but

Y. WANG et al. 1601December 2004

also modifying atmospheric circulation at thespatial scales not well represented by the hostGCM.

In principle, it would be possible to designa ‘‘two-way nested’’ regional modeling system,where the RCM is run simultaneously with thehost GCM, and regularly updates the hostGCM in the RCM region. This appears notto have been done yet; in fact it would be cum-bersome to fully couple a global and regionalmodel in this manner, because they are typi-cally developed using different numerics andphysical parameterizations. Note that similarbenefits to ‘‘two-way nesting’’ can be derivedfrom the use of a variable-resolution GCM, asdescribed in the next section. It should also benoted that variable resolution RCMs, as pro-posed by Qian et al. (1999), also have a goodpotential to ameliorate some of the boundarycondition problems found when the resolutionsof the nested RCM and driving GCM are verydifferent.

Another approach to enforce consistency be-tween the large-scale circulation simulated byan RCM and its host GCM is to employ broad-scale forcing in the RCM, whereby the broad-scale patterns of the RCM are forced to followthose of the host GCM. It was first used byKida et al. (1991), and further advocated by vonStorch et al. (2000), with the use of spectralnudging. This technique implies a modificationof the primitive equations throughout the RCMdomain, and thus represents a rather differentphilosophy of dynamical downscaling from the‘‘one-way nested’’ approach.

RCMs are being increasingly coupled withother component models. Several RCMs arecoupled with elaborate surface hydrologyschemes, which include river-routing (Leunget al. 1996; CRCM). Doscher et al. (2002)coupled their RCM to an ocean model. A fewmodels, such as ARCSyM (Lynch et al. 1995),are also coupled with sea-ice models. Mabuchiet al. (2000) included the CO2 cycle in an RCMcoupled with a biosphere atmosphere interac-tion model.

2.3 Variable-resolution global modelsA recent development of regional climate

modeling has been the application of variable-resolution global models to regional climatesimulations. Variable-resolution global models

have been used in numerical weather predic-tion (NWP) for some years, first in France withthe spectral ARPEGE model (Caian and Geleyn1997) based on the approach of Courtier andGeleyn (1988), then in Canada with the globalenvironmental multiscale (GEM) model (Coteet al. 1998). ARPEGE achieves its variableresolution by applying the Schmidt (1977)transformation to the primitive equations. TheSchmidt transformation has the unique prop-erty that if it is used to stretch a conformal/isotropic grid, the resulting grid is alsoconformal/isotropic. ARPEGE has beenadopted for regional climate modeling over Eu-rope by Deque and Piedlievre (1995), typicallyusing a stretching factor of 3.5. Recently, Dequeand Gibelin (2002) have addressed the concernthat systematic errors in the low-resolutionpart of the globe may contaminate the domainof interest. They compared 10-year simula-tions with prescribed sea surface temperatures(SSTs) for the variable-resolution model,against those for the same model run at uni-formly high resolution over the globe. Goodagreement between the model climatologiesis confirmed over the high resolution region.GEM is formulated on a latitude-longitudegrid, stretching is applied separately to the lat-itudinal and longitudinal directions to achievea region of high resolution. Fox-Rabinowitzet al. (2000) have applied similar stretchingtransformations to the Goddard Earth Observ-ing System GCM to produce their regional cli-mate modeling system.

Another global model that uses the Schmidttransformation to achieve variable resolutionis the Conformal-Cubic Atmospheric Model(CCAM), as described by McGregor and Dix(2001). CCAM has been used in climate changesimulations with a resolution of about 65 kmover Australia, employing a stretching factorof 3.3, for both 140 years and two 30 years(McGregor et al. 2002). CCAM has now re-placed DARLAM for modeling regional climateat CSIRO.

Variable-resolution global models essentiallyrequire no nesting data. For highly stretchedapplications of CCAM down to around 14 km,it has been considered prudent to also use far-field nudging by the ‘‘host’’ GCM winds to en-sure realistic climatology in the coarse partof the domain; the technique then has some


conceptual similarities to conventional nesteddownscaling. An intercomparison of severalvariable-resolution GCMs is currently beingundertaken by the Stretched Grid Model Inter-comparison Project (SGMIP). The intercom-parison is being run with prescribed SSTs andsea-ice for 12 years with a resolution of about50 km over North America.

3. Application to dynamical downscaling

3.1 Dynamical downscalingDynamical downscaling is the process of

deriving regional climate information basedon large-scale climate conditions using high-resolution RCMs. Numerous studies havedemonstrated that when driven by large-scaleconditions such as global analyses, RCMs canrealistically simulate regional climate featuressuch as orographic precipitation (Leung andGhan 1998; Kim et al. 2000; Frei et al. 2003),extreme climate events (Mearns et al. 1995;Kunkel et al. 2002; Wang et al. 2003), seasonaland diurnal variations of precipitation acrossdifferent climate regimes (Dai et al. 1999;Zhang et al. 2003), and regional scale climateanomalies such as that associated with theENSO (Leung et al. 2003a,b). When used as adynamical downscaling tool, RCMs are usuallydriven by large-scale lateral and lower bound-ary conditions provided by GCMs to obtain re-gional climate information that are then usedin impact assessment or resource management.

In dynamical downscaling, an importantquestion is whether the information generatedby this ‘‘extra step’’ (as opposed to directly us-ing GCM simulations), which is computation-ally demanding, really adds value. Since theability of RCMs to reproduce the observed re-gional climate depends strongly on the large-scale circulation that is provided throughthe lower and lateral boundary conditions, thisquestion is closely tied to the accuracy of thelarge-scale boundary conditions. Many studiessuggested that higher spatial resolution GCMsseem to provide more realistic simulationsof large-scale circulation (e.g., Pope and Strat-ton 2002; Duffy et al. 2003). However, theimportance of model evaluation should be em-phasized to help select GCMs that can pro-vide more realistic control simulations over thestudy region, or configure the RCM domain toavoid placing the lateral boundaries over areas

with known GCM biases, such as misplacedlarge scale features (e.g., jet stream and ITCZ),or erroneous moisture transport (e.g., Seth andRojas 2003; Rojas and Seth 2003). These, how-ever, are not always possible, because mostGCMs have yet to successfully reproducemany features such as those associated withthe Asian monsoon (e.g., Sperber and Palmer1996; Kato et al. 2001). Nevertheless, extensivemodel evaluation for both GCMs and RCMsshould be performed to provide guidance onuncertainty and improve their credible use inclimate research.

Even with improved boundary conditions, theskill and value of dynamical downscaling alsodepend strongly on the presence and strengthof the regional scale forcings. These forcingsmay include orography, land-sea contrast, veg-etation cover, lake effects, or they may be an-thropogenic such as air pollution, urban heatisland, and land and water management. Be-cause regional forcings may extend regionalclimate predictability, more skillful dynamicaldownscaling is often reported in studies of thewestern U.S., Europe, and New Zealand, wheretopographic effects on temperature and precipi-tation are prominent. On the contrary, regionalmodeling results are usually less realistic andmore sensitive to model configurations in re-gions such as the Great Plains in the U.S. andChina, especially during the warm season,where regional forcings are weak, and variousfeedbacks between cloud and radiation, or landand atmosphere, play a more dominant role indetermining regional climate and its sensitivityto greenhouse forcing.

In the following two subsections, we willprovide dynamical downscaling examples inthe areas of climate change and seasonal cli-mate predictions, and examine issues related towhether and how dynamical downscaling mayprovide added value in those applications.

3.2 Climate changeThe use of RCMs in climate change research

has grown rapidly over the last decade as in-dicated by the increasing volumes of literaturecited between the Second and Third Inter-governmental Panel on Climate Change (IPCC)reports (IPCC 1996; IPCC 2001). Because theapplication of RCMs to climate change hasbeen recently reviewed extensively by Giorgi


et al. (IPCC 2001), here only some ongoingissues and recent advancements are brieflydiscussed. Comparing studies that were per-formed in the early 1990s with those of the late1990s to early 2000s, gradual improvementsare evident in several areas: (1) GCMs are nowapplied at higher spatial resolution (approxi-mately 200 km rather than 400 km and morevertical levels) to provide more realistic large-scale boundary conditions for dynamical down-scaling; (2) RCMs are used more often to down-scale transient rather than equilibrium GCMclimate change scenarios to provide more use-ful information for assessing climate changeimpacts; (3) longer duration, higher spatialresolution (e.g., Christensen et al. 1998), andensemble RCM simulations (e.g., Leung et al.2004) are becoming more common to improvesignal detection and realism of the controlsimulations and enable the study of extremeevents; (4) both the strengths and weaknessesof dynamical downscaling are better under-stood through its applications to more diversegeographical regions and model intercompar-ison and diagnostic studies of current and fu-ture climate (e.g., Leung et al. 1999; Takle et al.1999; Christensen et al. 2001; Pan et al. 2001;Chen et al. 2003); and (5) downscaled climatechange scenarios have been used more oftenin impact assessment research (e.g., Thomsonet al. 2002; Bergstrom et al. 2001; Wood et al.2004), which has motivated more diagnosis andevaluation of regional climate simulations, andprovided an important framework for assess-ing the value of the downscaled climate infor-mation.

In the context of climate change, regionalforcings may interact with changes inducedby greenhouse warming, such as changes inlarge-scale circulation or direct radiative ef-fects, to generate more discernible regional cli-mate change signals or substantially alter theclimate change signals generated by the GCMs.This suggests that dynamical downscalingmay provide additional climate information, or‘‘added value’’ for the study of climate changeand its potential impacts. For example, severalRCM studies (e.g., Giorgi et al. 1994; Joneset al. 1997; Leung and Ghan 1999; Whettonet al. 2001; Kim et al. 2002; Synder et al. 2002;Leung et al. 2004) have illustrated how re-gional climate change signals can be signifi-

cantly different from that projected by GCMsbecause of orographic forcing and rainshadow-ing effect. The redistribution of precipitationand hence soil moisture by the complex terrain,and how this redistribution affects climatechange signals, has important implicationsfor climate change detection (Giorgi et al.1997), and assessing climate change impacts inmountains worldwide.

As an example, Fig. 1 shows the ensemblemean snowpack change in 2050 simulated byan RCM driven by the NCAR/DOE ParallelClimate Model (PCM, Leung et al. 2004). ThePCM was initialized in 1995, using assimilatedocean conditions and fixed greenhouse gasesand aerosols concentrations of 1995. The modelwas integrated forward with fixed greenhousegases and aerosols concentrations for the con-trol simulation. In the three ensemble PCMsimulations of the future climate, greenhouse

Fig. 1. Percentage change in mean an-nual snowpack in the western U.S.based on an ensemble of regional cli-mate simulations for the present andfuture climate (2040–2060) as projectedby a global climate model following abusiness as usual scenario. (Adoptedfrom Leung et al. 2004).


gases and aerosols concentrations follow thebusiness as usual scenario (Dai et al. 2004).The RCM was driven by the PCM control sim-ulation of 1995–2015, and each of the three en-semble PCM future simulations of 2040–2060.The model was applied using a nested configu-ration with the larger domain covering the con-tinental U.S. and the surrounding oceans at120 km resolution, and the nested domain cov-ering the western U.S. at 40 km resolution.Large changes in snowpack are found along thecoastal mountain ranges.

Figure 2 shows the temperature change si-mulated by the RCM and PCM. Becausethe global model does not resolve the coastalmountain ranges at the T42 resolution, largerwarming is found over the Rocky Mountains,where snowpack is reduced in the future cli-mate. In the RCM simulations, larger warmingis located along the coastal range where snow-pack reduction is the highest. These resultssuggest that snow-albedo feedback effects are

important, and they can cause an additionalwarming of near 1�C. In this example, topogra-phy is the main regional forcing that drives adiscernible climate change signal. Because ofdifferences in the temperature signals, assess-ment of hydrologic impacts based on PCM andRCM results can lead to different conclusions.Indeed, Wood et al. (2003) examined this issuewith a hydrologic model applied to the Colum-bia River basin in northwestern U.S. and Can-ada. They showed that water resources (e.g.,runoff and snowpack) display much highersensitivity to greenhouse warming when RCMrather than PCM simulations were used asinputs. Bias correction and statistical down-scaling were applied to both the RCM and PCMoutputs before they were used to drive the hy-drologic model, which was applied at 1/8 degree(@10 km) resolution. Interestingly, even afterthese procedures were applied to both the RCMand PCM outputs, rendering their control sim-ulations almost identical to the observed cli-

Fig. 2. Mean winter (December–January–February, DJF) surface temperature change in the west-ern U.S. simulated by a global climate model (the NCAR/DOE Parallel Climate Model, PCM, leftpanel), and a regional climate model based on the Penn State/NCAR MM5 (right panel), driven bythe PCM. Temperature changes were calculated as the difference between the ensemble simulationof the future climate (2040–2060) following a business as usual emission scenario and the controlclimate. (Adopted from Leung et al. 2004).


matology at 1/8 degree resolution, significantdifferences remain in the hydrologic impactswhen the RCM and PCM simulations wereused. In this case, the differences probably havea positive impact because the topographic rep-resentation, therefore the spatial distributionof warming signal, is clearly more realistic inthe RCM than in the PCM.

