​Community Land Model v5 runoff evaluation in small near-natural catchments in Fennoscandia

Validating model representations of land surface processes is crucial for reducing the uncertainty of future projections, especially at high latitudes where climate change is amplified. As part of a regional assessment of the latest version of the Community Land Model (CLM5) in cold environments, we compare simulated grid-scale runoff with discharge measurements in small near-natural catchments in Fennoscandia. CLM5 is the land component of the Norwegian Earth System Model. Evaluating land surface models involves a large set of state and flux variables, for many of which direct measurements are either not available or not representative at the typical modelling spatial scales (10⁰–10² km). In this context, discharge measurements provide valuable information that can be used to assess how well models are able to reproduce the downstream outcome of catchment hydrologic processes. We conduct two CLM5 simulations at 0.25° spatial resolution over Fennoscandia: one forced with the default 3-hourly 0.5° GSWP3v1 product (2000-2014) and another with the hourly 0.25° ERA5 near-surface atmospheric data (2000-2019). To characterise forcing uncertainty, precipitation and temperature forcing data are compared to the daily observational Nordic Gridded Climate Dataset (Norwegian Meteorological Institute), which covers Fennoscandia at 1 km resolution. Daily discharge and catchment information are obtained from the Norwegian Water Resources and Energy Directorate, the Swedish Meteorological and Hydrological Institute, and the Finnish Environment Institute. To avoid uncertainties due to human alterations and model representation of river routing, we select time series of unregulated catchments whose areas are smaller than 10³ km² and thus are compatible with single model grid-cells. Accordingly, we evaluate CLM5 daily total runoff, which is the sum of subsurface and surface runoff prior to channel routing, against observed discharge. We apply the following criteria: (1) bias, variance error and correlation, to assess the reproduction of the overall water balance and of the amplitude and shape of the hydrograph; (2) average seasonal cycles, to evaluate how runoff regimes are simulated; and (3) occurrence and persistence of low and high flow anomalies, to analyse the ability of the model to predict extremes. Further, we investigate whether spatio-temporal patterns of agreement/discrepancy between modelled and measured runoff correlate with atmospheric forcing uncertainties, land surface properties, or climatology. In particular, we try to detect model runoff errors prevailing in specific environmental conditions. This study aims to inform future regional CLM5 experiments that will test atmospheric forcing corrections and alternate parametrisations of hydrologic processes, in the framework of the Land-ATmosphere Interactions in Cold Environments (LATICE) research initiative.