The responses of atmospheric variability to Tibetan Plateau (TP) snow cover (TPSC) at seasonal, interannual and decadal time scales have been extensively investigated. However, the atmospheric response to faster subseasonal variability of TPSC has been largely ignored. Here, we show that the subseasonal variability of TPSC, as revealed by daily data, is closely related to the subsequent East Asian atmospheric circulation at medium-range time scales (approximately 3-8 days later) during wintertime. TPSC acts as an elevated cooling source in the middle troposphere during wintertime and rapidly modulates the land surface thermal conditions over the TP. When TPSC is high, the upper-level geopotential height is lower, and the East Asia upper-level westerly jet stream is stronger. This finding improves our understanding of the influence of TPSC at multiple time scales. Furthermore, our work highlights the need to understand how atmospheric variability is rapidly modulated by fast snow cover changes.

Tibetan Plateau snow cover (TPSC) has subseasonal variations and rapidly influences the atmosphere. In this study, we present the rapid response of the East Asian trough (EAT) within a week to subseasonal variations in TPSC during the boreal winter. Using snow cover analysis obtained from the daily interactive multisensor snow and ice mapping system and the ERA-Interim reanalysis, a considerable relationship between TPSC and 500-hPa geopotential height anomalies over the downstream EAT region is found. Significant negative (positive) 500-hPa geopotential height anomalies originating from the Tibetan Plateau and moving into the EAT region appear within a week following anomalous positive (negative) TPSC events, which lead to changes in EAT strength. Thus, a significantly enhanced (reduced) intensity of the EAT occurs approximately 5-6 days after increased (decreased) TPSC. Numerical experiments confirm the causality of this relationship. Further analysis of the quasi-geostrophic geopotential height tendency equations in numerical experiments indicates that such EAT variations result from anomalous thermal advection from the Tibetan Plateau forced by TPSC.

Satellite solar-induced chlorophyll fluorescence (SIF) products from the Global Ozone Monitoring Experiment 2 (GOME-2) and Orbiting Carbon Observatory 2 (OCO-2) are used to investigate the responses of vegetation growth to the 2019 heat wave in Australia. Both satellite SIF data sets are more sensitive to water and heat stress than is the greenness-based vegetation index (enhanced vegetation index). Moreover, the OCO-2 SIF observations show a more significant reduction and earlier response to the heat stress than does GOME-2 SIF, indicating that the two satellite SIF data sets differ in how they monitor the drought and heat wave event due to the different timing of observations. Eddy covariance measurements confirm the different responses of dryland vegetation to the 2019 heat wave at a subdaily time scale. The differences in the timing of the satellite SIF products can be used to assess different elements of the impact of heat and water stress on Australian dryland ecosystems.

To resolve a series of ecological and environmental problems over the Loess Plateau, the "Grain for Green Program" (GFGP) was initiated at the end of 1990s. Following the conversion of croplands and bare land on hill-slopes to forests, the Loess Plateau has displayed a significant greening trend, which has resulted in soil erosion being reduced. However, the GFGP has also affected the hydrology of the Loess Plateau, which has raised questions regarding whether the GFGP should be continued in the future. We investigated the impact of revegetation on the hydrology of the Loess Plateau using relatively high-resolution simulations and multiple realizations with the Weather Research and Forecasting (WRF) model. Results suggest that revegetation since the launch of the GFGP has reduced runoff and soil moisture due to enhanced evapotranspiration. Further revegetation associated with the GFGP policy is likely to further increase evapotranspiration, and thereby reduce runoff and soil moisture. The increase in evapotranspiration is associated with biophysical changes, including deeper roots that deplete deep soil moisture stores. However, despite the increase in evapotranspiration, our results show no impact on rainfall. Our study cautions against further revegetation over the Loess Plateau given the reduction in water available for agriculture and human settlements and the lack of any significant compensation from rainfall.

