An Improved Distributed Hydrologic Simulation Framework—Integrating Energy/Water Balances with Snow Processes

This paper describes efforts to extend land surface modeling capabilities, typically available at macro to moderate scales [Gent et al., 2011; Liang et al., 1994; Ludwig and Mauser, 2000], to high-resolution distributed hydrologic processes suitable for fine-scale flow simulations. The work extends the Coupled Routing and Excess STorage model (CREST) [Shen et al., 2015; Wang et al., 2011], which couples runoff-generation with a fully distributed routing module, by implementing new computation modules on vegetation, energy/water balance and snow processes. Similarly to [Anderson, 1976; Andreadis et al., 2009], the new runoff-generation module integrates snow process, canopy interception and evapotranspiration through coupling energy balances. Specifically, the soil-canopy-atmosphere structure is divided into several mediums (atmosphere, canopy surrounding air, canopy, snow pack, top soil and deep soil layers) to account for energy and water balances. Surface and deep soil layer thermal balances are solved to determine their temperatures, while thermal balances of the canopy and the surrounding air of canopy layers are solved only when snow is intercepted by the canopy. Energy balance in the surface layer of the snow pack is also solved if a snow pack is present on the ground. The energy balances are coupled with evaporation, transpiration, sublimation, refreezing and melting processes, which impacts the canopy through fall, outflow from snowpack and extractions of water from canopy, surface and sub-surface. These water flows are computed to finally account for the runoff, evapotranspiration and water storages. These energy balances are computed concurrently rather than sequentially. This attribute is computational more efficient and enables the model to comply with a uniform standard [Clark et al., 2015] that isolates the energy balance equations from a certain numerical solver [Liang et al., 1994]. This efficiency improvement finally helps keep CREST compatible with fine resolutions. In addition, a multi-objective optimization approach is utilized for calibration to account for multiple observations. In this study we used ET fluxes and USGS river discharge measurements from multiple hydrologic sites in Connecticut. Specifically, the newly developed model is tested based on 16 basins within Connecticut where annual snowfall and flood events during spring and early summer are considerable. The NASA Land Data Assimilation Systems (NLDAS) and the National Weather Service Stage IV multi-sensor precipitation analyses product are used for atmospheric forcing. High-resolution vegetation and land cover parameters are retrieved from Connecticut's Changing Landscape (CCL), while outside CT these datasets were supplemented with the National Land Cover Database (NLCD). Compared with the previous versions of CREST that uses only rainfall and Potential Evapotranspiration (PET) product, and involves neither energy balances nor snow processes, the improved framework significantly improves the overall performance, especially during the winter to spring periods. Compared to other models [Liang et al., 1994] that run at a coarser resolution, the proposed model gives improved simulations of discharge. Differences and improvements will be discussed based on 12 years (2002-2013) of observational data.