Besides adding value in climate change as-sessment in regions with strong regional scaleforcing such as topography, dynamical down-scaling can also provide improved simulation ofhigher moment climate statistics (e.g., extremeclimate such as intense precipitation, extremehigh/low temperatures and frost conditions,and diurnal, seasonal, and interannual vari-ability) for the control climate, and hence moreplausible climate change scenarios for extremeevents and climate variability at the regionalscale. A number of recent regional modelingstudies (e.g., Gao et al. 2002; Christensen andChristensen 2003; Leung et al. 2004) show thatboth the extreme summer and winter precipi-tation generally increases in a warmer futureclimate regardless of whether the seasonal/annual mean precipitation is enhanced or re-duced depending on changes in atmosphericmoisture, temperature, and large-scale circula-tion under greenhouse warming. Bell et al.(2004) examined a variety of indices to investi-gate the changes in frequency and intensity ofextreme heat waves and cold spells.

Recently, a large, coordinated effort hasbeen initiated by the Prediction of Regionalscenarios and Uncertainties for Defining Euro-peaN Climate change risks and Effects (PRU-DENCE) Project (http://prudence.dmi.dk /),with participation from over 20 research or-ganizations in Europe. The project aims to ad-dress and reduce deficiencies in projecting fu-ture climate change, quantify the uncertaintiesin predicting future climate and its impacts,and interpret the results for adapting and/ormitigating climate change effects. As part ofthis project, eight RCMs are being driven byboundary conditions from two GCMs to produceensemble simulations of current and future cli-mates for Europe. Through comparison of theRCM simulations driven by different boundaryconditions, an assessment can be made regard-ing the uncertainty and confidence level of thedownscaled climate change projections. When

the RCM simulations are used to drive impactmodels, this type of effort is very useful infurther clarifying the role of dynamical down-scaling in climate change projections and im-pact assessment.

3.3 Seasonal predictionDuring the last decade, significant progress

in dynamical seasonal prediction with globalAGCMs or coupled atmosphere-ocean GCMs (orAOGCMs) has been made by the meteorologicalservices and scientific community around theworld (Goddard et al. 2001). Numerous studieson seasonal predictability have shown thatseasonal climate is potentially predictable ifsignificant shifts, or anomalies in the proba-bilities of different weather regimes that occurover a season are produced when the boundaryconditions that force the atmosphere (e.g., SSTand land surface conditions such as soil mois-ture) are strongly perturbed (Palmer and An-derson 1994; Ji et al. 1998; Kusunoki et al.2001; Palmer et al. 2002). With AGCMs orAOGCMs potentially making skillful seasonalprediction of the large-scale fields, some in-vestigations have recently been undertaken toexamine the use of RCMs nested within theAGCMs or AOGCMs to improve seasonal cli-mate predictions at the regional scale. Leunget al. (2003c) recently suggested that seasonalclimate forecasting may be a useful frameworkfor testing the value of dynamical downscalingbecause unlike climate change projections, cli-mate forecasts can be verified.

The following procedures are usually takenin preparing seasonal prediction with a nestedRCM:

(1). Derivation of ‘‘correct’’ model climatology:In this step, the RCM is driven by observedboundary conditions, such as the ECMWF orNCEP/NCAR reanalyses. Multi-year to multi-decadal simulations must be carried out toprovide meaningful climate statistics and vari-ability, and to identify and possibly reduce sig-nificant systematic errors through sensitivityexperiments.

(2). Hindcasts for a multi-year period: Utiliz-ing the optimal experimental design suggestedfrom step (1), regional climate simulationsmust be performed with boundary conditionsprovided by the host AGCM or AOGCM hind-cast simulations for at least 10 years. Because


errors introduced by the AGCM or AOGCM aretransmittable to the RCM (Noguer et al. 1998),biases of the RCM hindcast simulation aregenerally larger than those driven by observedboundary conditions. To identify regionalbiases, various indices for evaluating predictiveskill may be used, such as the anomaly corre-lation coefficient (ACC), and the root meansquare skill score (RMSSS). One should furthercompare the systematic errors in the regionalsimulation and the driving AGCM or AOGCMsimulation to examine where errors are re-duced due to the use of the RCM.

(3). Develop methods to correct model biases:A system to reduce model systematic errors bystatistical methods may be developed and ap-plied to the seasonal prediction produced by thenested RCM based on analysis of error in thehindcasts. A number of statistical correctionmethods, such as singular value decomposi-tion analysis (SVDA) and canonical correlationanalysis (CCA), (Feddersen et al. 1999; Sperberet al. 2001) can be used to correct biases in theRCM hindcasts. This type of correction systemcan often improve prediction in some specificregions.

(4). Produce ensemble seasonal forecastswith the RCMs: Based on the ensemble sea-sonal predictions developed using the AGCM orAOGCM to provide initial and boundary con-ditions for the RCMs (Sivillo et al. 1997; Deque1997), dynamical seasonal predictions may bemade with the nested RCM and the correctionsystem.

To date, very limited works on the use ofRCMs in seasonal prediction have been re-ported. Cocke and LaRow (2000) used a re-gional spectral model (FSU-RSM) embeddedwithin an AOGCM to make seasonal predic-tions with the purpose of studying ENSO im-pacts on the Southeast United States andwestern North America. For the boreal winterof 1987 and 1988, when a significant El Ninoevent occurred, both the global and regionalmodels captured the precipitation differencebetween the two years, with the regional modelshowing more realistic spatial details. Fen-nessy and Shukla (2000) and Mitchell et al.(2001) also reported some successes of usingthe NCEP ETA model nested within the COLAAGCM to produce hindcast simulations withrealistic regional scale features of climate

anomalies that are comparable to observations.Nobre et al. (2001) performed a seasonal cli-mate forecast over Nordeste Brazil using pre-dicted SST over the tropical oceans, with theNCEP RSM nested in the ECHAM3 AGCM.They found that regional models could predictthe statistics of the weather phenomena dur-ing the rainy season of Nordeste, including theprobability distribution of area-averaged dailyrainfall and the spatial patterns of the fre-quency and duration of dry spells or heavy pre-cipitation periods. Recently Roads et al. (2003a)nested the NCEP RSM in the global spectralmodel to perform seasonal forecasts for thecontinental U.S. and found that the regionalmodel forecasts better depict the precipita-tion intensity. These results have demon-strated that skillful seasonal predictions arepossible using the nested RCMs for some indi-vidual seasons and years. However, no resultsof multi-year or interdecadal hindcasts or real-time seasonal predictions have been reportedso far.

Recently, Ding et al. (2003) presented someinitial results of the experimental use of thenested RCM developed at the China NationalClimate Center (NCC) for seasonal predic-tion since 2001. In their experiments, theNCC_AOGCM provides the boundary and ini-tial conditions for the RegCM_NCC developedbased on the NCAR RegCM2, with horizontalresolution of 60 km. To verify the performanceof the nested RCM, two 10-year (1991–2001)integrations for the summer (June–August)driven by observed and AOGCM simulatedlarge-scale conditions, respectively, were per-formed to produce model climatology andhindcasts. Preliminary results have shown thatRegCM_NCC has some skill in simulating andpredicting the seasonal rain belts, showingmore areas with positive ACC than theAOGCM simulations. The best predicted re-gions with high ACC are located in West China,Northeast China, and North China where theAOGCM also has maximum prediction skill(Fig. 3). One significant improvement derivedfrom RegCM_NCC is the increase of ACC in theYangtze River valley where the AOGCM showsa very low, or even negative ACC. This im-provement is likely related to the more realisticrepresentations of the large-scale terrains inthe regional model. Real-time experimental


predictions for the summers of 2001–2003 us-ing this nested RegCM_NCC were made onApril 1 of each year. As an example, Figure 4shows the predicted and observed precipitationanomaly patterns for the 2002 summer (June–August), which is characterized by severedroughts in North China, and the wet conditionin South China, typical of the mean anomalousprecipitation pattern during the last decade.The predicted precipitation pattern capturedthese major features reasonably well.

The above example indicates the potentialapplications of nested RCMs to operational dy-namical seasonal prediction. More work shouldbe done in the future to improve the represen-tation of physical parameterizations in RCMs,develop longer-range hindcasts, improve theinitialization of soil temperature and moisture,and develop new methods to correct model bias.In addition, seasonal prediction of tropical cy-clones in different ocean basins, including thefrequency and total number of landfalling trop-ical cyclones and preferred paths, is also de-sirable (Ahn and Lee 2002). A regional oceanmodel and oceanic data initialization should bedeveloped and coupled with the nested RCM toimprove SST forecast under highly perturbedweather conditions, when air-sea interactionbecomes important.

4. Application to climate process studies

RCMs have been widely used in climateprocess studies in the past decade. Indeed, theuse of high spatial resolution to resolve thecomplex lower boundary conditions and meso-scale weather systems with the improved rep-resentation of model physics makes RCMs idealfor improving our understanding of climateprocesses. Although this area has been quiteactive in climate research, it has not been em-phasized in any previous reviews. Here we pro-vide some examples to highlight the usefulnessof RCMs in climate process studies.

(a)

(b)

(c)

Fig. 3. The patterns of the 10-yr (1991–2000) mean temporal ACC (anomalycorrelation coefficient) for the simulatedprecipitations in China with NCEPdata as the boundary (a), the pre-dicted precipitation field in China byRegCM_NCC driven by the coupledGCM (T63L16/T63L30) (b), and thepattern of the 19-yr (1982–2000) mean

temporal ACC for the predicted pre-cipitation field in China by the coupledGCM (T63L16/T63L30) (c). The positivecorrelation is represented by shadedareas. Dark shaded areas denote thoseregions exceeding 90% significant level.(Adopted from Ding et al. 2003).


4.1 Land-atmosphere interactionThe land surface can exert a strong forcing

on climate at the global and regional scalesthrough the exchange of heat, moisture, andmomentum (e.g., Pielke and Avissar 1990). Inturn, it is affected by the feedback from theoverlying atmospheric conditions. This land-atmosphere interaction is very complex, and itsaccurate representation in a climate model iscritical to realistic simulation of the global andregional energy and water cycles. Because oftheir high spatial resolution, RCMs can better

resolve land surface heterogeneity and thusbetter represent feedback processes in theland-atmosphere system, although most RCMsadopt land surface models that were originallydeveloped for AGCMs (e.g., Paegle et al. 1996;Giorgi et al. 1996; Bosilovich and Sun 1999;Schar et al. 1999; Barros and Hwu 2002; Paland Eltahir 2001, 2002, 2003).

One extensively investigated land-atmo-sphere feedback process is the positive feed-back between soil moisture and precipitationanomalies. Schar et al. (1999) used an RCMto study soil moisture and precipitation feed-back and found three key feedback processes.(1) Wet soils with small Bowen ratios can leadto the buildup of a relatively shallow boundarylayer, capping the surface heat and moisturefluxes in a comparatively small volume of air,and building up high low-level moist entropy toprovide a source of convective instability. (2)The lowering of the level of free convection withwet soil facilitates the release of convective in-stability. (3) Wet soil decrease thermal emis-sion, increase cloud backscatter, and increasewater vapor greenhouse effect to reduce the netshortwave absorption at the surface, furtherincreasing the moist entropy flux into theboundary layer. These three processes act toincrease the potential for convective activityand thus precipitation. Using the RSM, Hongand Pan (2000) found that soil moisture modu-lates the partitioning of surface sensible andlatent heat fluxes. This partitioning and thedownward mixing of high moist static energywith dry air effectively affect the developmentof boundary layer and the convective availablepotential energy (CAPE). In dry (moist) soilconditions, the CAPE is smaller (larger) be-cause of strong (weak) turbulent mixing. As aresult, moist soil conditions favor convectiveactivity and precipitation in their simulation,which in turn increases the local soil moisture,thus a positive feedback between soil moistureand precipitation.

