FILLING IN MISSING PEAKFLOW DATA USING ARTIFICIAL NEURAL NETWORKS

作者：Steven K. Starrett, Shelli K. Starrett, Travis Heier, Yunsheng Su, Denny Tuan, Mark Bandurraga · 年份：2015 · 被引用次数：13 · 研究领域：Hermeneutics and Narrative Identity、Aging, Elder Care, and Social Issues、Health, Medicine and Society

The objectives of this study was to: i) use Artificial Neural Networks (ANNs) to fill-in missing data from the peak annual flow rate records for the Santa Clara river watershed, and ii) compare the ANN results with linear regression. Gauging station peaks were modeled with inputs consisting of: peak flows from nearby gauging stations, precipitation data, and temporal data. Model characteristics (number of nodes and layers, transfer functions, data pre-processing methods, etc.) were also studied to optimize the ability of the ANN to learn relationships between the inputs and the peak flows. In general, the models performed well with peak flows from one to four neighboring station, maximum annual 10-d precipitation total data, and the year (representing land use changes); and it was common for testing results to be within 20 % of the target. The ANN models had a sum squared error (SSE) value 2 to 400 times less than linear regression models.