- Email: [email protected]
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
Journal of Hydrology (2007) 333, 144– 154
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/jhydrol
Projection of global warming onto regional precipitation over Mongolia using a regional climate model Tomonori Sato a b c
a,*
, Fujio Kimura b, Akio Kitoh
c
Japan Science and Technology Agency, Kawaguchi, Japan Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Japan Meteorological Research Institute, Tsukuba, Japan
Received 31 August 2005; received in revised form 29 November 2005; accepted 24 July 2006
KEYWORDS RAISE; Global warming; Mongolia; Regional climate model; Dynamical downscaling
Summary Climate change due to global warming is of concern to the public and may cause significant changes in the hydrological regimes in arid/semi-arid areas including Mongolia, which locates at a boundary between arid and humid regions. However, general circulation models (GCMs) are not sufficient to evaluate climate change on a regional-scale. In this study, two kinds of dynamical downscaling (DDS), referred to as method-G and method-R, using a regional climate model (RCM) are applied to investigate the rainfall change over Mongolia in July due to the global warming. Method-G is a traditional DDS method in which an RCM is directly nested within a GCM, while method-R is newly suggested in this study and aims to improve the reproductivity of a regional climate. For current climate simulation, method-R uses reanalysis data as a boundary forcing of the RCM while a specially created boundary condition, in which projected changes of meteorological variables in a GCM simulation are added on reanalysis data, is used for global warming simulation. Compared with in situ observations, the rainfall amount for July is very well reproduced by the RCM, even in a smaller area of four subregions in Mongolia. Rainfall intensity by method-R is very close to actual observations; on the other hand, method-G fails to simulate heavy rainfall events stronger than 16 mm day 1. The two DDS methods show similar results with respect to the changes of precipitation in July due to the global warming, which are that precipitation decreases over northern and increases over southern Mongolia. In method-R, a decrease of precipitation of middle to heavy rainfall intensity, stronger than 4 mm day 1, contributes largely to the decreased July precipitation in northern Mongolia. Soil moisture over Mongolia also tends to decrease in July because of
* Corresponding author. Present address: Center for Climate System Research, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa 2778568, Japan. Tel.: +81 4 7136 4411; fax: +81 4 136 4375. E-mail address: [email protected] (T. Sato). 0022-1694/$ - see front matter ª 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2006.07.023
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
145
the combined effect caused by the decrease of precipitation and the increase of potential evaporation due to rising air temperature. This situation indicates that severe droughts may occur more frequently from the effects of global warming. ª 2006 Elsevier B.V. All rights reserved.
Introduction Climate change due to global warming is a worldwide public concern. In the Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report (Houghton et al., 2001), a warming trend of annual mean temperature was predicted for most parts of Northeast Asia by General Circulation Models (GCMs). However, the estimates of precipitation after global warming were quite different among GCMs, indicating the difficulties in predicting precipitation in this region. Mongolia is situated in an arid and semi-arid zone in Northeast Asia. The southern part of the territory has a desert type climate. The thermal effect of the Tibetan Plateau is the principle reason for the desert climate, remotely suppressing convective systems in the arid region (Sato and Kimura, 2005a). On the other hand, Taiga forest covers the northern part of Mongolia, extending to Siberia in Russia. More than half of the annual precipitation is observed during the summer season in Mongolia (Batima and Dagvadorj, 2000). Synoptic scale disturbances, e.g., cyclones, troughs, and cut-off lows, are likely to interact with the complicated topography in Mongolia, and generate convective precipitation systems. The territory has a prominent meridional contrast in its rainfall amount as well as in its surface conditions that changes from desert to grassland and to forest over a range of only several hundreds of kilometers from south to north. In general, such transition zones are very sensitive to climate change; as are the economic activities in this region (Sugita et al., this issue). Estimations of the effects of global warming on water resources in a region or in a river basin are required by decision makers to prevent such effects from spreading to a region’s economy. Currently, GCMs are the only tool to simulate global change due to greenhouse gas emission that causes the changes of precipitation on global and continental scales under the effects of global warming. However, it has been difficult to simulate the climate in smaller areas, ones on a national or even a river basin scale, due to the limitations in horizontal resolution as well as in physical parameterizations of GCMs. The dynamical downscaling (DDS) technique, which estimates higher-resolution climatic conditions in a physical model by taking account of detailed geographic information such as topography, allows us to obtain smaller scale prediction. A regional climate model (RCM) nested within GCM simulations (hereafter MethodG) has been a traditional tool to carry out DDS, although RCMs also contain some limitations in physical parameterizations. Reproduction of a regional climate system by method-G still has some difficulties since simulated climates in an RCM are strongly influenced by larger scale forcing given by a GCM, which may have strong model bias (Kato et al., 2001). On the other hand, an RCM allows reanalysis
data, which is computed using observed data, to be employed as a model forcing (hereafter Method-R). Method-R is known to more plausibly reproduce a regional climate system for current years (Wang et al., 2004). This study aims to predict rainfall after global warming utilizing two DDS methods using an RCM, by means of method-G and method-R. A comparison between the two methods is carried out to show the regional climate prediction ability of method-R, especially in and around the boundary between the humid and the arid areas in Northeast Asia. Finally, changes in the rainfall characteristics in subregions of Mongolia are discussed based on the results of method-R, as well as method-G.
Method Data Summertime (June, July and August) precipitation contributes more than 60% of the annual precipitation over Mongolia. We focus our analysis on the precipitation in July when approximately 20% of the annual precipitation falls. Horizontal resolutions of the compiled global precipitation datasets (e.g., Adler et al., 2003), are not sufficient to investigate rainfall in the study area due to their coarse horizontal resolution. Therefore, the daily precipitation data observed at 65 meteorological stations as shown in Fig. 1, provided by the Institute of Meteorology and Hydrology, Mongolia, are used to validate the simulated precipitation in recent years. In order to compare the model estimates
Figure 1 Calculation domain of regional climate model. Circles indicate the locations of meteorological stations. Colors indicate altitude ranges. Four subregions, NW, NE, SW, and SE, are indicated.
146
T. Sato et al.
with in situ observations, the nearest 65 grid points of each meteorological station are picked up from the numerical grid system.
Model The forcing data for the boundary condition of the RCM are given by MRI-CGCM2 (Yukimoto et al., 2001; Kitoh et al., 2005) with T42 (triangle truncation in the horizontal direction at the wave number of 42) which approximately corresponds to a 2.8 horizontal resolution. The control run of MRI-CGCM2 simulates the current climate condition, while the global warming run is performed based on the A2 scenario in the Special Report on Emission Scenarios (SRES) (Nakicenovic and Swart, 2000). This scenario yields that carbon dioxide concentration increases up to 800 ppm in the late 21st century by assuming slow technical change and economical growth in the future. Meteorological data for both integrations are recorded for every 6 h from 1991 through 2000 for the control run and from 2071 through 2080 for the A2 scenario run. The regional climate model, the Terrestrial Environment Research Center – Regional Atmospheric Modeling System (TERC-RAMS; Sato and Kimura, 2005b) is adopted for the DDS method. The role of TERC-RAMS is to estimate the regional scale atmospheric circulation from a low resolution GCM and global reanalysis dataset. An outline of the model settings is listed in Table 1. Arakawa–Schubert type convective parameterization (Arakawa and Schubert, 1974) and microphysics parameterization (Walko et al., 1995) are used to calculate precipitation in the model. Formation of the subgrid scale cumulus near the top of the convective boundary layer is parameterized with grid mean relative humidity. The concentration of carbon dioxide is assumed to be constant in all experiments by the TERC-RAMS since the effect of radiative forcing by increasing carbon dioxide concentration is negligibly small in a small region. The TERC-RAMS has two grid systems for two-level two-way nesting as shown in Fig. 1. The coarse grid system is centered on the Tibetan Plateau with a 150 km horizontal resolution covering an area of 12,000 · 9,000 km. The fine one covers the whole of Mongolia with a 30 km resolution. Both grid systems contain 30 vertical layers in a terrain following coordinate system.
Table 1
The thicknesses of the vertical layers vary from 110 m, at the lowest layer, to 800 m in the upper layers. The top of the model atmosphere is 17500 m. Meteorological variables in the coarse grid system are nudged to the forcing dataset of the MRI-CGCM2 or NCEP/NCAR reanalysis (Kalnay et al., 1996) with the time coefficient of 10 min in six grids from the lateral boundaries. The inner part of the domain is also nudged very weakly with the time constant of four days. The nudging of the inner domain produces similar effects to restarting the model with new initial conditions, provided by the forcing dataset, roughly every four days, and this helps to avoid accumulation of model bias in prognostic variables, such as temperature, specific humidity and so on, in TERC-RAMS. On the other hand, due to this procedure, the influence of land surface processes on the atmosphere tends to be reduced. Hereafter, the results in the fine grid system are mainly presented and discussed, since the target of this study is regional climate prediction. Surface conditions in the TERC-RAMS domains are given by a global land cover characterization dataset provided by the US geological survey (Loveland et al., 2000), which is based on satellite observations by an Advanced Very High Resolution Radiometer (AVHRR). The TERC-RAMS uses distributions of the Leaf Area Index (LAI), the vegetation albedo, the roughness height, and other parameters of vegetation determined in the Biosphere–Atmosphere Transfer Scheme (BATS; Dickinson et al., 1986). The soil texture is assumed uniformly as sandy clay loam type with saturated volumetric soil water content of 0.42. We performed four experiments as described in ‘‘Numerical experiments’’ and Table 2. Numerical integrations started on 26th June for each year. After a five-day spin up duration to accustom the forcing dataset to model physics, the model runs continued for another whole one month from the 1st to 31st July to obtain the data to be analyzed. A one month spin up run was carried out only for July 2003 in order to obtain the initial soil moisture distributions in 11 soil layers; they were adopted as the climatology in July in all years. This assumption means that the initial soil moisture exhibits the same distribution every year without interannual variation. The validity of such a simple treatment of initial soil moisture is discussed in ‘‘Reproduction of present climate’’.
