Verification of forecast rainfall anomalies

Abstract

Statistical downscaling is used to relate the large scale climate information with the local variables that is to find the relationship between the National Center of Environmental Prediction (NCEP) data with the ground data. This study examines the verification of forecast rainfall anomalies during November-December-January-February (NDJF). The ground data used is the 30 years NDJF rainfall for 40 stations while the NCEP data is the 20 grids point Sea Level Pressure (SLP). In this paper, Canonical correlation analysis (CCA) is used to find the maximum correlated pattern between two variables. CCA model is verified using the mean square error skill score and anomaly correlation coefficient and used to simulate the current rainfall using the General Circulation Model (GCM) data as predictors. This is so called the validation method. Due to appearance of some biases, the anomaly correlation coefficient is considerably higher than the skill score. These biases may relate to the penalty associated with retaining the Sea Level Pressure (SLP) in the meteorological features when such features are not predictable.