TY - JOUR
T1 - Uncertainty in determining extreme precipitation thresholds
AU - Liu, Bingjun
AU - Chen, Junfan
AU - Chen, Xiaohong
AU - Lian, Yanqing
AU - Wu, Lili
N1 - Funding Information:
The research in this paper is fully supported by the National Natural Science Foundation of China (Grant Nos. 50909106 and 51210013 ), The Foundation for Young Teachers Training from the Ministry of Education of China (Grant No. 3161399 ), and the Science and Technology Planning Project of Guangdong Province, China (Grant No. 2011B030800008 ). We thank the National Climatic Centre (NCC) of the China Meteorological Administration (CMA) for providing the valuable meteorological data. The authors also want to thank Lisa Sheppard for paper editing.
PY - 2013/10/30
Y1 - 2013/10/30
N2 - Extreme precipitation events are rare and occur mostly on a relatively small and local scale, which makes it difficult to set the thresholds for extreme precipitations in a large basin. Based on the long term daily precipitation data from 62 observation stations in the Pearl River Basin, this study has assessed the applicability of the non-parametric, parametric, and the detrended fluctuation analysis (DFA) methods in determining extreme precipitation threshold (EPT) and the certainty to EPTs from each method. Analyses from this study show the non-parametric absolute critical value method is easy to use, but unable to reflect the difference of spatial rainfall distribution. The non-parametric percentile method can account for the spatial distribution feature of precipitation, but the problem with this method is that the threshold value is sensitive to the size of rainfall data series and is subjected to the selection of a percentile thus make it difficult to determine reasonable threshold values for a large basin. The parametric method can provide the most apt description of extreme precipitations by fitting extreme precipitation distributions with probability distribution functions; however, selections of probability distribution functions, the goodness-of-fit tests, and the size of the rainfall data series can greatly affect the fitting accuracy. In contrast to the non-parametric and the parametric methods which are unable to provide information for EPTs with certainty, the DFA method although involving complicated computational processes has proven to be the most appropriate method that is able to provide a unique set of EPTs for a large basin with uneven spatio-temporal precipitation distribution. The consistency between the spatial distribution of DFA-based thresholds with the annual average precipitation, the coefficient of variation (CV), and the coefficient of skewness (CS) for the daily precipitation further proves that EPTs determined by the DFA method are more reasonable and applicable for the Pearl River Basin.
AB - Extreme precipitation events are rare and occur mostly on a relatively small and local scale, which makes it difficult to set the thresholds for extreme precipitations in a large basin. Based on the long term daily precipitation data from 62 observation stations in the Pearl River Basin, this study has assessed the applicability of the non-parametric, parametric, and the detrended fluctuation analysis (DFA) methods in determining extreme precipitation threshold (EPT) and the certainty to EPTs from each method. Analyses from this study show the non-parametric absolute critical value method is easy to use, but unable to reflect the difference of spatial rainfall distribution. The non-parametric percentile method can account for the spatial distribution feature of precipitation, but the problem with this method is that the threshold value is sensitive to the size of rainfall data series and is subjected to the selection of a percentile thus make it difficult to determine reasonable threshold values for a large basin. The parametric method can provide the most apt description of extreme precipitations by fitting extreme precipitation distributions with probability distribution functions; however, selections of probability distribution functions, the goodness-of-fit tests, and the size of the rainfall data series can greatly affect the fitting accuracy. In contrast to the non-parametric and the parametric methods which are unable to provide information for EPTs with certainty, the DFA method although involving complicated computational processes has proven to be the most appropriate method that is able to provide a unique set of EPTs for a large basin with uneven spatio-temporal precipitation distribution. The consistency between the spatial distribution of DFA-based thresholds with the annual average precipitation, the coefficient of variation (CV), and the coefficient of skewness (CS) for the daily precipitation further proves that EPTs determined by the DFA method are more reasonable and applicable for the Pearl River Basin.
KW - Detrended fluctuation analysis method
KW - Extreme precipitation threshold
KW - Non-parametric method
KW - Parametric method
KW - Pearl River Basin
UR - http://www.scopus.com/inward/record.url?scp=84884926532&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84884926532&partnerID=8YFLogxK
U2 - 10.1016/j.jhydrol.2013.09.002
DO - 10.1016/j.jhydrol.2013.09.002
M3 - Article
AN - SCOPUS:84884926532
SN - 0022-1694
VL - 503
SP - 233
EP - 245
JO - Journal of Hydrology
JF - Journal of Hydrology
ER -