Modeling and analysis of regional and global soil moisture variations.

摘要

Simulating land surface hydrological states and fluxes requires a comprehensive set of atmospheric forcing data at consistent temporal and spatial scales. At the continental-to-global scale, such data are not available except in reanalysis products. Unfortunately, reanalysis products are biased due to errors in the host weather forecast model. This research explores whether errors in model predictions of the initial soil moisture status and hydrological fluxes can be minimized by reducing bias to the European Centre for Medium Range Weather Forecast (ECMWF) and National Center for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis products. The bias reduction scheme uses both difference and ratio corrections based upon global observational data sets to create forcing data sets for North America and the global land surface for the time period 1979–1993.; The forcing data was used in a land surface model to simulate hydrological stores and fluxes for North America, and globally. In the North American simulations, soil moisture, snow depth and runoff output completed with bias reduced forcing are in much better agreement with observations than simulations completed with the raw reanalyses.; For the global scale simulations, soil moisture estimates were compared to in-situ, satellite observations and to the modeled estimates of Nijssen et al. [2001]. In general, agreements between anomalies in modeled and observed root zone soil moisture are high. Similarly for the surface soil wetness state, correlations between modeled estimates and satellite observations are also statistically significant; however, correlations decline with increasing sub-grid variability and vegetation amount. Comparisons to the data set of Nijssen et al., [2001] demonstrates that both simulations present complimentary estimates of wet and dry root zone soil moisture anomalies, despite being derived from different land surface models, and with different data sources for meteorological forcing.; Finally, the two bias reduced forcing data sets were used to simulate the initial soil moisture state for the North American continent (1985–1993). Differences between simulations were shown to persist over regions with the greatest soil memory, and were not strongly associated to patterns of difference in forcing fields. Therefore, the processes that contribute to soil memory may also limit our ability to accurately estimate its initial state.