Coupled hydrologic modeling for soil moisture initialization of high-resolution atmospheric simulations over complex terrain

Atmospheric flow over steep, mountainous terrain is particularly affected by variations in soil moisture due to thermal heating effects. Previous work using the Advanced Regional Prediction System (ARPS) for simulations of flow over Owens Valley in California have demonstrated the model's sensitivity to soil moisture and even snow cover. Standard initialization procedures for mesoscale models often interpolate surface conditions from coarse grids. This can lead to misrepresentation of surface variables due to the inability of the coarse grid to resolve finer scale topographic features, an issue which is especially important in complex terrain. For example, 32 km North American Regional Reanalysis (NARR) does not resolve the terrain of Owens Valley (width ~12 km), so that soil moisture and snow cover from NARR do not reflect the potentially drastic differences in surface conditions on the mountain tops versus the valley floor. In addition to resolution issues, most land-surface models used in mesoscale atmospheric modeling do not allow for lateral flow of water from grid cell to grid cell. This isolation of cells from their neighbors prohibits such models from including the effects of topography on the distribution of soil moisture. Groundwater recharge is also often neglected but can be an important factor nonetheless in determining the soil moisture distribution over complex terrain. We have run a 3D coupled hydrologic model in an offline multi-year spin-up procedure to obtain a more accurate initial soil moisture distribution for ARPS. The coupled hydrologic model, PF.CLM, consists of ParFlow, a variably saturated groundwater model with integrated overland flow, coupled to the Common Land Model (CLM). Our case study site is a 2,500 km 2 segment of the Owens Valley, a region of complex terrain east of the Sierras in California and the site of the Terrain-Induced Rotor Experiment (T-REX) in March and April, 2006. In previous work, PF.CLM was run with very simple subsurface geology for the valley region. The current study employs more accurate subsurface geology, complex terrain including steep mountain slopes and alluvial valleys, and fine spatial resolution that includes surface soil and vegetation types derived from ARPS. Results from the coupled model are compared to observations of soil moisture measured during the T-REX field campaign as well as to standard soil moisture initialization data from NARR. Due to the horizontal extent of the simulation domain (50 km by 50 km at 350 m resolution), as well as the steepness and complexity of the terrain (~4 km vertical extent and ~1 m resolution), the PF.CLM computational domain is very large, and requires high-performance parallel computing. Additionally, the ARPS simulations are run in a horizontal nested grid configuration, making this study a multi-step project that includes not only the challenge of accurately representing diverse physical processes in complex terrain, but also high-performance computing challenges.