Coupling groundwater and land surface models: Tradeoffs between physics and costs

In recent years, the importance of including lateral subsurface flow in the subsurface-land-surface-atmosphere continuum has been gradually recognized. There have been pioneer modeling efforts to couple groundwater models with land surface models, such as coupling Parflow and PFLOTRAN with the Common/Community Land Model (CLM), in which the one-dimensional (1D) vertical water flow in the CLM is replaced by the three-dimensional (3D) variably saturated flow through solving the Richards' Equation. However, the Richards' Equation is highly non-linear, which requires finer spatio-temporal resolution when solving it numerically. Therefore, the associated computational costs in such integrated models have hampered their utilization at large scales, though using terrain following grid in parflow.CLM can alleviate to a certain degree. Considering that the unsaturated flow in vadose zone is primarily vertical when averaged at the regional scale, we coupled the Common Land Model with the U.S. Geological Survey Modular Ground-Water Flow Model (MODFLOW). Thus, in subsurface, the unsaturated flow is conceptualized as 1D vertical flow by solving the UZF package in the MODFLOW, while the saturated flow is fully 3D. The simplification of the unsaturated flow in UZF and the availability of terrain following grid in MODFLOW has greatly improved the computation speed. As validations, regional-scale modeling based on both Parflow.CLM and the newly-developed Modflow.CLM were conducted in the North China Plain (NCP), simulated results of which (i.e., water and energy components) were compared to evaluate the applicability of using UZF in large-scale subsurface-land-surface integrated modeling. The computation efficiency of both integrated models was also analyzed. This study is expected to provide guidance to balance the physical reality and computational cost at the regional or even global scale.