APPLICATION OF STATISTICAL DOWNSCALING FOR SEASONAL RAINFALL FORECASTS IN CAMBODIA: A COMPARISON BETWEEN CONSTRUCTED ANALOGUE AND BIAS CORRECTION METHODS

• Main Author: Thean, THOEURN
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/43892
• Institution: Institut Teknologi Bandung

Seasonal rainfall prediction is a very useful product which is used for various purposes especially Agriculture in Cambodia. The forecast of the rain in Cambodia contains a high value of uncertainty due to high rainfall variability and complex geography. The level of uncertainty can be delivered quantitatively using probability. This study aims (1) to apply and evaluate two statistical downscaling methods; Constructed Analogue (CA) and Bias Correction (BC) Methods based on the Climate Forecast System version 2 (CFSv2) output for seasonal rainfall forecasts and (2) to assess the probabilistic forecasting skills of the forecasted rainfall over Cambodia. The predictors are stream function (?), velocity potential (?) and Geopotential Height at 850 hPa level (z850), whereas the predictand is rainfall for CA method. Based on the statistical performance, the predicted rainfall using wind fields show better performance compared with z850. The downscaled rainfall of combined ensembles still exhibits higher skill than wind fields (???? and ????) and z850. Bias Correction Method shows lower correlation compared to CA method. However, the Brier Score using to evaluate the seasonal extreme showed that the CA method was not able to capture the extreme event. Based on the reliability diagram, the CA method is more reliable than the BC method in the central area which can be considered as marginally useful for decision making.