Soil moisture is a key state variable in many hydrological processes. The Global Land Data Assimilation System (GLDAS) can produce global and continuous soil moisture data sets which have been used in many applications. In this study, simulated soil moisture from four land surface models (LSM) (Mosaic, Noah, Community Land Model, and Variable Infiltration Capacity) in GLDAS-1 and the more recent GLDAS-2 were evaluated against in situ soil moisture measurements collected from two soil moisture networks located on the Tibetan Plateau at different soil depths. The two networks provide a representation of different climates and land surface conditions on the Tibetan Plateau which can make the evaluation results more robust and reliable. The results show that all the LSMs can well capture the temporal variation of observed soil moisture with the correlation coefficients mostly being above 0.5. However, they all display biases with the surface soil moisture being systematically underestimated in both of two network regions, and the Mosaic model always shows the largest bias that even reaches 0.192m~3/m~3. The causes of the biases were investigated in detail, and we found that the biases may mainly be caused by the soil stratification phenomenon over the Tibetan Plateau. Moreover, errors in model parameters, especially the soil properties data, deficiencies in model structures, and mismatch of the spatial scale and soil depth between LSM simulations and in situ measurements may contribute to the biases as well. Additionally, it was found that GLDAS-2 nearly does not show superior performance than GLDAS-1 over the Tibetan Plateau.
展开▼
