PNNL

Abstract

Large-scale, high-resolution hydrologic modeling is an important tool to address questions of water quantity, availability, and potential recharge. Continental-to-Global scale models, particularly those that include groundwater, are growing in number. However, many of these approaches simplify aspects of the system and the connections between surface water and groundwater. The ParFlow CONUS modeling platform is a large-scale, hyper-resolution, hydrologic model that relies on the integrated solution of 3D partial differential equations that describe groundwater, soil, and 2D surface water flow. The prior version, ParFlow CONUS 1.0, was the first large-scale hydrologic model that included an explicit treatment of lateral groundwater flow for the contiguous US (CONUS). Here, we present the ParFlow CONUS 2.0 integrated hydrologic model. This model extends to the coastlines and contributing basins for CONUS and is consistent with the NOAA National Water Model. Here we document the roughly five years of technical development of this platform, present steady-state simulation results, rigorously compare these results to the prior ParFlow CONUS 1.0 simulations and evaluate the model performance based on observations. Simulated water table depth and streamflow were evaluated using more than 635K observations from USGS monitoring wells, other compiled groundwater datasets, and NHD and USGS streamflow gauges. Our results demonstrate improvement in both groundwater and surface water simulations over the prior generation model for all USGS Hydrologic Unit Code (HUC) basins. These results suggest that this current generation hydrologic model has good to excellent streamflow performance over the entire CONUS, with almost half of the HUC subbasins exhibiting excellent performance based on normalized root-square error (RSR). These results suggest that the current generation model approaches good performance for water table depth over the CONUS, a metric not usually compared directly at all in large-scale studies, with good-to-excellent performance exhibited over some HUC regions. We also delineate two regions that influence model performance, one where microtopography around streams dominates (D2), and another where a mix of subsurface heterogeneity and topographic gradients dominate (D1). Improvements in topography from CONUS1 to CONUS2 generally result in better streamflow and water table depth performances. Advancements in the subsurface depth and structure also produce better water table estimates.