COMPARING BIAS CORRECTION METHODS TO IMPROVE MODELLED PRECIPITATION EXTREMES

Abstract

This study aims to analyze and improve modelled extreme precipitation. It was conducted in the German Federal State of Saxony using the WEREX V data set. WEREX V is a model that statistically downscales Global Circulation Model (GCM) data. Inputs for the WEREX V model included GCMs ECHAM 5, HadCM3C and HadGEM2 (sometimes downscaled with Regional Climate Models RCMs REMO, RACMO and CCLM), SRES scenarios A1B and E1, and different model runs. The output of analysis was shown by a boxplot since the WEREX V data set has 120 future projections of precipitation. The model results were verified against observed data obtained from representative meteorological stations, and systematical deviations or biases were identified. To improve the model results, two bias correction methods were applied with special emphasis given to the reproduction of precipitation extremes. Empirical quantile mapping and gamma quantile mapping methods were applied. The ability of the WEREX V ensemble to capture extreme precipitation values varied; this was described in terms of biases. All of the identified correction methods were capable of reducing the bias related to the intensity of extreme precipitation occurrence during the calibration period. The performance of empirical quantile mapping is better than gamma quantile mapping to reduce biases (median value) and uncertainty (inter quartile range value).

Penelitian ini bertujuan untuk mengoreksi bias curah hujan ekstrim keluaran model. Wilayah kajian dalam penelitian ini adalah negara bagian Saxony, German sedangkan data model yang digunakan adalah data WEREX V. Dataset WEREX V adalah data GCM yang yang didownscale secara statistic. Adapun GMC yang digunakan adalah ECHAM 5, HadCM3, HadGem2 dan beberapa RCM (REMO, RAMCO, dan CCLM) dengan menggunakan skenarios SRES A1B and E1. Karena dataset WEREX V terdiri dari 120 data model, maka boxplot digunakan untuk menggambarkan hasil analisis baik untuk identifikasi maupun koreksi bias. Hasil keluaran model dibandingkan dengan data pengamatan (observasi) dari stasiun meteorologi. Dari hasil perbandingan ini, bias akan dideteksi. Untuk meningkatkan akurasi model, bias dikoreksi menggunakan dua metode yaitu Emperical quantile mappinh (EQM) dan Gamma quatile mapping (gamma). Kemampuan model (data WEREX V) untuk menggambarkan curah hujan ekstrim berbeda antar stasiun hal ini digambarkan dengan nilai bias yang berbeda. Metode EQM dan Gamma mampu mengurangi bias maupun ketidakpastian model (uncertainty). Performa EQM lebih baik dibandingkan Gamma. Secara umum EQM mampu mengurangi bias maupun ketidakpastian model.