Understanding of accuracy on calculated soil moisture field for the study of land-atmosphere interaction

Understanding the state of soil moisture is effective to enhance climate predictability on inter-seasonal or annual time scales. Thus, the Global Soil Wetness Project (GSWP) has been implemented as an environmental modeling research activity. The SiBUC (Simple Biosphere including Urban Canopy) land surface model is one of the participants of the 2nd GSWP, and it uses mosaic approach to incorporate all kind of land-use. In order to estimate the global soil moisture field as accurately as possible and to utilize the products of GSWP2 simulation more efficiently, SiBUC is run with irrigation scheme activated. Integration of one-way uncoupled SiBUC model from 1986 to 1995 have produced global soil moisture field. Both the model and forcing data may contain uncertainty. However, the SiBUC model is one of the few models which can consider irrigation effect. And also, the advantage of the meteorological forcing data provided from GSWP2 is hybridization among reanalysis products, observation data and satellite data. In this sense, it is assumed that GSWP2 products is the most accurate global land surface hydrological data set in available. Thus, these global products should be applied to land-atmosphere interaction study, if possible. To do this, it is important to understand inter-annual or much higher time scale accuracy on calculated soil moisture filed. In this study, calculated soil moisture field are validated with observation of soil moisture in five regions (Illinois:USA, China, India, Mongolia, Russia). The Russian data has two types data: one is located in spring wheat and another is located in winter wheat. These observation data are provided from Global Soil Moisture Data Bank (GSMDB). To understand the time scale accuracy on soil moisture field, three correlation coefficients are calculated between calculated soil moisture and observed soil moisture: inter-annual, inter-seasonal and monthly mean correlation, respectively. As a result, if the median value in each region is focused on, high monthly correlation are shown in Illinois (0.83) and India (0.75). In these regions, inter-seasonal correlation is also high, but inter-annual correlation becomes lower. On the other hand, in China or Mongolia, all median value of correlation is low. And, both types of monthly correlation in Russia are relatively high (0.69, 0.65). In addition, inter-seasonal and inter-annual correlation are almost same as monthly correlation. From the result, from the viewpoint of regional scale, calculated soil moisture field in Illinois and India have high accuracy on monthly and inter-seasonal time scales. In Russia, calculated soil moisture field has relatively high accuracy on any time scales. Low accuracy on monthly time scale in China and Mongolia correspond to the result of multi-model analysis and validation (Guo et al., 2007). From this concurrent result, it is assumed that it is difficult to estimate the soil moisture field in China and Mongolia for any models or meteorological forcing data have some uncertainty. Nevertheless these reason should be investigated.