China is several decades into large-scale afforestation programs to help address significant ecological and environmental degradation, with further afforestation planned for the future. However, the biophysical impact of afforestation on local surface temperature remains poorly understood, particularly in midlatitude regions where the importance of the radiative effect driven by albedo and the nonradiative effect driven by energy partitioning is uncertain. To examine this issue, we investigated the local impact of afforestation by comparing adjacent forest and open land pixels using satellite observations between 2001 and 2012. We attributed local surface temperature change between adjacent forest and open land to radiative and nonradiative effects over China based on the Intrinsic Biophysical Mechanism (IBM) method. Our results reveal that forest causes warming of 0.23 degrees C (+/- 0.21 ^oC) through the radiative effect and cooling of -0.74 degrees C (+/- 0.50 ^oC) through the nonradiative effect on local surface temperature compared with open land. The nonradiative effect explains about 79% (+/- 16%) of local surface temperature change between adjacent forest and open land. The contribution of the nonradiative effect varies with forest and open land types. The largest cooling is achieved by replacing grasslands or rain-fed croplands with evergreen tree types. Conversely, converting irrigated croplands to deciduous broadleaf forest leads to warming. This provides new guidance on afforestation strategies, including how these should be informed by local conditions to avoid amplifying climate-related warming.

Land cover type reconstructions, required in climate models, commonly utilize remote sensing products. There are inevitable misclassifications in land cover reconstructions due to the retrieval process. We use the Weather Research and Forecasting model to determine whether these misclassifications can affect the simulations of air temperature and rainfall over the Coordinated Regional Climate Downscaling Experiment (CORDEX) East Asia region, where the accuracy of the land cover classification is low. The Moderate Resolution Imaging Spectroradiometer land cover map is used for the control simulations and is then replaced by the most likely alternative land cover type at pixels where the classification confidence is below various threshold values. Results show that misclassification-induced land cover change can affect key biogeophysical characteristics (albedo, leaf area index, and roughness length) and these can affect the sensible and latent heat fluxes at regional scales. However, the impact on air temperature is very limited and is restricted to the Tibetan Plateau where warming of up to 2 ^oC occurs associated with the replacement of barren or sparsely vegetated land to grassland. The impact on rainfall is negligible, and most changes are likely caused by model internal variability rather than land cover change. Overall, uncertainties in the reconstruction of land cover have negligible impacts, and the Moderate Resolution Imaging Spectroradiometer land cover product can be used in regional simulations over East Asia. However, we note that land cover change experiments incorporating uncertainties must utilize large numbers of simulations if air temperature and rainfall changes are to be examined robustly.

Anthropogenic land use has a significant impact on climate change. Located in the typical East Asian monsoon region, the land-atmosphere interaction in the lower reaches of the Yangtze River is even more complicated due to intensive human activities and different types of land use in this region. To better understand these effects on micro-climate change, we compare differences in land surface temperature (T_s) for three land types around Nanjing from March to August, 2013, and then quantify the contribution of land surface factors to these differences (ΔT_s) by considering the effects of surface albedo, roughness length, and evaporation. The atmospheric background contribution to ΔT_s is also considered based on differences in air temperature ( ΔT_a. It is found that the cropland cooling effect decreases T_s by 1.76 ^oC and the urban heat island effect increases T_s by 1.25 ^oC. They have opposite impacts but are both significant in this region. Various changes in surface factors affect radiation and energy distribution and eventually modify T_s. It is the evaporative cooling effect that plays the most important role in this region and accounts for 1.40 ^oC of the crop cooling and 2.29 ^oC of the urban warming. Moreover, the background atmospheric circulation is also an indispensable part in land-atmosphere feedback induced by land use change and reinforces both these effects.

