How do model error and localization approaches affect model parameter estimation in the LETKF?

JUAN RUIZ; TAKEMASA MIYOSHI; MASARU KUNII

Montreal

Congreso; World Weather Open Science Conference; 2014

World Meteorological Organization

This study explores how localization methods affect model parameter estimation using the local ensemble transform Kalman filter (LETKF) with the Weather Research and Forecasting (WRF) model. The impact of different localization approaches is investigated in the real-world case of Typhoon Sinlaku (2008). Horizontal localization has been considered either as two dimensional fields or as global constants. The effect of applying vertical localization for the estimation of surface parameters is also explored. The effect of model error upon the estimation of the model parameters is investigated in a set of twin experiments with realistically distributed observations. In all cases, the LETKF successfully identified better parameter values of the sea-surface moisture exchange coefficients and improved the forecast of atmospheric variables near the surface, leading to a significant reduction of the moisture and temperature biases near the surface. The results indicate that the localization approach plays an essential role in parameter estimation and thus the analysis and forecast quality. Model error also produces an important impact upon the estimated parameters, reducing the improvement introduced by the estimation of the model parameters.