Blending of multi-source soil moisture data using Ensemble Bayesian model averaging

Soil moisture is an important hydro-meteorological parameter that affects energy budget and surface-atmosphere interaction. Soil moisture is estimated and observed through various source such as physical models, in-situ observations, and satellites. In-situ soil moisture data. These various soil moisture data may show different values because the size of the uncertainty of each data is different. The purpose of this study is to produce the blending map of the soil moisture obtained from various source. Ensemble Bayesian model averaging (EBMA) that is weighted-averaging method using posterior probability as weight is used as blending method. As the ensemble members, two satellite-based data and two model-based data were used. The satellite-based data are soil moisture data derived from Global Change Observation Mission-Water(GCOM-W1) and Soil Moisture Active Passive(SMAP). And the model-based data are one derived from ERA-interim of European Centre for Mediom-rage Weather Forecastes(ECMWF) and Global Land Data Assimilation System (GLDAS). To obtain the posterior probability and parameters and be used for error statistics, in-situ data measured by Time domain reflectrometry(TDR) sensor from Rural Development Administration(RDA) of South Korea was used as reference data. As the result, compared to each ensemble member, the EBMA soil moisture data showed an improvement in accuracy.