Pure Appl. Geophys. 174 (2017), 4251–4270 Ó 2017 Springer International Publishing AG DOI 10.1007/s00024-017-1641-8

Pure and Applied Geophysics

Coupling of Community Land Model with RegCM4 for Indian Summer Monsoon Simulation R. K. S. MAURYA,1 P. SINHA,1 M. R. MOHANTY,1 and U. C. MOHANTY1 Abstract—Three land surface schemes available in the regional climate model RegCM4 have been examined to understand the coupling between land and atmosphere for simulation of the Indian summer monsoon rainfall. The RegCM4 is coupled with biosphere–atmosphere transfer scheme (BATS) and the National Center for Atmospheric Research (NCAR) Community Land Model versions 3.5, and 4.5 (CLM3.5 and CLM4.5, respectively) and model performance is evaluated for recent drought (2009) and normal (2011) monsoon years. The CLM4.5 has a more distinct category of surface and it is capable of representing better the land surface characteristics. National Centers for Environmental Prediction (NCEP) and Department of Energy (DOE) reanalysis version 2 (NNRP2) datasets are considered as driving force to conduct the experiments for the Indian monsoon region (30°E– 120°E; 30°S–50°N). The NNRP2 and India Meteorological Department (IMD) gridded precipitation data are used for verification analysis. The results indicate that RegCM4 simulations with CLM4.5 (RegCM4–CLM4.5) and CLM3.5 (RegCM4–CLM3.5) surface temperature (at 2 ms) have very low warm biases (*1 °C), while with BATS (RegCM4–BATS) has a cold bias of about 1–3 °C in peninsular India and some parts of central India. Warm bias in the RegCM4–BATS is observed over the Indo-Gangetic plain and northwest India and the bias is more for the deficit year as compared to the normal year. However, the warm (cold) bias is less in RegCM4–CLM4.5 than other schemes for both the deficit and normal years. The model-simulated maximum (minimum) surface temperature and sensible heat flux at the surface are positively (negatively) biased in all the schemes; however, the bias is higher in RegCM4–BATS and lower in RegCM4–CLM4.5 over India. All the land surface schemes overestimated the precipitation in peninsular India and underestimated in central parts of India for both the years; however, the biases are less in RegCM4–CLM4.5 and more in RegCM4–CLM3.5 and RegCM4–BATS. During both the years, BATS scheme in RegCM4 failed to represent low precipitation over the leeward than windward side of the Western Ghats, while CLM schemes (both versions) in the RegCM4 are able to depict this feature. In the topographic regions, such as the Western Ghats, northeast India and state of Jammu and Kashmir, RegCM4–BATS overestimates the rainfall amount with higher

Electronic supplementary material The online version of this article (doi:10.1007/s00024-017-1641-8) contains supplementary material, which is available to authorized users. 1

School of Earth, Ocean and Climate Sciences (SEOCS), Indian Institute of Technology (IIT) Bhubaneswar, 309 Basic Science Building, Bhubaneswar, Odisha 752050, India. E-mail: [email protected]

bias. Statistical analysis using anomaly correlation coefficient, root mean square error, equitable threat score, and critical success index confirms that RegCM4–CLM performs better than RegCM4–BATS in the simulation of the Indian summer monsoon. Key words: Land surface schemes, CLM, BATS, RegCM4, Indian summer monsoon.

Abbreviations BATS Biosphere atmosphere transfer scheme CCM Community climate model CLM Community land model CSI Critical success index DOE Department of Energy ETS Equitable threat score GCM General circulation model GLCC Global land cover characterization GPP Gross primary product ICTP International Center for Theoretical Physics IMD India Meteorological Department ISMR Indian summer monsoon rainfall JJAS June July August September LSM Land surface model LSS Land surface schemes MM5 Mesoscale Model Version 5 NCAR National Center for Atmospheric Research NCEP National Center for Environmental Prediction NEI North East India NNRP2 NCEP/NCAR reanalysis 2 NWI North West India OISST Optimum interpolation sea surface temperature PSU Pennsylvania State University RCM Regional climate model RegCM4 Regional climate model version 4 T2M Surface air temperature

4252

SHF PI RH2M TIO USGS WG WH WRF

R. K. S. Maurya et al.

Sensible heat flux Peninsular India Surface relative humidity Tropical Indian ocean United States Geological Survey Western Ghats Western Himalayas Weather Research and Forecasting

1. Introduction The Indian summer monsoon rainfall (ISMR) contributes more than 70% of the total annual precipitation in India and controls the major agricultural productivity and economy of the country (Parthasarathy et al. 1994). However, the spatiotemporal variability of monsoon rainfall is high due to complex interactions between largescale flow and small-scale physical processes. The land surface feedback to the atmosphere is one of the major concerns in modulating local and regional-scale climate systems, as reported in various studies (Pielke et al. 2002, 2011; Roy et al. 2007; Niyogi et al. 2007; Douglas et al. 2009). Land surface acts as the lower boundary for the atmosphere and it regulates the energy partition and water balance and influences the regional climate. The land surface characteristics in India are highly heterogeneous and a complex system having almost all kinds of land properties such as desert, forests, plain land, rivers, semi-arid zones, mountains, lakes, etc. Thus, the rainfall over India is associated with a multiplex structural system of land–ocean–atmosphere interactions that have led to the difficulty in understanding and prediction of ISMR. In order to understand the role of land surface, various dynamic vegetation/land surface models have been coupled to global climate models to account for the impact of vegetation–climate feedback (Foley et al. 1998; Levis et al. 1999; Cox et al. 2000; Wang 2004; Delire et al. 2004; Crucifix et al. 2005; Bala et al. 2006). In the early 1980s, Shukla and Mintz (1982) and Yeh et al. (1984) studied and suggested a strong sensitivity of the climate response with respect to soil moisture anomalies. Zhang et al. (2008)

Pure Appl. Geophys.

studied the influence of land surface processes in the interannual scale over contiguous USA. The land– atmosphere coupling made a significant contribution to interannual summer temperature and precipitation variability. The coupling also captured the variability of total precipitation over northern USA. There are many other studies which imply the importance of land surface processes in the prediction of precipitation using atmospheric models (Fennessy and Shukla 1999; Koster et al. 2004, 2006, 2010, 2011; Guo et al. 2011; van den Hurk et al. 2012). Steiner et al. (2009) showed that land surface coupling can influence the regional circulation and precipitation over the regions exhibiting strong hydro-climatic gradients. The global coupling of land surface and atmosphere shows a wide spectrum of spatial and temporal variations. The global scale coupling of soil moisture and precipitation was investigated by Koster et al. (2004, 2006). They identified the global coupling hotspots, where soil moisture interaction directly influences summer precipitation variability. It was found that the Indian summer monsoon region is one of the hot spots of the land surface coupling. They have also shown that the land surface coupling is found to be strong over the transition zones between wet and dry climate regimes. A few global climate models have been validated at the regional scale especially for the Indian subcontinent. The representation of land surface characteristics is poor in the global climate models due to its coarse resolution (Wood et al. 2004); as a result, the feedback from vegetation/landcover to the atmosphere is less accurate. The land–air interaction plays a crucial role during the Indian summer monsoon (ISM) season when the climate is highly sensitive to vegetation and biogeophysical processes (Xue and Shukla 1993; Paeth et al. 2009; Xue et al. 2010). Therefore, there is a need to study the impacts of regional land surface feedback to understand the ISMR and its variability. Regional impact assessment requires climate information that is provided at a much finer spatial resolution and is accurate at the regional scale. Regional climate model is capable of representing finer scale land-surface features such as vegetation, soil moisture/temperature, sea surface temperature, sea ice and snow cover better than GCMs that significantly affect surface energy and moisture

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

equilibrium on global, regional and mesoscale, thereby an useful modeling framework to understand and simulate regional climate (Sellers et al. 1996; Dickinson et al. 1986; Noilhan and Planton 1989; Henderson-Sellers et al. 1993; Castro et al. 2005; Eden et al. 2014). The regional climate models coupled with dynamic vegetation models are desirable tools for climate prediction over India during the Indian summer monsoon period. Fewer attempts have been made to understand the influence of land surface characteristics to simulate ISMR using regional climate models such as Mesoscale Model version 5 (MM5) and Weather Research and Forecasting (WRF) (Singh et al. 2007; Dutta et al. 2009; Kar et al. 2014). Experiments with the MM5 model suggest that the performance of the model varies significantly with the changes in land surface schemes for ISMR simulations (Singh et al. 2007), while incorporation of satellite-derived vegetation covers and fractions improve the model skill in simulating ISMR (Dutta et al. 2009). Changes in the vegetation green fraction reasonably alter the PBL as well as heat fluxes to the atmosphere with the use of WRF model (Kar et al. 2014). However, these models are developed with the focus of mesoscale simulations, and there is a need to examine the performance of various land surface schemes coupled with a regional climate model for ISMR simulations. The regional climate model (RegCM) is widely used by the research communities for simulating the regional/local climate and it is stated that the performance of the RegCM model is satisfactory for simulation of ISM (Dash et al. 2006; Sinha et al. 2013). In order to understand the role of a heterogenetic structure of the land surface and its interaction with the atmosphere for the simulation of ISMR, an advanced regional climate model like RegCM4 is on demand for coupling with the better land surface model (LSM) that is yet to be evaluated for summer monsoon season. Section 2 briefly describes the model specifications with different land surface schemes and datasets used for model initial and boundary conditions and validation. Results and associated discussions are provided in Sect. 3 and conclusion is presented in Sect. 4.