The mid-to lower reaches of the Yangtze River valley are located within the typical East Asian monsoon zone. Rapid urbanization, industrialization, and development of agriculture have led to fast and complicated land use and land cover change in this region. To investigate land-atmosphere interaction in this region where human activities and monsoon climate have considerable interaction with each other, micrometeorological elements over four sites with different surface types around Nanjing, including urban surface at Dangxiao (hereafter DX-urban), suburban surface at Xianling (XL-suburb), and grassland and farmland at Lishui County (LS-grass and LS-crop), are analyzed and their differences are revealed. The impacts of surface parameters of different surface types on the radiation budget and land surface-atmosphere heat, water, and mass exchanges are investigated and compared. The results indicate the following. (1) The largest differences in daily average surface air temperature (T_a), surface skin temperature (T_s), and relative humidity (RH), which are found during the dry periods between DX-urban and LS-crop, can be up to 3.21 ^oC, 7.26 ^oC, and 22.79 %, respectively. The diurnal ranges of the above three elements are the smallest at DX-urban and the largest at LS-grass, XL-suburb, and LS-crop. (2) Differences in radiative fluxes are mainly reflected in upward shortwave radiation (USR) that is related to surface albedo and upward longwave radiation (ULR) that is related to T_s. When comparing four sites, it can be found that both the smallest USR and the largest ULR occur at the DX-urban site. The diurnal variation in ULR is same as that of T_s at all four sites. (3) The differences in daily average sensible heat (H) and latent heat (LE) between DX-urban and LS-crop are larger than 45 and 95 Wm^-2, respectively. The proportion of latent heat flux in the net radiation (LE/R_n) keeps increasing with the change in season from the spring to summer. (4) Human activities have obvious effects on microclimate. The urban heat island (UHI) effect results in a T_a 2 ^oC higher at the urban site than other sites in the nighttime. At the crop site, LE is dominant due to irrigation, and negative H is observed since evaporation cooling leads to low T_s. Although T_s is higher at XL-suburb than that at LS-grass, there is no large difference in T_a between the two sites due to the distinct effects of the planted forest.

Land surface processes play an important role in the East Asian Summer Monsoon (EASM) system. Parameterization schemes of land surface processes may cause uncertainties in regional climate model (RCM) studies for the EASM. In this paper, we investigate the sensitivity of a RCM to land surface parameterization (LSP) schemes for long-term simulation of the EASM. The Weather Research and Forecasting (WRF) Model coupled with four different LSP schemes (Noah-MP, CLM4, Pleim-Xiu and SSiB), hereafter referred to as Sim-Noah, Sim-CLM, Sim-PX and Sim-SSiB respectively, have been applied for 22-summer EASM simulations. The 22-summer averaged spatial distributions and strengths of downscaled large-scale circulation, 2-m temperature and precipitation are comprehensively compared with ERA-Interim reanalysis and dense station observations in China. Results show that the downscaling ability of RCM for the EASM is sensitive to LSP schemes. Furthermore, this study confirms that RCM does add more information to the EASM compared to reanalysis that imposes the lateral boundary conditions (LBC) because it provides 2-m temperature and precipitation that are with higher resolution and more realistic compared to LBC. For 2-m temperature and monsoon precipitation, Sim-PX and Sim-SSiB simulations are more consistent with observation than simulations of Sim-Noah and Sim-CLM. To further explore the physical and dynamic mechanisms behind the RCM sensitivity to LSP schemes, differences in the surface energy budget between simulations of Ens-Noah-CLM (ensemble mean averaging Sim-Noah and Sim-CLM) and Ens-PX-SSiB (ensemble mean averaging Sim-PX and Sim-SSiB) are investigated and their subsequent impacts on the atmospheric circulation are analyzed. It is found that the intensity of simulated sensible heat flux over Asian continent in Ens-Noah-CLM is stronger than that in Ens-PX-SSiB, which induces a higher tropospheric temperature in Ens-Noah-CLM than in Ens-PX-SSiB over land. The adaptive modulation of geopotential height gradients affects wind field (through geostrophic balance) simulation especially at lower levels, which subsequently affects the simulation of large-scale circulation, 2-m temperature and monsoon precipitation as well as RCM's downscaling ability.