Outline of regional climate model
Horizontal grids Coarse grid Fine grid
80 · 60, 150 km horizontal resolution, centered on 105E, 40N 102 · 57, 30 km horizontal resolution, centered on 104E, 47N
Vertical grids (terrain following coordinate system) Soil layers 30 layers with 110 m thickness in lowest layer, maximum thickness is 800 m 0.00, 0.06, 0.10, 0.18, 0.30, 0.50, 0.70, 0.90, 1.20, 1.50, 1.80 m below ground Vegetation type USGS land surface characterization (Loveland et al., 2000) Soil texture Sandy clay loam Sea surface temperature Start time Initial soil moisture
Monthly mean SST by Reynolds et al. (2002) 00Z 26 June for every integrations One month spin-up for July 2003
Projection of global warming onto regional precipitation over Mongolia using a regional climate model Table 2
147
List of numerical experiments
DDS
Run name
Forcing data
Period
Method-G
RCM.gcmCTL RCM.gcmA2
Six-hourly MRI-CGCM2 control run Six-hourly MRI-CGCM2 A2 scenario run
1991–2000 2071–2080
Method-R
RCM.ncepCTL RCM.ncepWM
Six-hourly NCEP/NCAR reanalysis Original dataset (six-hourly NCEP/NCAR reanalysis with GWMD, see ‘‘Numerical experiments’’)
1993–2003 1993–2003 (corresponding to 2070s)
Numerical experiments Method-G A list of the four numerical experiments is given in Table 2. The RCM.gcmCTL run uses products from 1991 through 2000 by the MRI-CGCM2 control run (Yukimoto et al., 2001), which simulates the present-day climate under a carbon dioxide concentration of 348 ppm, with forcing data corresponding to the period from 1991 to 2000. The RCM.gcmA2 run was carried out to estimate precipitation over Mongolia from 2071 through 2080 under global warming using the products of the MRI-CGCM2 SRES-A2 scenario run. Comparison between RCM.gcmCTL and RCM.gcmA2 gives a projection of the global warming onto the regional climate predicted by the DDS method. This method has been widely adopted to study the impacts of global warming on regional climate systems (e.g., Giorgi et al., 2001). Method-R Method-R is a proposal newly made in the current study, and was employed to prevent a bias in GCMs that has been the main concern in evaluations of regional climate predictions. The six-hourly NCEP/NCAR reanalysis data was used as a RCM forcing in the RCM.ncepCTL run for the period from 1993 to 2003, which represents the present climate experiment, and aims at demonstrating the ability of the TERCRAMS to reproduce the regional climate. In the RCM.ncepWM run, a new forcing dataset, as mentioned below, was used to simulate the regional climate influenced by global warming. Monthly averages of the ten-year mean difference between the control run (corresponding to 1991–2000) and the A2 scenario run (corresponding to 2071–2080) of MRI-CGCM2 was calculated for each 2.5 grid and 17 pressure levels (hereafter GWMD: Global Warming Monthly mean Difference) which represents the change of spatial structure in July due to global warming. The GWMD of the wind speed, the temperature, the geopotential height, the specific humidity, and the sea surface temperature are time-independently superimposed on each variable of the six-hourly NCEP/NCAR reanalysis data and the monthly mean SST data as a perturbation from the current weather conditions from 1993 through 2003. The forcing dataset specially prepared and generated by this process shows (but not shown here) very similar spatiotemporal variations with NCEP/NCAR reanalysis data, but it contains large-scale changes of temperature, humidity, static stability, zonal flow, and baroclinic instability in July as predicted by MRI-CGCM2. Hence, the RCM.ncepWM run is expected to simulate cyclones, troughs, and cut-off lows with basically the same structures during the period from
1993 through 2003 except that time-independent GWMD are added in the present climate as the perturbation induced by global warming. The advantages and disadvantages of method-R compared with method-G are addressed in ‘‘Discussion’’.
Results Reproduction of present climate Fig. 2 compares the mean rainfall distributions in July over Mongolia. July precipitation has its maximum in the northern part of Mongolia >100 mm, which gradually decreases going southward and westward with the minimum precipitation over the southern and western parts that have less than 40 mm. RCM.gcmCTL was able to simulate the meridional gradient of rainfall in the study area, although the amount of precipitation is underestimated in the southern and eastern parts of the domain, and overestimated in the western part of Mongolia (Fig. 2c). The horizontal distribution and the amount of the rainfall produced in RCM.ncepCTL (Fig. 2b), however, agrees very well with those in Fig. 2a in the southern desert region and also in the northern mountainous region, although rainfall is overestimated around the mountain peaks in a way similar to RCM.gcmCTL. These figures indicate that the nested RCM within the GCM still has some difficulties simulating the current regional climate, although it can properly simulate the larger scale characteristics of rainfall distribution. For the quantitative assessment of rainfall distribution, the Mongolian territory is divided into four subregions, the northwest (NW), the northeast (NE), the southwest (SW), and the southeast (SE) regions, by the lines along 47N and 104E as shown in Fig. 1. Fig. 3 shows the mean precipitation with the standard deviation for the calculated periods in individual subregions. RCM.ncepCTL reproduced the regional characteristics in Mongolia well, with more precipitation in the northern two subregions and less precipitation in the southern two subregions, although the model overestimates rainfall by about 30% in the NW and SW regions where the topography is complicated and also in the SE region where there is the Gobi desert. On the other hand, RCM.gcmCTL failed in a quantitative simulation of the subregional precipitation, although it quite well reproduced the total averaged amount of rainfall over the whole of Mongolia. The overestimation in the SW subregion comes from the result that too much precipitation occurs over mountains. In the MRI-CGCM2 control run, air temperature in the troposphere in July over eastern Mongolia is lower than that in the NCEP reanalysis, which is one reason for
148
T. Sato et al.
Figure 3 Mean precipitation in July for four subregions (NW, NE, SW, and SE) shown in Fig. 1 and stations in all of Mongolia, as also shown in Fig. 1 (ALL). The white bars depict observations while the light and dark shaded bars are results calculated by the RCM.ncepCTL and RCM.gcmCTL runs, respectively. The error bars show the standard deviation of precipitation in the observed 11-years and that as simulated in 11 years by the RCM.ncepCTL and 10 years by the RCM.gcmCTL.
Figure 2 Rainfall distributions in July. Contours are drawn for every 20 mm month 1. (a) Eleven-year mean by local meteorological stations from 1993 through 2003. Estimated precipitation by GPCP (Adler et al., 2003) was used to draw the contours outside the Mongolian boundary. Open circles represent stations used in this study. (b) Eleven-year mean by the RCM.ncepCTL run from 1993 through 2003. (c) Ten-year mean by the RCM.gcmCTL run from 1991 through 2000.
the underestimation in the SE and NE subregions in the RCM.gcmCTL run. Fig. 4 shows and compares the interannual variations of July precipitation obtained by observation and by RCM.ncepCTL. Only the ten-year average rainfall is shown for the RCM.gcmCTL run since the target years of the RCM.gcmCTL run do not necessarily correspond to a specific observed year. For example, the year 1994 in RCM.gcmCTL does not exactly correspond to the year 1994 observation, and so they cannot be compared with each other. The RCM.ncepCTL run simulated successfully the interannual variation of the July precipitation in the whole of Mongolia. In the years 1993, 1994 and 1998, more rainfall was observed than in the other years, while the precipitation was less in the dry years from 1999 to 2002. The RCM.ncepCTL overestimated rainfall by nearly 100% during the dry years except for 2002. Although one may attribute some parts of the error to the simple assumptions in the assumed initial soil moisture, we confirmed that the simulated overestimated precipitation still remains large even if a drier soil condition was used in the test. Additionally, the
interannual variation of precipitation was simulated well except for 1999, 2000, and 2001, despite the assumed simple initial soil moisture. This indicates that the initial soil moisture condition is not always important for simulations of the interannual variation of July precipitation. A possible cause for so much rainfall in 2001 might be in the reanalysis data; but further investigation is needed to clarify this point. A probability density distribution (PDD) of the daily rainfall in July is shown in Fig. 5. In Mongolia, July precipitation is mostly maintained by rainfall events stronger than 4 mm day 1. The contribution of the strongest rainfall events, over 32 mm day 1, is largest in the NE subregion. In the southern two subregions, all intensity categories are smaller than those in the northern two subregions. RCM.gcmCTL overestimated weak rainfalls, and underestimated heavy rainfalls. Such a tendency can be seen not only in the whole of Mongolia but also in all subregions. Underestimation of heavy rainfall events in the NE and SE subregions (Fig. 5c and e) by RCM.gcmCTL caused it to underestimation these for all of Mongolia (Fig. 5a) and an underestimation of the total subregional rainfalls in the NE and SE (Fig. 3). These results imply that rainfall intensity in Mongolia cannot be captured by method-G. On the other hand, the PDDs estimated by the RCM.ncepCTL agreed quite well with PDDs observed for all of Mongolia, although the frequency of the weak rainfalls of less than 8 mm day 1 was slightly overestimated. The RCM.ncepCTL very accurately simulates the largest three categories with intensities stronger than 8 mm day 1, and which contribute to a large part of the July precipitation in Mongolia. The estimated rainfall intensities in each of the four subregions also agreed with the observed intensities; meaning that TERC-RAMS can reproduce the precipitation properties even in a small target area, as shown on a basin scale in Mongolia.