4253

2. Data and Methodology The dataset and methodology used in the present study are described in this section. 2.1. Data National Centers for Environmental Prediction (NCEP) and Department of Energy (DOE) reanalysis 2 dataset (2.5° 9 2.5°) (hereafter referred to as NNRP2) (Kanamitsu et al. 2002) are used to provide the model initial and lateral boundary conditions. The lateral boundary conditions in the RegCM4 model are updated every 6 h from NNRP2. The RegCM4 simulations used sea surface temperatures (SSTs) acquired from National Oceanic and Atmospheric Administration (NOAA) optimal interpolation weekly SST (OISST) data (1° 9 1° resolution; Reynolds et al. 2002). The Global Land Cover Characterization (GLCC) (Loveland et al. 2000) is a series of global land cover classification datasets of the United States Geological Survey (USGS) at 30-arc second resolution (https://lta.cr.usgs.gov/GTOPO30) which are used for representing the geophysical characteristics in the model. The daily gridded rainfall dataset at spatial resolution 0.25° 9 0.25° provided by India Meteorological Department (IMD) (Pai et al. 2014) are used for verification of precipitation. The NNRP2 data are used for validation of large-scale circulation and surface parameters of RegCM4 model products. 2.2. Methodology The methodology of the present study contains two subsections in which the first subsection describes the main features of the regional climate model RegCM4 used for the simulation of the ISM. In addition, a brief description of the different land surface schemes is provided in this subsection to understand the interaction processes between land and air of these schemes. In the second subsection, the method for the designing of the experiments is described. 2.2.1 Regional Climate Model RegCM4 The latest version (version 4.4) of the ICTP Reginal climate model (RegCM4) is used and a detailed

4254

R. K. S. Maurya et al.

description of the model can be found in Giorgi et al. (2012). RegCM4 is a limited-area model using a terrain-following r-pressure vertical coordinate system and an Arakawa B-grid finite differencing algorithm. The model’s dynamical component is from the hydrostatic version of the Pennsylvania State University (PSU) Mesoscale Model version 5 (MM5, Grell et al. 1994). The radiation scheme is derived from the National Center for Atmospheric Research (NCAR) Community Climate Model version 3 (CCM3: Kiehl et al. 1996) and includes representation of aerosols following Solmon et al. (2006) and Zakey et al. (2006). The model applies the non-local boundary layer scheme of Holtslag et al. (1990). The convection parameterization scheme over the land regions is the Grell scheme (Grell 1993) and over the ocean is the MIT-Emanuel scheme (Emanuel 1991; Emanuel and Zivkovic-Rothman 1999), a configuration which is referred to as mixed convection (Giorgi et al. 2012). Previous studies suggested that this mixed convection scheme performs better in simulating the Indian summer monsoon season (Dash et al. 2015), so the same mixed convection scheme is used in the present study. The physics in RegCM4 includes a large-scale cloud and precipitation scheme that accounts for cloud sub-grid variability (Pal et al. 2000). In order to consider the interactions between land surface and atmosphere in the modeling framework, the current version (version 4) of RegCM model (hereafter referred to as RegCM4) supporting three types of land surface schemes namely, the Biosphere–Atmosphere Transfer Scheme (BATS) (Dickinson et al. 1993), the Community Land Model version 3.5 (CLM3.5; Steiner et al. 2009) and version 4.5 (CLM4.5; Oleson et al. 2013) are used. BATS scheme is a state-of-the-art land surface scheme that is designed to describe the role of vegetation and interactive soil moisture in modifying the surface– atmosphere exchanges of momentum, energy, and water vapor. The second major improvement in the RegCM4 model has been made by incorporating the CLM model (CLM3.5). It uses a series of biogeophysically based parameterization to describe the land–atmosphere exchanges of energy, momentum, water, and carbon. In addition to the different surface processes that are represented in the BATS, CLM3.5

Pure Appl. Geophys.

has more features in the land surface characteristics (Oleson et al. 2004, 2008). More recent model development efforts have led to key improvements in the land surface parameterizations by inclusion of the latest version of CLM (CLM4.5), in which the most relevant ones such as climate–vegetation interactions concerned parameterization related to gross primary production (GPP) including canopy radiative transfer, photosynthesis, and stomatal conductance are considered to represent the finer scale heterogeneity characteristics and interactions with the atmosphere (Bonan et al. 2011, 2012). A brief description of the three land–surface schemes (LSS) namely, BATS, CLM3.5 and CLM4.5 is given below: The BATS scheme is added as a default land surface scheme in RegCM4 model for simulations. The BATS scheme has a vegetation layer, a snow layer, a surface soil layer with a thickness of 10 cm, root zone layer of 1–2 m thickness, and a third deep soil layer 3 m thick. Prognostic equations are solved for the soil layer temperatures using a generalized force-restore method of Deardorff (1978). The temperature of the canopy and canopy foliage is calculated diagnostically via an energy balance formulation including sensible, radiative, and latent heat fluxes. BATS has 20 categories of vegetation types and 17 types of soil textures ranging from coarse (sand) to intermediate (loam), to find (clay). Figure 1a shows the soil texture categories as defined in RegCM4 represented for the Indian landmass. The major improvement in representing land surface characteristics in RegCM4 is the inclusion of community land model (CLM) version 3.5. The CLM3.5 scheme divides the cell area first into a subgrid hierarchy composed of land units (glacier, wetland, lake, urban, and vegetated land cover), and a second and third sub-grid hierarchy for vegetated land units, including different snow/soil columns for the different vegetation fractions, and plant functional types (Oleson et al. 2004). The scheme includes a vegetation layer, a snow layer, a force-restore model for soil temperatures, and a 3-layer soil scheme. The bio-geophysical calculations in CLM3.5 include a coupled photosynthesis–stomatal conductance model, in-canopy radiation schemes, revised multi-layer snow parameterizations, and surface hydrology including a distributed river runoff scheme (Oleson

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

4255

Figure 1 a Soil texture dominant category in RegCM4. b RegCM4 model domain with topographical height (in shaded)

et al. 2008). Soil temperature and water content are calculated with the use of a multi-layer model. CLM3.5 substantially represents the land–atmosphere exchanges of moisture and energy and the associated surface climate feedbacks better as compared to BATS (Steiner et al. 2009). The updated version 4.5 of CLM is CLM4.5 that includes modified canopy processes including a revised canopy radiation scheme and canopy scaling of leaf processes, co-limitations on photosynthesis, revisions to photosynthetic parameters (Bonan et al. 2011, 2012), temperature acclimation of photosynthesis, and improved stability of the iterative solution in the photosynthesis and stomatal conductance model (Sun and Wang 2012). Hydrology updates include modifications such that hydraulic properties of frozen soils are determined by liquid water content only rather than total water content and the introduction of an ice impedance function, and other corrections that increase the consistency between soil water state and water table position and allow a perched water table above icy permafrost ground (Swenson et al. 2012). A surface water store is introduced, replacing the wetland land unit and permitting prognostic wetland distribution modeling. The surface energy fluxes are calculated separately (Swenson and Lawrence 2012) for snow-covered, water-covered, and snow/water-free portions of

vegetated and cropland units, and snow-covered and snow-free portions of glacier land units. Globally constant river flow velocity is replaced with variable flow velocity based on mean grid cell slope. A vertically resolved soil biogeochemistry scheme is introduced with base decomposition rates modified by soil temperature, water, and oxygen limitations and also including vertical mixing of soil carbon and nitrogen due to bioturbation, cryoturbation, and diffusion (Koven et al. 2013). 2.2.2 Method In this study, one severe monsoon drought year (2009) and one normal monsoon year (2011) are considered to understand the role of different land surface characteristics for the different monsoonal phenomenon. The observed seasonal summer monsoon rainfall during 2009 for the country as a whole was 78% (deficient rainfall) of long period average (LPA) and was the lowest recorded rainfall in recent decades. The 2011 southwest monsoon season rainfall over the country as a whole was normal (102% of LPA). For a particular season, if the summer monsoon rainfall is more (less) than 10% of LPA for all India average, then it is considered as excess (deficit) monsoon year, while the monsoon is considered as normal monsoon year if the all India

4256

R. K. S. Maurya et al.

Pure Appl. Geophys.

model simulates the ISM with the use of different land surface scheme separately. Basic statistical approaches like root mean square error, correlation coefficient, and bias are used to evaluate the performance of the model. Further, the model skill is evaluated using equitable threat score (ETS), critical success index (CSI) to examine the reliability of the model performance for the prediction of ISM. A brief description of the ETS and CSI is provided in the associated section.

average rainfall is within the range 90–110% of LPA. It may be noted here that India did not experience excess monsoon rainfall during the past two decades, thus the present study is confined to the deficit and normal monsoon years to understand the contrasting features of the interaction process between land surface and atmosphere. The RegCM4 model domain that is considered to simulate the ISM is shown in Fig. 1b. Table 1 shows the RegCM4 model configuration that covers the Indian monsoon region extending from 30°E to 120°E in the east–west and from 30°S to 45°N in the north–south directions. The model horizontal resolution is 27 km, and the vertical grids are composed of 18-r levels stretching from near the surface to the model top (10 hPa). All the simulations are conducted from 1st May to 1st October and the first month (May) is considered as a model spin-up time. For each year, the RegCM4

The results obtained from the model simulations are discussed in two broad subsections namely, (1) upper air circulation and surface parameters, and (2) precipitation.

