Extrapolability of hydrological models across diverse geographies and hydrological conditions

Determine the extrapolation ability of process-based hydrological models and data-driven hydrological models under different geographical locations and hydrological conditions, including drought, given the uneven global distribution of surface hydrological observation stations such as runoff gauge stations and eddy-covariance flux towers.

Background

The paper emphasizes that hydrological observation networks are unevenly distributed globally, with dense coverage in developed regions and sparse coverage in remote areas. This limits the ability to validate and quantify how well models generalize to regions and conditions outside their calibration/training domains.

Clarifying extrapolability is central for global applications in water resource management and climate change studies, yet the authors note that this capability is presently unclear, especially under extremes such as drought.

References

However, the extrapolation ability of models under different geographical and hydrological conditions (such as drought) remains unclear due to the uneven distribution of current surface hydrological observation stations (e.g., runoff observation stations, flux towers measuring ET) worldwide.

— Approaches for enhancing extrapolability in process-based and data-driven models in hydrology  (2408.07071 - Shi, 2024) in Section 1 Introduction

The same apparatus can test the questions these conditions raise: whether a frozen knowledge package and experiment definition hold at a new site, with a different observation combination or under a different task, and whether a full seeded search can be reproduced from the archive.

— Delegating Earth science model calibration to knowledge-guided LLM agents  (2610.18378 - Zhu et al., 13 Oct 2026) in Section 4, Discussion and conclusion