However, in some other studies, the feed-back between soil moisture and precipitationbecomes negative or quite weak (Paegle et al.1996; Giorgi et al. 1996; Bosilovich and Sun1999; Kanamitsu and Mo 2003). Pan et al.(1995) demonstrated that the sensitivity ofprecipitation to initial soil moisture dependson the model convective parameterization. Seth

Fig. 4. Precipitation anomaly field for the2002 summer (June–August) in China(a) predicted by RegCM_NCC and (b)observed from 160 stations in China.Unit: mm day�1. (Provided by Y. Ding2003).


and Giorgi (1998) reported the effects of domainchoice on their 1993 summer precipitationsimulation, and found that the smaller domaincaptured observed precipitation better, but thesensitivity to initial soil moisture appeared tobe more realistic in the larger domain. Hongand Pan (2000) found, however, weak depen-dency of the response of precipitation to initialsoil moisture on the domain size in the RSM.They suggested that the dependency in Sethand Giorgi (1998) could be related to the de-pendency of the model simulated low-level jeton model domain size due to model systematicerrors. The land-atmosphere interaction is nota local phenomenon; it may be affected signifi-cantly by the prevailing large-scale circulationto induce remote effects (Pal and Eltihir 2002).The use of a relatively large model domain,with good model physics parameterizations,should be encouraged in future assessmentto thoroughly understand the complex land-atmosphere interactions.

4.2 Topographic effects on regional climateAn original application of RCMs is to im-

prove climate simulation in regions where forc-ing due to orography regulates the regional cli-mate (Dickinson 1989; Giorgi and Bates 1989;Semazzi and Sun 1997; Indeje et al. 2001).Many previous studies have focused on theimportant effect of orography on the formationand maintenance of the low-level jets such asthat over the U.S. Southern Great Plains thataffect moisture transports and regional precipi-tation. To improve the simulation of orographicprecipitation that controls the hydrological cy-cle in regions with complex terrain, Leungand Ghan (1995, 1998) developed a subgridorographic precipitation parameterization, andtested it in an RCM for regions with strongsubgrid variations in surface elevation andland cover. Their studies show significant im-pacts of subgrid scale orography on precipita-tion, snowpack, and streamflow in mountain-ous regions.

In a recent study, Xu et al. (2004) inves-tigated the effect of the Andes on the easternPacific climate using the RCM (IPRC-RegCM)developed at the International Pacific ResearchCenter (Wang et al. 2003). In the SouthernHemisphere cold season, the model reproduceskey climatic features including the intertropical

convergence zone (ITCZ) north of the equatorand an extensive low-level cloud deck cappedby a strong temperature inversion to the south(Wang et al. 2004b). In a sensitivity experimentwith the Andes artificially removed, the warmadvection from the South American continentlowers the inversion height, and reduces thelow-level divergence off shore, leading to a sig-nificant reduction in clouds and an increasein solar radiation at the sea surface (Fig. 5). In

Fig. 5. Differences of vertically in-tegrated liquid water content (10�2

mm, upper panel) and downward short-wave radiation flux at the sea surface(W m�2, lower panel) between thecontrol and No-Andes runs, averagedfor August–October 1999. Contour in-terval is 10�2 mm and values greaterthan 10�2 mm are shaded in the upperpanel. Contour interval is 15 W m�2

and values less than �15 W m�2 areshaded in the lower panel. (Adoptedfrom Xu et al. 2004).


the warm season, the model simulates a doubleITCZ in response to the seasonal warming onand south of the equator, in agreement withobservations. Under the same SST forcing, theremoval of the Andes prolongs the existence ofthe southern ITCZ for three weeks, because theintrusion of the easterlies from South Americaenhances the convergence in the lower tropo-sphere over the local warm SST, and the tran-sient disturbances travel freely westward fromthe continent, both favoring deep convectionsouth of the equator (not shown). The same sen-sitivity experiments were repeated with orog-raphy used in the T42 AGCMs; the results con-firm that an under-representation of the Andesreduces the stratus clouds in the cold seasonand prolongs the southern ITCZ in the warmseason, both acting to weaken the latitudinalasymmetry of the eastern Pacific climate.

Yoshikane et al. (2001) investigated the oro-graphic effects of the Tibetan Plateau and land/ocean heat contrast on the Baiu/Meiyu frontusing RAMS. Zonally uniform and temporallyconstant atmospheric fields from the ECMWFanalysis were used as initial and lateralboundary conditions. Their results showed thatthe Baiu/Meiyu front was simulated even whenthe initial and boundary conditions did not in-clude any signal of regional scale disturbances.A low-level jet is formed along the eastern coastof the Asian continent due to the orographiceffects of the Tibetan Plateau and land/oceanheat contrast. The low-level jet transports agreat amount of moisture and forms a precip-itation zone by the interaction with the upper-level jet streak in the middle latitude. Sensi-tivity experiments show that the Baiu/Meiyufront forms mainly due to the zonal mean flowand the land/ocean heat contrast, while theorographic effect intensifies the low-level jetand precipitation over the Baiu/Meiyu front.The mechanisms have some similarity to thoseresponsible for the formation of the SouthPacific convergence zone (SPCZ), although theheat contrast is much more important forthe Baiu/Meiyu front (Yoshikane and Kimura2003).

4.3 Effect of land use change on regionalclimate

Land use changes modify the exchange ofenergy, momentum, moisture, and trace gases,

affecting the earth’s climate (Charney et al.1977). Desertification and deforestation aretwo types of land cover change, which haveexpanded rapidly during the last century, andhave been a major research focus over the lasttwo decades (Nobre et al. 1991; Henderson-Sellers et al. 1993; Polcher and Laval 1994; Xue1996, among others). Deforestation substan-tially increases surface albedo and decreasesurface roughness. Increased albedo reducessurface net radiation, which then leads to areduction in evapotranspiration and precipita-tion. Reduced surface roughness has a similareffect, but through reducing the surface ex-change and drag coefficients. Desertificationmainly increases surface albedo similar to de-forestation and reduces the net radiation atthe surface, and cools the surface. This surfacecooling usually results in sinking motion anddecreases precipitation (Wang and Jenkins2002).

Land use changes are highly inhomogeneousspatially (Pielke 2001). The high-resolution ofRCMs is ideal for assessing the effects of landuse changes at different scales on regional cli-mate (Copeland et al. 1996; Pan et al. 1999; Fu2003; Wang et al. 2003; Sen et al. 2004a,b).Using RAMS, Copeland et al. (1996) assessedthe impact of a natural versus current vegeta-tion distribution on the weather and climate ofJuly 1989, and found coherent regions of sub-stantial changes of both positive and negativesigns in many surface parameters, such assurface air temperature, humidity, winds, andprecipitation throughout the U.S. as a result ofland use change. They concluded that currentland use in the U.S. has caused summertimesurface conditions to be warmer and wetterthan what the natural landscape would indi-cate. Using the NCAR RegCM2, Pan et al.(1999) carried out a similar assessment forthree summer months during normal, drought,and flood years and found that the response ofprecipitation and surface temperature to landuse change has strong interannual variations.

Kanae et al. (2001) found a long-term de-creasing trend in the 40 years precipitationobserved in Thailand, and speculated that thechange might be a response of the regional cli-mate to the deforestation in the Indochina pen-insula. However, the trend is apparent only inthe monthly mean precipitation during Sep-


tember. In order to estimate the effect of thedeforestation, they carried out some sensitivityexperiments using RAMS. Results showed that,consistent with observations, the effect of de-forestation is significant only in September,when the synoptic wind is mild.

In a recent study, Sen et al. (2004a) studiedboth the local and remote effects of the Indo-china deforestation on the East-Asian summermonsoon in the flood year 1998 using the IPRC-RegCM. Since the deforested Indochina penin-

sula is subject to strong monsoonal flow duringsummer months, in addition to the local effect,there is a strong remote effect on downstreammonsoon rainfall over East-Asia (Fig. 6). In an-other study, Sen et al. (2004b) investigated thelocal and regional effects of vegetation restora-tion in northern China (90�–110�E, 36�–42�N)using the same RCM, and evaluated whetherthe changes in rainfall induced by landscapechange are large enough to support a restoredvegetation cover.

Fig. 6. Spatial distribution of the absolute rainfall change (mm/month; shaded) in June, July, andAugust 1998 between ensembles with current and reforested vegetation cover in the Indochinapeninsula (94�–109�E, 9�–19�N). The hatching is for statistically significant areas at 90% confi-dence level. The contours show the orography at the 500, 1000, 1500, 2000, 3000, 4000, and 5000 mheights. (Adopted from Sen et al. 2004a).


Different from other works mentioned above,the studies of Sen et al. (2004a,b) took advan-tage of ensemble simulations to remove thechaotic noise in the simulations, and performeda significance check for the correspondingresponse to the land use change. However,similar to other studies, the sensitivity experi-ments suffer from small sample sizes of severalmonths long in the 1998 flooding year. There-fore, extension to larger samples will be im-portant for assessing the impact of land coverchanges on regional climate in future studies.

4.4 Cloud-aerosol-climate interactionTo study climate and climate change, the

presence of natural and anthropogenic aerosolsin the atmosphere needs to be considered,particularly over regions where aerosol load-ings are substantial, or regions that are locateddownwind (Husar et al. 1997). Aerosols affectclimate directly through scattering and absorp-tion of the solar radiation and, to a less extent,trapping the thermal radiation, thus inducing astrong diurnal variation of the radiative forcingon the surface (Russell et al. 1999). In additionto the direct effect, aerosols affect climate in-directly through changing cloud droplet num-ber concentration (CDNC) and cloud liquidwater path, lifetime and geographical extent.Although the direct effect is relatively known,the indirect effect of cloud-aerosol-climate in-teraction is much less well understood, which isan area in climate modeling that requires fur-ther improvement (IPCC 2001).

Figure 7 illustrates the processes and vari-ables of the cloud-aerosol-climate connection.Climate models need to include the factors thatcontrol the cloud condensation nuclei (CCN)which depends on the size distribution ofwater-soluble species (sulfates, organics, seasalt and nitrates), and the degree of solubilityand the mixing ratio of individual species with-in a given size fraction. Although considerableprogress has been made in recent years toinclude in GCM parameterizations for aerosol-cloud droplet interaction and explicit micro-physics for cloud water/ice content, inadequateunderstanding of the processes, in particularthose affecting the CDNC, contributes signifi-cantly to uncertainties in estimating aerosol ef-fects on climate (IPCC 2001). The individualcomponents in Fig. 7 are currently being devel-

oped, such as the chemical transport model(CTM), to simulate the aerosol mass and sizedistribution, and the warm and cold cloudmodels, to simulate the cloud droplet and icenuclei concentrations (see NACIP 2002). How-ever, a lack of computational efficiency andaccurate parameterizations results in variousinconsistent and disjoint parameterizationscurrently used in GCMs.

Efforts to include the interactive coupling ofthe climate and aerosols were made by Qianand Giorgi (1999), Qian et al. (2003), and Giorgiet al. (2002, 2003) using the NCAR RegCM2and a simple radiatively active sulfate aerosolmodel for climate simulations over East Asia.Both direct and indirect aerosol effects are rep-resented and evaluated. It is found that theaerosol distribution and cycling processes showsubstantial spatial and temporal variability,and that both direct and indirect aerosol forc-ings have regional effects on surface climate,with the indirect effect dominating in inhibitingprecipitation. Because of the use of a one-moment parameterization for cloud micro-physics, which predicts mixing ratios of watervapor, cloud liquid and ice waters, rainwater,snow and graupel, and number concentration ofcloud ice (e.g., Reisner et al. 1998), the size dis-tribution of the CCN is not explicitly simulated;consequently its effect on the cloud radiativeforcing is heavily parameterized, based on anempirical relationship between CDNC and themass concentration of the sulfate aerosol. Inaddition, although cloud water/ice content ispredicted from cloud microphysical scheme inmost RCMs, cloud cover is generally diagnosed.Consequently, the diagnosed cloud cover mightnot be realized in the radiation calculation.Therefore, a consistent cloud-aerosol-climateinteractive scheme is required to simulate thecloud cover, and cloud liquid/ice water pathand number concentration that are the pri-mary parameters for cloud radiative forcingcalculations.

5. Regional climate predictability

5.1 Uncertainties in driving fieldsThe climatology of an RCM is determined

by a dynamical equilibrium between the large-scale forcing provided through the lateralboundary conditions and regional character-istics produced by internal physics and dynam-


ics of the regional model (Giorgi and Mearns1999). As already discussed in section 3.1,biases of a nested RCM may be partly attrib-uted to the inaccuracy of the large-scale drivingfields. Therefore, the skill of an RCM in dy-namical downscaling applications is highly de-pendent upon the skill of the driving GCM.

In general, the results of an RCM are betterwhen it is forced by the reanalysis data thanembedded in a GCM (e.g., Mo et al. 2000; Rojasand Seth 2003). However, substantial differ-ences exist among several reanalysis datasets,in particular, in the lower-atmospheric circu-lation and water vapor flux. For example, di-agnosis of the mean behavior of the Asiansummer monsoon, and its interannual and in-traseasonal variabilities, shows significant dif-

ferences between the ECMWF and NCEP/NCAR reanalysis data in several aspects (An-namalai et al. 1999). Therefore, it is a neces-sary step to evaluate the skill of an RCM usingrealistic large-scale boundary conditions beforeit is nested into a GCM.

