PIDuaL: Physics-Informed Dual-Output LSTM

PIDuaL is an extension of the Mass-Conserving Long Short-Term Memory (MC-LSTM) model for hydrological modeling. It enables the simultaneous prediction of streamflow and evapotranspiration through a modified water-balance structure. This work builds upon the MC-LSTM implementation, as published by Hoedt et al. (2021) and hosted on the Neural Hydrology library (https://github.com/neuralhydrology/neuralhydrology). Modification SummaryCompared to the original MC-LSTM, the following key modifications have been introduced in PIDuaL:- Extension of the MC-LSTM architecture to a dual-output framework- Simultaneous prediction of streamflow and evapotranspiration- Modified water balance structure to support multiple hydrological flux outputs- Additional PIDuaL variants by modifying loss function with water and energy balance penaltiesIf you find this code is useful for your research, please cite the below papers.Kratzert, F., Gauch, M., Nearing, G., Klotz, D., 2022. NeuralHydrology — A Python library for Deep Learning research in hydrology. J. Open Source Softw. 7, 4050. https://doi.org/10.21105/joss.04050Hoedt, P.-J., Kratzert, F., Klotz, D., Halmich, C., Holzleitner, M., Nearing, G.S., Hochreiter, S., Klambauer, G., 2021. MC-LSTM: Mass-Conserving LSTM, in: Proceedings of the 38th International Conference on Machine Learning. p. PMLR 139:4275-4286.