当前位置: X-MOL 学术Water Resources Management › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Estimation of Regional Sub-Daily Rainfall Ratios Using SKATER Algorithm and Logistic Regression
Water Resources Management ( IF 4.3 ) Pub Date : 2021-01-07 , DOI: 10.1007/s11269-020-02730-1
Mohamed M. Fathi , Ayman G. Awadallah , Nabil A. Awadallah

Developing Intensity-Duration-Frequency (IDF) curves is a paramount input in stormwater systems design. To construct these IDF curves, rainfall records at sub-daily durations, provided by continuous rainfall recorders, are required; however, these recorders are seldom available in many locations of interest. To fill this gap, available meteorological and topographical information for a study area in Saudi Arabia are investigated to get an estimate of the ratios of sub-daily rainfall depths to the 24-h depths (sub-daily ratios or SDRs), via applying the following methodology. A spatially constrained regionalization approach is implemented, using the SKATER algorithm, based on 60 gauging stations, to form regions of contiguous stations, based on the similarities of their SDRs. Four different regions are formed, where each region shows consistent SDRs; yet distinctly different from other regions. Subsequently, a multinomial logistic regression model is built and trained, with commonly available meteorological and topographical information as explanatory variables, to determine to which region a specific location belongs. The model is validated based on a hold-out validation method and assessed through confusion matrix statistics to evaluate the model performance. The model shows high performance in predicting the correct regional SDR and it is extended to produce a gridded map covering ungauged areas. Based on this procedure, one can develop the IDF curve for any location within the study area, even if there is no rainfall recorder in that location.



中文翻译:

基于SKATER算法和Logistic回归的区域次每日降雨率估算。

在雨水系统设计中,开发强度-持续时间-频率(IDF)曲线是至关重要的输入。要构建这些IDF曲线,需要连续降雨记录仪提供的次日降雨记录。但是,这些记录器很少在许多感兴趣的位置可用。为了填补这一空白,通过应用以下方法,对沙特阿拉伯某研究区域的可用气象和地形信息进行了调查,以估算次日降雨深度与24小时深度之比(次日比例或SDR)。以下方法。使用基于60个测量站的SKATER算法,实施空间受限的区域化方法,以基于其SDR的相似性来形成连续站的区域。形成四个不同的区域,每个区域显示一致的SDR;但与其他地区却截然不同。随后,以常用的气象和地形信息作为解释变量,建立并训练了多项式逻辑回归模型,以确定特定位置属于哪个区域。该模型基于保持验证方法进行验证,并通过混淆矩阵统计信息进行评估,以评估模型的性能。该模型在预测正确的区域SDR方面显示出高性能,并且可以扩展为生成覆盖未覆盖区域的网格地图。基于此过程,即使研究区域内没有降雨记录仪,也可以为研究区域内的任何位置绘制IDF曲线。建立和训练多项式逻辑回归模型,并以常用的气象和地形信息作为解释变量,以确定特定位置属于哪个区域。该模型基于保持验证方法进行验证,并通过混淆矩阵统计信息进行评估,以评估模型的性能。该模型在预测正确的区域SDR方面显示出高性能,并且可以扩展为生成覆盖未覆盖区域的网格地图。基于此过程,即使研究区域内没有降雨记录仪,也可以针对研究区域内的任何位置绘制IDF曲线。建立和训练多项式逻辑回归模型,并以常用的气象和地形信息作为解释变量,以确定特定位置属于哪个区域。该模型基于保持验证方法进行验证,并通过混淆矩阵统计信息进行评估,以评估模型的性能。该模型在预测正确的区域SDR方面显示出高性能,并且可以扩展为生成覆盖未覆盖区域的网格地图。基于此过程,即使研究区域内没有降雨记录仪,也可以针对研究区域内的任何位置绘制IDF曲线。该模型基于保持验证方法进行验证,并通过混淆矩阵统计信息进行评估,以评估模型的性能。该模型在预测正确的区域SDR方面显示出高性能,并且可以扩展为生成覆盖未覆盖区域的网格地图。基于此过程,即使研究区域内没有降雨记录仪,也可以针对研究区域内的任何位置绘制IDF曲线。该模型基于保持验证方法进行验证,并通过混淆矩阵统计信息进行评估,以评估模型的性能。该模型在预测正确的区域SDR方面显示出高性能,并且可以扩展为生成覆盖未覆盖区域的网格地图。基于此过程,即使研究区域内没有降雨记录仪,也可以针对研究区域内的任何位置绘制IDF曲线。

更新日期:2021-01-07
down
wechat
bug