Entropy Based Multicriterion Evaluation for Rainfall Monitoring Networks under the Impact of Discretization

Abstract

Rainfall monitoring networks provide fundamental input for hydrological models. Entropy, as a measure of uncertainty or information, is widely used in network evaluation or optimization. Computing entropy requires data discretization with methods like floor function, whereas the parameter selected is crucial and influential. This paper proposed an entropy based multicriterion method for evaluation of rainfall monitoring networks. Two indexes, separately account for information content and redundancy, were integrated with ideal point method. Values of the objective function were then computed to rank the stations and identify the significant ones. To find out the effect of discretization, parameter of the floor function was altered to get different schemes. A rainfall monitoring network containing 95 stations in the western Taihu Lake basin of China was analyzed as case study. Results showed that stations in the northern hilly area are more prominent in the network. Impact of the parameter in floor function is non-negligible as it determines entropy values, including its ranging scale and distribution pattern. Location of the stations rank extremely high and low also varies. As discretization process has an impact on the evaluation, it should be carefully used and sensitivity analysis is needed to avoid subjectivity and arbitrariness.