Comparisons of Simulations of Soil Moisture Variations in the Yellow River Basin Driven by Various Atmospheric Forcing Data Sets

Based on station observations, The European Centre for Medium-Range Weather Forecasts reanalysis (ERA40), the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis and Princeton University's global meteorological forcing data set (Princeton), four atmospheric forcing fields were constructed for use in driving the Community Land Model version 3.5 (CLM3.5). Simulated soil moisture content throughout the period 1951-2000 in the Yellow River basin was validated via comparison with corresponding observations in the upper, middle, and lower reaches. The results show that CLM3.5 is capable of reproducing not only the characteristics of intra-annual and annual variations of soil moisture, but also long-term variation trends, with different statistical significance in the correlations between the observations and simulations from different forcing fields in various reaches. The simulations modeled with station-based atmospheric forcing fields are the most consistent with observed soil moisture, and the simulations based on the Princeton data set are the second best, on average. The simulations from ERA40 and NCEP/NCAR are close to each other in quality, but comparatively worse to the other sources of forcing information that were evaluated. Regionally, simulations are most consistent with observations in the lower reaches and less so in the upper reaches, with the middle reaches in between. In addition, the soil moisture simulated by CLM3.5 is systematically greater than the observations in the Yellow River basin. Comparisons between the simulations by CLM3.5 and CLM3.0 indicate that simulation errors are primarily caused by deficiencies within CLM3.5 and are also associated with the quality of atmospheric forcing field applied.