Enhanced Artificial Neural Networks for Salinity Estimation and Forecasting in the Sacramento-San Joaquin Delta of California

Abstract

Domain-specific architectures of artificial neural networks (ANNs) have been developed to estimate salinity levels for planning at key monitoring stations in the Sacramento-San Joaquin Delta, California. In this work, we propose three major enhancements to existing ANN architectures for purposes of training time reduction, estimation error reduction, and better feature extraction. Specifically, we design a novel multitask ANN architecture with shared hidden layers for joint salinity estimation at multiple stations, achieving a reduction of 90% training and inference time. As another major structural redesign, we replace predetermined preprocessing on input data by a trainable convolution layer. We further enhance the multitask ANN design and training for salinity forecasting. Test results indicate that these enhancements substantially improve the efficiency and expand the capacity of the current salinity modeling ANNs in the Delta. Our enhanced ANN design methodologies have the potential for incorporation into the current modeling practice and provide more robust and timely information to guide water resource planning and management in the Delta.