Although current generation GCMs haveshown improved skills in simulating thepresent-day climate compared with previousgenerations (IPCC 2001), the performance ofGCMs in reproducing the observed monthly,seasonal, and interannual variabilities variesfrom region to region, and across models. Anumber of RCM experiments clearly show thatspatial patterns produced by RCMs are inbetter agreement with observations than thedriving fields because of the better representa-

Fig. 7. Flow chart showing the processes linking aerosol emissions/production, cloud condensationnuclei (CCN), cloud droplet number concentration (CDNC), ice nuclei (IN), ice particles (IP), opticaldepth (OD), hydrometeor (HC), albedo (A), cloud fraction ð fcÞ, cloud optical depth ðtcÞ and radiativeforcing ðDFÞ. (Adopted from IPCC 2001, Fig. 5.5).


tion of orographic forcing and land surface con-ditions; however, the biases of an RCM arenot necessarily smaller than that of the drivingGCM (Laprise et al. 2000; Achberger et al.2003). In this regard, it is still a priority for theclimate modeling community to reduce modelbias through improving physics representationsin both GCMs and RCMs.

5.2 Uncertainties in the nested RCMGiven perfect driving fields, uncertainties in

the nested RCM still exist and may result fromunphysical treatment of lateral boundary con-ditions, inconsistency in dynamics and physicsbetween the regional model and the large-scalemodel that provides the driving fields, unreal-istic representation of physical parameteriza-tions, and internal flow-dependent instabilitiesof the chaotic climate system at the regionalscales.

To increase the consistency between thelarge-scale driving fields and the RCM solu-tions, and ensure a smooth transition from thedriving fields at the lateral boundary to theregional model interior, most RCMs have em-ployed the relaxation method in a buffer zonenext to the lateral boundaries (see section2.2). Marbaix et al. (2003) provide practicalguidelines for choosing relaxation coefficientsthrough sensitivity experiments. Orographicblending in the buffer zone, proposed by Joneset al. (1995) and Hong and Juang (1998), isuseful and effective in reducing systematicbiases of the regional model influenced by dy-namical uncertainty when steep orography ap-pears in the buffer zone. The big-brother exper-iment (BBE) is a methodology for testing thedownscaling abilities of nested RCMs and canbe used to identify impacts and/or sensitivityof various parameters, such as domain size andlocation, resolution jump, and update frequencyfor minimizing most of the model errors relatedto the one nesting approach (Denis et al. 2001).

RCMs are increasingly using higher resolu-tions (e.g., Christensen et al. 1998; Leung et al.2003a,b). Generally, RCMs produce, to a cer-tain degree, better results with increased spa-tial resolution (e.g., Mo et al. 2000), especiallyfor extreme events (e.g., Christenson et al.1998; Christensen and Christenson 2003; Wanget al. 2003; Zhang et al. 2003). On the otherhand, model resolution can modulate the effects

of both forcings and physical parameterizations(Laprise et al. 1998; Kato et al. 1999). Theincrease of resolution does not necessarily leadto the improved model performance. With in-creasing grid resolution, RCMs become sensi-tive to internal/external forcings representedby model physical parameterizations (Nobreet al. 2001). Therefore, careful selections or im-provements of physical parameterizations, suchas convective parameterization, are necessaryfor meaningful simulations at very high reso-lutions to reduce the uncertainties due to thescale dependency of physical parameterizations(Molinari and Dudek 1992). However, this is achallenging task since the performance of dif-ferent parameterization schemes usually showsregional and seasonal dependencies, and sen-sitivity to a combination of other physical pa-rameterizations in the model (e.g., Giorgi andShields 1999; Gochis et al. 2002).

5.3 Regional model predictabilitySignificant progress has been made in the

development and improvement of the regionalclimate modeling technique, including not onlya number of newly developed RCMs but alsocoupling with various components of the cli-mate system such as ocean, hydrology andchemistry/aerosols. These achievements haveconsiderably increased RCM predictability andcredibility. Driven by reanalysis large-scalefields, present RCMs at about 50 km horizontalresolution have temperature biases generallywithin 2�C, and precipitation biases withinabout 50% of observation, respectively (IPCC2001, p. 601). The predictability of RCMs islimited mainly by uncertainties in drivingfields and those in the nested regional modelparameterizations and chaotic nature of theclimate system.

Since RCMs share many features of limitedarea models (LAMs) used in weather predictionmode, predictability of LAMs is related to thatof RCMs to a certain degree, although thereare fundamental differences. One of the impor-tant topics concerning predictability of LAMs iswhether the regional models generate mean-ingful small-scale features that are absent inthe initial and lateral boundary conditions(Laprise 2000; De Elıa et al. 2002). With a so-called perfect-model approach similar to theBBE, De Elıa et al. (2002) found that LAMs


could recreate the right amplitude of small-scale variability that might be absent in drivingfields, but is incapable of reproducing it withprecision required by a root-mean-square mea-sure of error, implying that RCMs can addvalues to climate statistics rather than to dailyweather events (Denis et al. 2001). Vidale et al.(2003) studied the predictability and uncer-tainty in an RCM using multiyear ensemblesimulations over Europe to assess the abilityof the RCM in representing the natural inter-annual variability on monthly and seasonaltime scales. They showed that the RCM hasskill in reproducing interannual variabilityin precipitation and surface temperature, whilethe predictability varies strongly between sea-sons and regions. In general, the predictabilityis weakest during summer and over continentalregions due to the weak large-scale forcing inthe summer season, and discrepancies in theland surface model.

Another related topic is the study of internalvariability of an RCM, which is related to non-linearity of the model physics and dynamics.Giorgi and Bi (2000) investigated the internalvariability of an RCM using a random pertur-bation that was applied to the initial and lat-eral boundary conditions. They found that theresponse was not sensitive to the origin, loca-tion, and magnitude of the perturbation, butmostly tied to synoptic conditions, seasons andregions. Christensen et al. (2001) performedmultiyear RCM simulations, in which one yearof lateral boundary condition was used as thedriving forcing, while the soil moisture was al-lowed to retain its memory of initial conditions.They found that in spite of the same large-scaleforcing, the climatologies showed substantialdifferences among the simulations, due to themodel internal variability in response to land-atmosphere interactions even after severalyears of simulation. This long-lasing effect ofsoil moisture might have two consequences: onone hand, the long memory of initial conditionmight extend the predictability; on the otherhand, uncertainty in the initial soil moisturemight cause persistent errors, limiting the po-tential predictability.

The predictability of RCMs can be improvedusing ensemble methods with perturbed initialconditions, or model physics (Yang and Arritt2002) or different models such as the super-

ensemble approach proposed by Krishnamurtiet al. (1999). Wandishin et al. (2001) showedthat an ensemble with five members at 80 kmhorizontal resolution could significantly out-perform a single higher-resolution 29-km sim-ulation in precipitation forecasts. The super-ensemble technique can yield forecasts withconsiderable reduction in forecast error com-pared to the errors of the member models, orthe simple ensemble mean (Krishnamurti et al.1999). The super-ensemble technique has po-tential to be adopted to increase the RCM pre-dictability and thus the credibility for RCMs tobe used for climate change assessment.

Currently model-produced climate variabilitygenerally increases with increasing spatial res-olution in both GCMs and RCMs. Regional cli-mate predictability may increase up to a cer-tain regional resolution, or at least similar tolarge-scale predictability (Leung et al. 2003c).It is generally believed that ever-increasingspatial resolution with current RCMs could notsimply increase regional climate predictability,which may strongly depend on how accurateare the representation of clouds and precipita-tion processes in RCMs. Research aiming at notonly understanding natural climate variability,but also reducing systematic errors in bothGCMs and RCMs, will continuously increaseregional climate predictability.

6. Challenges and prospects

6.1 Dynamical downscalingWhen used as a downscaling tool, RCMs are

usually driven by large-scale lateral and lowerboundary conditions provided by GCMs with aone-way nesting approach. By this nesting ap-proach, using similar physics representationsin the RCM and GCM can reduce inconsistencybetween the large-scale conditions simulatedby the two models, and minimize ambiguityin interpreting differences between the RCMand GCM simulations. This approach has beenadopted in the development of some RCMs in-cluding the NCAR RegCM2, which uses physicsparameterizations of the NCAR CommunityClimate Model, the NCEP RSM, which is a re-gional extension of the NCEP global spectralmodel.

However, because current physics parame-terizations may not scale properly over a largerange of spatial scales, one may argue that


sharing the same physics parameterizations,particularly for clouds and convection, may notyield the best downscaling skill. Studies haveshown that the large-scale circulation simu-lated by the RCM and GCM are generally ingood agreement, whether or not the two modelsshare identical physics parameterizations. Thisis particularly true when the RCM is appliedto relatively small domains, where lateralboundary conditions exert major control overthe large-scale features simulated by theRCM. However, differences between the RCMand GCM physics parameterizations, especiallyland surface models and cumulus convectionschemes, can still complicate the interpretationof differences between the GCM simulated anddownscaled climate variables such as surfacetemperature and precipitation, and the cor-responding climate change signals. That is, onecannot easily distinguish whether the differ-ences arise because of regional forcing or dif-ference in physics representations. Arguably,even if the same physics representationsare used in both models, it remains unclearwhether differences between the GCM andRCM simulations are related to regional forc-ings or sensitivity of the physics parameter-izations to the spatial resolution at which theyare applied.

The utility of RCMs in downscaling researchis still being argued. Two issues remain criticalat the center of the debate. First, with allknown and hidden biases in GCM simulations,can RCMs be expected to provide realistic re-gional climate information for global changeresearch or improve seasonal climate predic-tion? That is, does dynamical downscaling re-ally add valuable information or merely addspatial details that are intricately tied to theGCM and RCM model biases that render themuseless? Second, with or without biases inthe GCMs, errors will be introduced in the dy-namical downscaling, because of limitations inthe model physics, numerics, and nesting tech-niques. How should regional simulations, withknown biases, be used to advance global changeresearch and seasonal climate prediction? Al-though this latter issue is not unique to re-gional modeling (GCMs face the same problembecause of inherent model biases), it has beenmore relevant to dynamical downscaling be-cause regional simulations are often used as

inputs by other models for assessing climatechange impacts, or using seasonal predictionfor managing resources. The ability to repro-duce the observed conditions lends credibilityto the end-to-end approach for impact assess-ments or resource management.

The issue of climate change impacts, adapta-tion, and mitigation requires the use of regionalclimate change scenarios. RCMs will remain animportant dynamical downscaling tool for pro-viding the needed information. Since society ismore vulnerable to changes in the frequency orintensity of extreme events (e.g., drought andflood, extreme high/low temperature) ratherthan the mean climate states, future applica-tions of RCMs in climate change study will re-quire demonstration of skill in simulating ex-treme events. The use of ensemble simulationtechnique will be more important to establishthe statistical significance of changes asso-ciated with events that have low probability ofoccurrence. With advances in high performancecomputing, both GCMs and RCMs are beingapplied at increasingly higher spatial resolu-tion. More studies need to be performed tounderstand and document model behaviors athigher spatial resolutions, and address possibleissues that could arise in applying physicsparameterizations beyond the spatial scales in-tended. It is also important that any improve-ment in skill with increasing spatial resolutionbe carefully evaluated to avoid creating a falsesense of advancement. Lastly, current assess-ments of climate change effects are done mostlyusing an offline modeling approach where cli-mate simulations are used to drive processmodels, such as hydrologic models, to estimateclimate change impacts. Because feedback ef-fects are important (as shown in the snowpackexample in section 3.2), coupled regional mod-eling systems are more useful for examiningclimate impacts and assessing adaptation andmitigation strategies. More research is neededin developing these regional modeling systemsthat represent feedback effects at the properspatial and temporal scales.

6.2 RCM development and improvementWith increasing computer power and in-

creased confidence in the applicability of RCMs,several trends are becoming apparent in thenew developments. There is a general move


to even finer resolution, both horizontally andvertically. With the finer resolution, thereis also a trend to further adoption of non-hydrostatic formulations, which could allow in-creasingly accurate simulation of phenomenasuch as tropical cyclones. Another issue relatedto the increase in model resolution is the appli-cability of current physics parameterizations atvery high resolution, as indicated in section 6.1.Even with the resolution currently used inmany regional climate modeling studies, physi-cal parameterizations for subgrid scale pro-cesses need to be improved.

As in most GCMs, the treatment of cloudprocesses, both at grid-resolved and subgridscales, is still a difficult problem in RCMs.In particular, applicability of a particular cu-mulus parameterization scheme may be region-dependent, indicating the potential dependenceof convective activities upon their large-scaleforcing with different triggering mechanisms.This however is not well understood, andtherefore will remain an unsolved problem forclimate modeling studies. While some modelerschoose the best schemes for the interestedregions through comprehensive sensitivity ex-periments, development of new cumulus pa-rameterization schemes that are not stronglyscale/region dependent will remain an impor-tant area where the regional climate modelingcommunity can make significant contributions.

