Improving the quality of simulated soil moisture with a multi-model ensemble approach

Multi-model ensembles have been found to perform significantly better than a single-model system in weather and seasonal climate forecasts. This paper explores the possibility of applying the multi-model ensemble approach to the land surface component in order to improve the quality of simulated soil moisture. The simple average of 17 multiyear global soil moisture products is validated with long-term in situ data in five regions (Illinois, USA; China; India; Mongolia and the former Soviet Union). The results show that in all regions the multi-model analysis is clearly better than most individual products in simulating the phasing of the annual cycle, interannual variability, and magnitudes in observed soil moisture. The sensitivity of the performance of the multi-model analysis to ensemble member size and composition is also examined. It is found that there is usually a clear improvement when a product of higher correlation to observations or lower error is included in the multi-model ensemble, while there is no apparent degradation when a product with relatively poor skill is included. Copyright (C) 2007 Royal Meteorological Society.