Summary
This study evaluated long short-term memory (LSTM) neural networks enhanced with lagged observational data to improve snow water equivalent predictions across diverse snow regimes in the western United States. Integrating 30-day-lagged SWE observations substantially improved prediction accuracy (NSE from 0.92 to 0.97) and reduced peak SWE error by 57%, with effectiveness varying by site characteristics and snow persistence. The analysis reveals how data integration can diagnose and mitigate accumulated hydrological model errors over different temporal scales, offering benchmarks for future hydrological forecasting.
Regional applicability
This study concerns the western United States and is not directly applicable to United Kingdom hydrology, which has markedly different snow regimes and precipitation patterns. However, the methodology of using LSTM networks with data integration for hydrological variable prediction could transfer to UK water resource management applications, particularly for streamflow or rainfall forecasting where lagged observations are available.
Key measures
Nash–Sutcliffe model efficiency coefficient (NSE), root-mean-square error (RMSE), difference between estimated and observed peak SWE values (d_max), temporal testing validation
Outcomes reported
The study tested LSTM neural networks with data integration to improve snow water equivalent (SWE) predictions across the western United States. It measured prediction accuracy improvements using Nash–Sutcliffe efficiency, RMSE, and peak SWE estimation error when integrating lagged observations of SWE or satellite-derived snow cover fraction.
Topic tags
Dig deeper with Pulse AI.
Pulse AI has read the whole catalogue. Ask about this record, its theme, or how the findings apply to UK farming and policy — every answer cites the underlying studies.