Subgrid-scale stratiform cloud process is alsoan area that needs to be improved (Pal et al.2000). For most RCMs, only grid resolved cloudliquid/ice water content is predicted by the ex-plicit cloud microphysics scheme, while cloudcover and cloud optical properties are generallydiagnosed. As such, inconsistency betweencloud cover and cloud water/ice content usuallyexists, and studies of Leung et al. (1999) andWang et al. (2000) have shown that differenttreatments of clouds and their representationsin the radiative transfer schemes can result inlarge differences in the regional climate simu-lations through various feedback mechanisms.Specifically, the prognostic cloud microphysicalscheme implies grid-scale saturation, while thediagnostic or empirical schemes for cloud coverdo not necessarily require such a condition(Wang et al. 2003). Therefore, introducing cloudfraction as a prognostic variable in current cloudmicrophysics schemes needs to be examined for

RCMs, as was previously investigated forGCMs (e.g., Bechtold et al. 1993; Tiedtke 1993).

The inclusion of cloud-aerosol interactionbecomes possible using the two-moment cloudmicrophysics scheme, which predicts both themixing ratio and total number concentration ofliquid/ice water as prognostic variables. Chenand Liu (2004) recently developed the warmcloud parameterization based on statisticalanalysis of results from a detailed model con-sidering cloud drop activation, drop growth byvapor diffusion, and drop collision, coalescence,and breakup. This scheme responds sensitivelyto the effect of aerosol types on CCN and thetiming of rain initiation. Most importantly, theeffective radii of cloud drops and raindropsare simulated, so that their effects on radiationcan be included more realistically. Precipitationdevelopment is thought to be more efficientin cold-cloud compared to that in warm clouds,particularly for mid-latitude weather systemsthat have a deeper mixed-phase zone. Yet, themodeling of mixed-phase cloud microphysics israther difficult, due to its great complexities.The ice-phase cloud processes, which is toocomplicated to handle with the traditional bulkparameterization scheme, can be resolved viathe physical approach of Chen and Liu (2004),in which two more moments to resolve theshape and density of ice particles should beadded. Nevertheless, significant effort is re-quired before it can be used in RCMs.

6.3 Model evaluation and diagnosticsMuch has been learnt about RCMs through

previous evaluation and diagnostic studies ap-plied to many different regions around theworld. As discussed throughout this review, theimportance of model evaluation and diagnosticscannot be over-emphasized. However, limitedby available observational data and datastorage capacity for model outputs, previousstudies have focused more on evaluating gen-eral aspects of regional climate such as regionalmean precipitation and temperature, and theirseasonal variations and spatial distributions.Few studies have examined the 3-D structuresof the atmosphere, and diagnosed the relation-ships among various variables at differenttemporal and spatial scales to provide morecomplete and process-based understanding ofwhen and why models fail to capture certain


climate features. Model diagnostic approachesthat have been developed by the mesoscalemodeling community can be applied, withemphasis on statistical behaviors, rather thancase by case model-observation comparison.These types of investigations need to be pur-sued more extensively and in depth in the fu-ture for model improvement and better charac-terization of model uncertainties.

Future studies are also required to investi-gate climate aspects that are more region-specific. Examples are orographic precipitation,regional hydrologic cycle, monsoon features,and tropical cyclone and thunderstorm activ-ities. In addition, similar to the fingerprintingmethod used in climate change detection andattribution, model evaluation and diagnosticscan make use of inter-variable relationships orfingerprints in the model outputs and observa-tions to better characterize model biases. Re-gional climate information will likely be used ina larger variety of applications, such as assess-ment or seasonal prediction of air quality andactivity of tropical cyclones. Climate featuresthat are important for realistic simulations ofthese phenomena will need to be more carefullyevaluated at the appropriate temporal andspatial scales. Furthermore, additional insightscan be gained by examining and evaluating theend products of these applications.

6.4 Modeling extreme climate eventsAs indicated in the IPCC 2001 report, previ-

ous studies have shown the capability of RCMsin reproducing monthly to seasonal mean cli-mate and interannual variability, when drivenby good quality driving fields. However, moreanalysis and improvements are needed ofmodel performance in simulating climate vari-ability at daily to sub-daily time scales. Inparticular, the increased resolution of RCMscan allow simulation of a broader spectrumof weather events to improve simulation ofdaily precipitation intensity distributions. Sucha skill is extremely important to give confi-dence of the model-simulated climate sensitiv-ity or climate change scenarios. This is a greatvalue that can be added to both climate model-ing and prediction from the dynamical down-scaling approach.

Although most current RCMs can producemore realistic statistics of heavy precipitation

events than the driving GCMs (e.g., Chris-tensen and Christensen 2003; Frei et al. 2003;Huntinford et al. 2003; Wang et al. 2003), theyare still suffering relatively low skills in re-producing the daily precipitation intensity dis-tributions. This deficiency seems not to be dueto the uncertainties of the driving fields, butmost likely model resolution and physics pa-rameterizations of the RCM. An example is theinability of most climate models in simulatingthe diurnal cycle of precipitation and the parti-tioning of precipitation between convective andstratiform (Dai et al. 1999). Future work thusshould focus on the improvement of modelphysics, so that the daily intensity distributionand diurnal cycle of precipitation can be simu-lated realistically. This is especially importantfor assessment studies of global climate changeimpacts.

6.5 RCM intercomparisonIn the last decade, several RCM intercom-

parison projects have been carried out to iden-tify different or common model strengths andweaknesses. These include Modeling EuropeanRegional Climate Understanding and ReducingErrors (MERCURE) over Europe (Christensenet al. 1997), Project to Intercompare RegionalClimate Simulation (PIRCS) over the UnitedStates (Takle et al. 1999), Regional Model In-tercomparison Project (RMIP) over Asia (Fuet al. 2003), International Research Institute/Applied Research Centers (IRI/ARCs) regionalmodel intercomparison over South America(Roads et al. 2003b), and the Arctic RegionalClimate Model Intercomparison (ARCMIP,http://curry.eas.gatech.edu/ARCMIP). Thesestudies show that synoptic-scale systems aresimulated in good agreement with observationsby the better models, while there exists signifi-cant disagreement among models and from re-gion to region, and season to season.

Although some aspects have been learntfrom different model intercomparison projects,the achievements might not be as originallyexpected. The common weaknesses in the pre-vious intercomparison studies lie in severalaspects. First, only limited regions with verylimited models were involved and some of theparticipating models were not developed in-dependently. Second, most variables comparedwere model outputs generated by complex


physics parameterizations, such as precipita-tion and surface air temperature; less attentionwas given to diagnose the processes leading tothe differences among different models. Third,there have not been any efforts to perform sen-sitivity experiments by replacement of one ormore components from one model to another toexamine the strength and weakness of differentindividual physics parameterization schemes.Finally, few studies compared the surface en-ergy balance, cloud radiation forcing, and diur-nal cycle of simulated clouds and precipitation.Therefore, it is expected that these weaknessesshould be avoided in the future in order tomaximize the results from different model in-tercomparison projects, and to improve RCMsand regional climate modeling studies.

6.6 Climate process studiesApplication of RCMs to climate process

studies is an active area of research in thecommunity. However, there are several limi-tations in previous studies. First, when drivenby reanalysis data or GCM output with the one-way nesting approach, the RCM response toany internal forcing in the model domain doesnot affect large-scale fields significantly, in par-ticular when a relatively small model domainis used. The use of a large integration domainmay mitigate this effect if the RCM can producean accurate large-scale response in the largemodel domain. However, some studies haveindicated that the use of a large domain coulddegrade the skill of the RCM in reproducing thelarge-scale circulation due to model deficien-cies. Further improvements of RCMs, includingvariable-resolution global models, may par-tially reduce such uncertainties. Second, mostprevious studies generalize the finding basedon case studies of particular years and seasons.This is problematic since the large-scale circu-lation varies from year to year and from seasonto season. Future studies need to be performedfor multiple years to examine the ability of themodel in reproducing intraseasonal, inter-annual, interdecadal variabilities. Finally, sin-gle simulation could not distinguish the physi-cal response reliably, because of the chaoticnature of the atmospheric motion. Ensemblesimulations are therefore strongly recom-mended in future sensitivity studies to under-stand regional climate processes.

RCMs will be a very useful tool for under-standing climate processes in the future be-cause of its use of reanalysis data as drivingfields, high spatial resolution and advancedmodel physics representations. In addition, thefeasibility for the next generation GCMs to beapplied at spatial resolution comparable to thatcurrently used in RCM simulations creates anurgent need to develop physical parameter-izations for high spatial resolution simulations.Such new parameterizations are expected to bescaleable for applications at different spatialresolutions. RCMs can be used as test beds forsuch developments, since they can cover rela-tively small domain driven by observed lateralboundary conditions and they can be used totest the suitability of new development in dif-ferent geographical regions. In this sense, theregional climate modeling community can leadthe way in parameterization development forGCMs, as already suggested by Leung et al.(2003d).

6.7 Limited-area high-resolution versusvariable-resolution global modeling

Variable-resolution global models provide anumber of advantages as RCMs. Because sucha model is an atmospheric GCM in its ownright, it provides a straightforward alternativeto the concept of a ‘‘two-way nested’’ RCM. Itcan thus avoid, in a dynamically consistentmanner, the dilemma of whether the daily syn-optic patterns of the RCM and host GCM needto be similar. It also avoids the potential forincompatibilities between the physical parame-terizations of the host and nested models.An obvious advantage of a variable-resolutionglobal model is the greatly reduced amountof data needed from a previous coupled GCMsimulation. A further advantage is that it isfairly easy to impose conservation of massand moisture in a global model, whereas it isextremely difficult to properly achieve in alimited-area RCM.

There are a few disadvantages of variable-resolution global models, as compared tolimited-area RCMs. The traditional concern re-garding variable-resolution climate models iswhether the model parameterizations, in par-ticular those for cumulus convection and cloudcover, can be applied over the range of gridsizes. Although such difficulties have not been


found to occur in practice, this issue needs to beinvestigated systematically in future studies.We believe that both high-resolution RCMs andvariable-resolution global models will continueto be widely used in regional climate studiesand seasonal climate predictions.

6.8 Towards integrated regional earth systemmodeling

Coupled GCMs are increasingly evolving into‘‘Earth Systems Models’’. It is inevitable thatmany RCMs will also follow this path (Giorgi1995). As the validity of the regional climatemodeling approach has been increasingly es-tablished through model evaluation and appli-cations, RCMs are now being used to investi-gate more diverse climate and environmentalchange and prediction issues. When applied atthe appropriate spatial scales, RCMs can moreaccurately represent spatial variations of cli-mate forcings such as topography, lakes, andland-sea contrast, and human influence suchas air pollution and land/water use. Therefore,RCMs can be an integral component of regionalearth system models to study climate change,aerosol effects and air pollution, sea level riseand storm surge, and land, water, crop, andcarbon management strategies.

There has been significant effort in recentyears in developing regional modeling systemswhere RCMs are coupled in core or offline withhydrological models (Leung et al. 1996; Millerand Kim 1996), lake models (Bates et al. 1995),crop models (Thomson et al. 2002), ocean andsea ice models (Lynch et al. 1995; Doscher et al.2002), terrestrial ecosystem models (Mabuchiet al. 2000; Lu et al. 2001), chemistry/aerosolmodels (Qian and Giorgi 1999; Qian et al. 2003;Giorgi et al. 2002, 2003), and air quality models(Grell et al. 2000) with encouraging results.Due to the complexity of various feedbacks,and often mismatch of temporal/spatial scalesamong various parts of the earth system, inte-gration of different model components needsto proceed expeditiously to determine the ap-propriate temporal/spatial scales for integra-tion, and evaluate offline versus online couplingstrategies.

7. Concluding remarks

Regional climate modeling has proven to beable to improve climate simulation at the re-

gional scales, especially in regions where forc-ings due to complex orographic effect, land-seacontrast, and land use, regulate the regionaldistribution of climate variables and variations.The regional climate modeling approach hasalso been shown to be useful for improvingour understanding of various climate processes,such as land-atmosphere interaction, cloud-radiation feedback, topographic forcing, andland use change, as discussed in section 4. Sig-nificant progress has been made in the areaof the application of RCMs to global changeresearch and seasonal climate predictions asdynamical downscaling tools during the lastdecade. Progress has also been made in bothunderstanding and improving the regional cli-mate predictability.

