Keywords: calibration, validation, decomposition approach, coupled surface-subsurface model, watershed model, GMS, WASH123D.

Numerical modelling codes and computational resources have progressed to a point where accurate and useful physics-based models can now be developed to evaluate regional watershed issues in a cost-effective manner. Effective and efficient model calibration and validation for large-scale, physics-based watershed model is thus needed. It is often more efficient to sub-divide the large scale models into smaller sub-domains for calibration and validation. This paper presents a decomposition approach for large-scale watershed model calibration and validation, where the watershed system can be conceptualized as a combination of one-dimensional channel networks, two-dimensional overland regimes, and three-dimensional subsurface media. This decomposition approach subdivides the whole-domain model into several sub-domain models by using existing channels as the "cut" boundary to separate sub-domains. Four steps are included in the model calibration and validation process. In step one, the coupled two-three dimensional sub-domain models are constructed and calibrated, where the historical one-dimensional channel stages are used as boundary conditions on ground surface. In step two, the coupled one-two-three dimensional sub-domain models are constructed and calibrated, where the finalized overland and subsurface model parameters from step one were fixed such that only the channel model parameters are adjustable in this step. In step three, the sub-domain models are stitched together into the whole-domain model, and the model parameters calibrated from the previous steps are fixed for the coupled one-two-three dimensional whole-domain model calibration. In this step, only the channel model parameters associated with the "cut" boundary are adjustable. In step four, the calibrated coupled one-two-three dimensional whole-domain model from the former steps is validated against a set of field data other than that used for calibration. This decomposition approach allows the modeller to conduct more model runs at the sub-domain level, rather than at the whole-domain level, which helps generate a better calibrated and validated model within the given modelling time.

In this paper, a hypothetical example is employed as proof of concept to verify the proposed decomposition approach. This paper also describes the application of this modelling process to a regional-scale watershed system in South Florida that covers an area over 8,000 square miles as part of a research demonstration project in the Corps of Engineers Civil Work's System-Wide Water Resources Program at the US Army Engineer Research and Development Centre. (ERDC). Through the verification and the demonstration, the following points were drawn:

• The proposed decomposition approach is adequate for large-scale watershed model calibration and validation.
• The hydro-static assumption used to set up boundary conditions on the "cut" boundary for sub-domain models may introduce source error in the sub-domain model calibration.
• For a coupled surface-subsurface watershed model, daily canal stage data may be used to calibrate the subsurface flow model. However, sufficient temporal resolution of the canal structure flow rate data (e.g., hourly or every 15 minutes) is essential for calibrating canal flow models.

The US Department of Defence Groundwater Modelling System and the ERDC in-house WASH123D numerical model were used to construct and perform computer simulations, respectively. All the model runs were conducted on a PC cluster machine in ERDC's Major Shared Resource Centre for high performance computing.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £95 +P&P)