SSM 기법 기반 정량적 일 강수량 예보의 검증에 관한 연구

Abstract

Recently, severe weather phenomenon such as heavy rainfall and typhoon occurred frequently due to the climate change. So, the water management is more important than before. The efficient water management accompanied by meteorological information is needed. Especially, quantitative precipitation prediction which is the unit of monthly or seasonal is necessary for long-term water management. For this, In previous study, daily precipitation with unit of one month was predicted and verified from June 2009 to August 2009 (Moon, 2010). But, verification period of previous study is limit to the summer. So, we need to reverify during longer period. In this study, we forecasted and reverified from January 2012 to December 2012 using the SSM(Slice Step Method) technique which is forecast system of daily precipitation with unit of one month based on short-range forecast and medium-range forecast using NWP model and long-range forecast using statistical model. To verify the high-resolution precipitation data, we used the Neighborhood which is verification method (Theis et al, 2005; Ebert, 2008) and checked the change of FSS (Fractions Skill Score) (Roberts and Lean, 2008) according to percentile thresholds by the Neighborhood size. We used a contingency table to verify whether precipitation occurred or not and category divided by precipitation thresholds. Using a contingency table, we calculated skill score which is ACC (Accuracy), CSI (Critical Success Index), POD (Probability of Detection), SR (Success Ratio), FAR (False Alarm Rate), and ETS (Equivalent Threat Score) to evaluate the accuracy. The results were expressed as the Performance Diagram and ROC (Relative Operating Characteristic) curve. Using monthly skill score, the hypothesis test was performed (Hamill, 1999; Jolliffe, 2007; Mason, 2008). Through the hypothesis test, we determined the distinction of skill scores according to the forecast process. To check the difference between observation and forecast, we expressed as the Histogram (Brooks and Doswell, 1996). In the results of verification, through the analysis of the Performance diagram, ROC curve and Histogram, the accuracy of short-range forecast process was the highest. The hypothesis test results showed that there is some difference between short and long-range forecast process but no difference has been found between medium and long-range forecast process in the 95% confidence level. From this study, it shows that daily precipitation forecast based on the SSM technique has high accurate forecast before 10 days and in spite of over a period of two weeks that is forecast limit proposed by Lorenz (1963, 1982), maintains confidence of medium-range forecast process.