It is generally believed that with both thedemonstrated credibility of RCMs’ capability inreproducing not only monthly to seasonal meanclimate and interannual variability, but alsothe extreme climate events when driven bygood quality reanalysis and the continuous im-provements in the skill of GCMs in simulatinglarge-scale atmospheric circulation, regionalclimate modeling will remain an importantdynamical downscaling tool for providing theneeded information for assessing climatechange impacts and seasonal climate predic-tion, and continue to serve as a powerful toolfor improving our understanding of regionalclimate processes.

Several areas need to be further developed orimproved, such as (1) the inclusion of cloud-aerosol interactions based on higher momentsmixed-phase cloud microphysics parameter-ization so that more accurate cloud-radiation-climate interactions can be modeled with high-resolution RCMs; (2) understanding the modelbehaviors at high-spatial resolutions to addressissues that could arise in applying physics pa-rameterizations, in particular the cumulus con-vective parameterization schemes beyond thespatial scales originally intended; (3) examina-tion of the relationships among various vari-ables at different temporal and spatial scalesto provide more complete and process-basedunderstanding of the model biases and the fail-ure to capture certain climate features; (4) im-provement of model physics parameterizationsso that the daily precipitation intensity distri-bution and diurnal cycle of clouds and precipi-


tation can be simulated realistically; (5) useof ensemble simulations to improve signal de-tection of climate change or climate sensitivity;and (6) development and application of regionalearth system modeling systems, so that thecomplex feedbacks among various climate sys-tem components can be considered consistentlyand interactively. It is our belief that interna-tionally coordinated efforts can be developed toadvance regional climate modeling studies. Fi-nally, since the final quality of the results fromnested RCMs depends in part on the realismof the large-scale forcing provided by GCMs,the reduction of errors and improvement inphysics parameterizations in both GCMs andRCMs remain a priority for the climate model-ing community.

Acknowledgments

The preparation of this review article isstimulated by a group discussion at The 2nd

Workshop on Regional Climate Modeling forMonsoon System held in Yokohama, Japan,March 4–6, 2003, co-sponsored by the FrontierResearch System for Global Change (FRSGC)and GAME International Science Panel (GISP).International Pacific Research Center Contri-bution Number 297 and School of Ocean andEarth Science and Technology ContributionNumber 6477.

References

Achberger, C., M.-L. Linderson, and D. Chen, 2003:Performance of the Rossby Center regionalatmospheric model in southern Sweden: Com-parison of simulated and observed precipita-tion. Theor. Appl. Climatol., 76, 219–234.

Ahn, Y.-I., and D.-K. Lee, 2002: Impact of bogustropical cyclones on summertime circulation inregional climate simulation. J. Geophys. Res.,107 (D16), 4303, doi:10.1029/2001JD000416.

Annamalai, H., J.M. Slingo, K.R. Sperber, and K.Hodges, 1999: The mean evolution and vari-ability of the Asian summer monsoon: Com-parison of ECMWF and NCEP-NCAR re-analyses. Mon. Wea. Rev., 127, 1157–1186.

Barros, A.P., and W. Hwu, 2002: A study of land-atmosphere interactions during summertimerainfall using a mesoscale model. J. Geo-phys. Res., 107 (D14), 4227, doi:10.1029/2000JD000254.

Bates, G.T., S.W. Hostetler, and F. Giorgi, 1995:Two-year simulation of the Great Lakes region

with a coupled modeling system. Mon. Wea.Rev., 123, 1505–1522.

Bechtold, P., J.P. Pinty, and P. Mascart, 1993: Theuse of partial cloudiness in a warm-rain pa-rameterization: A subgrid-scale precipitationscheme. Mon. Wea. Rev., 121, 3301–3311.

Bell, J.L., L.C. Sloan, and M.A. Synder, 2004: Re-gional changes in extreme climatic events: Afuture climate scenario. J. Climate, 17, 81–87.

Bergstrom, S., B. Carlsson, M. Gardelin, G. Lind-strom, A. Pettersson, and M. Rummukainen,2001: Climate change impacts on runoff inSweden—assessments by global climate mod-els, dynamical downscaling and hydrologicalmodeling. Clim. Res., 16, 101–112.

Bosilovich, M.G., and W.-Y. Sun, 1999: Numericalsimulation of the 1993 Midwestern flood: Land-atmosphere interactions. J. Climate, 12, 1490–1505.

Caian, M., and J.F. Geleyn, 1997: Some limits tothe variable-mesh solution and comparisonwith the nested-LAM solution. Quart. J. Roy.Meteor. Soc., 123, 743–766.

Caya, D., and R. Laprise, 1999: A semi-implicit semi-Lagrangian regional climate model: the Cana-dian RCM. Mon. Wea. Rev., 127, 341–362.

Charney, J., W.J. Quirk, S.-H. Chow, and J. Korn-field, 1977: A comparative study of the effectsof albedo change on drought in semi arid re-gions. J. Atmos. Sci., 34, 1366–1385.

Chen, J.-P., and S.-T. Liu, 2004: Physically-basedtwo-moment Bulkwater parameterization forwarm-cloud microphysics. Quart. J. Roy. Me-teor. Soc., 130, 51–78.

Chen, M., D. Pollard, E.J. Barron, 2003: Comparisonof future climate change over North Americasimulated by two regional climate model. J.Geophys. Res., 108 (D12), 4348, doi:10.1029/2002JD002738.

Christensen, J.H., B. Machenhauer, R.G. Iones, C.Schar, P.M. Ruti, M. Castro, and G. Visconti,1997: Validation of present-day regional cli-mate simulations over Europe: LAM simula-tions with observed boundary conditions. Clim.Dyn., 13, 498–506.

Christensen, J.H., J. Raisanen, T. Iversen, D. Bjorge,O.B. Christensen, and M.A. Rummukainen,2001: Synthesis of regional climate changesimulations—A Scandinavian perspective. Ge-ophys. Res. Lett., 28, 1003–1006.

Christensen, J.H., and O.B. Christensen, 2003: Cli-mate Modeling: Severe summertime flooding inEurope. Nature, 421, 805–806.

Christensen, O.B., J.H. Christensen, B. Machen-hauer, and M. Botzet, 1998: Very high-resolu-tion regional climate simulations over Scandi-


navia-Present climate. J. Climate, 11, 3204–3229.

Christensen, O.B., M.A. Gaertner, J.A. Prego, and J.Polcher, 2001: Internal variability of regionalclimate models. Clim. Dyn., 17, 875–887.

Cocke, S., and T.E. LaRow, 2000: Seasonal predic-tions using a regional spectral model em-bedded within a coupled ocean-atmospheremodel. Mon. Wea. Rev., 128, 689–708.

Copeland, J.H., R.A. Pielke, and T.G.F. Kittel, 1996:Potential climatic impacts of vegetationchange: A regional modeling study. J. Geophys.Res., 101 (D3), 7409–7418.

Cote, J., S. Gravel, A. Methot, A. Patoine, M. Roch,and A. Staniforth, 1998: The operational CMC-MRB global environmental multiscale (GEM)model. Part I: design considerations and for-mulation. Mon. Wea. Rev., 126, 1373–1395.

Courtier, P., and J.-F. Geleyn, 1988: A globalnumerical weather prediction model with vari-able resolution: Application to the shallow-water equations. Quart. J. Roy. Meteor. Soc.,114, 1321–1346.

Dai, A., F. Giorgi, and K.E. Trenberth, 1999:Observed and model-simulated diurnal cyclesof precipitation over the contiguous UnitedStates. J. Geophys. Res., 104, 6377–6402.

Dai, A., W.M. Washington, and G.A. Meehl, 2004:The ACPI climate change simulations. Cli-matic Change, (in press).

Davies, H.C., 1976: A lateral boundary formula-tion for multi-level prediction models. Quart.J. Roy. Meteor. Soc., 102, 405–418.

De Elia, R., R. Laprise, and B. Denis, 2002: Fore-casting skill limits of nested, limited-areamodels: a perfect-model approach. Mon. Wea.Rev., 130, 2006–2023.

Denis, B., R. Laprise, D. Caya, and J. Cote, 2001:Downscaling ability of one-way nested regionalclimate models: The big-brother experiments.Clim. Dyn., 18, 627–646.

Deque, M., 1997: Ensemble size for numerical sea-sonal forecasts. Tellus, 49A, 74–86.

Deque, M., and J.P. Piedlievre, 1995: High resolutionclimate simulation over Europe. Clim. Dyn.,11, 321–339.

Deque, M., and A.L. Gibelin, 2002: High versus vari-able resolution in climate modelling. ResearchActivities in Atmospheric and Oceanic Model-ling Report No. 32 (ed. H. Ritchie), WMO/TD-No. 1105, 7.4–7.5.

Dickenson, R.E., R.M. Errico, F. Giorgi, and G.T.Bates, 1989: A regional climate model forwestern United States. Clim. Change, 15, 383–422.

Ding, Y.H., Y.M. Liu, X.L. Shi, and Q.Q. Li, 2003:

The experimental use of the regional climatemodel in the seasonal prediction in China Na-tional Climate Center. Proceedings of the 2nd

Workshop on Regional Climate Model, March3–6, 2003, Yokohama, Japan, 9–14.

Doscher, R., U. Willen, C. Jones, A. Rutgersson,H.E.M. Meier, U. Hansson, and L.P. Graham,2002: The development of the coupled regionalocean-atmosphere model RCAO. Boreal Envi-ron. Res., 7, 183–192.

Dudek, M.P., X.-Z. Liang, and W.-C. Wang, 1996:A regional climate model study of the scaledependence of cloud-radiation interaction. J.Climate, 9, 1221–1234.

Duffy, P.B., B. Govindasamy, J. Milovich, K. Taylor,M. Wehner, A. Lamont, and S. Thompson,2003: High resolution simulations of global cli-mate. Part I: Present climate. Clim. Dyn., 21,371–390.

Feddersen, H., A. Navarra, and M.N. Ward, 1999:Reduction of model systematic error by statis-tical correction for dynamical seasonal predic-tions. J. Climate, 12, 1974–1989.

Fennessy, M.J., and J. Shukla, 2000: Seasonal pre-diction over North America with a regionalmodel nested in a global model. J. Climate, 13,2605–2627.

Fox-Rabinovitz, M.S., G.L. Stenchikov, M.J. Suarez,L.L. Takacs, and R.C. Govindaraju, 2000: Auniform- and variable-resolution stretched-gridGCM dynamical core with realistic orography.Mon. Wea. Rev., 128, 1883–1898.

Frei, C., J.H. Christensen, M. Deque, D. Jacob,R.G. Jones, and P.L. Vidale, 2003: Daily pre-cipitation statistics in regional climate models:Evaluation and intercomparison for the Euro-pean Alps. J. Geophys. Res., 108, 2287–2306.

Fu, C., 2003: Potential impacts of human-inducedland cover change on East Asia monsoon. Glob.Planet. Change, 37, 219–229.

Fu, C., S.Y. Wang, Z. Xiong, W.J. Gutowsky, D.-K.Lee, J.L. McGregor, Y. Sato, H. Kato, J.-W.Kim, and M.-S. Suh, 2004: Regional ClimateModel Intercomparison Project (RMIP) forAsia. Bull. Amer. Meteor. Soc., (submitted).

Gao, X., Z.-C. Zhao, and F. Giorgi, 2002: Changes inextreme events in regional climate simulationsover East Asia. Adv. Atmos. Sci., 19, 927–942.

Giorgi, F., 1990: Simulation of regional climate usinga limited area model nested in general circula-tion model. J. Climate, 3, 941–963.

Giorgi, F., 1995: Perspectives for regional earth sys-tem modeling. Glob. Planet. Change, 10, 23–42.

Giorgi, F., and G.T. Bates, 1989: The climatologicalskill of a regional model over complex terrain.


Mon. Wea. Rev., 117, 2325–2347.Giorgi, F., and L.O. Mearns, 1991: Approaches to

the simulation of regional climate change: areview. Rev. Geophys., 29, 191–216.

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

Giorgi, F., M.R. Marinucci, G.T. Bates, and DeCanio, 1993b: Development of a second gener-ation regional climate model (RegCM2). PartII: Convective processes and assimilation oflateral boundary conditions. Mon. Wea. Rev.,121, 2814–2832.

Giorgi, F., C. Shields Brodeur, and G.T. Bates, 1994:Regional climate change scenarios over theUnited States produced with a nested regionalclimate model. J. Climate, 7, 375–399.

Giorgi, F., L.O. Mearns, C. Shields, and L. Mayer,1996: A regional model study of the importanceof local versus remotely controls of the 1988drought and the 1993 flood over central UnitedStates. J. Climate, 9, 1150–1162.

Giorgi, F., J.W. Hurrell, M.R. Marinucci, and M.Beniston, 1997: Elevation Dependency of theSurface Climate Signal: A Model Study. J. Cli-mate, 10, 288–296.

