In the analytic hierarchy process (AHP) based flood risk estimation models, it is widely acknowledged that different weighting criteria can lead to different results. In this study, we evaluated and discussed the sensitivity of flood risk estimation brought by judgment matrix definition by investigating the performance of pixel-based and sub-watershed-based AHP models. Taking a flood event that occurred in July 2020 in Chaohu basin, Anhui province, China, as a study case, we used the flood areas extracted from remote sensing images to construct ground truth for validation purposes. The results suggest that the performance of the pixel-based AHP model fluctuates intensively given different definitions of judgment matrixes, while the performance of sub-watershed-based AHP models fluctuates considerably less than that of the pixel-based AHP model. Specifically, sub-watershed delimitated via multiple flow direction (MFD) always achieves increases in the correct ratio and the fit ratio by >35% and >5% with the pixel-based AHP model, respectively.

翻译：在基于分析层次的洪水风险估计模型(AHP)中,人们普遍承认,不同的加权标准可能导致不同的结果。在这项研究中,我们通过调查以像素为基础的和以分流域为基础的AHP模型的性能,评估和讨论了根据判断矩阵定义产生的洪水风险估计的敏感性。以2020年7月在中国安徽省Chaohu盆地发生的洪水事件为研究案例,我们利用从遥感图像中提取的洪泛区为验证目的构建地面真相。结果显示,基于像素的AHP模型的性能会因判断矩阵的不同定义而剧烈波动,而以次流域为基础的AHP模型的性能比以像素为基础的AHP模型的性能波动要小得多。具体地说,通过多流方向划定的亚水系,其比例和适合率分别高于35%和大于5%,与以像素为基础的AHP模型相匹配。