Technical note: Data assimilation and autoregression for using near-real-time streamflow observations in long short-term memory networks

Abstract. Ingesting near-real-time observation data is a critical component of many operational hydrological forecasting systems. In this paper, we compare two strategies for ingesting streamflow observations into long short-term memory (LSTM) rainfall–runoff models: autoregression (a forward method) and variational assimilation. Autoregression both more accurate computationally efficient than sensitive to missing data, however an appropriate (and simple) training strategy mitigates problem. We introduce assimilation procedure recurrent deep learning models that uses backpropagation make the state updates.