Giorgi, F., and L.O. Mearns, 1999: Introduction tospecial section: Regional climate modeling re-visited. J. Geophys. Res., 104, 6335–6352.

Giorgi, F., and C. Shields, 1999: Tests of precipita-tion parameterizations available in the latestversion of the NCAR Regional Climate Model(RegCM) over the continental United States. J.Geophys. Res., 104, 6353–6376.

Giorgi, F., X. Bi, 2000: A study of internal variabilityof a regional climate model. J. Geophys. Res.,105, 29503–29521.

Giorgi, F., and co-authors, 2001: Regional climateinformation–Evaluation and projections.Chapter 10 in Climate Change 2001: TheScientific Basis. Cambridge University Press,Houghton et al., (Eds.), 881pp.

Giorgi, F., X. Bi, and Y. Qian, 2002: Direct radiativeforcing and regional climatic effects of anthro-pogenic aerosols over East Asia: A regionalcoupled climate-chemistry/aerosol model study.J. Geophys. Res., 107, 4439, doi:10.1029/2001JD001066.

Giorgi, F., X. Bi, and Y. Qian, 2003: Indirect vs.direct effects of anthropogenic sulfate on theclimate over east Asia as simulated with aregional coupled climate-chemistry/aerosolmodel. Clim. Change, 58, 345–376.

Gochis, D.J., W.J. Shuttleworth, and Z.-L. Yang,

2002: Sensitivity of the modeled North Ameri-can monsoon regional climate to convectiveparameterization. Mon. Wea. Rev., 130, 1282–1298.

Goddard, L., S.J. Mason, S.E. Zebiak, C.F. Rope-lewski, R. Basher, and M.A. Cane, 2001:Current approaches to seasonal-to-interannualclimate predictions. Intl. J. Climatol., 21,1111–1152.

Grell, G.A., L. Schade, R. Knoche, A. Pfeiffer, andJ. Egger, 2000: Nonhydrostatic climate simu-lations of precipitation over complex terrain. J.Geophys. Res., 105 (D24), 29595–29608.

Grell, G.A., S. Emeis, W.R. Stockwell, T. Schoene-meyer, R. Forkel, J. Michalakes, R. Knoche,and W. Seidl, 2000: Application of a multiscale,coupled MM5/chemistry model to the complexterrain of the VOTALP valley campaign. At-mos. Environ., 34, 1435–1453.

Henderson-Sellers, A., R.E. Dickinson, T.B. Dur-bidge, P.J. Kennedy, K. McGuffie, and A.J.Pitman, 1993: Tropical deforestation: Modelinglocal- to regional-scale climate change. J. Geo-phys. Res., 98, 7289–7315.

Hong, S.-Y., and H.-M.H. Juang, 1998: Orographyblending in the lateral boundary of a regionalmodel. Mon. Wea. Rev., 126, 1714–1718.

Hong, S.-Y., and H.-L. Pan, 2000: Impact of soilmoisture anomalies on seasonal, summertimecirculation over North America in a regionalclimate model. J. Geophys. Res., 105, 29625–29634.

Huntingford, C., R.G. Jones, C. Prudhomme, R.Lamb, J.H.C. Gash, and D.A. Jones, 2003:Regional climate model predictions of extremerainfall for a changing climate. Quart. J. Roy.Meteor. Soc., 123, 265–292.

Husar, R.B., J.M. Prospero, and L.L. Stowe, 1997:Characteristization of tropospheric aerosolsover the oceans with the NOAA advanced veryhigh resolution radiometer optical thicknessoperational product. J. Geophys. Res., 102,16889–16909.

Indeje, M., F.H.M. Semazzi, and L. Xie, 2001: Mech-anistic model simulations of the East Africanclimate using NCAR regional climate model:Influence of large-scale orography on the Tur-kana low-level jet. J. Climate, 14, 2710–2724.

IPCC, 1996: Climate Change 1996, The Science ofClimate Change, Cambridge University Press,Houghton et al. (Eds.), 752pp.

IPCC, 2001: Climate Change 2001: The ScientificBasis. Cambridge University Press, Houghtonet al., (Eds.), 881pp.

Ji, M., A. Leetmaa, and V.E. Kousky, 1998: Animproved coupled model for ENSO prediction


and implication for ocean initialization. PartII: The coupled model. Mon. Wea. Rev., 126,1022–1034.

Jones, R.G., J.M. Murphy, and M. Noguer, 1995:Simulation of climate change over Europe us-ing a nested regional-climate model. Part I:Assessment of control climate, including sensi-tivity to location of lateral boundaries. Quart.J. Roy. Meteor. Soc., 121, 1413–1449.

Jones, R.G., J.M. Murphy, M. Noguer, and M. Keen,1997: Simulation of climate change over Eu-rope using a nested regional climate model. I:Comparison of driving and regional model re-sponses to a doubling of carbon dioxide. Quart.J. Roy. Meteor. Soc., 123, 265–292.

Juang, H.-M.H., and M. Kanamitsu, 1994: The NMCnested regional spectral model. Mon. Wea.Rev., 122, 3–23.

Kanae, S., T. Oki, and K. Musiake, 2001: Impact ofdeforestation on regional precipitation over theIndochina peninsula, J. Hydrometeor., 2, 51–70.

Kanamitsu, M., and K.C. Mo, 2003: Dynamical effectof land surface processes on summer precipita-tion over the southwestern United States. J.Climate, 16, 496–509.

Kato, H., H. Hirakuchi, K. Nishizawa, and F. Giorgi,1999: Performance of the NCAR RegCM in thesimulations of June and January climates overEastern Asia and the high-resolution effects ofthe model. J. Geophys. Res., 104, 6455–6476.

Kato, H., K. Nishizawa, H. Hirakuchi, S. Kadokura,N. Oshima, and F. Giorgi, 2001: Performanceof RegCM2.5/NCAR-CSM nested system forthe simulation of climate change in East Asiacaused by global warming. J. Meteor. Soc. Ja-pan, 79, 99–121.

Kida, H., T. Koide, H. Sasaki, and M. Chiba, 1991: Anew approach to coupling a limited area modelwith a GCM for regional climate simulation. J.Meteor. Soc. Japan, 69, 723–728.

Kim, J., N.L. Miller, J.D. Farrara, S.-Y. Hong, 2000:A seasonal precipitation and stream flowhindcast and prediction study in the WesternUnited States during the 1997/98 winter sea-son using a dynamic downscaling system. J.Hydrometeor., 1, 311–329.

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

Krishnamurti, T.N., C.M. Kishtawal, T.E. LaRow,D.R. Bachiochi, Z. Zhang, C.E. Williford, S.Gadgil, and S. Surendran, 1999: Improvedweather and seasonal climate forecasts frommultimodel superensemble. Science, 285,

1548–1550.Kunkel, K.E., K. Andsager, X.-Z. Liang, R.W. Arritt,

E.S. Takle, W.J. Gutowski Jr., Z. Pan, 2002:Observations and regional climate model sim-ulations of heavy precipitation events andseasonal anomalies: A comparison. J. Hydro-meteor., 3, 322–334.

Kusunoki, S., M. Sugi, A. Kitoh, C. Kobayashi, andK. Takano, 2001: Atmospheric seasonal pre-dictability experiments by the JMA AGCM. J.Meteor. Soc. Japan, 79, 1183–1206.

Laprise, R., D. Caya, M. Giguere, G. Bergeron,H. Cote, J.-P. Blanchet, G.T. Boer, and N.A.McFarlane, 1998: Climate and climate changein western Canada as simulated by the Cana-dian regional climate mode. Atmosphere-Ocean, 36, 119–167.

Laprise, R., M.R. Varma, B. Denis, D. Caya, andI. Zawadzki, 2000: Predictability of a NestedLimited-Area Model. Mon. Wea. Rev., 128,4149–4154.

Lee, D.-K., and M.-S. Suh, 2000: Ten-year East Asiansummer monsoon simulation using a regionalclimate model (RegCM2). J. Geophys. Res.,105, 29565–29577.

Leung, L.R., and S.J. Ghan, 1995: A subgrid param-eterization of orographic precipitation. Theor.Appl. Climatol., 52, 95–118.

Leung, L.R., M.S. Wigmosta, S.J. Ghan, D.J.Epstein, and L.W. Vail, 1996: Application ofa subgrid orographic precipitation/surface hy-drology scheme to a mountain watershed. J.Geophys. Res., 101, 12803–12818.

Leung, L.R., and S.J. Ghan, 1998: Parameterizingsubgrid orographic precipitation and surfacecover in climate models. Mon. Wea. Rev., 126,3271–3291.

Leung, L.R., and S.J. Ghan, 1999: Pacific Northwestclimate sensitivity simulated by a regional cli-mate model driven by a GCM. Part II: 2xCO2simulations. J. Climate, 12, 2031–2053.

Leung, L.R., S.J. Ghan, Z.-C. Zhao, Y. Luo, W.-C.Wang, and H. Wei, 1999: Intercomparison ofregional climate simulations of the 1991 sum-mer monsoon in East Asia. J. Geophys. Res.,104, 6425–6454.

Leung, L.R., Y. Qian, and X. Bian, 2003a: Hydro-climate of the western United States based onobservations and regional climate simulationof 1981–2000. Part I: Seasonal statistics. J.Climate, 16, 1892–1911.

Leung, L.R., Y. Qian, X. Bian, and A. Hunt, 2003b:Hydroclimate of the western United Statesbased on observations and regional climatesimulation of 1981–2000. Part II: MesoscaleENSO anomalies. J. Climate, 16, 1912–1928.


Leung, L.R., L.O. Mearns, F. Giorgi, and R.L. Wilby,2003c: Regional climate research—needs andopportunity. Bull. Amer. Meteor. Soc., 84, 89–95.

Leung, L.R., Y. Qian, X. Bian, W.M. Washington,J. Han, and J.O. Roads, 2004: Mid-centuryensemble regional climate change scenarios forthe western United States. Clim. Change, 62,75–113.

Liston, G.E., and R.A. Pielke, 2000: A climate ver-sion of the Regional Atmospheric ModelingSystem. Theor. Appl. Climatol., 66, 29–47.

Lu, L., R.A. Pielke Sr., G.E. Liston, W.J. Parton, D.Ojima, M. Hartman, 2001: Implementation of atwo-way interactive atmospheric and ecologicalmodel and its application to the central UnitedStates. J. Climate, 4, 900–919.

Lynch, A.H., W.L. Chapman, J.E. Walsh, G. Weller,1995: Development of a regional climate modelof the western Arctic. J. Climate, 8, 1555–1570.

Mabuchi, K., Y. Sato, and H. Kida, 2000: Numericalstudy of the relationships between climate andthe carbon dioxide cycle on a regional scale. J.Meteor. Soc. Japan, 78, 25–46.

Marbaix, P., H. Gallee, O. Brasseur, and J.-P. vanYpersele, 2003: Lateral boundary conditionsin regional climate models: A detailed study ofthe relaxation procedure. Mon. Wea. Rev., 131,461–479.

McGregor, J.L., 1997: Regional climate modelling.Meteor. Atmos. Phys., 63, 105–117.

McGregor, J.L., and K. Walsh, 1994: Climate changesimulations of Tasmanian precipitation usingmultiple nesting. J. Geophys. Res., 99, 20889–20905.

McGregor, J.L., and M.R. Dix, 2001: The CSIROconformal-cubic atmospheric GCM. In IUTAMSymposium on Advances in MathematicalModelling of Atmosphere and Ocean Dynamics,P.F. Hodnett (Ed.), Kluwer, Dordrecht, 197–202.

McGregor, J.L., K.C. Nguyen, and J.J. Katzfey,2002: Regional climate simulations using astretched-grid global model. Research Activitiesin Atmospheric and Oceanic Modelling ReportNo. 32 (ed. H. Ritchie), WMO/TD-No. 1105, 3.15–3. 16.

Mearns, L.O., F. Giorgi, L. McDaniel, and C. Shields,1995: Analysis of daily precipitation variabilityin a nested regional climate model: Compari-son with observations and 2XCO2 results.Glob. Planet. Change, 10, 55–78.

Miller, N.L., and J. Kim, 1996: Numerical predictionof precipitation and river flow over the RussianRiver Watershed during the January 1995

California storms. Bull. Amer. Meteor. Soc., 77,101–106.

Mitchell, K., M.J. Fennessy, E. Rogers, J. Shukla,T. Black, J. Kinter, F. Mesinger, Z. Janjic, andE. Altshuler, 2001: Simulation of North Amer-ican summertime climate with the NCEP ETAmodel nested in the COLA GCM. Global En-ergy and Water Cycle Experiment (GEWEX)Newsletter, 11, 3–6.

Mo, K., M. Kanamitsu, H.-M.H. Juang, and S.-Y.Hong, 2000: Ensemble regional and globalclimate prediction for the 1997/1998 winter. J.Geophys. Res., 105, 29609–29623.