Table 1

3.1. Upper Air Circulation and Surface Parameters

RegCM4 configuration used in this study

The large-scale circulation patterns, as well as surface meteorological variables obtained from the model simulations are compared to the NNRP2 and associated discussions are made in various subsections below:

No. of horizontal grid points

380 (along y-direction) 300 (along x-direction)

Model domain Initial condition No. of vertical levels Horizontal resolution Central longitude and latitude Time step Map-projection Model boundary conditions Cumulus convection scheme

30°E–120°E and 30°S–45°N 1st May 2009, 2011 18 r levels 27 km 77°E and 10°N

Convective closure assumption Radiation scheme Boundary layer scheme Moist physics scheme Ocean flux scheme Land surface physics

Sea surface temperature Data type

40 s Rotated Mercator (ROTMER) Relaxation, exponential technique Land: Grell scheme (Grell 1993) Ocean: MIT-Emanuel scheme (Emanuel 1991) Arakawa and Schubert (1974) Modified CCM3 (Kiehl et al. 1996) Holtslag PBL (Holtslag et al. 1990) Explicit moisture (SUBEX; Pal et al. 2000) Zeng et al. (1998) BATS (Dickinson et al. 1993) CLM3.5 (Steiner et al. 2009) CLM4.5 (Oleson et al. 2013) OISST (NOAA optimal interpolation weekly sea surface temperature) NNRP2 [National Centers for Environmental Prediction (NCEP) and Department of Energy (DOE) reanalysis]

3. Results and Discussion

3.1.1 Wind The low-level summer monsoon seasonal (June– September; JJAS) mean wind fields at 850 hPa pressure level from the verification analysis (NNRP2) and RegCM4 simulations with different LSS for 2009 and 2011 monsoon years are shown in Fig. 2. The important characteristics of winds at 850 hPa are the cross-equatorial southwest monsoon flows, i.e., trade winds from the southern hemisphere cross the equator near the Somali coast, forming the low-level jet stream known as Somali Jet. Previous studies have indicated that the strength of the southwest wind over the Arabian Sea (AS), specifically Somali Jet is positively correlated with the ISMR (Ju and Slingo 1995; Halpern and Woiceshyn 2001). The wind pattern having a maximum strength of the wind at Somali coast at 850 hPa is similar to that of NNRP2 in all the RegCM4–LSS simulations. The wind speeds (850 hPa) near the Somali coast are weaker

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

Figure 2 Seasonal (JJAS) mean wind magnitude (m s-1) and wind vector at 850 hPa pressure level obtained from a NNRP2, b RegCM4 simulations with BATS, c RegCM4 simulations with CLM3.5, and d RegCM4 simulations with CLM4.5 for the year 2009. Panels e–h are same as a–d, respectively, but for 2011

in 2009 drought year (Fig. 2a–d) as compared to 2011 normal year (Fig. 2e–h) and this variation is well depicted by the RegCM4 simulations. During the drought year (2009), the highest wind speeds (southwesterly) as observed over the AS are 16 m s-1 in NNRP2 (Fig. 2a), and 12, 14 and 16 m s-1 in RegCM4 simulations with BATS, CLM3.5, CLM4.5 (Fig. 2b–d), respectively, in 2009. For the year 2011, the wind speeds are 12–18 m s-1 in NNRP2 (Fig. 2e), 6–14 m s-1 in RegCM4–BATS (Fig. 2f), 6–14 m s-1 in RegCM4– CLM3.5 (Fig. 2g) and 10–16 m s-1 in RegCM4– CLM4.5 (Fig. 2h). The normal year (2011) has stronger southwesterly over the AS compared to deficient year (2009). The NNRP2 has stronger southwesterly winds at 850 hPa over the AS and

4257

Bay of Bengal (BoB) as compared with RegCM4– LSS, whereas RegCM4–CLM4.5 winds simulation at 850 hPa shows closest wind magnitudes to NNRP2 as compared to other RegCM4–LSS simulations. The JJAS mean wind field model bias of RegCM4–LSS from NNRP2 at 850 hPa in 2009 and 2011 (provided as Supplementary Figure 1) suggests that the RegCM4–LSSs have a negative bias over the Indian peninsular region and adjoining seas of AS and BoB in both the years. The bias using CLM4.5 is lower than BATS and CLM3.5 in low-level wind magnitude over the Indian and its adjoining regions in both the years. The mean wind pattern at 200 hPa as represented in RegCM4–LSS in 2009 and 2011 are nearly same throughout the domain as compared to NNRP2 for the corresponding years (provided as Supplementary Figure 2). The south peninsular India and adjoining seas have strong easterlies in the NNRP2 of 15–18 m s-1, whereas easterlies in RegCM4 simulated land-surface schemes are 9–12 m s-1 in both BATS and CLM3.5, and 12–15 m s-1 in CLM4.5 for both the years (Supplementary Figure 2). The seasonal mean wind bias at 200 hPa in RegCM4–LSS simulations from NNRP2 is shown in Fig. 3. The biases in the upper air wind are positive over the latitudes 30°S and 30°N (except for 30°N in RegCM4–CLM4.5) and negative bias in the latitudinal belts extending from the equator to 25°N for all the simulated land-surface schemes for both the years. However, the biases are lesser in RegCM4– CLM4.5 as compared to the other two RegCM–LSS. Overall, the RegCM4–CLM4.5 represents a better pattern of 850 hPa and 200 hPa wind direction and magnitude followed by RegCM4–CLM3.5 and RegCM4–BATS. 3.1.2 Surface Parameters/Characteristics (a) Surface air temperature (mean, maximum and minimum). The mean JJAS surface air temperature (hereafter referred to as T2M) obtained from the NNRP2, and RegCM4 simulations with BATS, CLM3.5, and CLM4.5 for both the years are represented in Fig. 4. The RegCM4–LSS biases for the mean

4258

R. K. S. Maurya et al.

Pure Appl. Geophys.

Figure 3 Spatial distributions of differences in JJAS mean wind at 200 hPa (m s-1) between a RegCM4–BATS minus NNRP2 (BATSNNRP2), b RegCM4–CLM3.5 minus NNRP2 (CLM3.5-NNRP2), c RegCM4–CLM4.5 minus NNRP2 (CLM4.5-NNRP2) for the year 2009. Panels d–f are same as a–c but for year 2011

(Fig. 5), maximum (figure not shown) and minimum (figure not shown) T2M with respect to NNRP2 are computed, for both the years 2009 and 2011. The RegCM4 model using all LSSs could represent JJAS mean T2M pattern close to NNRP2 over India. The RegCM4–BATS (Fig. 5a, d) simulation shows a cold bias of 1–3 °C in the foothills of the Himalayas, northeast India (NEI) and southern India and warm bias of 1–2 °C in the northwest India (NWI) and central northeast India (CNEI) as compared to NNRP2 in both the years. The computed biases in RegCM4–CLMs (Fig. 5b–c, e–f) compared to NNRP2 have shown an almost similar pattern over India as seen in RegCM4–BATS in both the years 2009 and 2011; however, the bias is less in CLM4.5. The highest cold bias is of 1–3 °C over the foothills of the Himalayas and cold/warm bias of 1 °C over most of the parts of India in both RegCM4–CLMs for both the years. The cold bias over foothills of the Himalayas of 1–3 °C may have been contributed due to high range of mountain orography impact with

Figure 4 Mean surface air temperature (°C) obtained from a NNRP2, b RegCM4 simulations with BATS, c RegCM4 simulations with CLM3.5, and d RegCM4 simulations with CLM4.5 for the year 2009. Panels e–h are same as a–d, respectively, but for 2011

sigma level coordinate used for RegCM4.4 simulation. The maximum/minimum T2M (figure not shown) shows a warm/cold bias of 2–6 °C in the RegCM4–BATS and 2 °C in the RegCM4–CLMs over the most parts of India in both the years 2009 and 2011. The mean, maximum and minimum T2M shows lowest cold/warm biases over the most parts of India in RegCM4–CLM4.5 followed by RegCM4– CLM3.5 and RegCM4–BATS for both the years 2009 and 2011. The CLM leads to lesser cold bias over the most parts of India in summer monsoon season as a compared to BATS. The main effect of the land surface scheme is on the simulation of the surface hydrologic cycle; a result noted and validated in other CLM studies (Bonan et al. 2002; Zeng et al. 2002; Dai et al. 2003). Soil moisture content in CLMs is

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

4259

Table 2 The mean, bias, root mean square error (RMSE) and correlation coefficients (CCs) of JJAS (122 days) surface temperature (°C) at 2-m height computed between NNRP2 and RegCM4 simulations with –BATS, –CLM3.5, and –CLM4.5 for 2009 and 2011 Years