The Tibetan Plateau (TP), known as the third pole of the Earth, has snow cover with intraseasonal to decadal variability that affects weather and climate both inside and outside the TP. However, the factors that generate the TP snow cover (TPSC) anomalies at the intraseasonal time-scale are unclear. This report reveals the influence of the Madden. Julian oscillation (MJO), which is the most dominant component of the tropical intraseasonal variability, on TPSC. We focus on wintertime snow cover over the central and eastern TP, where the intraseasonal variability is large. TPSC increases/ decreases in the MJO phases 8-1/4-5, when the eastward-propagating MJO suppressed/ enhanced convection locates over the Maritime Continent. Such a change in TPSC leads to the most dominant positive/ negative anomalies of TPSC in the following phases 2-3/6-7 due to the non-significant change of TPSC in these phases. There is anomalous moisture advection over the upstream of the TP caused by MJO-excited large-scale atmospheric circulation. The advection process generates the low-frequency eastward-propagating anomalous water vapour from upstream to the TP that influences precipitation and, eventually, TPSC.

The impacts of land-use and land-cover change (LULCC) on tropospheric temperatures are investigated in this study using the fully coupled Community Earth System Model. Two simulations are performed using potential and current vegetation cover. The results show that LULCC can induce detectable changes in the tropospheric air temperature. Although the influence of LULCC on tropospheric temperature is weak, a significant influence can still be found below 300 hPa in summer over land. Compared to the global-mean temperature change, LULCC-induced changes in the regional-mean air temperature can be 2-3 times larger in the middle-upper troposphere and approximately 8 times larger in the lower troposphere. In East Asia and South Asia, LULCC is shown to produce significant decreases (0.2 to 0.4 ^oC) in air temperature in the middle-upper troposphere in spring and autumn due to the largest decrease in the latent heat release from precipitation. In Europe and North America, the most significant tropospheric cooling occurs in summer, which can be attributed to the significant decrease in the absorbed solar radiation and sensible heat flux during this season. In addition to local effects, LULCC also induces nonlocal responses in the tropospheric air temperature that are characterized by significant decreases over the leeward sides of LULCC regions, which include East Asia-western North Pacific Ocean, Mediterranean Sea-North Africa, North America-Atlantic Ocean, and North America-eastern Pacific. Cooling in the leeward sides of LULCC regions is primarily caused by an enhanced cold advection induced by LULCC.

Loess Plateau is one of the dust aerosol source regions featured by its sandy underlying surface and affected significantly by dust events. In order to investigate climatic forcing of dust aerosols in semiarid region, continuous observations of particulate matter (PM10 concentration), meteorological elements, and energy fluxes were collected at Semiarid Climate Observatory and Laboratory in northwestern China from March to May during 2007-2010. The result shows that dusty days are often evoked under the condition when a strengthening trough is combined with the development of strong convection. During dusty days, the frequency of northerly winds increases significantly with the average wind velocity to be around 4.0 m/s; temperature is low during the daytime and high at nighttime. Relative humidity and surface pressure, however, are about 15% and 70% lower than average in dusty days, respectively. Energy balance closure is typically poor in dusty days. During daytime, the downward/upward solar radiation at land surface is less in dusty days than in nondusty days with the largest difference of 206.7W/m² and 33.25 W/m², respectively. Difference in downward longwave radiation between dusty and nondusty days is 35 W/m², accounting for 11.7% and 14% of the daily mean for dusty and nondusty days, respectively. The net radiation flux, as well as sensible/latent heat fluxes at surface is smaller during the daytime but larger at nighttime in dusty days. The maximum differences of sensible/latent heat fluxes between dusty and nondusty days can reach for 41.9% and 12.1% of the maximum net radiation, respectively.

