A process-based, distributed hydrologic model based on a large-scale method for surface–subsurface coupling

Chaopeng Shen, Mantha S. PhanikumarAdvances 2010 · DOI

TL;DR

This paper introduces PAWS, a novel process-based, distributed hydrologic model designed for large-scale watershed simulations. PAWS addresses the computational burden of traditional models by employing an efficient, non-iterative coupling method between surface and subsurface flow, enabling accurate predictions across vast areas.

What Problem Does This Paper Address?

Accurately modeling hydrological processes at watershed scales is crucial for understanding water resource availability, predicting the impacts of climate change, and assessing pollutant transport. However, existing process-based hydrologic models often struggle with computational demands, particularly when simulating the complex, three-dimensional subsurface flow governed by the Richards equation. This limitation has historically restricted their application to smaller catchments and shorter time frames. The authors aim to overcome these computational constraints by developing a model that can efficiently handle large-scale simulations while maintaining physical realism.

Methods

The core innovation of PAWS lies in its novel, non-iterative method for coupling surface and subsurface flow. Instead of solving the full 3D Richards equation for the entire domain, PAWS separates the problem into:

  • 1D Richards Equation for Unsaturated Zone: This equation governs water movement in the soil column.
  • 2D Groundwater Flow Equation for Saturated Zone: This equation describes flow in aquifers.

The key to efficiency is the coupling strategy between these two domains. PAWS conceptualizes the unsaturated zone as an array of 1D soil columns connected to the unconfined aquifer. It makes two main assumptions: water moves only vertically in the unsaturated zone, and lateral flow in the saturated zone is uniformly distributed. This allows for a quasi-steady state approximation of the groundwater head at the interface, avoiding computationally expensive iterative solutions between the Richards and groundwater flow equations.

Other key methodological aspects include:

  • Spatial Discretization: A structured grid is used for the watershed, with rivers modeled as separate network objects.
  • Surface Flow: Modeled using the 2D Diffusive Wave Equation (DWE), with a Runge-Kutta Finite Volume (RKFV) scheme for efficient routing.
  • Channel Flow: Modeled using 1D Diffusive Wave Equations.
  • Snowpack Model: The UEB (Utah Energy Balance) snow model is adapted.
  • Vegetation Module: A simple vegetation growth module is included.
  • Infiltration: A modified Generalized Green and Ampt (GGA) method is used for heavy rain events.

Datasets

The paper uses publicly available datasets for model input and evaluation:

  • National Elevation Dataset (NED): Used for land grid elevation (30 m resolution).
  • IFMAP Land Use Land Cover Data: Used for land use information (30 m resolution).
  • SSURGO Data: Used for initial soil properties.
  • National Hydrography Dataset (NHD): Used to define the modeled river network.
  • National Climatic Data Center (NCDC) and Michigan Automated Weather Network (MAWN): Provided daily or sub-daily weather data (precipitation, etc.).
  • Michigan Wellogic Database: Used for groundwater-related data (hydraulic conductivities, heads, thicknesses) for the Red Cedar River watershed.
  • USGS Streamflow Data: Used for model calibration and verification at two gages (04112500 at East Lansing and 04111379 at Williamston).

The paper also references and compares its results to established test cases and experimental data from other studies, such as the "V-catchment" problem and experimental data from Vauclin et al. [55].

Key Insights and Ideas

  • Efficient Surface-Subsurface Coupling: The non-iterative coupling method is the paper's most significant contribution, enabling large-scale simulations that were previously computationally prohibitive. This approach significantly reduces computational burden without sacrificing physical realism.
  • Hierarchical Modeling: By separating the problem into 1D unsaturated flow and 2D saturated flow, and then coupling them efficiently, PAWS effectively decomposes a complex 3D problem into more manageable, lower-dimensional subsystems.
  • Importance of Vadose Zone Dynamics: The paper emphasizes the critical role of the vadose zone in linking surface and subsurface hydrology and influencing ET, highlighting the need for accurate representation.
  • Model Versatility: PAWS is designed to be a flexible and modular model, capable of interacting with GIS datasets and allowing for future expansion with fate and transport modules.

Key Results

  • Accurate Streamflow Prediction: PAWS achieved high performance metrics (NASH, RMSE, R², ME) when applied to the Red Cedar River watershed in Michigan, demonstrating comparable or superior results to established models in various test cases.
  • V-Catchment Test: The RKFV scheme for overland and channel flow showed good agreement with analytical and finite element solutions, with slightly better performance on the declining limb of the hydrograph.
  • Infiltration Test: The model accurately predicted infiltration into dry soils, with results showing good agreement with analytical solutions, though vertical resolution remains a factor.
  • Coupled Saturated/Unsaturated Flow: The novel coupling method showed good agreement with experimental data and demonstrated stability, outperforming traditional Dirichlet boundary condition approaches.
  • Surface-Subsurface Coupling: Comparisons with full 3D models (CATHY, ParFlow) for various test cases showed that PAWS's simplified subsurface formulation reasonably approximates the complex 3D physics.
  • Computational Efficiency: The model demonstrated efficient performance, with a 10-year simulation of the Red Cedar River watershed taking approximately 0.85 hours on a standard PC.

Limitations and Future Work

  • Lateral Flow in Unsaturated Zone: The current model assumes no lateral flow in the unsaturated zone. This limitation might affect performance in catchments dominated by perched water table dynamics. Future work will include macropore flow and lateral flow in the unsaturated zone.
  • Vegetation Module: The current vegetation module is simple. A more comprehensive module will be developed.
  • Spatial Resolution: The effects of spatial resolution on model results will be examined in future work.
  • Transport Modules: Future papers will describe the development and application of fate and transport modules.
  • Model Application: The authors plan to apply PAWS to larger watersheds with different land use, geology, soil, and climate conditions.

Summary updated

BibTeX
@inproceedings{shen2010process,
  author = {Chaopeng Shen and Mantha S. Phanikumar},
  title = {{A process-based, distributed hydrologic model based on a large-scale method for surface–subsurface coupling}},
  booktitle = {Advances},
  year = {2010},
  doi = {10.1016/j.advwatres.2010.09.002},
}
Figure