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
149
Figure 4 Year to year comparison of precipitation in July between observations and RCM.ncepCTL averaged over all of Mongolia. Error bars indicate the standard deviation for the 65 stations shown in Fig. 1. Bars on the right side of the figure show the 11-year mean precipitation from observation, and the RCM.ncepCTL with a standard deviation during 11 years and RCM.gcmCTL with a standard deviation during 10 years.
Figure 5 Probability density distribution of daily precipitation intensity obtained by observations (solid line with open square), the RCM.ncepCTL (broken line with light-shaded square) and the RCM.gcmCTL (dot-dash-line with dark-shaded square). Bars show
150
T. Sato et al.
Rainfalls of heaviest intensity class of >32 mm day 1 occur more frequently in the NE subregion than in the other subregions. However, the categories from 4 to 32 mm day 1 contribute to most of the total rainfall in the southern two subregions. These features of rainfall intensity and regional distribution are well simulated in the RCM.ncepCTL run, its biggest advantage being its excellent ability to reproduce the current regional climate, whereas only the averaged rainfall amount for all of Mongolia is valid in the product of the RCM.gcmCTL run. Although method-G has widely applied in previous studies, a method for the correct reproductivity of the present-day climate needs to be investigated before addressing the changes due to global warming. In the RCM.ncepCTL run, daily or longer temporal variations of meteorological variables such as temperature, humidity, wind speed, and pressure in the atmosphere are basically dependent on reanalysis data. The RCM.ncepCTL estimated the monthly mean surface air temperature with less than 1 C error as compared by the average of all stations and the corresponding grid values in the model. Daily maximum and minimum temperature were underestimated by approximately 3 C and 1 C, respectively.
Changes of spatial distribution by two DDS methods The predicted changes in the amount of precipitation from the impact of global warming in Mongolia are shown in Fig. 6 (future minus present). Both DDS methods produced results that indicate a statistically significant decrease of precipitation in the northern part of Mongolia. Especially, rainfall especially decreases over the Hangai and the Khenti Mountains. Contrarily, rainfall slightly increases around the Gobi Desert in southern Mongolia. The signs indicating precipitation changes are different in western Mongolia when comparing results between the two methods; method-R produced a decreasing precipitation, while methodG predicted an increase with small areas of decreasing trends, like a complex mosaic. The two methods show a different trend in eastern Mongolia where method-G indicates very little change in the July precipitation, but a decrease can be seen in the results of method-R. Fig. 7 shows the rainfalls estimated for the four subregions by the two DDS methods. The effects of global warming on precipitation are negligibly small in the SE and the SW subregions in both methods. In method-R, rainfall decreases by 15% in the NW and the NE subregions leading to a decrease in the averaged rainfall for all of Mongolia. Method-G also predicts decreases in the NW and the NE subregions, which are mountainous areas (Fig. 1) where the sources of most of Mongolia’s rivers are situated (Tsujimura et al., this issue). Therefore, the decrease of precipitation in these subregions could present very serious concerns for river water management.
Figure 6 Horizontal distribution of the precipitation difference (future minus present) in July by the two DDS methods (mm month 1). Negative and positive values indicate decreases and increases of precipitation. Solid contours and broken contours are drawn for increases and decreases of July precipitation with every 20 mm month 1 interval. (a) Method-G (RCM.gcmA2 minus RCM.gcmCTL). (b) Method-R (RCM.ncepWM minus RCM.ncepCTL).
Changes of rainfall characterization
Figure 7 Same as in Fig. 3 but for the RCM.ncepCTL (light shade), the RCM.ncepWM (hatched light shade), the RCM.gcmCTL (dark shade), and the RCM.gcmA2 (hatched dark shade).
With method-R, as shown in ‘‘Reproduction of present climate’’, the model bias for precipitation is small enough and is comparable with the amplitude of changes caused by global warming, although the former is usually much lar-
ger than the latter because of limitations in the accuracy of the model. This section describes how rainfall properties are changed due to global warming.
Projection of global warming onto regional precipitation over Mongolia using a regional climate model Method-R allows us to test the impact of global warming for the past several years by giving GWMD as perturbation components to the reanalysis data. Fig. 8 compares the interannual variations of the averaged precipitation for all of Mongolia for both RCM.ncepCTL and RCM.ncepWM. Due to the impact of the large-scale changes caused by global warming, precipitation decreases in most years. In particular, the decrease is very large in 1993 and 1994 which are wet years. The standard deviation of the interannual variation of July precipitation tends to be smaller than that in the present climate. According to the IPCC report (Houghton et al., 2001), the frequencies of extreme events such as drought and flood, are likely to increase worldwide due to the impact of global warming. Theoretically, method-R is not expected to be able to precisely estimate changes in frequency since only GWMD, which was calculated as the difference of the ten-year averages of the monthly means between the present climate in 1990s and the future climate in 2070s, was taken into account without consideration of nonlinear effects such as the change in the range of interannual variability and the shift of storm tracks. Some of the limitations of method-R are addressed in the next section. Estimated projections of global warming onto rainfall intensities by method-R and method-G are shown in Fig. 9. Changes of the rainfall intensity by method-R can be described as follows. The frequency of weaker rainfall does not change, while that of the rainfall events stronger than 4 mm day 1 tends to decrease. In all subregions, the estimated changes in frequency of the weak rainfall categories are very small. Rainfalls in all categories exceeding 4 mm day 1 decrease uniformly in the NW subregion. In the NE and SE subregions, decreases of heavier rainfall events, of over 16 mm day 1 are prominently. On the other hand, rainfalls in the heaviest category of over 32 mm day 1 increase from the effects of the global warming in just the SW subregion. On the other hand, with method-G results, the increases in rainfall in the heavier rainfall categories of more than 8 mm day 1, were found when the rainfall intensity was investigated for all of Mongolia (Fig. 9a). The increases in
151
those categories are prominent in the NE and SE subregions, which may indicate changes of the storm tracks due to global warming. However, reproductivity of the present-day climate in these subregions is not very good in RCM.gcmCTL, especially for strong rainfall events (Fig. 5c and e). Thus, method-G probably cannot evaluate any increases of strong rainfall events in the NE and the SE subregions. Potential evaporation tends to increase from the effects of global warming. This leads to reduced soil moisture, when the increment of evapotranspiration can be larger than that of precipitation (Manabe and Wetherald, 1987). In Fig. 6, an increase of July precipitation was estimated by method-R and method-G over southern Mongolia. However, in both methods the potential evapotranspiration also increases due to rising air temperatures. As a result, soil moisture in July decreases in most parts of Mongolia (Fig. 10). Especially in northern Mongolia, the combined effects of increased evapotranspiration and decreased precipitation cause a severe decrease in soil moisture in July.
Discussion Interannual variations of precipitation over Mongolia were discussed in ‘‘Changes of rainfall characterization’’. As presented in Fig. 8, the amplitude of the variations, which exceeded 50 mm during 1993 and 2003, is much larger than the signals that we focus on, since the 11-year mean precipitation difference caused by global warming, as estimated by both methods, is below 10 mm. In order to address this problem by method-G, several ten-year averages are required for both the present climate and that resulting after global warming. In method-R, on the other hand, the complexity of such treatments can be much reduced. Precipitation changes can be estimated by a single projection of global warming onto just one typical year, e.g., a comparison between the simulated precipitation in 1993 and in ‘‘pseudo’’ 1993 under global warming, as a sensitivity experiment. In this study, the GWMD variables, which were calculated as the differences of the monthly averages of the ten-year means between 1991– 2000 and 2071–2080, were given in the RCM.ncepWM run
Figure 8 Same as in Fig. 4 but for the RCM.ncepCTL (light shade), the RCM.ncepWM (hatched light shade), the RCM.gcmCTL (dark shade), and the RCM.gcmA2 (hatched dark shade).
152
T. Sato et al.
Figure 9 Same as in Fig. 5 but for the RCM.ncepCTL (light-shaded square, light-shaded bar), the RCM.ncepWM (light-shaded triangle, hatched light-shaded bar), the RCM.gcmCTL (dark-shaded square, dark-shaded bar), and RCM.gcmA2 (dark-shaded triangle, hatched dark-shaded bar). (a) All Mongolia, (b) NW, (c) NE, (d) SW, (e) SE.
by assuming that the global warming trend would be a given constant from 1993 to 2003. However, as mentioned in ‘‘Changes of rainfall characterization’’, the sensitivity of the GWMD on precipitation is not a constant but varies with the year, even though a constant GWMD is given for all years. Precipitation decreased greatly in years 1993 and 1994, while the decrement was very small in years 1995 and 2003, and precipitation tended to increase in year 2002. These results indicate that the impact of global warming upon regional precipitation is quite complicated, having a different sensitivity for each year, which may have contributed to the changes of interannual variability. RCMs are capable of reproducing synoptic- and meso-scale cyclogeneses that develop and decay in the numerical domain. Therefore, method-R can to some extent simulate the effect of global warming on interannual variation, by taking account of the changes in the structures of cyclogeneses by temperature rises and increased moisture.