Plant leaves play an important role in water, carbon, and energy exchanges between terrestrial ecosystems and atmosphere. Assimilating remotely sensed leaf area index (LAI) into land surface models is a promising approach to improve our understanding of those processes. Toward this goal, this study uses the Community Land Model with carbon and nitrogen components (CLM4CN) coupled with the Data Assimilation Research Testbed (DART). Global Land Surface Satellite (GLASS) LAI data are assimilated via the Ensemble Adjustment Kalman Filter. A random 40-member atmospheric forcing ensemble is used to drive the CLM4CN to provide background error covariance. The results show that assimilating GLASS LAI and updating both LAI and leaf C/N is an effective method to provide a high-accuracy estimate of LAI. The simulations always systematically overestimate LAI, especially in low-latitude regions, with the largest bias up to 5 m²/m², which are effectively corrected in the analyzed LAI, with the bias reduced to +/- 1 m²/m². Significantly improved regions are located in central Africa, Amazonia, southern Eurasia, northeastern China, and western Europe, where evergreen/deciduous forests and mixed forests are dominant. Except for the temperate zone in the Southern Hemisphere, the analyzed LAI can well represent seasonal variations. The most pronounced assimilation impact in low-latitude regions is attributed to large initial forecast error covariance and sufficient background errors. The MOD 16 evapotranspiration estimates and upscaled gross primary production have been used to evaluate the assimilation impact, which highlight neutral to highly positive improvement.

The leaf area index (LAI) is a crucial parameter for understanding the exchanges of mass and energy between terrestrial ecosystems and the atmosphere. In this study, the Data Assimilation Research Testbed (DART) has been successfully coupled to the Community Land Model with explicit carbon and nitrogen components (CLM4CN) by assimilating Global Land Surface Satellite (GLASS) LAI data. Within this framework, four sequential assimilation algorithms, including the kernel filter (KF), the ensemble Kalman filter (EnKF), the ensemble adjust Kalman filter (EAKF), and the particle filter (PF), are thoroughly analyzed and compared. The results show that assimilating GLASS LAI into the CLM4CN is an effective method for improving model performance. In detail, the assimilation accuracies of the EnKF and EAKF algorithms are better than those of the KF and PF algorithm. From the perspective of the average and RMSD, the PF algorithm performs worse than the EAKF and EnKF algorithms because of the gradually reduced acceptance of observations with assimilation steps. In other words, the contribution of the observations to the posterior probability during the assimilation process is reduced. The EAKF algorithm is the best method because the matrix is adjusted at each time step during the assimilation procedure. If all the observations are accepted, the analyzed LAI seem to be better than that when some observations are rejected, especially in low-latitude regions.

The terrestrial carbon and water cycles are coupled through a multitude of connected processes among soil, roots, leaves, and the atmosphere. The strength and sensitivity of these couplings are not yet well known at the global scale, which contributes to uncertainty in predicting the terrestrial water and carbon budgets. We now have synchronous, global-scale satellite observations of critical terrestrial carbon and water cycle components: solar-induced chlorophyll fluorescence (SIF) and soil moisture. We used these observations within the framework of a global terrestrial biosphere model (Simplified Simple Biosphere Model version 2.0, SSiB2) to investigate carbon-water coupling processes. We updated SSiB2 to include a mechanistic representation of SIF and tested the sensitivity of model parameters to improve the simulation of both SIF and soil moisture with the ultimate objective of improving the first-order terrestrial carbon component, gross primary production. Although several vegetation parameters, such as leaf area index and the green leaf fraction, improved the simulated SIF, and several soil parameters, such as hydraulic conductivity, improved simulated soil moisture, their effects were mainly limited to their respective cycles. One root-mean-square error parameter emerged as the key coupler between the carbon and water cycles: the wilting point. Updates to the wilting point significantly improved the simulations for SIF and gross primary production although substantial mismatches with the satellite data still existed. This study demonstrates the value of synchronous global measurements of the terrestrial carbon and water cycles in improving the understanding of coupled carbon-water cycles.

