Contrasting Local Ecosystem Water and Energy Balances with a Regional Land Surface Model in the Southeastern United States

Abstract Land use and land cover changes, especially urbanization, can strongly alter coupled water and energy cycles, intensifying urban heat island effects, stormwater runoff, and flood risk. Process-based land surface models provide an important tool for quantifying these interactions in heterogeneous landscapes with mixed urban, grassland, and forest cover where direct observations are limited. However, few studies have evaluated how models designed for regional-scale applications perform across scales from small watersheds to a broader urban–rural interface. Here, we diagnostically assessed the upgraded Common Land Model -Urban (CoLM-U) parameterization in the Nashville metropolitan area in the State of Tennessee, southern United States. We integrated local precipitation observations, streamflow records from five small watersheds, and remote sensing-derived evapotranspiration (ET) products from 2000 to 2020. CoLM-U reproduced the broad magnitude of annual ET and runoff but showed substantial watershed-to-watershed and seasonal discrepancies. At the monthly scale, runoff root mean square error ranged from 14.4 to 27.4 mm month −1 and Nash–Sutcliffe Efficiency ranged from 0.3 to 0.6. The model generally underestimated winter runoff and overestimated summer runoff, reflecting the combined effects of precipitation forcing uncertainty, growing-season ET underestimation, and limitations in representing soil water storage and runoff generation. Long-term ET estimated by CoLM-U was broadly consistent with watershed water balance estimates, while remote sensing ET was generally over estimated ET, especially over forested areas. CoLM-U modeling results indicated that forests exhibited the highest ET and lowest sensible heat fluxes during the growing season, whereas urban cores showed lower ET and higher sensible heat. This study demonstrates what a regional land surface model can and cannot capture of ecohydrological processes at local watershed-to-regional scales. Improved local model climate input data and refined soil parameterizations are essential for accurate application in estimating the environmental effects of land cover change.