On the other hand, method-R is only partially applicable for estimations of long-term trends in interannual variability, because it does not fully consider the nonlinear interaction between global warming and atmospheric disturbances. Therefore, this method can not estimate the projection of global warming onto such factors as storm frequency, storm intensity, and the positions of storm tracks, which all interacts with the large-scale climate system beyond the model boundary. Such limitations of method-R may also explain the contradictory results from the two DDS methods. In ‘‘Results’’, we showed an excellent performance in reproducing the present-day precipitation by a RCM, which is a major advantage of method-R. Model bias in the GCM, that often reduces reproductivity of regional climate, is limited to the GWMD variables estimated using the ten-year averages of the monthly means of GCM products. Thus, the meteorological variables in method-R, such as precipitation and temperature, are more credible under the effects of global warming in their absolute value than those in method-G.
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
153
Conclusion
Figure 10 Change of soil water content at the shallowest layer in July (%; future minus present) by (a) method-G and (b) method-R. Shaded regions indicate decrease of soil moisture.
Projections of global warming onto regional precipitation in Mongolia should differ according to the choice of GCM, greenhouse gas emission scenarios, and analyzed periods. Model ensemble has been regarded as the most reliable prediction in the Northeast Asia region. In consideration of ensemble experiments for regional climate simulation, method-R has an advantage in its data handling of GCMs products. To carry out a DDS projection of global warming, method-R requires GCM data only in monthly-mean form, while an at least twelve-hourly GCM product is demanded for method-G. Accordingly, method-R is much easier to compare DDSs products across many GCMs and/or several emission scenarios, and thus, more applicable to an ensemble experiment. According to Kitoh et al. (2005), precipitation increases in spring while it decreases in summer over northern China (35–45N, 100–120E) under the effects of global warming during 2081 to 2100 as revealed by MRI-CGCM2. They indicated that the GWMD of soil moisture had a seasonal cycle. Regional projection of global warming onto seasonal cycles of precipitation and soil moisture will be carried out in a future study in a further assessment of water resources. The high-resolution regional projection obtained by the RCM is applicable to a hydrological model that aims to study river runoff and groundwater regimes. Additionally, the roles of the high latitude ecosystem in the global scale carbon cycle have recently been focused on (White et al., 2000). A projection, as given in this study, can also be used for a terrestrial carbon cycle model, which allows investigation of the influences of activities such as grazing upon the terrestrial environments in Mongolia.
A dynamical downscaling (DDS) method using a regional climate model (RCM) was applied to investigate rainfall changes across Mongolia due to global warming. The DDS was carried out by two methods; method-R that uses sixhourly reanalysis data for a present-day simulation and newly prepared six-hourly dataset in which the changes of the monthly averages of ten-year mean variables by GCM due to global warming were added into the reanalysis data to prevent GCM bias in the regional climate. MethodG uses six-hourly GCM calculations for both the present and the future climates under the effects of global warming. Method-R successfully reproduced rainfall distribution in July even in the individual subregions in Mongolia. Rainfall intensity was found to be very similar to the in situ observations even for the four subregions in Mongolia. The method-G reproduced large-scale features of rainfall distribution; however, it failed to simulate stronger rainfall events. Projections of global warming onto regional precipitation levels obtained using the two DDS methods were compared. Changes in regional precipitation in both methods agreed in their spatial patterns that indicated decreasing precipitation over northern Mongolia and slight increases over southern Mongolia. Under method-R, rainfall tended to decrease for most of the recent years as shown by the sensitivity experiments of RCM.ncepWM in which large scale changes of meteorological variables (GWMD) are considered. In method-R, the reduction in middle to heavy rainfall events, over 4 mm day 1, dominated over northern Mongolia but caused a remarkable decrease of July precipitation in this region. Soil moisture was found to decrease in July over most of all Mongolia, because of the coupling effect of less precipitation and more evaporation due to the rising air temperature; a fact that implies severe droughts may occur more frequently in the future. Detailed evaluations on changes in hydrological regimes including river runoff and ground water are needed to further address this issue using a downscaled climate dataset.
Acknowledgments The authors appreciate the help of M. Hara of the Frontier Research Center for Global Change for data processing, and T. Kurokawa of the Terrestrial Environment Research Center, University of Tsukuba for assistance with RCM simulation. Appreciation is also extended to the ICCAP (Impact of Climate Changes on Agricultural Production System in the Arid Areas) Project of Research Institute for Humanity and Nature for the improvement of TERC-RAMS. Precipitation data at meteorological stations have been provided by the Institute of Meteorology and Hydrology, Mongolia. The authors thank anonymous reviewers for their suggestions and comments which greatly improve the manuscript. This study was supported by a CREST project (RAISE, Rangelands Atmosphere–hydrosphere–biosphere Interaction Study Experiment in Northeastern Asia) of JST (Japan Science and Technology Agency).
154
References Adler, R.F., Huffman, G.J., Chang, A., Ferraro, R., Xie, P., Janowiak, J., Rudolf, B., Schneider, U., Curtis, S., Bolvin, D., Gruber, A., Susskind, J., Arkin, P., 2003. The version 2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979-present). J. Hydrometeorol. 4, 1147–1167. Arakawa, A., Schubert, W.H., 1974. Interaction of a cumulus cloud ensemble with the large-scale environment, Part I. J. Atmos. Sci. 31, 674–701. Batima, P., Dagvadorj, D. (Eds.), 2000. Climate Change and Its Impacts in Mongolia. JEMR Publishing, Mongolia, p. 227. Dickinson, R.E., Henderson-Sellers, A., Kennedy, P.J., Wilson M.F., 1986. Biosphere atmosphere transfer scheme (BATS) for the NCAR community climate model. NCAR Technical Note NCAR/ TN275+STR, 69p. Giorgi, F., Hewitson, B., 2001. Regional climate information – evaluation and projections. In: Houghton, J.T. et al. (Eds.), Climate Change 2001: The Scientific Basis. Cambridge University Press, p. 881 (Chapter 10). Houghton, J.T. Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., Dai, X., Maskell, K., Johnson, C.A., 2001. Climate change 2001: the scientific basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 881p. Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Leetmaa, A., Reynolds, B., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K.C., Ropelewski, C., Wang, J., Jenne, R., Joseph, D., 1996. The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc. 77, 437–472. Kato, H., Nishizawa, K., Hirakuchi, H., Kadokura, S., Oshima, N., 2001. Performance of RegCM2.5/NCAR-CSM nested system for the simulation of climate change in East Asia caused by global warming. J. Meteorol. Soc. Jpn. 79, 99–121. Kitoh, A., Hosaka, M., Adachi, Y., Kamiguchi, K., 2005. Future projections of precipitation characteristics in East Asia simulated by the MRI CGCM2. Adv. Atmos. Sci. 22, 467–478. Loveland, T.R., Reed, B.C., Brown, J.F., Ohlen, D.O., Zhu, J., Yang, L., Merchant, J.W., 2000. Development of a global land cover
T. Sato et al. characteristics database and IGBP DISCover from 1-km AVHRR data:. Int. J. Remote Sensing 21, 1303–1330. Manabe, S., Wetherald, R.T., 1987. Large-Scale changes of soil wetness induced by an increase in atmospheric carbon dioxide. J. Atmos. Sci. 44, 1211–1236. Nakicenovic, N., Swart, R., 2000. Special Report on Emissions Scenarios: A Special Report of Working group III of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 599p. Reynolds, R.W., Rayner, N.A., Smith, T.M., Stokes, D.C., Wang, W., 2002. An improved in situ and satellite SST analysis for climate. J. Climate 15, 1609–1625. Sato, T., Kimura, F., 2005a. Impact of diabatic heating over the Tibetan plateau on subsidence over Northeast Asian arid region. Geophys. Res. Lett. 32, L05809. doi:10.1029/2004GL02208. Sato, T., Kimura, F., 2005b. Diurnal cycle of convective instability around the central mountains in Japan during the warm season. J. Atmos. Sci. 62, 1626–1636. Sugita, M., Asanuma, J., Tsujimura, M., Mariko, S., Lu, M., Kimura, F., Azzaya, D., Adyasuren, Ts., this issue. An overview of the rangelands atmosphere-hydrosphere–biosphere interaction study experiment in northeastern Asia (RAISE). J. Hydrol. Tsujimura, M., Tanaka, T., Abe, Y., Shimada, J., Higuchi, S. Yamanaka, T., Saito T., this issue. Interaction between groundwater and river water in a semi-arid region – a case in Kherlen River basin, eastern Mongolia. J. Hydrol. Walko, R.L., Cotton, W.R., Meyers, M.P., Harrington, J.Y., 1995. New RAMS cloud microphysics parameterization. Part 1: the single-moment scheme. Atmos. Res. 38, 29–62. Wang, Y., Leung, L.R., McGregor, J.L., Lee, D.-K., Wang, W.-C., Ding, Y., Kimura, F., 2004. Regional climate modeling: progress, challenges, and prospects. J. Meteorol. Soc. Jpn. 82, 1599– 1628. White, A., Cannell, M.G.R., Friend, A.D., 2000. The high-latitude terrestrial carbon sink: a model analysis. Global Change Biol. 6, 227–245. Yukimoto, S., Noda, A., Kitoh, A., Sugi, M., Kitamura, Y., Hosaka, M., Shibata, K., Maeda, S., Uchiyama, T., 2001. The new Meteorological Research Institute coupled GCM (MRI-CGCM2) – model climate and variability. Pap. Meteorol. Geophys. 51, 47–88.