The process of radiative transfer over vegetated areas has a profound impact on energy, water, and carbon balances over the terrestrial surface. In this paper, a generalized radiative transfer scheme (GRTS) within canopy is implemented in the Simplified Simple Biosphere land surface model (SSiB). The main concept and structure of GRTS and its coupling methodology to a land model are presented. Different from the two-stream method, the GRTS takes into account the effects of complex canopy morphology and inhomogeneous optical properties of leaves on radiative transfer process within the canopy. In the offline SSiB/GRTS simulation for the period of 2001-2012, the nonuniform leaf angle distribution within canopy layers is considered in SSiB/GRTS in the areas of evergreen broadleaf trees. Compared with the SSiB/two stream method, SSiB/GRTS produces lower canopy reflectance and higher transmittance, which leads to more realistic albedo simulation. The canopy-absorbed radiation flux in SSiB/GRTS simulation is lower than that in SSiB/two stream method simulation throughout the year in the areas of evergreen broadleaf trees. The largest difference of -18.4W/m² occurs in the Amazon region in the autumn. The ground-absorbed radiation flux increases in the SSiB/GRTS simulation, especially in the spring and autumn. The largest difference in the ground-absorbed radiation flux between SSiB/GRTS simulation and SSiB/two stream method simulation is 25.45W/m². In the boreal winter season, compared with the two-stream method in the SSiB, the GRTS gives higher surface albedo in the areas with high snow cover fraction over leaf.

Uncertainties in some key parameters in land surface models severely restrict the improvement of model capacity for successful simulation of surface-atmosphere interaction. These key parameters are related to soil moisture and heat transfer and physical processes in the vegetation canopy as well as other important aerodynamic processes. In the present study, measurements of surface-atmosphere interaction at two observation stations that are located in the typical semi-arid region of China, Tongyu Station in Jilin Province and Yuzhong Station in Gansu Province, are combined with the planetary boundary layer theory to estimate the value of two key aerodynamic parameters, i.e., surface roughness length z_0m and excess resistance kB^-1. Multiple parameterization schemes have been used in the study to obtain values for surface roughness length and excess resistance kB^-1 at the two stations. Results indicate that z_0m has distinct seasonal and inter-annual variability. For the type of surface with low-height vegetation, there is a large difference between the default value of z_0m in the land surface model and that obtained from this study. kB^-1 demonstrates a significant diurnal variation and seasonal variability. Using the modified scheme for the estimation of z_0m and kB^-1 in the land surface model, it is found that simulations of sensible heat flux over the semi-arid region have been greatly improved. These results suggest that it is necessary to further evaluate the default values of various parameters used in land surface models based on field measurements. The approach to combine field measurements with atmospheric boundary layer theory to retrieve realistic values for key parameters in land surface models presents a great potential in the improvement of modeling studies of surface-atmosphere interaction.

Precipitation is one of the major challenges in climate modeling. Among various factors, the large-scale atmospheric circulation plays an important role in modulating regional precipitation through dynamic processes that has been widely discussed in previous studies. However, few efforts have been made to investigate the relationship of model abilities to simulate precipitation and vector winds. Such an investigation may help to understand the source of uncertainty of precipitation simulation. Here, we examined the relationship between model performances in simulating precipitation with that in simulating vector winds by using the Coupled Model Intercomparison Project Phase 5 (CMIP5) models. Our results suggest that the model biases/uncertainties in simulating climatological mean precipitation often accompanied by the biases/uncertainties in vector wind fields. Model ability to simulate precipitation is closely related to the ability to simulate vector winds, especially over monsoon regions and the regions with warm and moist advection or high terrains, such as the South Asian and East Asian summer monsoon region, the Alaskan region, the Rocky Mountain, etc. Over these regions, the models with higher horizontal resolution tends to generate improved simulations in both the vector winds and precipitation relative to the models with coarse horizontal resolution. Besides, the model's ability to simulate vector winds, compared to simulate the zonal wind, meridional wind, and skin temperature, is more closely related to the ability to simulate precipitation. This indicates that it is more meaningful to evaluate the vector winds than the zonal or meridional wind from the perspective of improving regional precipitation simulation.