2009 NCEP-DOE RegCM4–BATS RegCM4–CLM3.5 RegCM4–CLM4.5 2011 NCEP-DOE RegCM4–BATS RegCM4–CLM3.5 RegCM4–CLM4.5

Figure 5 Spatial distributions of mean surface air temperature differences between a RegCM4–BATS minus NNRP2 (BATS-NNRP2), b RegCM4–CLM3.5 minus NNRP2 (CLM3.5-NNRP2), c RegCM4–CLM4.5 minus NNRP2 (CLM4.5-NNRP2) for the year 2009. Panels d–f are same as a–c but for year 2011

higher than the BATS with much reduced rates of ground evaporation. Therefore, BATS has a different partitioning of moisture and has different surface water budget as compared to CLMs. On the other hand, BATS scheme produces a larger snow accumulation rate on the surface than CLM. The reduced CLM snow cover causes a reduction in surface albedo and increases the amount of radiation absorbed. This, in turn, increases the surface radiative energy fluxes, leading to higher seasonal T2M in CLM. The correlation coefficients (CCs) and root mean square error (RMSE) are two basic statistical methods for understanding the model’s simulation ability (Zhao and Luo 1997). Table 2 presents the mean, bias, RMSE and CCs of JJAS T2M in RegCM4–LSS simulations in comparison with the NNRP2 dataset for both years. The computations have been carried out each day after taking the area average of T2M over India using RegCM4–LSS simulated and NNRP2 datasets and then considering the mean of all 122 days (the whole JJAS period). The RegCM4

Temperature (°C) Mean

Bias

CCs

RMSE

26.75 25.62 26.49 26.27

-1.14 -0.26 -0.48

0.72 0.78 0.79

1.64 1.20 1.29

25.61 24.77 25.46 25.26

-0.84 -0.14 -0.35

0.62 0.70 0.78

1.35 1.04 0.94

shows a cold bias of 1.14 and 0.84 °C using BATS, 0.26 and 0.48 °C using CLM3.5, 0.14 and 0.35 °C using CLM4.5 for 2009 and 2011 years, respectively. Therefore, model evaluation using important statistical methods indicate that the RegCM4 simulation has the highest cold bias with BATS and lowest cold bias with CLM4.5. The CCs are 0.72/0.62, 0.78/0.70 and 0.79/0.78 in BATS, CLM3.5, and CLM4.5 for years 2009 and 2011, respectively. The correlation (bias) is higher (lower) in CLM4.5 than other land surface schemes in RegCM4. CLM3.5 performs better than BATS when compared the CCs and bias for both the years. The results discussed above indicate that both RegCM4–CLM3.5 and RegCM4–CLM4.5 have better simulated results for T2M as compared to RegCM4–BATS. (b) Surface relative humidity (%) and sensible heat flux (Wm-2). The surface (at 2 m above the surface) relative humidity (RH2M) differences between RegCM4– LSS from NNRP2 are shown in Fig. 6 for both years. The dry season has lowest (positive or negative) bias as compared to the normal year in summer monsoon season over India. The RH2M shows the highest positive bias of 10–20% over the western Himalaya (WH) and negative bias of 10–20% over the most parts of India in RegCM4–BATS and RegCM4– CLM3.5 for both the years. In general, BATS has the highest negative bias of 10–20% in 2009 and above

4260

R. K. S. Maurya et al.

Figure 6 Same as Fig. 5 but for near-surface relative humidity (%)

20% in 2011 over the most parts of India (except WH) as shown in Fig. 6a, d. The second highest negative bias of 10–20% is seen for CLM3.5 over the CNEI, NEI, east WCI and PI in 2009 (Fig. 6b). Negative biases with greater than 20% are seen over the CNEI, NEI, and PI in 2011 (Fig. 6e). However, CLM4.5 has the lowest positive/negative bias, with a negative bias of 10% over the south NWI and east CNEI in 2009 and 10–20% in 2011 over central India (Fig. 6c, f). The analysis of sensible heat flux (SHF) near the surface clearly indicates an overestimation of 10–50 Wm-2 over most of the Indian region in RegCM4–LSS simulations in comparison to NNRP2 (Fig. 7). The high biases in SHF are noticed in the RegCM4–BATS with an overestimation of 30–50 Wm-2 over central India, Western Ghats (WG), western Rajasthan and most parts of WH for both the years. The RegCM4–CLMs approximately have a similar pattern of biases as noticed in the RegCM4–BATS over same regions; however, the magnitude of biases are lesser in the RegCM4–

Pure Appl. Geophys.

Figure 7 Same as Fig. 5 but for sensible heat flux (Wm-2)

CLMs. Results illustrated that the RegCM4–CLMs have a positive bias of 10–30 Wm-2 over the WCI and PI in 2009; whereas, about 30 Wm-2 over WCI and 30–50 Wm-2 over the PI in 2011. RegCM4– CLMs show a negative bias of 10–30 Wm-2 over the state of Gujarat in 2009. (c) Surface moisture flux (g kg-1 ms-1). The surface moisture flux (SMF) is the product of surface-specific humidity and surface horizontal wind. The JJAS mean SMF obtained from the NNRP2, and RegCM4 simulations with BATS, CLM3.5, and CLM4.5 for both years 2009 and 2011 are represented in Fig. 8. The JJAS mean SMF (NNRP2) of 60–120 g kg-1 ms-1 is distributed over the southern India, 10–60 g kg-1 ms-1 over the northern India and 10 g kg-1 ms-1 over the WH for both years as shown in Fig. 8a, e. The mean JJAS SMF is less for the deficient year (2009) as compared to the normal year (2011). The RegCM4–BATS (Fig. 8b, f) simulates reduced SMF while RegCM4– CLM3.5 (Fig. 8c, g) and RegCM4–CLM4.5 (Fig. 8d,

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

4261

RegCM–CLM4.5 captures the soil moisture pattern better than other two RegCM–LSS. Over WGs, RegCM4–LSS have more reduced SMF as compared to NNRP2. On a large scale, the RegCM4–CLM has a better representation of SMF as compared to RegCM4–BATS. (d) Evapotranspiration, Soil Moisture and Surface Albedo.

Figure 8 Same as Fig. 4 but for surface moisture flux (g kg-1 ms-1)

h) simulate similar SMF as a compared to NNRP2 over the southern parts of India for both years. The soil moisture is the maximum near the coastal region which gradually reduces towards the northern India. The differences in soil moisture between the dry and normal monsoon year can be seen clearly over south PI. The soil moisture gradient is from the southern India to the tip of Himalayas. For the dry year, both RegCM–BATS and RegCM–CLM3.5 fail to capture the pattern of soil moisture during monsoon season. The soil moisture flux in the RegCM–CLM4.5 simulation is closer to NNRP2 as compared to the other two RegCM–LSS during the year 2009. The soil moisture flux over southern India is captured better in CLM4.5. During the year 2011, there is high soil moisture flux in southern India as compared to 2009. During the normal monsoon year,

It is of interest to conduct further investigation for understanding better performance of CLMs than BATS in representing surface variables. For this purpose, differences between CLMs and BATS are computed for evapotranspiration, soil moisture content, and surface albedo for 2009 and 2011 years. Analysis of the surface evapotranspiration consists of evaporation from wet stems and leaves, transpiration through the plant, and initial evaporation from the ground (i.e., bare soil or snow surfaces). The analysis of evapotranspiration used in BATS is similar to the one used in CLMs, while Philip’s (1957) formulation is used for the computation of evaporation from the ground. The stomatal resistance in the formulation of transpiration through the plant is directly adopted in CLM. Canopy water is a simple mass balance determined by gains from the interception of precipitation and dew condensation and loss from evaporation; that is precipitation arriving at the vegetation top is either intercepted by foliage or directly falls through the gaps of leaves to the ground. Figure 9a, b shows surface evapotranspiration of CLMs which is reduced from BATS over India except over the desert and snowy region of WH. The soil moisture is governed by infiltration, runoff, gradient diffusion and soil water extraction through roots for canopy transpiration. Steiner et al. (2009) analyzed the effect of soil texture on soil moisture and precipitation and concluded that although the soil texture could trigger land–atmosphere coupling differences between the two models, the physical parameterization schemes allow the soil moisture changes to interact effectively with the atmosphere. The soil moisture–precipitation feedback is an important element of Earth’s climate system (Pal and Eltahir 2001; Koster et al. 2003; Seneviratne et al. 2010). The JJAS mean soil moisture content (SMC) by CLMs is 100–200 kg m-2 and higher than

4262

R. K. S. Maurya et al.

Pure Appl. Geophys.

layer. In India, during the JJAS period there is high soil moisture content as compared to other seasons because during this season India receives the maximum percentage of precipitation. Snow albedos are inferred from the calculations of Warren and Wiscombe (1980) snow model and snow data of Anderson (1976) which is a function of snow age, grain size, solar zenith angle, pollution, and the amount of fresh snow. The surface albedo decreases in both CLMs from BATS, while CLM4.5 shows a further reduction in surface albedo as compared to CLM3.5 over the semi-desert/desert and central regions of India (Fig. 9e, f). 3.2. Precipitation

Figure 9 JJAS difference in surface evapotranspiration computed between a CLM4.5 and BATS, and b CLM3.5 and BATS; difference in moisture content of the soil layers (mm day-1) computed between c CLM4.5 and BATS, and d CLM3.5 and BATS; difference in albedo e CLM4.5 and BATS and f CLM3.5 and BATS

that of BATS, except in the south peninsular India (SPI), the state of Rajasthan (semi-desert/desert) and east-northeast India as shown in Fig. 9c, d. The soil moisture flux is an important parameter that regulates the precipitation over land. The enhanced soil moisture fluxes in CLMs as compared to BATS may be the reason for enhanced precipitation over most parts of the land during summer monsoon season. The differences between CLM3.5 and CLM4.5 from BATS are analyzed in Fig. 9e, f. Figure 9e, f shows that surface albedo in CLM3.5 and CLM4.5 have decreased from BATS. Surface albedo is a function of soil color and moisture in the surface soil

Spatial distributions of JJAS summer monsoon accumulated precipitation simulated by RegCM4 with different land-surface schemes (BATS, CLM3.5 and CLM4.5) and as observed (IMD) are shown in Fig. 10. During the summer monsoon season, the WGs and NEI receive the highest amount of rainfall, while the WH, northwest India (NWI) and east PI receive the least rainfall as seen in the IMD observations. The zones of the highest and lowest precipitations are well represented by the RegCM4 simulations with all the land surface schemes. During 2009, the IMD dataset shows highest seasonal precipitation of 3000 mm in WGs, 1500 mm in NEI, 600 mm in central India and lowest precipitation of 400 mm in the WH, NWI and PI (Fig. 10a). The RegCM4–BATS shows the highest precipitation of 2000–3000 mm over the PI and NEI, 600 mm in central India and below 400 mm in the WH and NWI. The RegCM4–BATS overestimates precipitation over PI, NEI, and Eastern India as compared to RegCM4–CLMs. The rainfall over CI and eastern India are similar to the IMD dataset (Fig. 10c–d, g– h). For the year 2011, the IMD dataset shows the highest seasonal precipitation of 3000 mm in WGs, 2000 mm in NEI, 1000 mm in central India and lowest precipitation of 600 mm in the WH, NWI and PI (Fig. 10e). The RegCM4–BATS shows the highest precipitation of 2000–3000 mm over the PI, 1000 mm in central India and below 400 mm in the WH and NWI. The RegCM4–CLMs simulate 2000 mm over PI, 1000 mm over CI and NEI.