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/jhydrol
Projection of global warming onto regional precipitation over Mongolia using a regional climate model Tomonori Sato a b c
a,*
, Fujio Kimura b, Akio Kitoh
c
Japan Science and Technology Agency, Kawaguchi, Japan Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Japan Meteorological Research Institute, Tsukuba, Japan
Received 31 August 2005; received in revised form 29 November 2005; accepted 24 July 2006
KEYWORDS RAISE; Global warming; Mongolia; Regional climate model; Dynamical downscaling
Summary Climate change due to global warming is of concern to the public and may cause significant changes in the hydrological regimes in arid/semi-arid areas including Mongolia, which locates at a boundary between arid and humid regions. However, general circulation models (GCMs) are not sufficient to evaluate climate change on a regional-scale. In this study, two kinds of dynamical downscaling (DDS), referred to as method-G and method-R, using a regional climate model (RCM) are applied to investigate the rainfall change over Mongolia in July due to the global warming. Method-G is a traditional DDS method in which an RCM is directly nested within a GCM, while method-R is newly suggested in this study and aims to improve the reproductivity of a regional climate. For current climate simulation, method-R uses reanalysis data as a boundary forcing of the RCM while a specially created boundary condition, in which projected changes of meteorological variables in a GCM simulation are added on reanalysis data, is used for global warming simulation. Compared with in situ observations, the rainfall amount for July is very well reproduced by the RCM, even in a smaller area of four subregions in Mongolia. Rainfall intensity by method-R is very close to actual observations; on the other hand, method-G fails to simulate heavy rainfall events stronger than 16 mm day 1. The two DDS methods show similar results with respect to the changes of precipitation in July due to the global warming, which are that precipitation decreases over northern and increases over southern Mongolia. In method-R, a decrease of precipitation of middle to heavy rainfall intensity, stronger than 4 mm day 1, contributes largely to the decreased July precipitation in northern Mongolia. Soil moisture over Mongolia also tends to decrease in July because of
* Corresponding author. Present address: Center for Climate System Research, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa 2778568, Japan. Tel.: +81 4 7136 4411; fax: +81 4 136 4375. E-mail address: [email protected] (T. Sato). 0022-1694/$ - see front matter ª 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2006.07.023
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
145
the combined effect caused by the decrease of precipitation and the increase of potential evaporation due to rising air temperature. This situation indicates that severe droughts may occur more frequently from the effects of global warming. ª 2006 Elsevier B.V. All rights reserved.
Introduction Climate change due to global warming is a worldwide public concern. In the Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report (Houghton et al., 2001), a warming trend of annual mean temperature was predicted for most parts of Northeast Asia by General Circulation Models (GCMs). However, the estimates of precipitation after global warming were quite different among GCMs, indicating the difficulties in predicting precipitation in this region. Mongolia is situated in an arid and semi-arid zone in Northeast Asia. The southern part of the territory has a desert type climate. The thermal effect of the Tibetan Plateau is the principle reason for the desert climate, remotely suppressing convective systems in the arid region (Sato and Kimura, 2005a). On the other hand, Taiga forest covers the northern part of Mongolia, extending to Siberia in Russia. More than half of the annual precipitation is observed during the summer season in Mongolia (Batima and Dagvadorj, 2000). Synoptic scale disturbances, e.g., cyclones, troughs, and cut-off lows, are likely to interact with the complicated topography in Mongolia, and generate convective precipitation systems. The territory has a prominent meridional contrast in its rainfall amount as well as in its surface conditions that changes from desert to grassland and to forest over a range of only several hundreds of kilometers from south to north. In general, such transition zones are very sensitive to climate change; as are the economic activities in this region (Sugita et al., this issue). Estimations of the effects of global warming on water resources in a region or in a river basin are required by decision makers to prevent such effects from spreading to a region’s economy. Currently, GCMs are the only tool to simulate global change due to greenhouse gas emission that causes the changes of precipitation on global and continental scales under the effects of global warming. However, it has been difficult to simulate the climate in smaller areas, ones on a national or even a river basin scale, due to the limitations in horizontal resolution as well as in physical parameterizations of GCMs. The dynamical downscaling (DDS) technique, which estimates higher-resolution climatic conditions in a physical model by taking account of detailed geographic information such as topography, allows us to obtain smaller scale prediction. A regional climate model (RCM) nested within GCM simulations (hereafter MethodG) has been a traditional tool to carry out DDS, although RCMs also contain some limitations in physical parameterizations. Reproduction of a regional climate system by method-G still has some difficulties since simulated climates in an RCM are strongly influenced by larger scale forcing given by a GCM, which may have strong model bias (Kato et al., 2001). On the other hand, an RCM allows reanalysis
data, which is computed using observed data, to be employed as a model forcing (hereafter Method-R). Method-R is known to more plausibly reproduce a regional climate system for current years (Wang et al., 2004). This study aims to predict rainfall after global warming utilizing two DDS methods using an RCM, by means of method-G and method-R. A comparison between the two methods is carried out to show the regional climate prediction ability of method-R, especially in and around the boundary between the humid and the arid areas in Northeast Asia. Finally, changes in the rainfall characteristics in subregions of Mongolia are discussed based on the results of method-R, as well as method-G.
Method Data Summertime (June, July and August) precipitation contributes more than 60% of the annual precipitation over Mongolia. We focus our analysis on the precipitation in July when approximately 20% of the annual precipitation falls. Horizontal resolutions of the compiled global precipitation datasets (e.g., Adler et al., 2003), are not sufficient to investigate rainfall in the study area due to their coarse horizontal resolution. Therefore, the daily precipitation data observed at 65 meteorological stations as shown in Fig. 1, provided by the Institute of Meteorology and Hydrology, Mongolia, are used to validate the simulated precipitation in recent years. In order to compare the model estimates
Figure 1 Calculation domain of regional climate model. Circles indicate the locations of meteorological stations. Colors indicate altitude ranges. Four subregions, NW, NE, SW, and SE, are indicated.
146
T. Sato et al.
with in situ observations, the nearest 65 grid points of each meteorological station are picked up from the numerical grid system.
Model The forcing data for the boundary condition of the RCM are given by MRI-CGCM2 (Yukimoto et al., 2001; Kitoh et al., 2005) with T42 (triangle truncation in the horizontal direction at the wave number of 42) which approximately corresponds to a 2.8 horizontal resolution. The control run of MRI-CGCM2 simulates the current climate condition, while the global warming run is performed based on the A2 scenario in the Special Report on Emission Scenarios (SRES) (Nakicenovic and Swart, 2000). This scenario yields that carbon dioxide concentration increases up to 800 ppm in the late 21st century by assuming slow technical change and economical growth in the future. Meteorological data for both integrations are recorded for every 6 h from 1991 through 2000 for the control run and from 2071 through 2080 for the A2 scenario run. The regional climate model, the Terrestrial Environment Research Center – Regional Atmospheric Modeling System (TERC-RAMS; Sato and Kimura, 2005b) is adopted for the DDS method. The role of TERC-RAMS is to estimate the regional scale atmospheric circulation from a low resolution GCM and global reanalysis dataset. An outline of the model settings is listed in Table 1. Arakawa–Schubert type convective parameterization (Arakawa and Schubert, 1974) and microphysics parameterization (Walko et al., 1995) are used to calculate precipitation in the model. Formation of the subgrid scale cumulus near the top of the convective boundary layer is parameterized with grid mean relative humidity. The concentration of carbon dioxide is assumed to be constant in all experiments by the TERC-RAMS since the effect of radiative forcing by increasing carbon dioxide concentration is negligibly small in a small region. The TERC-RAMS has two grid systems for two-level two-way nesting as shown in Fig. 1. The coarse grid system is centered on the Tibetan Plateau with a 150 km horizontal resolution covering an area of 12,000 · 9,000 km. The fine one covers the whole of Mongolia with a 30 km resolution. Both grid systems contain 30 vertical layers in a terrain following coordinate system.
Table 1
The thicknesses of the vertical layers vary from 110 m, at the lowest layer, to 800 m in the upper layers. The top of the model atmosphere is 17500 m. Meteorological variables in the coarse grid system are nudged to the forcing dataset of the MRI-CGCM2 or NCEP/NCAR reanalysis (Kalnay et al., 1996) with the time coefficient of 10 min in six grids from the lateral boundaries. The inner part of the domain is also nudged very weakly with the time constant of four days. The nudging of the inner domain produces similar effects to restarting the model with new initial conditions, provided by the forcing dataset, roughly every four days, and this helps to avoid accumulation of model bias in prognostic variables, such as temperature, specific humidity and so on, in TERC-RAMS. On the other hand, due to this procedure, the influence of land surface processes on the atmosphere tends to be reduced. Hereafter, the results in the fine grid system are mainly presented and discussed, since the target of this study is regional climate prediction. Surface conditions in the TERC-RAMS domains are given by a global land cover characterization dataset provided by the US geological survey (Loveland et al., 2000), which is based on satellite observations by an Advanced Very High Resolution Radiometer (AVHRR). The TERC-RAMS uses distributions of the Leaf Area Index (LAI), the vegetation albedo, the roughness height, and other parameters of vegetation determined in the Biosphere–Atmosphere Transfer Scheme (BATS; Dickinson et al., 1986). The soil texture is assumed uniformly as sandy clay loam type with saturated volumetric soil water content of 0.42. We performed four experiments as described in ‘‘Numerical experiments’’ and Table 2. Numerical integrations started on 26th June for each year. After a five-day spin up duration to accustom the forcing dataset to model physics, the model runs continued for another whole one month from the 1st to 31st July to obtain the data to be analyzed. A one month spin up run was carried out only for July 2003 in order to obtain the initial soil moisture distributions in 11 soil layers; they were adopted as the climatology in July in all years. This assumption means that the initial soil moisture exhibits the same distribution every year without interannual variation. The validity of such a simple treatment of initial soil moisture is discussed in ‘‘Reproduction of present climate’’.
