Source Code for: Diagnosing Cryospheric Runoff Dynamics: A Distributed Differentiable Hydrological Model with Global Transfer Learning

This repository contains the source code for the differentiable Coupled Routing and Excess STorage model (dCREST) presented in the WRR paper titled "**Diagnosing Cryospheric Runoff Dynamics: A Distributed Differentiable Hydrological Model with Global Transfer Learning**". dCREST is a distributed differentiable physics-informed hydrological model for alpine catchments. It combines a bottom-up hydrological process backbone with top-down global transfer learning, so that frozen-ground structural hypotheses, dynamic cryospheric parameterizations, and operational runoff forecasting skill can be evaluated within one trainable framework. The code includes: 1. a single-basin pretraining and fine-tuning workflow used to test local model structures;2. a multi-basin runoff simulation workflow used for global transfer learning without explicit river-network routing;3. a topology-aware runoff generation and routing workflow used for distributed operational evaluation. The code was built on PyTorch.