Interpreting the internals of Long Short-Term Memory networks for rainfall-runoff forecasting
NeuralHydrology -- Interpreting LSTMs in Hydrology
We look at the application of long short-term memory networks for rainfall-runoff forecasting, one of the central tasks in the field of hydrology, in which the river discharge has to be predicted from meteorological observations.
On basis of two different catchments, one with snowinfluence and one without, we demonstrate how the trained model can be analyzedand interpreted.
In the process, we show that the network internally learns to represent patterns that are consistent with our qualitative understanding of the hydrological system.
Frederik Kratzert, Mathew Herrnegger, Daniel Klotz, Sepp Hochreiter, Günter Klambauer