Outline of regional climate model
Horizontal grids Coarse grid Fine grid
80 · 60, 150 km horizontal resolution, centered on 105E, 40N 102 · 57, 30 km horizontal resolution, centered on 104E, 47N
Vertical grids (terrain following coordinate system) Soil layers 30 layers with 110 m thickness in lowest layer, maximum thickness is 800 m 0.00, 0.06, 0.10, 0.18, 0.30, 0.50, 0.70, 0.90, 1.20, 1.50, 1.80 m below ground Vegetation type USGS land surface characterization (Loveland et al., 2000) Soil texture Sandy clay loam Sea surface temperature Start time Initial soil moisture
Monthly mean SST by Reynolds et al. (2002) 00Z 26 June for every integrations One month spin-up for July 2003
Projection of global warming onto regional precipitation over Mongolia using a regional climate model Table 2
147
List of numerical experiments
DDS
Run name
Forcing data
Period
Method-G
RCM.gcmCTL RCM.gcmA2
Six-hourly MRI-CGCM2 control run Six-hourly MRI-CGCM2 A2 scenario run
1991–2000 2071–2080
Method-R
RCM.ncepCTL RCM.ncepWM
Six-hourly NCEP/NCAR reanalysis Original dataset (six-hourly NCEP/NCAR reanalysis with GWMD, see ‘‘Numerical experiments’’)
1993–2003 1993–2003 (corresponding to 2070s)
Numerical experiments Method-G A list of the four numerical experiments is given in Table 2. The RCM.gcmCTL run uses products from 1991 through 2000 by the MRI-CGCM2 control run (Yukimoto et al., 2001), which simulates the present-day climate under a carbon dioxide concentration of 348 ppm, with forcing data corresponding to the period from 1991 to 2000. The RCM.gcmA2 run was carried out to estimate precipitation over Mongolia from 2071 through 2080 under global warming using the products of the MRI-CGCM2 SRES-A2 scenario run. Comparison between RCM.gcmCTL and RCM.gcmA2 gives a projection of the global warming onto the regional climate predicted by the DDS method. This method has been widely adopted to study the impacts of global warming on regional climate systems (e.g., Giorgi et al., 2001). Method-R Method-R is a proposal newly made in the current study, and was employed to prevent a bias in GCMs that has been the main concern in evaluations of regional climate predictions. The six-hourly NCEP/NCAR reanalysis data was used as a RCM forcing in the RCM.ncepCTL run for the period from 1993 to 2003, which represents the present climate experiment, and aims at demonstrating the ability of the TERCRAMS to reproduce the regional climate. In the RCM.ncepWM run, a new forcing dataset, as mentioned below, was used to simulate the regional climate influenced by global warming. Monthly averages of the ten-year mean difference between the control run (corresponding to 1991–2000) and the A2 scenario run (corresponding to 2071–2080) of MRI-CGCM2 was calculated for each 2.5 grid and 17 pressure levels (hereafter GWMD: Global Warming Monthly mean Difference) which represents the change of spatial structure in July due to global warming. The GWMD of the wind speed, the temperature, the geopotential height, the specific humidity, and the sea surface temperature are time-independently superimposed on each variable of the six-hourly NCEP/NCAR reanalysis data and the monthly mean SST data as a perturbation from the current weather conditions from 1993 through 2003. The forcing dataset specially prepared and generated by this process shows (but not shown here) very similar spatiotemporal variations with NCEP/NCAR reanalysis data, but it contains large-scale changes of temperature, humidity, static stability, zonal flow, and baroclinic instability in July as predicted by MRI-CGCM2. Hence, the RCM.ncepWM run is expected to simulate cyclones, troughs, and cut-off lows with basically the same structures during the period from
1993 through 2003 except that time-independent GWMD are added in the present climate as the perturbation induced by global warming. The advantages and disadvantages of method-R compared with method-G are addressed in ‘‘Discussion’’.
Results Reproduction of present climate Fig. 2 compares the mean rainfall distributions in July over Mongolia. July precipitation has its maximum in the northern part of Mongolia >100 mm, which gradually decreases going southward and westward with the minimum precipitation over the southern and western parts that have less than 40 mm. RCM.gcmCTL was able to simulate the meridional gradient of rainfall in the study area, although the amount of precipitation is underestimated in the southern and eastern parts of the domain, and overestimated in the western part of Mongolia (Fig. 2c). The horizontal distribution and the amount of the rainfall produced in RCM.ncepCTL (Fig. 2b), however, agrees very well with those in Fig. 2a in the southern desert region and also in the northern mountainous region, although rainfall is overestimated around the mountain peaks in a way similar to RCM.gcmCTL. These figures indicate that the nested RCM within the GCM still has some difficulties simulating the current regional climate, although it can properly simulate the larger scale characteristics of rainfall distribution. For the quantitative assessment of rainfall distribution, the Mongolian territory is divided into four subregions, the northwest (NW), the northeast (NE), the southwest (SW), and the southeast (SE) regions, by the lines along 47N and 104E as shown in Fig. 1. Fig. 3 shows the mean precipitation with the standard deviation for the calculated periods in individual subregions. RCM.ncepCTL reproduced the regional characteristics in Mongolia well, with more precipitation in the northern two subregions and less precipitation in the southern two subregions, although the model overestimates rainfall by about 30% in the NW and SW regions where the topography is complicated and also in the SE region where there is the Gobi desert. On the other hand, RCM.gcmCTL failed in a quantitative simulation of the subregional precipitation, although it quite well reproduced the total averaged amount of rainfall over the whole of Mongolia. The overestimation in the SW subregion comes from the result that too much precipitation occurs over mountains. In the MRI-CGCM2 control run, air temperature in the troposphere in July over eastern Mongolia is lower than that in the NCEP reanalysis, which is one reason for
148
T. Sato et al.
Figure 3 Mean precipitation in July for four subregions (NW, NE, SW, and SE) shown in Fig. 1 and stations in all of Mongolia, as also shown in Fig. 1 (ALL). The white bars depict observations while the light and dark shaded bars are results calculated by the RCM.ncepCTL and RCM.gcmCTL runs, respectively. The error bars show the standard deviation of precipitation in the observed 11-years and that as simulated in 11 years by the RCM.ncepCTL and 10 years by the RCM.gcmCTL.
Figure 2 Rainfall distributions in July. Contours are drawn for every 20 mm month 1. (a) Eleven-year mean by local meteorological stations from 1993 through 2003. Estimated precipitation by GPCP (Adler et al., 2003) was used to draw the contours outside the Mongolian boundary. Open circles represent stations used in this study. (b) Eleven-year mean by the RCM.ncepCTL run from 1993 through 2003. (c) Ten-year mean by the RCM.gcmCTL run from 1991 through 2000.
the underestimation in the SE and NE subregions in the RCM.gcmCTL run. Fig. 4 shows and compares the interannual variations of July precipitation obtained by observation and by RCM.ncepCTL. Only the ten-year average rainfall is shown for the RCM.gcmCTL run since the target years of the RCM.gcmCTL run do not necessarily correspond to a specific observed year. For example, the year 1994 in RCM.gcmCTL does not exactly correspond to the year 1994 observation, and so they cannot be compared with each other. The RCM.ncepCTL run simulated successfully the interannual variation of the July precipitation in the whole of Mongolia. In the years 1993, 1994 and 1998, more rainfall was observed than in the other years, while the precipitation was less in the dry years from 1999 to 2002. The RCM.ncepCTL overestimated rainfall by nearly 100% during the dry years except for 2002. Although one may attribute some parts of the error to the simple assumptions in the assumed initial soil moisture, we confirmed that the simulated overestimated precipitation still remains large even if a drier soil condition was used in the test. Additionally, the
interannual variation of precipitation was simulated well except for 1999, 2000, and 2001, despite the assumed simple initial soil moisture. This indicates that the initial soil moisture condition is not always important for simulations of the interannual variation of July precipitation. A possible cause for so much rainfall in 2001 might be in the reanalysis data; but further investigation is needed to clarify this point. A probability density distribution (PDD) of the daily rainfall in July is shown in Fig. 5. In Mongolia, July precipitation is mostly maintained by rainfall events stronger than 4 mm day 1. The contribution of the strongest rainfall events, over 32 mm day 1, is largest in the NE subregion. In the southern two subregions, all intensity categories are smaller than those in the northern two subregions. RCM.gcmCTL overestimated weak rainfalls, and underestimated heavy rainfalls. Such a tendency can be seen not only in the whole of Mongolia but also in all subregions. Underestimation of heavy rainfall events in the NE and SE subregions (Fig. 5c and e) by RCM.gcmCTL caused it to underestimation these for all of Mongolia (Fig. 5a) and an underestimation of the total subregional rainfalls in the NE and SE (Fig. 3). These results imply that rainfall intensity in Mongolia cannot be captured by method-G. On the other hand, the PDDs estimated by the RCM.ncepCTL agreed quite well with PDDs observed for all of Mongolia, although the frequency of the weak rainfalls of less than 8 mm day 1 was slightly overestimated. The RCM.ncepCTL very accurately simulates the largest three categories with intensities stronger than 8 mm day 1, and which contribute to a large part of the July precipitation in Mongolia. The estimated rainfall intensities in each of the four subregions also agreed with the observed intensities; meaning that TERC-RAMS can reproduce the precipitation properties even in a small target area, as shown on a basin scale in Mongolia.
