Soil moisture is a key variable of land surface hydrology, and its correct representation in land surface models is crucial for local to global climate predictions. The errors may come from the model itself (structure and parameterization) but also from the meteorological forcing used. In order to separate the two source of errors, four atmospheric forcing datasets, GSWP3 (Global Soil Wetness Project Phase 3), PGF (Princeton Global meteorological Forcing), CRU-NCEP (Climatic Research Unit-National Center for Environmental Prediction), and WFDEI (WATCH Forcing Data methodology applied to ERA-Interim reanalysis data), were used to drive simulations in China by the land surface model ORCHIDEE-MICT(ORganizing Carbon and Hydrology in Dynamic EcosystEms: aMeliorated Interactions between Carbon and Temperature). Simulated soil moisture was compared with in situ and satellite datasets at different spatial and temporal scales in order to (1) estimate the ability of ORCHIDEE-MICT to represent soil moisture dynamics in China; (2) demonstrate the most suitable forcing dataset for further hydrological studies in Yangtze and Yellow River basins; and (3) understand the discrepancies of simulated soil moisture among simulations. Results showed that ORCHIDEE-MICT can simulate reasonable soil moisture dynamics in China, but the quality varies with forcing data. Simulated soil moisture driven by GSWP3 and WFDEI shows the best performance according to the root mean square error (RMSE) and correlation coefficient, respectively, suggesting that both GSWP3 and WFDEI are good choices for further hydrological studies in the two catchments. The mismatch between simulated and observed soil moisture is mainly explained by the bias of magnitude, suggesting that the parameterization in ORCHIDEE-MICT should be revised for further simulations in China. Underestimated soil moisture in the North China Plain demonstrates possible significant impacts of human activities like irrigation on soil moisture variation, which was not considered in our simulations. Finally, the discrepancies of meteorological variables and simulated soil moisture among the four simulations are analyzed. The result shows that the discrepancy of soil moisture is mainly explained by differences in precipitation frequency and air humidity rather than differences in precipitation amount.
Climate change strongly influences the hydrological cycle,
which in turn affects ecosystems services, food security, and water resources
SM indeed plays a crucial role in adjusting local climate
In the investigation of spatial and temporal SM variations, in situ
measurements
Land surface models (LSMs) are able to simulate the short- and long-term SM
dynamics consistently with atmospheric forcing and surface information
In this study, the land surface model ORCHIDEE-MICT (ORganizing Carbon
and Hydrology in Dynamic EcosystEms: aMeliorated Interactions between Carbon
and Temperature;
Four global atmospheric forcing datasets are chosen to drive the simulations
in China, including GSWP3 (Global Soil Wetness Project Phase 3), PGF
(Princeton Global meteorological Forcing), CRU-NCEP (Climatic Research
Unit-National Center for Environmental Prediction), and WFDEI (WATCH Forcing
Data methodology applied to ERA-Interim reanalysis data), due to their wide
application in numerous hydrological studies
Our SM simulations are evaluated with different SM datasets including in situ
data, remote sensing measurements, and reanalysis. In situ measurements
including ISMN (International Soil Moisture Network;
Finally, GLEAM SM data (Global Land Evaporation Amsterdam Model;
Through the simulations and comparisons, three questions will be
addressed:
Is the model able to provide a reasonable estimation of SM dynamics in China, as a prerequisite for further hydrological studies? Which atmospheric forcing gives the best SM simulation
according to the comparisons with available observations? Which meteorological variable drives the differences of SM among the simulations?
The study area, atmospheric forcing, and SM datasets used in this study are
described in Sect.
China has multiple climate regimes, which creates hydrological
situations influenced by different variables in different regions. The land
water budget in China is affected by anthropogenic factors, such as irrigation
Four important regions mentioned in this paper (green rectangles). The grey background is the cropland fraction.
Four gridded atmospheric forcing datasets are used to force the model over
China: GSWP3, PGF, CRU-NCEP, and WFDEI. All input variables needed are the
air temperature at 2 m (
General information of the climate forcing datasets. “Reanalysis”
and “Observations” are corresponding datasets used in producing the
atmospheric forcing. Detailed description can be found in
Sect.
The GSWP3 v0 (
The PGF (
The CRU-NCEP v6.1
(
The WFDEI forcing (version 31 July 2012) is generated by applying the WATCH
Forcing Data methodology (
The ISMN is an international cooperative project providing a global gauged SM
database
In spite of the long length of this dataset, the data availability and monitoring period among stations vary widely. Some stations only recorded SM during the growing season, while others have a full year record. Furthermore, the measurements including the five deep layers (below 50 cm) are fewer in number than those including the top six layers. Only stations with more than 15 years of data were selected, which at least cover the same period (1984–1999). To make sure that at least half of the data are available in the 15-year time series, stations with fewer than 270 measurement points in the top six layers are removed. This leads to selecting a subset of 20 stations, and given the sparseness of data below 50 cm, only SM in the top six layers is used for model evaluation.
