Deep learning reveals future streamflow characteristics change and climate sensitivity

Abstract

This study deploys the potential of Deep Learning (DL) technique for an improved future streamflow projection from General Circulation Model (GCM) simulations, by developing a reliable association between the observed streamflow and a set of primary meteorological variables at monthly scale over a historical period. Towards this, a DL-based Long Short-Term Memory (LSTM) framework is developed to capture the hidden complex dynamics between streamflow and its two primary hydrometeorological precursors – precipitation and temperature, identified through Kendall’s partial correlation analysis. After ensuring model stability through various combinations of hypothesized climatic forcings, the developed model is used for long-term projection of basin-scale streamflow characteristics, utilizing future-projected bias-corrected temperature and precipitation data from six state-of-the-art GCMs following two emission scenarios. In general, the proposed DL-based approach is found to outperform two benchmark machine learning models in identifying the basin-specific climate sensitivity controlling streamflow variation. The efficacy of the proposed model is demonstrated over four rain-fed tropical river basins in India, located in different climate zones, namely Bhadra, Netravati, Tenughat, and Upper Narmada river basins. All four diverse basins showcase more than 90% correlation, with Netravati basin achieving an impressive 0.98 correlation coefficient over the testing period. Likewise, the Nash–Sutcliffe model efficiency values, evaluated over the testing period, range from 0.83 to 0.95 across these basins affirming the model’s robust and efficient performance. This multi-model, multi-scenario analysis reveals an increased streamflow variability in all the basins under a wetter and hotter climate in future, which gets more pronounced with time and higher emission scenario. The flow is projected to increase in the monsoon months, along with a practically unchanged or marginally less flow over the dry months. Such redistribution of streamflow pattern in near future definitely requires suitable management strategies and their implementation in well advance.