Premium
Predictions of Diffuse Pollution by the HSPF Model and the Back‐Propagation Neural Network Model
Author(s) -
Chang ChiaLing,
Li MengYuan
Publication year - 2017
Publication title -
water environment research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.356
H-Index - 73
eISSN - 1554-7531
pISSN - 1061-4303
DOI - 10.2175/106143017x14902968254665
Subject(s) - probabilistic logic , surface runoff , watershed , artificial neural network , environmental science , pollution , computer science , water quality , simulation modeling , flexibility (engineering) , backpropagation , statistical model , data mining , hydrology (agriculture) , machine learning , statistics , artificial intelligence , engineering , mathematics , ecology , geotechnical engineering , mathematical economics , biology
Watershed models are important tools for predicting the possible change of watershed responses. Environmental models comprise the deterministic model and the probabilistic model. This study discusses the Hydrological Simulation Program Fortran (HSPF) and the Back‐Propagation Neural Network (BPNN); these two models represent the deterministic model and the probabilistic model, respectively. As the properties of the two models are distinct, they have differing abilities to predict surface‐runoff pollution. For the two models, the runoff simulation results are satisfactory. However, due to the limitation of the water quality monitoring records, pollution simulation is more difficult than runoff simulation. The results indicate that the prediction accuracy in the pollution simulation can be improved by adjusting the BPNN neurons. On the contrary, improving the prediction accuracy is limited by HSPF. Although the flexibility of BPNN is higher than HSPF, sufficient historical monitoring records are important for both of these models.