The SM was measured over 778 stations of agro-meteorological stations over
China by the Chinese Meteorological Administration and collected and
harmonized by the research team in Peking University (PKU;
The ESA CCI SM is a multi-satellite-based product
GLEAM v3.0 is a multiple-algorithm, observation-based model reconstructing
the components of the land evaporation process, including daily SM,
evapotranspiration, and interception at 0.25
Both surface and root-zone SM from GLEAM, which has been validated by
ORCHIDEE (Organizing Carbon and Hydrology In Dynamic EcosystEms;
There are two main outputs of SM in ORCHIDEE. The total SM
(
Four simulations were performed driven by different forcing datasets
described in Sect.
The temporal resolution of forcing datasets is either 3-hourly or 6-hourly (CRU-NCEP), which is larger than the simulation time step of SECHIBA (30 min). To have a reasonable precipitation intensity, and thus a good infiltration of water in the soil, the default precipitation splitting algorithm of ORCHIDEE is applied in our simulations. At the beginning of each forcing time step, if precipitation occurred, the precipitation amount (precipitation rate multiplied by the time interval of specific forcing) will be uniformly distributed to the first half of the forcing time step.
Summary of the SM datasets for validation. “M+RS+RA”
indicates that the dataset is a model output driven by both remote sensing
and reanalysis data. More details can be found in
Sect.
As the soil depths, periods, and spatiotemporal resolutions are different in
the four SM datasets (Sect.
According to the sampled depth of the ESA CCI SM, the daily top four-layer
(2.2 cm) averaged SM from ORCHIDEE is used. Regarding the definition of
GLEAM SM (Sect.
Pearson's correlation coefficient (
The root mean square error (RMSE) is applied in order to estimate the temporal differences between simulation and observation. The same data pairs are used for RMSE calculation as the correlation coefficient except for PKU due to the normalization. Note that RMSE is related to the magnitude of SM, which varies significantly in China. To make it comparable in space, the relative RMSE is calculated by dividing the mean of the simulated and observed SM.
According to
Finally, to evaluate the characteristic timescale of modeled SM response to
hydrological processes, the lag-
The linear trend of SM change in the 29 years is of interest as well. The
Mann–Kendall test
Pearson's correlation coefficients of modeled and measured SM at each
gauging station from ISMN (triangles) and PKU (circles). Symbols with a dark
border indicate significant correlations (
Time series of 10-day SM from ORCHIDEE and ISMN at three stations.
The station locations are shown in Fig.
In our simulations, the difference in atmospheric forcing is the only source
of difference in simulated SM. We look at different climate variables to
explain SM differences among simulations. These variables include monthly
precipitation amount (
In most cases, the correlations between modeled and measured SM at ISMN
stations (see Sect.
The correlation coefficients of
According to the
Figure
Figures S2 and
The availability and uncertainty of CCI SM vary with space and time
(Sect.
The left panel of Fig.
Figure S6 and the right panel of Fig.
Figure
Median of metrics in specific comparisons. The subscripts of correlation coefficients indicate the quantile of stations (samples) with significant correlation (
NA: not available.
The trend of ORCHIDEE
Evaluation of the forcing datasets for simulating SM dynamics in
China, YZRB, and YLRB.
To find the most realistic forcing dataset for SM performance given the
ORCHIDEE model, several metrics were calculated and are shown in
Fig.
Thus we conclude that both GSWP3 and WFDEI are suitable to simulate SM dynamics in China with ORCHIDEE. The best choice can be made based on the main focus of specific research. For estimating magnitude of SM, GSWP3 is preferable; for investigating SM variation, WFDEI is the best choice. Note that this study only provides the evaluation of SM, but other hydrological components should be compared with observations to confirm the superiority of GSWP3 and WFDEI.
Matrix of correlation coefficients between the
By investigating the
Figure S11 shows maps of
To look for clearer links between input and SM
Due to the spatiotemporal complexity of SM and its vertical profile, four
datasets were selected to drive the simulations, and modeled SM at different
depths was validated against multiple datasets. The results showed that
ORCHIDEE SM coincides well with CCI (median
Higher
Because the in situ SM measurements were only collected for croplands and
grasslands
In the comparison to CCI and GLEAM SM, low
From the results we conclude that ORCHIDEE provides a satisfactory simulation
of SM dynamics in China, except in areas subject to irrigation. This calls
for inclusion of irrigation and realistic crop phenology
In Sect.
Differences in
Estimating impacts of meteorological factors on SM dynamics is difficult.
