Hydrologically informed machine learning for rainfall–runoff modelling: towards distributed modelling

Abstract. Despite showing great success of applications in many commercial fields, machine learning and data science models generally show limited scientific including hydrology (Karpatne et al., 2017). The approach is often criticized for its lack interpretability physical consistency. This has led to the emergence new modelling paradigms, such as theory-guided (TGDS) physics-informed learning. motivation behind approaches improve meaningfulness by blending existing knowledge with algorithms. Following same principles our prior work (Chadalawada 2020), a model induction framework was founded on genetic programming (GP), namely Machine Learning Rainfall–Runoff Model Induction (ML-RR-MI) toolkit. ML-RR-MI capable developing fully fledged lumped conceptual rainfall–runoff watershed interest using building blocks two flexible frameworks. In this study, we extend towards inducing semi-distributed models. reliability hydrological inferences gained from may tend deteriorate within large catchments where spatial heterogeneity forcing variables properties significant. distributed titled Knowledge Augmented – System Hydrologique Asiatique (MIKA-SHA). MIKA-SHA captures variabilities automatically induces catchment without any explicit user selections. Currently, learns utilizing components However, proposed can be coupled internally coherent collection blocks. MIKA-SHA's capabilities have been tested Rappahannock River basin near Fredericksburg, Virginia, USA. builds tests configurations frameworks quantitatively identifies optimal concern. utilized identify (one each framework) capture runoff dynamics basin. Both achieve high-efficiency values hydrograph predictions (both at subcatchment outlets) good visual matches observed response catchment. Furthermore, resulting architectures are compatible previously reported research findings fieldwork insights readily interpretable hydrologists. MIKA-SHA-induced performances were compared against outperform used study terms efficiency while benefitting hydrologists more meaningful about