Molinari, J., and M. Dudek, 1992: Parameteriza-tion of convective precipitation in mesoscalenumerical models: A critical review. Mon. Wea.Rev., 120, 326–344.

NACIP, 2002: National Aerosol Climate InteractionProgram. [Available online from http://www.nacip.ucsd.edu/].

Nobre, C.A., P.J. Sellers, and J. Shukla, 1991: Ama-zonian deforestation and regional climatechange. J. Climate, 4, 957–988.

Nobre, P., A.D. Moura, and L. Sun, 2001: Dynamicaldownscaling of seasonal climate predictionover Nordeste Brazil with ECHAM3 andNCEP’s regional spectral models at IRI. Bull.Amer. Meteor. Soc., 82, 2787–2796.

Noguer, M., R.G. Jones, and J.M. Murphy, 1998:Sources of systematic errors in climatology ofa nested regional climate model (RCM) overEurope. Clim. Dyn., 14, 691–712.

Paegle, J., K.C. Mo, and J.N. Paegle, 1996: Depen-dence of simulated precipitation on surfaceevaporation during the 1993 United Statessummer floods. Mon. Wea. Rev., 124, 345–361.

Pal, J.S., and E.A.B. Eltahir, 2001: Pathways relat-ing soil moisture conditions to future summerrainfall within a model of the land-atmospheresystem. J. Climate, 14, 1227–1242.

Pal, J.S., and E.A.B. Eltahir, 2002: Teleconnectionsof soil moisture and rainfall during the 1993midwest summer flood. Geophys. Res. Lett., 29,1865, doi:10.1029/2002GL014815.

Pal, J.S., and E.A.B. Eltahir, 2003: A feedbackmechanism between soil-moisture distributionand storm tracks. Quart. J. Roy. Meteor. Soc.,129, 2279–2297.

Pal, J.S., E.E. Small, and E.A.B. Eltahir, 2000: Sim-ulation of regional-scale water and energybudgets: Representation of subgrid cloud andprecipitation processes within RegCM. J. Geo-phys. Res., 105, 29,579–29,594.

Palmer, T.N., and D.L.T. Anderson, 1994: The pros-pects for seasonal forecasting. Quart. J. Roy.Meteor. Soc., 120, 755–793.


Palmer, T.N., C. Brankovic, and D.S. Richardson,2002: A probability and decision-model analy-sis of PROVOST seasonal multi-model en-semble integrations. Quart. J. Roy. Meteor.Soc., 126, 2013–2033.

Pan, Z., M. Segal, R. Turner, and E. Takle, 1995:Model simulation of impacts of transient sur-face wetness on summer rainfall in the UnitedStates Midwest during drought and floodyears. Mon. Wea. Rev., 123, 1575–1581.

Pan, Z., E. Takle, M. Segal, and R. Arritt, 1999:Simulation of potential impacts of man-madeland use changes on U.S. summer climate un-der various synoptic regimes. J. Geophys. Res.,104 (D6), 6515–6528.

Pan, Z., J.H. Christensen, R.W. Arritt, W.J.Gutowski, E.S. Takle, and F. Otieno, 2001:Evaluation of uncertainties in regional climatechange simulations. J. Geophys. Res., 106,17,735–17,751.

Pielke, R.A., and R. Avissar, 1990: Influence of land-scape structure on local and regional climate.Landscape Ecol., 4, 133–155.

Pielke, R.A. Jr., 2001: Influence of the spatial distri-bution of vegetation and soils on the predictionof cumulus convective rainfall. Rev. Geophys.,39, 151–177.

Polcher, J., and K. Laval, 1994: The impact of Afri-can and Amazonian deforestation on tropicalclimate. J. Hydrol., 155, 389–405.

Pope, V.D., and R.A. Stratton, 2002: The processesgoverning horizontal resolution sensitivity in aclimate model. Clim. Dyn., 19, 211–236.

Qian, J.-H., F. Giorgi, and M. Fox-Rabinowitz, 1999:Regional stretched grid generation and itsapplication to the NCAR RegCM. J. Geophys.Res., 104, 6501–6513.

Qian, Y., and F. Giorgi, 1999: Interactive couplingof regional climate and sulfate aerosol modelsover eastern Asia. J. Geophys. Res., 104, 6477–6499.

Qian, Y., L.R. Leung, S.J. Ghan, and F. Giorgi, 2004:Regional climate effects of aerosols over China:Observation and modeling. Tellus B, 55, 914–934.

Rasch, P.J., and J.E. Kristjansson, 1998: A Compari-son of the CCM3 Model Climate Using Diag-nosed and Predicted Condensate Parameter-izations. J. Climate, 11, 1,587–1,614.

Reisner, J., R.M. Rasmussen, and R.T. Bruintjes,1998: Explicit forecasting of supercooled liquidwater in winter storms using the MM5 meso-scale model. Quart. J. Roy. Meteor. Soc., 124,1,071–1,107.

Roads, J., S.-C. Chen, M. Kanamitsu, 2003a: U.S.regional climate simulations and seasonal

forecasts. J. Geophys. Res., 108 (D16), 8606,doi:10.1029/2002JD002232.

Roads, J., S. Chen, S. Cocke, L. Druyan, M. Fula-keza, T. LaRow, P. Longergan, J.-H. Qian,and S. Zebiak, 2003b: International ResearchInstitute/Applied Research Centers (IRI/ARCs)regional model intercomparison over SouthAmerica. J. Geophys. Res., 108 (D14), 4425,doi:10.1029/2002JD003201.

Rodwell, M.J., D.P. Rowell, and C.K. Folland, 1999:Oceanic forcing of the wintertime North Atlan-tic Oscillation and European climate. Nature,398, 320–323.

Rojas, M., and A. Seth, 2003: Simulation and sensi-tivity in a nested modeling system for SouthAmerica. Part II: GCM boundary forcing. J.Climate, 16, 2454–2471.

Russell, P.B., P.V. Hobbs, and L.L. Stowe, 1999:Aerosol properties and radiative effects in theUnited States east coast haze plume: An over-view of the tropospheric aerosol radiative forc-ing observational experiment (TARFOX). J.Geophys. Res., 104, 2213–3333.

Schar, C., D. Luthi, and U. Beyerle, 1999: The soil-precipitation feedback: A process study witha regional climate model. J. Climate, 12, 722–741.

Schmidt, F., 1977: Variable fine mesh in spectralglobal model. Beitr. Phys. Atmos., 50, 211–217.

Semazzi, H.F.M., and L. Sun, 1997: The role of orog-raphy in determining the Sahelian climate.Intl. J. Climatol., 17, 581–596.

Sen, O.L., Y. Wang, and B. Wang, 2004a: Impactof Indochina deforestation on the East-Asiansummer monsoon. J. Climate, 17, 1366–1380.

Sen, O.L., B. Wang, and Y. Wang, 2004b: Re-greening the desertification lands in northernChina: Implications from a regional climatemodel experiment. J. Meteor. Soc. Japan, (thisissue).

Seth, A., and F. Giorgi, 1998: The effects of domainchoice on summer precipitation simulation andsensitivity in a regional climate model. J. Cli-mate, 11, 2698–2712.

Seth, A., and M. Rojas, 2003: Simulation and sensi-tivity in a nested modeling system for SouthAmerica. Part I: Reanalyses boundary forcing.J. Climate, 16, 2437–2453.

Sivillo, J.K., J.E. Ahlquist, and Z. Toth, 1997: Anensemble forecasting primer. Wea. Forecasting,12, 809–818.

Sperber, K.R., and T.N. Palmer, 1996: Interannualtropical rainfall variability in general circula-tion simulations associated with the Atmo-spheric Model Intercomparison Project. J. Cli-mate, 9, 2727–2750.


Sperber, K.R., C. Brankovic, M. Deque, C.S. Freder-iksen, R. Graham, A. Kitoh, C. Kobayashi,T. Palmer, K. Puri, W. Tennant, and E. Volo-din, 2001: Dynamic seasonal predictability ofthe Asian summer monsoon. Mon. Wea. Rev.,129, 2226–2248.

Synder, M.A., J.L. Bell, and L.C. Sloan, 2002: Cli-mate responses to a doubling of atmosphericcarbon dioxide for a climatically vulnerableregion. Geophys. Res. Lett., 29, 1514, doi:10.1029/2001GL014431.

Takle, E., and co-authors, 1999: Project to inter-compare regional climate simulations (PIRCS):description and initial results. J. Geophys.Res., 104, 19443–19461.

Thomson, A.M., R.A. Brown, S.J. Ghan, R.C. Izaur-ralde, N.J. Rosenberg, and L.R. Leung, 2002:Elevation dependence of winter wheat produc-tion in eastern Washington State with climatechange: A methodological study. Clim. Change,4, 141–164.

Tiedtke, M., 1993: Representation of clouds in large-scale models. Mon. Wea. Rev., 121, 3040–3061.

Vidale, P.L., D. Luthi, C. Frei, S.I. Seneviratne, andC. Schar, 2003: Predictability and uncertaintyin a regional climate model. J. Geophys. Res.,108 (D18), 4586, doi:10.1029/2002JD002810.

von Storch, H., H. Langenberg, and F. Feser, 2000:A spectral nudging technique for dynamicaldownscaling purposes. Mon. Wea. Rev., 128,3664–3673.

Walsh, K., and J.L. McGregor, 1995: January andJuly climate simulations over the Australianregion using a limited-area model. J. Climate,8, 2387–2403.

Wandishin, M.S., S.L. Mullen, D.J. Stensrud, andH.E. Brooks, 2001: Evaluation of a short-rangemulti-model ensemble system. Mon. Wea. Rev.,129, 729–747.

Wang, G., and G.S. Jenkins, 2002: Deserts anddesertification. Encyclopedia of AtmosphericScience, Eds., J.R. Holton, J.A. Curry, and J.A.Pyle, Academic Press, 633–640.

Wang, H., A.J. Pitman, M. Zhao, and R. Leemans,2003: The impact of land-cover modificationon the June meteorology of China since 1700,simulated using a regional climate model. Intl.J. Climatol., 23, 511–527.

Wang, W.-C., W. Gong, and H. Wei, 2000: A regionalmodel simulation of the 1991 severe precipita-tion event over the Yangtze-Huai river valley.

Part I: Precipitation and circulation statistics.J. Climate, 13, 74–92.

Wang, Y., O.L. Sen, B. Wang, 2003: A highly resolvedregional climate model (IPRC-RegCM) andits simulation of the 1998 severe precipitationevent over China. Part I: Model descriptionand verification of simulation. J. Climate, 16,1721–1738.

Wang, Y., S.-P. Xie, B. Wang, and H. Xu, 2004a:Large-scale atmospheric forcing by SoutheastPacific boundary-layer clouds: A regionalmodel study. J. Climate, (in press).

Wang, Y., S.-P. Xie, H. Xu, and B. Wang, 2004b:Regional model simulations of boundary layerclouds over the Southeast Pacific off SouthAmerica. Part I: Control experiment. Mon.Wea. Rev., 132, 274–296.

Whetton, P.H., J.J. Katzfey, K.J. Hannesey, X. Wu,J.L. McGregor, and K. Njuyen, 2001: Usingregional climate models to develop fine resolu-tion scenarios of climate change: An examplefor Victoria, Australia. Clim. Res., 16, 181–201.

Wood, A.W., L.R. Leung, V. Sridhar, and D.P.Lettenmaier, 2004: Hydrologic implications ofdynamical and statistical approaches to down-scaling climate model outputs. Clim. Change,62, 233–256.

Xu, H., Y. Wang, and S.-P. Xie, 2004: Effects ofthe Andes on eastern Pacific climate: A re-gional atmospheric model study. J. Climate,17, 589–602.

Xue, Y.-K., 1996: The impact of deforestation in theMongolian and inner Mongolian grassland onthe regional climate. J. Climate, 9, 2173–2189.

Yang, Z., and R.W. Arritt, 2002: Tests of a perturbedphysics ensemble approach for regional climatemodeling. J. Climate, 15, 2881–2896.

Yoshikane, T., F. Kimura, and S. Emori, 2001:Numerical study on the Baiu front genesis byheating contrast between land and ocean. J.Meteor. Soc. Japan, 79, 671–686.

Yoshikane, T., and F. Kimura, 2003: Formationmechanism of the simulated SPCZ and Baiufront using a regional climate model. J. Atmos.Sci., 60, 2612–2632.

Zhang, D.-L., W.-Z. Zheng, and Y.-K. Xue, 2003:A numerical study of early summer regionalclimate and weather over LSA-East. Part I:Model implementation and verification. Mon.Wea. Rev., 131, 1895–1909.


regional climate modeling: progress, challenges, and...

Documents