Forecast errors in land-surface variables over South America

PABLO C. SPENNEMANN; A. CELESTE SAULO

Exeter

Workshop; 4th WGNE Workshop on Systematic Errors in Weather and Climate Models; 2013

Land surface variables (e.g. soil moisture, evapotranspiration, etc.) play an important role in the energy balance between the land surface and the atmosphere as well as in the evolution of the hydrological cycle. A correct representation of these variables in numerical weather forecasts is essential to improve weather forecasting. The goal of this work is to evaluate how the numerical prediction model WRF-ARW (Weather Research and Forecasting) represents these soil variables over South America. A special emphasis is set on soil moisture, both in the initial condition and in the forecasts with different lead times (up to 96 hours) for the 2002-2003 summer period. The initial condition and forecasts were compared with the GLDAS (Global Land Data Assimilation System) data set, through mean fields and time evolution of areal averages over two sub-regions of interest: Southeastern South America and the South Atlantic Convergence Zone. Forecasts show similar patterns to those obtained with the GLDAS, although there are some evidences of an error increase (both systematic and non-systematic) with forecast length. These errors are related with the uncertainty in the initial condition and also with errors in the representation of soil moisture model driver, mainly rainfall. The presence of non-systematic errors, suggest a major problem related with the implementation of methods for reducing these errors. However, a proper bias correction method is proposed to reduce soil moisture initial condition systematic errors, in order to improve forecast quality.