First of all, the importance of a meteorological variable on SM may vary with
climate regimes. For instance, the importance of precipitation and radiation
on SM changes from water- to energy-limited regions. Secondly, impacts of
meteorological variables can be nonlinear through interactions with local
ecosystems
Indeed this approach is not able to demonstrate explicit links between
meteorological variables and SM. We underlined the impacts of
Simulations in China were performed in ORCHIDEE-MICT driven by different forcing datasets, GSWP3, PGF, CRU-NCEP, and WFDEI. Simulated soil moisture was compared to several datasets to evaluate the ability of ORCHIDEE-MICT in reproducing soil moisture dynamics in China. Results showed that ORCHIDEE soil moisture coincided well with other datasets in wet areas and in non-irrigated areas. It was suggested that ORCHIDEE-MICT is suitable for further hydrological studies in China. However, the abnormal variation of observed SM in North China Plain implied potential impacts of irrigation, which was recommended to be considered in further simulations. Moreover, results showed that bias was mainly from model parameterization and atmospheric forcing. Thus parameterizations in ORCHIDEE-MICT should be calibrated, and atmospheric forcing should be carefully selected to reflect the situation of China.
Several criteria were chosen and compared among the four simulations in China, YZRB, and YLRB. Results showed that GSWP3 and WFDEI, which had the best performances in correlation coefficients and RMSE, respectively, were ideal choices for hydrological study in China. However, higher MSD in the Yellow River basin reflected the complicated climate conditions in northern China, which might be significantly influenced by human activities as well. Finally, we used the differences of simulated soil moisture and meteorological variables to simply investigate the linkage between them. Results showed that the differences of simulated soil moisture were mainly explained by the differences of air humidity and precipitation frequency among the four atmospheric forcing datasets. However, this coarse analysis cannot give explicit explanations about related mechanisms. Further study is needed to discover the interactions between soil water and climate through tracing the surface hydrological cycles and energy balances.
The SVN version of ORCHIDEE-MICT used in this study is
3952, which is available at
Philippe Ciais (Laboratoire des Sciences du Climat et de l'Environnement, CNRS-CEA-UVSQ, Gif-sur-Yvette 91191, France), Patrice Dumas (Centre de Coopération Internationale en Recherche Agronomique pour le Développement, Avenue Agropolis, 34398 Montpellier CEDEX 5, France), Xiaoming Feng (State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China), Matthieu Guimberteau (Laboratoire des Sciences du Climat et de l'Environnement, CNRS-CEA-UVSQ, Gif-sur-Yvette 91191, France; UMR 7619 METIS, Sorbonne Universités, UPMC, CNRS, EPHE, 4 place Jussieu, Paris 75005, France), Laurent Li (Laboratoire de Météorologie Dynamique, UPMC/CNRS, IPSL, Paris 75005, France), Catherine Ottlé (aboratoire des Sciences du Climat et de l'Environnement, CNRS-CEA-UVSQ, Gif-sur-Yvette 91191, France), Shushi Peng (Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), Shilong Piao (Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), Jan Polcher (Laboratoire de Météorologie Dynamique, UPMC/CNRS, IPSL, Paris 75005, France), Pengfei Shi (State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Center for Global Change and Water Cycle, Hohai University, Nanjing 210098, China), Shuai Wang (State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China), Xuhui Wang (Laboratoire des Sciences du Climat et de l'Environnement, CNRS-CEA-UVSQ, Gif-sur-Yvette 91191, France; Laboratoire de Météorologie Dynamique, UPMC/CNRS, IPSL, Paris 75005, France; Laboratoire de Météorologie Dynamique, UPMC/CNRS, IPSL, Paris 75005, France; Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), Yi Xi (Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), Hui Yang (Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), Tao Yang (State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Center for Global Change and Water Cycle, Hohai University, Nanjing 210098, China), Zun Yin (Laboratoire des Sciences du Climat et de l'Environnement, CNRS-CEA-UVSQ, Gif-sur-Yvette 91191, France), Xuanze Zhang (Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), Feng Zhou (Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China), and Xudong Zhou (State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Center for Global Change and Water Cycle, Hohai University, Nanjing 210098, China).
PC, CO, and ZY designed the research. ZY performed the research, analyzed the data, and wrote the draft; all authors contributed to interpreting results, discussing findings, and improving the manuscript.
The authors declare that they have no conflict of interest.
This study was supported by the National Natural Science Foundation of China (grant number 41561134016) and by the CHINA-TREND-STREAM French national project (ANR grant no. ANR-15-CE01-00L1-0L). Matthieu Guimberteau and Philippe Ciais acknowledge support from the European Research Council Synergy grant ERC-2013-SyG-610028 IMBALANCE-P. Hyungjun Kim was supported by Japan Society for the Promotion of Science KAKENHI (16H06291). We thank Brecht Martens and Suxia Liu for helpful discussion about GLEAM and ISMN datasets. We gratefully acknowledge two anonymous referees and the editor for their helpful comments and efforts. Edited by: Miriam Coenders-Gerrits Reviewed by: two anonymous referees