Vector wind plays a crucial role in shaping regional climate through transferring energy and moisture. In this study, we evaluate 37 Coupled Model Intercomparison Project Phase 5 (CMIP5) models and multi-model ensembles (MME) in terms of the climatological mean state, annual cycle, and interannual variability of vector winds in the Asian-Australian monsoon (A-AM) region. Unlike most previous studies those assessed meridional and zonal wind separately, we treat vector wind as a whole by employing a recently developed vector field evaluation method. The results are summarized as follows: (1) MME exhibits the best performance in reproducing the climatological mean of vector winds, followed by CESM1-CAM5 and three MPI-ESM models. However, models still show significant biases characterized by overestimated lower level vector winds and its spatial variation. The biases are mainly rooted in the anomaly components of vector winds and are observed in the regions with complex topography. (2) CMIP5 models can well simulate the annual cycle of upper-tropospheric vector winds, especially in the extratropical regions, but show large biases and dispersion over complex terrains in the lower troposphere. (3) MME still outperforms individual model for the simulation of interannual variance of vector winds, although most CMIP5 models overestimate the strength of vector wind variability in the lower troposphere. (4) Model skills in simulating climatological means, annual cycle, and interannual variability are positively correlated with each other to a certain degree over the A-AM region, suggesting an improvement in climatological mean may lead to a better simulation in the annual cycle or interannual variability of vector winds.

An integrated evaluation of monthly mean land surface energy fluxes over China in seven reanalysis and land model products during the period 1979-2015 is conducted. Observations from seven field sites are used to evaluate these flux products, including four reanalysis data sets and three produced by off-line land surface models. In general, the expected seasonal variations and spatial patterns in major climatic regimes are well reproduced by all reanalysis and modeling products. However, large differences among the four reanalysis products are found, while the three off-line land surface modeling products correlate well with each other. Looking at the Bowen ratio, it is found that the off-line land surface models convert a larger fraction of surface available energy into sensible heat flux compared to the reanalysis products in all climatic regimes. There are three centers of high interannual variability in sensible heat located in West China, Northeast China, and the eastern Inner Mongolia, respectively. In addition, the sensible heat flux agrees better with observations at grassland sites than at forest sites, while the latent heat flux and net radiation are significantly overestimated at forest sites in all the flux products. Besides, mean square errors of the fluxes are decomposed into biases, correlations, and differences in standard deviation. Finally, based on a ranking system adopted to quantitatively evaluate the performance of each data set, it is found that the surface energy fluxes in ERA-Interim and JRA-25 agree well with observations and the ensemble mean of all these products remains reasonably realistic as well.

This study systematically evaluates simulations of near-surface temperature and precipitation using the station observations collected in the semi-arid region of China during the Coordinated Enhanced Observing Period (CEOP) from October 2002 to December 2004 (EOP3 and EOP4). The outputs being evaluated are from eight general circulation models (GCMs) archived by the Coordinated Energy and Water Cycle Observations Project (CEOP), as well as a multi-model ensemble based on these eight models. We find that the multi-model ensemble has a better performance than most of the individual models. Our results show that all individual models and the Model Analysis Comparison (MAC) ensemble mean perform much better when simulating regionally averaged temperature than precipitation. For most models, a systematically low bias is identified in the regionally averaged simulated temperatures, while a high bias exists in the simulated precipitation except in summer. For the simulated temperatures, the lowest and largest rRMSE are found in JMA and BMRC, respectively. Furthermore, temperature is always overestimated when it is between -18 and -10 ^oC, while the temperature is underestimated when it is greater than 6 ^oC; the best performance lies between -10 and 2 ^oC for all the models except BMRC. For the simulated precipitation, excessive rainfall is reproduced at all intervals except in ECPC-SFM, and the largest deviation is identified at the interval of 2-5 mm with a bias of 18.3%. With respect to sub-regions, the simulated temperatures are better in eastern China, but the simulated precipitation is better in the transition zone from the semi-arid region to the arid region. However, the simulation bias increases west of 100 ^oE, which may be associated with the complex and steep topography there. We want to stress that the MAC ensemble mean is superior to any individual models.