Improving model prediction reliability through enhanced representation of wetland soil processes and constrained model auto calibration – A paired watershed study Academic Article uri icon

abstract

  • © 2016 Elsevier B.V. Process based, distributed watershed models possess a large number of parameters that are not directly measured in field and need to be calibrated, in most cases through matching modeled in-stream fluxes with monitored data. Recently, concern has been raised regarding the reliability of this common calibration practice, because models that are deemed to be adequately calibrated based on commonly used metrics (e.g., Nash Sutcliffe efficiency) may not realistically represent intra-watershed responses or fluxes. Such shortcomings stem from the use of an evaluation criteria that only concerns the global in-stream responses of the model without investigating intra-watershed responses. In this study, we introduce a modification to the Soil and Water Assessment Tool (SWAT) model, and a new calibration technique that collectively reduce the chance of misrepresenting intra-watershed responses. The SWAT model was modified to better represent NO3 cycling in soils with various degrees of water holding capacity. The new calibration tool has the capacity to calibrate paired watersheds simultaneously within a single framework. It was found that when both proposed methodologies were applied jointly to two paired watersheds on the Delmarva Peninsula, the performance of the models as judged based on conventional metrics suffered, however, the intra-watershed responses (e.g., mass of NO3 lost to denitrification) in the two models automatically converged to realistic sums. This approach also demonstrates the capacity to spatially distinguish areas of high denitrification potential, an ability that has implications for improved management of prior converted wetlands under crop production and for identifying prominent areas for wetland restoration.

author list (cited authors)

  • Sharifi, A., Lang, M. W., McCarty, G. W., Sadeghi, A. M., Lee, S., Yen, H., ... Yeo, I.

citation count

  • 14

publication date

  • October 2016