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

Figure 10 Spatial distributions of seasonal mean summer monsoon rainfall obtained from a IMD, b RegCM4 simulations with BATS, c RegCM4 simulations with CLM3.5, and d RegCM4 simulations with CLM4.5 for the year 2009. Panels e–h are same as a–d, respectively, but for 2011

Figure 11 shows the differences in JJAS precipitation between the RegCM4 simulation with respective land-surface schemes and observations for the years, 2009 and 2011. The total JJAS precipitation in RegCM4–BATS (Fig. 11a, d) shows an overestimate of 500 mm over the PI and NEI, 300 mm over the northeast India (NEI) and underestimate of 100–300 mm over central India as compared to IMD dataset for both the years. However, RegCM4–CLM3.5 (Fig. 11b, e) and RegCM4– CLM4.5 (Fig. 11c, f) simulate an overestimate of 300–500 mm over the PI and an underestimate of 100–300 mm over the north India. The RegCM4– CLM4.5 has lowest positive/negative precipitation biases over most parts of India followed by RegCM4–CLM3.5 and RegCM4–BATS for both years 2009 and 2011. The model biases are similar

4263

Figure 11 Spatial distributions of differences in JJAS precipitation (mm/ JJAS) computed between a RegCM4 simulations with BATS and observations (IMD) (referred as BATS-IMD), b RegCM4 simulations with CLM3.5 and observations (IMD) (referred as CLM3.5IMD), and c RegCM4 simulations with CLM4.5 and observations (IMD) (referred as CLM4.5-IMD for 2009. Panels d–f are same as a–c, respectively, but for 2011

for both the deficient and normal years. It is of interest to investigate the higher precipitation over PI and NEI regions in RegCM4–BATS simulations. For this, seasonal mean pressure vertical velocity at 500 hPa is examined in all the simulations of RegCM4 with –BATS, –CLM3.5 and –CLM4.5 schemes (Shown in Supplementary Figure 3). It is found that the downward velocity is higher over PI and NEI regions in RegCM4–BATS than other simulations. This is probably one probable reason to have overestimated precipitation over these regions in the RegCM4–BATS. Overall, RegCM4–CLM4.5 has a better simulation of JJAS precipitation over India. The seasonal rainfall variation between the two monsoon seasons is clearly evident from the spatial distribution as per the observed IMD dataset. The difference between total rainfall is depicted over most regions of India. Major regions of variation in seasonal rainfall are observed over CI, NEI, WG,

4264

R. K. S. Maurya et al.

and NWI. On the same hand, comparing RegCM– BATS for both the years, the variability is not captured well. Both the years simulate nearly equal seasonal rainfall. The variations between the two contrasting monsoon seasons are not well represented by RegCM–BATS. Comparing for the two seasons using RegCM–CLMs, differences over NWI, CI, and Eastern India are depicted well in CLM4.5. During the drought season, rainfall over NWI and CI is less than the normal season, and the same feature is represented better in CLM4.5 than CLM3.5 and BATS. Overall, the RegCM–CLM4.5 captures the seasonal variability better than the other two RegCM–LSSs. Since the ISM rainfall has a large spatial variability, the analysis of RegCM4 simulations is carried out for different rainfall homogeneous regions to understand the performance of the model with the various land surface schemes. The seasonal mean precipitation (mm day-1) obtained from IMD, and RegCM4 experiments with BATS, with CLM3.5, and with CLM4.5 calculated for six rainfall homogeneous regions (Parthasarathy et al. 1995) for 2009 and 2011 are presented in Table 3. It is seen from the Table 3 that the RegCM4 coupled with CLM provides better results than the RegCM4–BATS over all the homogeneous regions for both the years. It is also noticed that the RegCM4 model with BATS scheme has a large positive bias over peninsular India and northeast India. The positive biases over these regions in RegCM4–BATS probably due to downward pressure vertical velocity are higher over the same regions (Supplementary Figure 3). During the deficit monsoon year (2009), observed seasonal mean rainfall is 4.89, 2.91, 6.42, 9.46, 5.81, and 4.13 mm day-1 for

Pure Appl. Geophys.

Hilly Region of India (HRI), North West India (NWI), Central Northeast India (CNEI), North East India (NEI), West Central India (WCI), and Peninsular India (PI), respectively. The RegCM4 simulated precipitation for the corresponding regions are 4.67, 3.49, 5.96, 9.52, 6.81, and 9.56 mm day-1, respectively, using CLM4.5 scheme, while it is 6.24, 4.02,8.01, 13.23, 7.99, and 13.63 mm day-1, respectively, with the use of BATS scheme. The RegCM4– CLM4.5 simulated results are the closest to IMD observations followed by RegCM4–CLM3.5 and RegCM4–BATS. The analysis revealed that the performance of different schemes RegCM4 in representing homogeneous region rainfall during 2011 is in a similar fashion as seen in 2009, i.e., bias is less in RegCM4–CLM4.5 and more in RegCM4–BATS. It is seen that the simulated precipitation is higher in RegCM4 than the observations. However, the positive bias is more in RegCM4–BATS than in other schemes over all the regions for both the years except NWI region in 2011. The RegCM4 model with BATS scheme has a large positive bias over peninsular India and northeast India probably due to downward pressure vertical velocity is higher over these regions (Supplementary Figure 3). It can be illustrated that the performance of RegCM4 with CLM4.5 is the best in representing rainfall intensity over homogeneous regions. 3.2.1 Statistical Evaluation of Model-Simulated Precipitation Several statistical analyses are used to understand the performance of different land surface schemes in terms of precipitation simulation. Before doing any

Table 3 -1

The seasonal mean precipitation (mm day ) obtained from IMD, and RegCM4 experiments with BATS, (referred as BATS), with CLM3.5 (referred as CLM3.5), and CLM4.5 (referred as CLM4.5) calculated for six rainfall homogeneous regions for 2009 and 2011 2009

HRI NWI CNEI NEI WCI PI

2011

IMD

BATS

CLM3.5

CLM4.5

IMD

BATS

CLM3.5

CLM4.5

4.89 2.91 6.42 9.46 5.81 4.13

6.24 4.02 8.01 13.23 7.99 13.63

5.58 4.39 6.65 9.18 6.94 10.99

4.67 3.49 5.96 9.52 6.81 9.56

5.37 5.94 8.43 9.86 7.60 5.82

6.88 5.13 10.50 14.80 9.61 16.65

5.97 6.18 7.75 10.64 7.67 9.85

5.21 5.08 8.00 10.29 7.59 8.92

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

Table 4 The mean, bias, root mean square error (RMSE) and correlation coefficients (CCs) of JJAS (122 days) daily precipitation (mm day-1) computed between IMD and RegCM4 simulations with –BATS, –CLM3.5, and –CLM4.5 for 2009 and 2011 Years

Precipitation (mm day-1) Mean

2009 Observed RegCM4–BATS RegCM4–CLM3.5 RegCM4–CLM4.5 2011 Observed RegCM4–BATS RegCM4–CLM3.5 RegCM4–CLM4.5

Bias

CCs

RMSE

5.92 8.28 6.32 6.34

2.36 0.39 0.41

0.59 0.58 0.61

3.72 3.13 2.83

7.62 9.89 7.47 7.80

2.27 -0.16 -0.35

0.49 0.53 0.78

4.14 3.05 2.75

statistical analysis, the bilinear interpolation technique (Wilks 1995) is used to interpolate RegCM4 simulated precipitation to the IMD grid points. Table 4 presents the mean, bias, RMSE and pattern correlation coefficient between IMD and RegCM4– BATS, –CLM3.5, and –CLM4.5 simulations, respectively, for JJAS precipitation during 2009 and 2011 monsoon seasons. Over India, the RegCM4–BATS overestimates mean precipitation of an about 2.36 and 2.27 mm day-1 in 2009 and 2011, respectively. While RegCM4–CLM3.5 and RegCM4–CLM4.5

