การปรับแก้ความคลาดเคลื่อนของปริมาณน้ำฝนจากแบบจำลองภูมิอากาศภูมิภาค WRF-CFSR ภาคเหนือตอนบนของประเทศไทย

Abstract

The objective of this study aims to find and test the precipitation bias correction methods or techniques and evaluate the efficiency of the bias correction methods by applying them to the WRF-CFSR regional climate model output. In this study were selected two methods for bias correction that include: 1) Cumulative Distribution Function (CDF) and 2) Empirical Orthogonal Function (EOF) to correct the discrepancy in monthly precipitation in upper northern Thailand covering 1980-2010 (31 years) and grided observation data (APHRODITE, CRU, GPCC, and GPCP) were used to compare the WRF-CFSR regional climate model precipitation data. The study found that both bias correction methods were able to correct the precipitation values of the WRF-CFSR regional climate model with the consistent and match for each grid observation dataset (APHRODITE, CRU, GPCC, and GPCP). The CDF bias correction method is suitable for a station or point-to-point that is spatially specific, which can reduce the error from the processing model, and reduce the systematic error in the model. As for the bias correction result using the EOF method, which calculated the spatial and temporal relative precipitation data that able to maintain spatial continuity of precipitation anomalies and able to maintain patterns of precipitation between the observation data and the data from the model as well. Therefore, the EOF analysis result is appropriate for the variance and anomalies analysis of long term precipitation. The correlation (r) validation results of bias correction using CDF and EOF methods found the precipitation values from the WRF-CFSR regional climate model were correlated at improved compared to uncorrected. The validation of the effectiveness of the correction method with the Root Mean Square Error (RMSE) found that the CDF bias correction method can reduce the RMSE between the observed and the model data to be more reliable and accurate while the EOF bias correction method is still unable to reduce the error of precipitation.