Relative importance of uncertain model parameters driving water fluxes in a land surface model

We analyze the way temporal distributions of key components of the water cycle are influenced by typically uncertain parameters embedded in a Land Surface Model (LSM). The main objective of this study is to gain a clearer quantitative understanding of how uncertainty in model parameters affects land surface model outputs is critical for improving water balance assessment and supporting resource management. We explore the sensitivity of transpiration, evaporation, and groundwater recharge dynamics to uncertain parameters in the modular NIHM (Normally Integrated Hydrological Model) LSM. The latter is employed to simulate realistic field conditions (in terms of, e.g., climate, vegetation, and soil type) across a one-year period associated with two contrasting watersheds in the Vosges region (France), differing in vegetation and soil characteristics. A key novelty of our work lies in the simultaneous application of multiple global sensitivity analysis metrics (including moment-based and moment-independent indices) to enable a richer and multi-faceted evaluation of how input uncertainty propagates to various statistical aspects (mean, variance, or full distribution) of model outputs. This multi-metric approach reveals temporal dynamics of parameter sensitivity, also depending on the model output (statistical) moment considered. Our results suggest that evaporation is primarily controlled by energy transfer through the canopy and drainage properties of the top litter layer. Transpiration drivers differ across sites, vegetation traits, albedo, and canopy radiation attenuation playing central roles. Groundwater recharge appears to be sensitive to only a limited subset of parameters, such as root zone drainage and rainfall interception. These types of insights are valuable in the context of future model calibration phases, as they enable one to prioritize parameters requiring detailed field characterization and to support simplification of model structures without hampering accuracy.