4265

overestimate precipitation of an about 0.39 and 0.41 mm day-1, respectively, in 2009 and underestimate precipitation of an about 0.16 and 0.35 mm day-1, respectively, in 2011. The error is less in CLM than BATS in replicating the precipitation intensity. The correlation coefficient is computed between RegCM4–LSS and IMD daily JJAS precipitation averaged over all India. The CCs have 0.59/0.49, 0.58/0.53 and 0.61/0.78 values in RegCM4–BATS, RegCM4–CLM3.5 and RegCM4– CLM4.5, respectively, for years 2009/2011. The RegCM–CLM4.5 has a high temporal correlation as compared to IMD followed by RegCM4–CLM3.5 and RegCM4–BATS. The RMSE has low values of 2.83/2.75 mm day-1 in RegCM4–CLM4.5 followed by 3.13/3.05 mm day-1 in RegCM4–CLM3.5 and 3.72/4.14 mm day-1 in RegCM4–BATS for 2009/2011, respectively. Overall, RegCM4–CLM4.5 simulation has a low bias, high CCs and low RMSE in simulating JJAS precipitation over all India. The results discussed above clearly indicate that RegCM4–CLM4.5 has a better simulation of JJAS precipitation as compared to other two RegCM4– LSS. Figure 12 shows the spatial pattern of CCs computed between the three LSSs with IMD observation for mean JJAS precipitation over the Indian landmass in both the years. The RegCM4–LSSs show a significant (at 99% confidence level) positive

Figure 12 Spatial correlation coefficients of RegCM4–BATS, RegCM4–CLM3.5 and RegCM4–CLM4.5 with IMD precipitation for 2009 and 2011 summer monsoon seasons

4266

R. K. S. Maurya et al.

correlations of values 0.311/0.213, 0.335/0.226 and 0.348/0.231 in RegCM4–BATS, RegCM4–CLM3.5 and RegCM4–CLM4.5 in 2009/2011, respectively. On a comparison of land surface schemes in the RegCM4, CLM4.5 has shown better spatial relationship over the Indian landmass as compared with BATS and CLM3.5 for both the years. It is noted that the CLM4.5 scheme in the RegCM4 has a significant relationship for the drought year. The reduced biases in CLMs may be attributed to the improved representation of soil moisture and surface relative humidity that affect in depicting better precipitation distribution in CLMs. For better understanding the performances of different LSS, further evaluation has been made using equitable threat score (ETS) and critical success index (CSI) which have been carried out between RegCM4–LSS and IMD datasets for 2009 and 2011. 3.2.1.1 Equitable Threat Score (ETS) The ETS is a skill score that measures the fraction of correct forecasts with adjustment to hits by random chance (Gilbert 1884; Wilks 1995). The ETS is calculated for each day for the individual category from six different categories (0–1, 1–5, 5–10, 10–30, 30–50, [50 mm day-1) of rainfall intensities considering all

Pure Appl. Geophys.

grid points in the Indian landmass. The final value is obtained by computing the time average of JJAS season, i.e., 122 days. Same calculations are done for both the seasons and all the RegCM4–LSS combinations. Mathematically ETS is calculated by: ETS ¼

H  Hk ; ðH þ M þ F  Hk Þ

ð1Þ

where Hk ¼

ðH þ M ÞðM þ FÞ : T

M, H, and F are the number of misses, hits, and false alarms, respectively, for each category and these numbers are identified based on contingency table. Hits due to random chance are denoted by Hk and T is the total number of events. ETS varies from -0.33 to 1 with ETS = 0 indicating no skill and ETS = 1 indicating perfect skill in prediction. Figure 13a, c represents the ETS of RegCM4–LSS and IMD mean JJAS precipitation for 2009 and 2011, respectively. The RegCM4–LSS have high ETS values for rainfall intensities 0–1 and 10–30 mm day-1 and low values for high rainfall intensities (30–50 and [50 mm day-1) for both years. RegCM4–CLMs have better skill than BATS in predicting the rainfall

Figure 13 Skill scores based on daily precipitation distribution of IMD and RegCM4 model simulations with BATS (red dashed line), CLM3.5 (blue solid line), and CLM4.5 (black dotted line) using method: a equitable threat score (ETS), and b critical success index (CSI) for the year 2009. Panels c, d are same as a and b, respectively, but for the year 2011

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

intensity events in the low rainfall intensity. Overall, the RegCM4–LSS have an almost similar skill for high rainfall categories; the skill is less for very heavy rainfall events or extreme rainfall events. 3.2.1.2 Critical Success Index (CSI) CSI is a skill score that measures the fraction of events that were correctly predicted. It is a metric that is concerned with the forecast that counts (Schaefer 1990). CSI is computed for both the years 2009 and 2011. Mathematically CSI is given by: CSI ¼

H ; HþMþF

ð2Þ

where M, H, and F are the same as defined in the previous section. The CSI tells us how well did the forecast ‘yes’ events correspond to the observed ‘yes’ events. This score ranges from 0 to 1, where 0 indicates no skill and 1 indicates a perfect score. The CSI is only concerned with forecasts that count and is accurate when the correct negatives have been removed. This score is sensitive to hits and penalizes both misses and false alarms. It removes the hits that occur purely by chance. Figure 13b, d shows the CSI computed for the RegCM4–LSSs with IMD rainfall over the Indian landmass for 2009 and 2011, respectively. For the dry year (2009), CSI values are similar for all the RegCM–LSS. Higher CSI values are observed for low- and medium- rainfall intensities (0–30 mm day-1). The CSI values decrease for heavy rainfall events. A similar pattern is observed for the year 2011. Overall RegCM4–CLMs and RegCM4–BATS have almost similar skills for most of the categories. The CSI of the model decreases with increase in rainfall intensities for all the RegCM4–LSSs that indicates that regional climate model till needs to be improved in order to simulate extreme weather associated with the heavy to very heavy rainfall events

4. Conclusions The performance of the coupled land–atmosphere processes is evaluated based on simulations of three land-surface schemes (BATS, CLM3.5 and CLM4.5) in RegCM4 for two different summer monsoon

4267

(2009–dry and 2011–normal monsoons years). The RegCM4 driven by NNRP2 datasets and modelsimulated upper air and surface parameters are compared with NNRP2, while precipitation is compared with IMD. Several statistical techniques are used to confirm the model performance. The major conclusions from the present study are summarized below: The CLMs represent surface albedo and surface evapotranspiration closer to the verification analysis NNRP2 and produce the moisture content of the soil layers better than BATS. The CLMs have added to addressing an overestimation of forest and other inconsistencies in model physics (Bonan et al. 2011, 2012) causing a systematic decrease of the evapotranspiration, which causes a decrease of precipitation in CLMs over BATS. The CLMs have different types of vegetation and qualitatively including interactive vegetation leads to significant differences in the model’s physical climate. These differences are consistent with our theoretical understanding of how vegetation-induced changes of surface bio-geophysical properties may influence regional climate. These results on the role of land– atmosphere coupling are more important for regional prediction of summer monsoon season over India. In the JJAS simulation, CLMs have reduced evapotranspiration and increased sensible heat flux compared to BATS. In fact, CLMs increase sensible heat flux by increasing the surface drag coefficient. The RegCM4–CLM4.5 is able to depict the overlying surface meteorological parameters and surface fluxes than other two land surface schemes, which in turn, indicate that the land–air interaction is more realistic in nature with the use of the CLM4.5 land surface scheme. Qualitative and quantitative studies confirm that CLM4.5 shows relatively low bias in simulating near-surface relative humidity, surface air temperature, and sensible heat flux over other RegCM4–LSS. The RegCM4 with all land surface schemes underestimate precipitation over central India and overestimate precipitation over the PI, specifically leeward sides during both the years. The skill of RegCM4 for central India is less probably due to an inability of depicting the strong convective activities known as ‘‘monsoon depression’’ that generates over

4268

R. K. S. Maurya et al.

the North Bay of Bengal and plays a dominant role in the occurrence of rainfall over that region. In the leeward side of the Western Ghats over PI, topographical uplift may be one cause to have more precipitation in the RegCM4 (Sinha et al. 2014). The bias in precipitation intensity is more in BATS than CLMs and qualitative and quantitative representation of precipitation in CLM4.5 is better than BATS and CLM3.5. Statistical scores (ETS and CSI) also infer that RegCM4–CLMs have better performance over RegCM4–BATS in the simulation of the ISMR.

Acknowledgements The authors acknowledge the financial support given by the Department of Agriculture and Cooperation and Farmer Welfare (DAC&FW), Government of India, to carry out the present research work. The authors acknowledge ICTP for providing the RegCM4 model source code and required input datasets through the website at http://clima-dods. ictp.it/regcm4/. The authors sincerely thank NCEP for NCEP_Reanalysis 2 data provided by the NOAA/ OAR/ESRL PSD, Boulder, Colorado, USA, from their website at http://www.esrl.noaa.gov/psd/. The authors are grateful to India Meteorological Department (IMD) for providing high spatial resolution daily precipitation data at 0.25° 9 0.25° resolution. The authors sincerely acknowledge the anonymous reviewers for their valuable suggestions.

REFERENCES Anderson, E. A. (1976). A point energy and mass balance model of a snow cover. NOAA Tech. Rep. NWS 19, Office of Hydrology, National Weather Service, Silver Spring, MD, pp. 150. Arakawa, E. A., & Schubert, W. H. (1974). Interaction of a cumulus cloud ensemble with the large-scale environment, part I. Journal Atmospheric Science, 31, 674–701. Bala, G., Caldiera, K., Mirin, A., Wickett, M., Delire, C., & Phillips, T. J. (2006). Biogeophysical effects of CO2 fertilization on global climate. Tellus B, 58, 620–627. doi:10.1111/j.1600-0889. 2006.00210.x. Bonan, G. B., Lawrence, P. J., Oleson, K. W., Levis, S., Jung, M., Reichstein, M., et al. (2011). Improving canopy processes in the Community Land Model version 4 (CLM4) using global flux fields empirically inferred from FLUXNET data. Journal Geophysical Research, 116, G02014.