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
149
Figure 4 Year to year comparison of precipitation in July between observations and RCM.ncepCTL averaged over all of Mongolia. Error bars indicate the standard deviation for the 65 stations shown in Fig. 1. Bars on the right side of the figure show the 11-year mean precipitation from observation, and the RCM.ncepCTL with a standard deviation during 11 years and RCM.gcmCTL with a standard deviation during 10 years.
Figure 5 Probability density distribution of daily precipitation intensity obtained by observations (solid line with open square), the RCM.ncepCTL (broken line with light-shaded square) and the RCM.gcmCTL (dot-dash-line with dark-shaded square). Bars show
150
T. Sato et al.
Rainfalls of heaviest intensity class of >32 mm day 1 occur more frequently in the NE subregion than in the other subregions. However, the categories from 4 to 32 mm day 1 contribute to most of the total rainfall in the southern two subregions. These features of rainfall intensity and regional distribution are well simulated in the RCM.ncepCTL run, its biggest advantage being its excellent ability to reproduce the current regional climate, whereas only the averaged rainfall amount for all of Mongolia is valid in the product of the RCM.gcmCTL run. Although method-G has widely applied in previous studies, a method for the correct reproductivity of the present-day climate needs to be investigated before addressing the changes due to global warming. In the RCM.ncepCTL run, daily or longer temporal variations of meteorological variables such as temperature, humidity, wind speed, and pressure in the atmosphere are basically dependent on reanalysis data. The RCM.ncepCTL estimated the monthly mean surface air temperature with less than 1 C error as compared by the average of all stations and the corresponding grid values in the model. Daily maximum and minimum temperature were underestimated by approximately 3 C and 1 C, respectively.
Changes of spatial distribution by two DDS methods The predicted changes in the amount of precipitation from the impact of global warming in Mongolia are shown in Fig. 6 (future minus present). Both DDS methods produced results that indicate a statistically significant decrease of precipitation in the northern part of Mongolia. Especially, rainfall especially decreases over the Hangai and the Khenti Mountains. Contrarily, rainfall slightly increases around the Gobi Desert in southern Mongolia. The signs indicating precipitation changes are different in western Mongolia when comparing results between the two methods; method-R produced a decreasing precipitation, while methodG predicted an increase with small areas of decreasing trends, like a complex mosaic. The two methods show a different trend in eastern Mongolia where method-G indicates very little change in the July precipitation, but a decrease can be seen in the results of method-R. Fig. 7 shows the rainfalls estimated for the four subregions by the two DDS methods. The effects of global warming on precipitation are negligibly small in the SE and the SW subregions in both methods. In method-R, rainfall decreases by 15% in the NW and the NE subregions leading to a decrease in the averaged rainfall for all of Mongolia. Method-G also predicts decreases in the NW and the NE subregions, which are mountainous areas (Fig. 1) where the sources of most of Mongolia’s rivers are situated (Tsujimura et al., this issue). Therefore, the decrease of precipitation in these subregions could present very serious concerns for river water management.
Figure 6 Horizontal distribution of the precipitation difference (future minus present) in July by the two DDS methods (mm month 1). Negative and positive values indicate decreases and increases of precipitation. Solid contours and broken contours are drawn for increases and decreases of July precipitation with every 20 mm month 1 interval. (a) Method-G (RCM.gcmA2 minus RCM.gcmCTL). (b) Method-R (RCM.ncepWM minus RCM.ncepCTL).
Changes of rainfall characterization
Figure 7 Same as in Fig. 3 but for the RCM.ncepCTL (light shade), the RCM.ncepWM (hatched light shade), the RCM.gcmCTL (dark shade), and the RCM.gcmA2 (hatched dark shade).
With method-R, as shown in ‘‘Reproduction of present climate’’, the model bias for precipitation is small enough and is comparable with the amplitude of changes caused by global warming, although the former is usually much lar-
ger than the latter because of limitations in the accuracy of the model. This section describes how rainfall properties are changed due to global warming.
Projection of global warming onto regional precipitation over Mongolia using a regional climate model Method-R allows us to test the impact of global warming for the past several years by giving GWMD as perturbation components to the reanalysis data. Fig. 8 compares the interannual variations of the averaged precipitation for all of Mongolia for both RCM.ncepCTL and RCM.ncepWM. Due to the impact of the large-scale changes caused by global warming, precipitation decreases in most years. In particular, the decrease is very large in 1993 and 1994 which are wet years. The standard deviation of the interannual variation of July precipitation tends to be smaller than that in the present climate. According to the IPCC report (Houghton et al., 2001), the frequencies of extreme events such as drought and flood, are likely to increase worldwide due to the impact of global warming. Theoretically, method-R is not expected to be able to precisely estimate changes in frequency since only GWMD, which was calculated as the difference of the ten-year averages of the monthly means between the present climate in 1990s and the future climate in 2070s, was taken into account without consideration of nonlinear effects such as the change in the range of interannual variability and the shift of storm tracks. Some of the limitations of method-R are addressed in the next section. Estimated projections of global warming onto rainfall intensities by method-R and method-G are shown in Fig. 9. Changes of the rainfall intensity by method-R can be described as follows. The frequency of weaker rainfall does not change, while that of the rainfall events stronger than 4 mm day 1 tends to decrease. In all subregions, the estimated changes in frequency of the weak rainfall categories are very small. Rainfalls in all categories exceeding 4 mm day 1 decrease uniformly in the NW subregion. In the NE and SE subregions, decreases of heavier rainfall events, of over 16 mm day 1 are prominently. On the other hand, rainfalls in the heaviest category of over 32 mm day 1 increase from the effects of the global warming in just the SW subregion. On the other hand, with method-G results, the increases in rainfall in the heavier rainfall categories of more than 8 mm day 1, were found when the rainfall intensity was investigated for all of Mongolia (Fig. 9a). The increases in
151
those categories are prominent in the NE and SE subregions, which may indicate changes of the storm tracks due to global warming. However, reproductivity of the present-day climate in these subregions is not very good in RCM.gcmCTL, especially for strong rainfall events (Fig. 5c and e). Thus, method-G probably cannot evaluate any increases of strong rainfall events in the NE and the SE subregions. Potential evaporation tends to increase from the effects of global warming. This leads to reduced soil moisture, when the increment of evapotranspiration can be larger than that of precipitation (Manabe and Wetherald, 1987). In Fig. 6, an increase of July precipitation was estimated by method-R and method-G over southern Mongolia. However, in both methods the potential evapotranspiration also increases due to rising air temperatures. As a result, soil moisture in July decreases in most parts of Mongolia (Fig. 10). Especially in northern Mongolia, the combined effects of increased evapotranspiration and decreased precipitation cause a severe decrease in soil moisture in July.
Discussion Interannual variations of precipitation over Mongolia were discussed in ‘‘Changes of rainfall characterization’’. As presented in Fig. 8, the amplitude of the variations, which exceeded 50 mm during 1993 and 2003, is much larger than the signals that we focus on, since the 11-year mean precipitation difference caused by global warming, as estimated by both methods, is below 10 mm. In order to address this problem by method-G, several ten-year averages are required for both the present climate and that resulting after global warming. In method-R, on the other hand, the complexity of such treatments can be much reduced. Precipitation changes can be estimated by a single projection of global warming onto just one typical year, e.g., a comparison between the simulated precipitation in 1993 and in ‘‘pseudo’’ 1993 under global warming, as a sensitivity experiment. In this study, the GWMD variables, which were calculated as the differences of the monthly averages of the ten-year means between 1991– 2000 and 2071–2080, were given in the RCM.ncepWM run
Figure 8 Same as in Fig. 4 but for the RCM.ncepCTL (light shade), the RCM.ncepWM (hatched light shade), the RCM.gcmCTL (dark shade), and the RCM.gcmA2 (hatched dark shade).
152
T. Sato et al.
Figure 9 Same as in Fig. 5 but for the RCM.ncepCTL (light-shaded square, light-shaded bar), the RCM.ncepWM (light-shaded triangle, hatched light-shaded bar), the RCM.gcmCTL (dark-shaded square, dark-shaded bar), and RCM.gcmA2 (dark-shaded triangle, hatched dark-shaded bar). (a) All Mongolia, (b) NW, (c) NE, (d) SW, (e) SE.
by assuming that the global warming trend would be a given constant from 1993 to 2003. However, as mentioned in ‘‘Changes of rainfall characterization’’, the sensitivity of the GWMD on precipitation is not a constant but varies with the year, even though a constant GWMD is given for all years. Precipitation decreased greatly in years 1993 and 1994, while the decrement was very small in years 1995 and 2003, and precipitation tended to increase in year 2002. These results indicate that the impact of global warming upon regional precipitation is quite complicated, having a different sensitivity for each year, which may have contributed to the changes of interannual variability. RCMs are capable of reproducing synoptic- and meso-scale cyclogeneses that develop and decay in the numerical domain. Therefore, method-R can to some extent simulate the effect of global warming on interannual variation, by taking account of the changes in the structures of cyclogeneses by temperature rises and increased moisture.
