资源环境科技发展态势分析平台

Precipitation data of finer timescale and higher spatial density are crucial for continuous hydrological modelling and flood risk assessment. Disaggregation methods are often used to transform the coarser-timescale precipitation data into finer resolutions. The nonparametric approach based on method of fragments (MOF) has received broad attention in precipitation disaggregation literature, given its superiority in reproducing the at-site statistical attributes. However, a detailed literature review has shown that the MOF-based resampling approaches are mainly focused in the single-site precipitation disaggregation context, which may subject to limitations such as the unavailability of at-site precipitation records and the incapability to preserve the inter-site correlation structure. To address these issues, we propose three resampling approaches based on MOF. The first approach is a single-site interval-based resampling approach which only draws subdaily fragment vectors from at-site record. The second one extends the first one to a regionalized version where subdaily fragment vectors are drawn from both the at-site and neighbouring stations. The third one is a multi-site approach developed to preserve the observed inter-site correlation. The performances of the three methods are evaluated with applications to daily-to-hourly precipitation disaggregation at six rain gauges in Singapore and eight precipitation stations in Wangkuai reservoir catchment in northern China. An elaborate list of performance measures, including standard validation statistics, spatial correlation, inter-day connectivity, annual extreme analysis, and intra-day dry and wet spell characteristics are used to assess the performance. The proposed three methods are shown to be effective in reproducing the at-site attributes, and no significant deterioration of performance is observed when moving from the single-site method to the regionalized and multi-site versions. As expected, the multi-site approach is the only one method that is able to reconstruct the spatial correlation in the disaggregated precipitation field. The approaches can be applied for daily-to-subdaily precipitation disaggregation in different regions.