Pure Appl. Geophys. Bonan, G. B., Oleson, K. W., Fisher, R. A., Lasslop, G., & Reichstein, M. (2012). Reconciling leaf physiological traits and canopy flux data: Use of the TRY and FLUXNET databases in the Community Land Model version 4. Journal Geophysical Research, 117, G02026. Bonan, G. B., Oleson, K. W., Vertenstein, M., Levis, S., Zeng, X., Dai, Y. J., et al. (2002). The land surface climatology of the community land model coupled to the NCAR Community Climate Model. Journal of Climate, 15, 3123–3149. Castro, C. L., Pielke, R. A., & Leoncini, G. (2005). Dynamical downscaling: Assessment of value retained and added using the Regional Atmospheric Modeling System (RAMS). Journal Geophysical Research, 110, D05108. Cox, P. M., Betts, R. A., Jones, C. D., Spall, S. A., & Totterdell, I. J. (2000). Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model. Nature, 408, 184–187. Crucifix, M., Betts, R. A., & Cox, P. M. (2005). Vegetation and climate variability: A GCM modeling study. Climate Dynamics, 24, 457–467. Dai, Y. J., Zeng, X., Dickinson, R. E., Baker, I., Bonan, G. B., Bosilovich, M. G., et al. (2003). The common land model. Bulletin of the American Meteorological Society, 84, 1013–1023. Dash, S. K., Pattnayak, K. C., Panda, S. K., Vaddi, D., & Mamgain, A. (2015). Impact of domain size on the simulation of Indian summer monsoon in RegCM4 using mixed convection scheme and driven by HadGEM2. Climate Dynamics, 44, 961–975. Dash, S. K., Shekhar, M. S., & Singh, G. P. (2006). Simulation of Indian summer monsoon circulation and rainfall using RegCM3. Theoretical and Applied Climatology, 86, 161–172. Deardorff, J. W. (1978). Efficient prediction of ground surfacetemperature and moisture, with inclusion of a layer of vegetation. Journal Geophysical Research Atmospheric, 83, 1889–1903. Delire, C., Foley, J. A., & Thompson, S. (2004). Long-term variability in a coupled atmosphere –biosphere model. Journal of Climate, 17, 3947–3959. Dickinson, R. E., Henderson-Sellers, A., & Kennedy, P. J. (1993). Biosphere–atmosphere transfer scheme (BATS) version 1e as coupled to the NCAR community climate model. 80, National Center for Atmospheric Research, Boulder, CO, p 422. NCAR. doi:10.5065/D67W6959 Dickinson, R., Henderson-Sellers, A., Kennedy, P., & Wilson, M. (1986). Biosphere–atmosphere transfer scheme (BATS) for the NCAR community climate model. NCAR Tech Note, NCAR/TN275 ? STR. Douglas, E. M., Beltra’n-Przekurat, A., Niyogi, D., Pielke, R. A., & Vo¨ro¨smarty, C. J. (2009). The impact of agricultural intensification and irrigation on land–atmosphere interactions and Indian monsoon precipitation: A mesoscale modeling perspective. Global and Planetary Change, 67, 117–128. Dutta, S. K., Das, S., Kar, S. C., Mohanty, U. C., & Joshi, P. C. (2009). Impact of vegetation on the simulation of seasonal monsoon rainfall over the Indian sub-continent using a regional model. Journal Earth System Science, 118, 413–440. Eden, J. M., Widmann, M., Maraun, D., & Vrac, M. (2014). Comparison of GCM- and RCM-simulated precipitation following stochastic post processing. Journal of Geophysical Research, 119, 11040–11053. doi:10.1002/2014JD021732. Emanuel, K. A. (1991). A scheme for representing cumulus convection in large-scale models. Journal Atmospheric Sciences, 48, 2313–2335.

Vol. 174, (2017)

Coupling of Community Land Model with RegCM4

Emanuel, K. A., & Zivkovic-Rothman, M. (1999). Development and evaluation of a convection scheme for use in climate models. Journal of Atmospheric Sciences, 56, 1766–1782. Fennessy, M. J., & Shukla, J. (1999). Impact of initial soil wetness on seasonal atmospheric prediction. Journal of Climate, 12, 3167–3180. Foley, J. A., Levis, S., Prentice, I. C., Pollard, D., & Thompson, S. L. (1998). Coupling dynamic models of climate and vegetation. Global Change Biology, 5, 561–579. Gilbert, G. F. (1884). Finley’s tornado predictions. American Meteorological Journal, 1, 166–172. Giorgi, F., Coppola, E., Solmon, F., Mariotti, L., Sylla, M. B., Bi, X., et al. (2012). RegCM4: Model description and preliminary tests over multiple CORDEX domains. Climate Research, 52, 7–29. Grell, G. A. (1993). Prognostic evaluation of assumptions used by cumulus parameterization. Monthly Weather Review, 12, 764–787. Grell, G., Dudhia, J., & Stauffer, D. R. (1994). A description of fifth generation Penn state/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398 ? STR. Guo, Z., Dirmeyer, P. A., & DelSole, T. (2011). Land surface impacts on subseasonal and seasonal predictability. Geophysical Research Letters, 38, L24812. doi:10.1029/2011GL049945. Halpern, D., & Woiceshyn, P. M. (2001). Somali Jet in the Arabian Sea, El Nin˜o, and India Rainfall. Journal of Climate, 14, 434–441. Henderson-Sellers, A., Yang, Z. L., & Dickinson, R. E. (1993). The project for inter-comparison of land-surface parameterization schemes. Bulletin of American Meteorological Society, 74, 1335–1349. Holtslag, A., de Bruijn, E., & Pan, H. L. (1990). A high resolution air mass transformation model for short-range weather forecasting. Monthly Weather Review, 118, 1561–1575. Ju, J., & Slingo, J. (1995). The Asian summer monsoon and ENSO. Quarterly Journal Royal Meteorological Society, 121, 1133–1168. Kanamitsu, M., Ebisuzaki, W., Woollen, J., Yang, S-K, Hnilo, J. J., Fiorino, M., & Potter, G. L. (2002). NCEP-DOE AMIP-II Reanalysis (R-2). Bulletin of the American Meteorological Society, 1631–1643. Kar, S. C., Mali, P., & Routray, A. (2014). Impact of land surface processes on the South Asian monsoon simulations using WRsF modeling system. Pure and Applied Geophysics, 171, 2461–2484. Kiehl, J., Hack, J., Bonan, G., Boville, B., Breigleb, B., Williamson, D., & Rasch, P. (1996). Description of the NCAR Community Climate Model (CCM3). National Center for Atmospheric Research Tech Note NCAR/TN-420 ? STR, NCAR, Boulder, CO. Koster, R. D., Dirmeyer, P. A., Guo, Z., Bonan, G., Chan, E., Cox, P., et al. (2004). Regions of strong coupling between soil moisture and precipitation. Science, 305, 1138–1140. Koster, R. D., Guo, Z., Dirmeyer, P., Sud, Y., Bonan, G., Oleson, K., et al. (2006). GLACE: The Global land-atmosphere coupling experiment. Part II: Analysis. Journal of Hydrometeorology, 7, 611–625. doi:10.1175/JHM511.1. Koster, R. D., Mahanama, S. P. P., Yamada, T. J., Balsamo, G., Berg, A. A., Boisserie, M., et al. (2010). Contribution of land surface initialization to subseasonal forecast skill: First results from a multi-model experiment. Geophysical Research Letters, 37(6), L02402. doi:10.1029/2009gl041677.