On the other hand, method-R is only partially applicable for estimations of long-term trends in interannual variability, because it does not fully consider the nonlinear interaction between global warming and atmospheric disturbances. Therefore, this method can not estimate the projection of global warming onto such factors as storm frequency, storm intensity, and the positions of storm tracks, which all interacts with the large-scale climate system beyond the model boundary. Such limitations of method-R may also explain the contradictory results from the two DDS methods. In ‘‘Results’’, we showed an excellent performance in reproducing the present-day precipitation by a RCM, which is a major advantage of method-R. Model bias in the GCM, that often reduces reproductivity of regional climate, is limited to the GWMD variables estimated using the ten-year averages of the monthly means of GCM products. Thus, the meteorological variables in method-R, such as precipitation and temperature, are more credible under the effects of global warming in their absolute value than those in method-G.
Projection of global warming onto regional precipitation over Mongolia using a regional climate model
153
Conclusion
Figure 10 Change of soil water content at the shallowest layer in July (%; future minus present) by (a) method-G and (b) method-R. Shaded regions indicate decrease of soil moisture.
Projections of global warming onto regional precipitation in Mongolia should differ according to the choice of GCM, greenhouse gas emission scenarios, and analyzed periods. Model ensemble has been regarded as the most reliable prediction in the Northeast Asia region. In consideration of ensemble experiments for regional climate simulation, method-R has an advantage in its data handling of GCMs products. To carry out a DDS projection of global warming, method-R requires GCM data only in monthly-mean form, while an at least twelve-hourly GCM product is demanded for method-G. Accordingly, method-R is much easier to compare DDSs products across many GCMs and/or several emission scenarios, and thus, more applicable to an ensemble experiment. According to Kitoh et al. (2005), precipitation increases in spring while it decreases in summer over northern China (35–45N, 100–120E) under the effects of global warming during 2081 to 2100 as revealed by MRI-CGCM2. They indicated that the GWMD of soil moisture had a seasonal cycle. Regional projection of global warming onto seasonal cycles of precipitation and soil moisture will be carried out in a future study in a further assessment of water resources. The high-resolution regional projection obtained by the RCM is applicable to a hydrological model that aims to study river runoff and groundwater regimes. Additionally, the roles of the high latitude ecosystem in the global scale carbon cycle have recently been focused on (White et al., 2000). A projection, as given in this study, can also be used for a terrestrial carbon cycle model, which allows investigation of the influences of activities such as grazing upon the terrestrial environments in Mongolia.
A dynamical downscaling (DDS) method using a regional climate model (RCM) was applied to investigate rainfall changes across Mongolia due to global warming. The DDS was carried out by two methods; method-R that uses sixhourly reanalysis data for a present-day simulation and newly prepared six-hourly dataset in which the changes of the monthly averages of ten-year mean variables by GCM due to global warming were added into the reanalysis data to prevent GCM bias in the regional climate. MethodG uses six-hourly GCM calculations for both the present and the future climates under the effects of global warming. Method-R successfully reproduced rainfall distribution in July even in the individual subregions in Mongolia. Rainfall intensity was found to be very similar to the in situ observations even for the four subregions in Mongolia. The method-G reproduced large-scale features of rainfall distribution; however, it failed to simulate stronger rainfall events. Projections of global warming onto regional precipitation levels obtained using the two DDS methods were compared. Changes in regional precipitation in both methods agreed in their spatial patterns that indicated decreasing precipitation over northern Mongolia and slight increases over southern Mongolia. Under method-R, rainfall tended to decrease for most of the recent years as shown by the sensitivity experiments of RCM.ncepWM in which large scale changes of meteorological variables (GWMD) are considered. In method-R, the reduction in middle to heavy rainfall events, over 4 mm day 1, dominated over northern Mongolia but caused a remarkable decrease of July precipitation in this region. Soil moisture was found to decrease in July over most of all Mongolia, because of the coupling effect of less precipitation and more evaporation due to the rising air temperature; a fact that implies severe droughts may occur more frequently in the future. Detailed evaluations on changes in hydrological regimes including river runoff and ground water are needed to further address this issue using a downscaled climate dataset.
Acknowledgments The authors appreciate the help of M. Hara of the Frontier Research Center for Global Change for data processing, and T. Kurokawa of the Terrestrial Environment Research Center, University of Tsukuba for assistance with RCM simulation. Appreciation is also extended to the ICCAP (Impact of Climate Changes on Agricultural Production System in the Arid Areas) Project of Research Institute for Humanity and Nature for the improvement of TERC-RAMS. Precipitation data at meteorological stations have been provided by the Institute of Meteorology and Hydrology, Mongolia. The authors thank anonymous reviewers for their suggestions and comments which greatly improve the manuscript. This study was supported by a CREST project (RAISE, Rangelands Atmosphere–hydrosphere–biosphere Interaction Study Experiment in Northeastern Asia) of JST (Japan Science and Technology Agency).
154
References Adler, R.F., Huffman, G.J., Chang, A., Ferraro, R., Xie, P., Janowiak, J., Rudolf, B., Schneider, U., Curtis, S., Bolvin, D., Gruber, A., Susskind, J., Arkin, P., 2003. The version 2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979-present). J. Hydrometeorol. 4, 1147–1167. Arakawa, A., Schubert, W.H., 1974. Interaction of a cumulus cloud ensemble with the large-scale environment, Part I. J. Atmos. Sci. 31, 674–701. Batima, P., Dagvadorj, D. (Eds.), 2000. Climate Change and Its Impacts in Mongolia. JEMR Publishing, Mongolia, p. 227. Dickinson, R.E., Henderson-Sellers, A., Kennedy, P.J., Wilson M.F., 1986. Biosphere atmosphere transfer scheme (BATS) for the NCAR community climate model. NCAR Technical Note NCAR/ TN275+STR, 69p. Giorgi, F., Hewitson, B., 2001. Regional climate information – evaluation and projections. In: Houghton, J.T. et al. (Eds.), Climate Change 2001: The Scientific Basis. Cambridge University Press, p. 881 (Chapter 10). Houghton, J.T. Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., Dai, X., Maskell, K., Johnson, C.A., 2001. Climate change 2001: the scientific basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 881p. Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Y., Leetmaa, A., Reynolds, B., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K.C., Ropelewski, C., Wang, J., Jenne, R., Joseph, D., 1996. The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc. 77, 437–472. Kato, H., Nishizawa, K., Hirakuchi, H., Kadokura, S., Oshima, N., 2001. Performance of RegCM2.5/NCAR-CSM nested system for the simulation of climate change in East Asia caused by global warming. J. Meteorol. Soc. Jpn. 79, 99–121. Kitoh, A., Hosaka, M., Adachi, Y., Kamiguchi, K., 2005. Future projections of precipitation characteristics in East Asia simulated by the MRI CGCM2. Adv. Atmos. Sci. 22, 467–478. Loveland, T.R., Reed, B.C., Brown, J.F., Ohlen, D.O., Zhu, J., Yang, L., Merchant, J.W., 2000. Development of a global land cover
T. Sato et al. characteristics database and IGBP DISCover from 1-km AVHRR data:. Int. J. Remote Sensing 21, 1303–1330. Manabe, S., Wetherald, R.T., 1987. Large-Scale changes of soil wetness induced by an increase in atmospheric carbon dioxide. J. Atmos. Sci. 44, 1211–1236. Nakicenovic, N., Swart, R., 2000. Special Report on Emissions Scenarios: A Special Report of Working group III of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 599p. Reynolds, R.W., Rayner, N.A., Smith, T.M., Stokes, D.C., Wang, W., 2002. An improved in situ and satellite SST analysis for climate. J. Climate 15, 1609–1625. Sato, T., Kimura, F., 2005a. Impact of diabatic heating over the Tibetan plateau on subsidence over Northeast Asian arid region. Geophys. Res. Lett. 32, L05809. doi:10.1029/2004GL02208. Sato, T., Kimura, F., 2005b. Diurnal cycle of convective instability around the central mountains in Japan during the warm season. J. Atmos. Sci. 62, 1626–1636. Sugita, M., Asanuma, J., Tsujimura, M., Mariko, S., Lu, M., Kimura, F., Azzaya, D., Adyasuren, Ts., this issue. An overview of the rangelands atmosphere-hydrosphere–biosphere interaction study experiment in northeastern Asia (RAISE). J. Hydrol. Tsujimura, M., Tanaka, T., Abe, Y., Shimada, J., Higuchi, S. Yamanaka, T., Saito T., this issue. Interaction between groundwater and river water in a semi-arid region – a case in Kherlen River basin, eastern Mongolia. J. Hydrol. Walko, R.L., Cotton, W.R., Meyers, M.P., Harrington, J.Y., 1995. New RAMS cloud microphysics parameterization. Part 1: the single-moment scheme. Atmos. Res. 38, 29–62. Wang, Y., Leung, L.R., McGregor, J.L., Lee, D.-K., Wang, W.-C., Ding, Y., Kimura, F., 2004. Regional climate modeling: progress, challenges, and prospects. J. Meteorol. Soc. Jpn. 82, 1599– 1628. White, A., Cannell, M.G.R., Friend, A.D., 2000. The high-latitude terrestrial carbon sink: a model analysis. Global Change Biol. 6, 227–245. Yukimoto, S., Noda, A., Kitoh, A., Sugi, M., Kitamura, Y., Hosaka, M., Shibata, K., Maeda, S., Uchiyama, T., 2001. The new Meteorological Research Institute coupled GCM (MRI-CGCM2) – model climate and variability. Pap. Meteorol. Geophys. 51, 47–88.