Developing a Deep Learning Framework to Build Temporally Continuous Daily Nitrate Time-series Data for Agricultural Watersheds

Nitrate contamination to streams, lakes, and estuaries is a critical problem in many agricultural watersheds of the United States. The US Environmental Protection Agency (USEPA), US Geological Survey (USGS), and several state agencies monitor water quality data. The majority of nitrate data collection efforts are done at a coarser temporal scale of biweekly or monthly scale. USGS continuous nitrate monitoring data for selected stations were available from 2014. Conversion of coarser temporal scale data to continuous data and aggregating to annual scale is challenging. The goal of this research is to utilize deep learning techniques to generate temporally continuous nitrate concentration data by utilizing existing nitrate data, streamflow, weather, and biophysical characteristics of the watershed. The study objectives are to i) build a database of hydrology, water quality, and watershed's biophysical data required for developing the deep learning model framework; ii) explore deep learning methods to identify best suitable model for representing nitrate data; and ii) develop temporally continuous nitrate datasets for several agricultural watersheds of Iowa State. Studies showed the extensive use of the recurrent neural network (RNN) structure to analyze time-series data. The well-established RNN structure long short-term memory (LSTM) was used in this study. We utilized ArcGIS to process input datasets, including land use, topography, and soil characteristics at the catchment scale for each nitrate station in the watershed. Statistical approaches cluster analysis and feature selection were used to identify correlations among input data and to make the data consistent. The preliminary results of the modeling efforts are highly promising. The temporally continuous daily nitrate time-series from the study could be used for high-resolution physically-based modeling research, making management decisions, and developing nitrate related policies.