Application and assessment of entropy-based criterions for hydrometric network optimization

This study applied four classes of entropy-based optimization criterions for hydrometric networks, i.e., entropy-transinformation (H-T), entropy-total correlation (H-C), entropy-transinformation-total correlation (H-T1-C/H-T2-C), and transinformation index (TI), to a rainfall monitoring network in the southwest hilly area in the Taihu Lake Basin, and made comparison and assessment of the criterions. Using daily precipitation of 2007—2016 as samples, we calculated the stations' ranks with three data discretization methods respectively, then made correlation analysis of ranks, sensitivity analysis concerning the index weight of multi-objective criterions, and investigated the inter-annual variance of ranks. Results showed that ranks obtained with H-C was the most representative, the least sensitive to the index weight, and had insignificant inter-annual variance; H-T2-C was the most sensitive to the index weight, with significant inter-annual variance. H-C criterion can best reflect the basic entropy based optimization principles (increasing information content and reducing redundancy), while H-T2-C criterion can best reflect decision preferences.