Assessment of Simulated Water Balance from Four Land Surface Models using the NLDAS Test Bed
Land surface models (LSMs) are a key component in weather and climate models. A team of scientists, including U.S. Department of Energy researchers at Pacific Northwest National Laboratory, assessed the performance of four LSMs for hydrologic simulations over the continental United States. using the North American Land Data Assimilation System (NLDAS) test bed.
The four LSMs are as follows: the baseline community Noah LSM (Noah, version 2.8); the Variable Infiltration Capacity (VIC, version 4.0.5) model; the substantially augmented Noah LSM with multi-parameterization options (Noah-MP); and the Community Land Model version 4 (CLM4). The team found that compared to Noah, the other three models show significant improvements in simulating terrestrial water storage and streamflow, and moderate improvements in estimating evapotranspiration and soil moisture. Among these LSMs, Noah-MP shows the best performance in simulating soil moisture and is among the best in simulating terrestrial water storage; CLM4 shows the best performance in simulating evapotranspiration; and VIC shows the best performance in simulating streamflow. The Noah-MP-simulated evapotranspiration grows too fast in the spring, which coincides with the fact that its modeled leaf area index peaks too soon. Although the CLM4 model produced the highest-correlation for terrestrial water storage for the entire continental United States, it produced either too high or too low amplitude of the annual terrestrial water storage variation. In addition, CLM4 produced much weaker soil moisture variability than both Soil Climate Analysis Network Observations and other LSMs. Finally, the VIC model constantly overestimated evapotranspiration compared to observations from Moderate-Resolution Imaging Spectroradiometer and eddy covariance flux tower sites.
The team demonstrated that, by providing reliable atmospheric forcing data, four LSMs, most appropriate observational data, and necessary tools, the NLDAS test bed is a valid platform for evaluating land models on continental or large river basin scales in the United States.
Land surface models (LSMs) are a key component in weather and climate models. A team of scientists, including U.S. Department of Energy researchers at Pacific Northwest National Laboratory, assessed the performance of four LSMs for hydrologic simulations over the continental United States. using the North American Land Data Assimilation System (NLDAS) test bed. The four LSMs are as follows: the baseline community Noah LSM (Noah, version 2.8); the Variable Infiltration Capacity (VIC, version 4.0.5) model; the substantially augmented Noah LSM with multi-parameterization options (Noah-MP); and the Community Land Model version 4 (CLM4). The team found that compared to Noah, the other three models show significant improvements in simulating terrestrial water storage and streamflow, and moderate improvements in estimating evapotranspiration and soil moisture. Among these LSMs, Noah-MP shows the best performance in simulating soil moisture and is among the best in simulating terrestrial water storage; CLM4 shows the best performance in simulating evapotranspiration; and VIC shows the best performance in simulating streamflow. The Noah-MP-simulated evapotranspiration grows too fast in the spring, which coincides with the fact that its modeled leaf area index peaks too soon. Although the CLM4 model produced the highest-correlation for terrestrial water storage for the entire continental United States, it produced either too high or too low amplitude of the annual terrestrial water storage variation. In addition, CLM4 produced much weaker soil moisture variability than both Soil Climate Analysis Network Observations and other LSMs. Finally, the VIC model constantly overestimated evapotranspiration compared to observations from Moderate-Resolution Imaging Spectroradiometer and eddy covariance flux tower sites. The team demonstrated that, by providing reliable atmospheric forcing data, four LSMs, most appropriate observational data, and necessary tools, the NLDAS test bed is a valid platform for evaluating land models on continental or large river basin scales in the United States.
This work is supported by the NASA grant NNX11AJ43G, the National Natural Science Foundation of China grant 41375088, and the Ronald K. DeFord Field Scholarship (2012) of the University of Texas at Austin. Y.X. was sponsored by NOAA/CPO/MAPP. M.H. and L.R.L. were supported by the Office of Science of the U.S. DOE through the Earth System Modeling program. PNNL is operated for the U.S. DOE by Battelle Memorial Institute under contract DE-AC05-76RLO1830. The authors would like to thank Huilin Gao for her help in calculating the modeled TWS anomaly, Qiaozhen Mu for providing 1/8 degree monthly MODIS ET data over CONUS, Rolf Reichle for providing the quality-controlled SCAN soil moisture observations, and three anonymous reviewers for their constructive comments. The observational data used in this study are available from the data sources described in the paper. The model output data presented here are available upon request to the corresponding author (liang@jsg.utexas.edu).