4269

Koster, R., Mahanama, S., Yamada, T., Balsamo, G., Berg, A., Boisserie, M., et al. (2011). The second phase of the global land– atmosphere coupling experiment: Soil moisture contributions to subseasonal forecast skill. Journal of Hydrometeorology, 12, 805–822. doi:10.1175/2011JHM1365.1. Koster, R. D., Suarez, M. J., Higgins, R., & van den Dool, H. (2003). Observational evidence that soil moisture variations affect precipitation. Geophysical Research Letter, 30(5), 1241. doi:10.1029/2002GL016571. Koven, C. D., Riley, W. J., Subin, Z. M., Tang, J. Y., Torn, M. S., Collins, W. D., et al. (2013). The effect of vertically resolved soil biogeochemistry and alternate soil C and N models on C dynamics of CLM4. Biogeosciences, 10, 7109–7131. doi:10. 5194/bg-10-7109-2013. Levis, S., Foley, J. A., & Pollard, D. (1999). Potential high-latitude vegetation feedbacks on CO2-induced climate change. Geophysical Research Letter, 26, 747–750. Loveland, T. R., Reed, B. C., Brown, J. F., Ohlen, D. O., Zhu, J., Yang, L., et al. (2000). Development of a global land cover characteristics database and IGBP DISCover from 1-km AVHRR data. International Journal of Remote Sensing, 21, 1303–1330. Niyogi, D., Chang, H.-I., Chen, F., Gu, L., Kumar, A., Menon, S., et al. (2007). Potential impacts of aerosol–land–atmosphere interaction on the Indian monsoonal rainfall characteristics. Natural Hazards, 42, 345–359. doi:10.1007/s11069-006-9085-y. Noilhan, J., & Planton, S. (1989). A simple parameterization of land surface processes for meteorological models. Monthly Weather Review, 117(3), 536–549. Oleson, K.W., Dai, Y., Bonan, G., Bosilovich, M., Dickinson, R., Dirmeyer, P., Hoffman, F., Houser, P., Levis, S., Niu, G-Y, Thornton, P., Vertenstein, M., Yang, Z-L, & Zeng, X. (2004). Technical description of the community land model. National Center for Atmospheric Research Tech Note NCAR/TN461 ? STR, NCAR, Boulder, CO. Oleson, K. W., Gy, N., Yang, Z. L., Lawrence, D. M., Thornton, P. E., Lawrence, P. J., et al. (2008). Improvements to the community land model and their impact on the hydrologic cycle. Journal of Geophysical Research, 113, G01021. doi:10.1029/ 2007JD000563. Oleson, K.W., Lawrence, D.M., & Bonan, G.B., et al (2013). Technical description of version 4.5 of the community land model (CLM). NCAR technical note NCAR/TN-503 ? STR. National Center for Atmospheric Research, Boulder. Paeth, H., Born, K., Girmes, R., Podzun, R., & Jacob, D. (2009). Regional climate change in tropical and northern Africa due to greenhouse forcing and land use changes. Journal of Climate, 22, 114–132. Pai, D., Sridhar, S., Badwaik, L., & Rajeevan, M. R. (2014). Development of a new high spatial resolution (0.25 9 0.25) Long period (1901–2010) daily gridded rainfall data set over India and its comparison with existing datasets over the region. Mausam, 65, 1–18. Pal, J. S., & Eltahir, E. A. B. (2001). Pathways relating soil moisture conditions to future summer rainfall within a model of the land-atmosphere system. Journal of Climate, 14, 1227–1242. Pal, J. S., Small, E., & Eltahir, E. (2000). Simulation of regionalscale water and energy budgets: Representation of subgrid cloud and precipitation processes within RegCM. Journal Geophysical Research, 105, 29579–29594. Parthasarathy, B., Munot, A. A., & Kothawale, D. R. (1994). All India monthly and seasonal rainfall series: 1871–1993. Theoretical and Applied Climatology, 49, 217–224.

4270

R. K. S. Maurya et al.

Parthasarathy, B., Munot, A.A., & Kothawale, D.R., (1995). Monthly and seasonal rainfall series for All-India homogeneous regions and meteorological subdivisions: 1871–1994. Contributions from Indian Institute of Tropical Meteorology, Research Report RR-065 (ISSN 0252–1075), August 1995, Pune 411 008 India. Philip, J. R. (1957). Evaporation and moisture and heat fields in the soil. Journal of Meteorology, 14, 354–366. Pielke, R. A., Marland, G., Betts, R. A., Chase, T. N., Eastman, J. L., Niles, J. O., et al. (2002). The influence of land-use change and landscape dynamics on the climate system: Relevance to climate change policy beyond the radiative effect of greenhouse gases. Philosophical Transactions of Royal Society A, 360, 1705–1719. Pielke, R. A., Pitman, A., Niyogi, D., Mahmood, R., McAlpine, C., Hossain, F., et al. (2011). Land use/land cover changes and climate: Modeling analysis and observational evidence. WIREs Climate Change, 2, 828–850. doi:10.1002/wcc.144. 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. Journal of Climate, 15, 1609–1625. Roy, S. S., Mahmood, R., Niyogi, D., Lei, M., Foster, S. A., Hubbard, K. G., et al. (2007). Impacts of the agricultural green revolutioninduced land use changes on air temperatures in India. Journal of Geophysical Research, 112, D21108. doi:10.1029/2007JD008834. Schaefer, J. (1990). The critical success index as an indicator of warning skill. Weather and Forecasting, 5, 570–575. doi:10. 1175/1520-0434(1990)005\0570:TCSIAA[2.0.CO;2. Sellers, P. J., Randall, D. A., Collatz, G. J., Berry, J. A., Field, C. B., Dazlich, D. A., et al. (1996). A revised land surface parameterization (SiB2) for atmospheric GCMs. Part 1: Model formulation. Journal of Climate, 9, 676–705. Seneviratne, S. I., Corti, T., Davin, E. L., Hirschi, M., Jaeger, E. B., Lehner, I., et al. (2010). Investigating soil moisture–climate interactions in a changing climate: A review. Earth-Science Reviews, 99, 125–161. doi:10.1016/J.Earscirev.2010.02.004. Shukla, J., & Mintz, Y. (1982). The influence of land surface evapotranspiration on the earth’s climate. Science, 214, 1498–1501. Singh, A. P., Mohanty, U. C., Sinha, P., & Mandal, M. (2007). Influence of different land surface processes on Indian summer monsoon circulation. Natural Hazards, 42, 423–438. Sinha, P., Mohanty, U. C., Kar, S. C., Dash, S. K., & Kumari, S. (2013). Sensitivity of the GCM driven summer monsoon simulations to cumulus parameterization schemes in nested RegCM3. Theoretical and Applied Climatology, 112, 285–306. doi:10. 1007/s00704-012-0728-5. Sinha, P., Mohanty, U. C., Kar, S. C., & Kumari, S. (2014). Role of the Himalayan orography in simulation of the Indian summer monsoon using RegCM3. Pure and Applied Geophysics, 171, 1385–1407. doi:10.1007/s00024-013-0675-9. Solmon, F., Giorgi, F., & Liousse, C. (2006). Aerosol modeling for regional climate studies: Application to anthropogenic particles and evaluation over a European/African domain. Tellus B, 58, 51–72. Steiner, A. L., Pal, J. S., Rauscher, S. A., Bell, J. L., Diffenbaugh, S., Boone, A., et al. (2009). Land surface coupling in regional climate simulations of the West African monsoon. Climate Dynamics, 33, 869–892. doi:10.1007/s00382-009-0543-6.

Pure Appl. Geophys. Sun, S., & Wang, G. (2012). The complexity of using a feedback parameter to quantify the soil moisture–precipitation relationship. Journal of Geophysical Research, 117, D11113. doi:10. 1029/2011JD017173. Swenson, S. C., & Lawrence, D. M. (2012). A new fractional snow covered area parameterization for the community land model and its effect on the surface energy balance. Journal of Geophysical Research, 117, D21107. doi:10.1029/2012JD018178. Swenson, S. C., Lawrence, D. M., & Lee, H. (2012). Improved simulation of the terrestrial hydrological cycle in permafrost regions by the community land model. Journal of Advances in Modeling Earth Systems, 4, M08002. doi:10.1029/2012MS000165. van den Hurk, B., Doblas-Reyes, F., Balsamo, G., Randall, D., Koster, D., Seneviratne, S. I., et al. (2012). Soil moisture effects on seasonal temperature and precipitation forecast scores in Europe. Climate Dynamic, 38, 349–362. doi:10.1007/s00382010-0956-2. Wang, G. L. (2004). A conceptual modeling study on biosphere– atmosphere interactions and its implications for physically based climate modeling. Journal of Climate, 17, 2572–2583. Warren, S. G., & Wiscombe, W. J. (1980). A model for the spectral albedo of snow. II: Snow containing atmospheric aerosols. Journal of Atmospheric Science, 37, 2734–2745. Wilks, D. S. (1995). Statistical methods in the atmospheric sciences (p. 467). San Diego: Academic Press. Wood, A. W., Leung, L. R., Sridhar, V., & Letternmaier, D. P. (2004). Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs. Climate Change, 62, 189–216. Xue, Y., Sales, F. De., Vasic, R., Mechooso, C. R., Prince, S. D., & Arakawa, A. (2010). Global and temporal characteristics of seasonal climate/vegetation biophysical process (VBP) interactions. Journal of Climate, 23, 1411–1433. Xue, Y., & Shukla, J. (1993). The influence of land surface properties on Sahel climate. Part I Desertification. Journal of Climate, 6, 2232–2245. Yeh, T. C., Wetherald, R., & Manabe, S. (1984). The effect of soilmoisture on the short-term climate and hydrology change—A numerical experiment. Monthly Weather Review, 112, 474–490. Zakey, A. S., Solmon, F., & Giorgi, F. (2006). Implementation and testing of a desert dust module in a regional climate model. Atmospheric Chemistry and Physics, 6, 4687–4704. Zeng, X., Shaikh, M., Dai, Y. J., Dickinson, R. E., & Myneni, R. (2002). Coupling of the common land model to the NCAR community climate model. Journal of Climate, 15, 1832–1854. Zeng, X., Zhao, M., & Dickinson, R. E. (1998). Intercomparison of bulk aerodynamic algorithms for the computation of sea surface fluxes using TOGA COARE and TAO data. Journal of Climate, 11, 2628–2644. Zhang, J., Wang, W.-C., & Leung, L. R. (2008). Contribution of land–atmosphere coupling to summer climate variability over the contiguous United States. Journal of Geophysical Research, 113, D22109. doi:10.1029/2008JD010136. Zhao, Z., & Luo, Y. R. (1997). Simulation of summer monsoon over East Asia: Inter comparison of three regional climate models. Journal of Applied Meteorology, 8(Supplementary), 116–123.

(Received March 31, 2017, revised June 29, 2017, accepted July 31, 2017, Published online August 30, 2017)