Author(s): Chulsang Yoo; Byoungsoo Kim; Kyoungjun Kim; Jeongho Choi
Linked Author(s):
Keywords: Radar rainfall; Ground-truth; Design bias; Retrieval bias
Abstract: This study evaluated three possible ground-truths for the radar rainfall. Three ground-truth designs of comparing radar rainfall and rain gauge rainfall in this study are: (1) the design 1 is to use all the datasets available including zero radar rainfall and zero rain gauge rainfall, (2) the design 2 is to use the data sets of positive radar rainfall and zero or positive rain gauge rainfall, (3) and the design 3 is to use the data sets of positive radar rainfall and positive rain gauge rainfall. Theoretical review of these three designs show that the ground-truth designs 1 and 2 do not cause any design bias but the ground-truth design 3 can cause a negative design bias. This theoretical result was also assured by applying these three designs to the rainfall field generated by the WGR model, as well as to that of the Mt. Gwanak radar system. In conclusion, the ground-truth design 3, which considers only positive rainfalls for both radar and rain gauges, causes a serious design bias to be inappropriate as a ground-truth design.
Year: 2007