Get access

Improved SWAT Model Performance With Time-Dynamic Voronoi Tessellation of Climatic Input Data in Southern Africa

Authors

  • Jafet C.M. Andersson,

    1. Respectively, Doctoral Student, Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland, and Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, Zurich, Switzerland
    Search for more papers by this author
  • Alexander J.B. Zehnder,

    1. Scientific Director, Alberta Water Research Institute (AWRI), Edmonton, Alberta, Canada, and Nanyang University of Technology, Singapore
    Search for more papers by this author
  • Bernhard Wehrli,

    1. Professor, Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, Zurich, Switzerland
    Search for more papers by this author
  • Hong Yang

    1. Senior Scientist, Eawag, Swiss Federal Institute of Aquatic Science and Technology, Dübendorf, Switzerland
    Search for more papers by this author

  • Paper No. JAWRA-10-0200-P of the Journal of the American Water Resources Association (JAWRA). Discussions are open until six months from print publication.

(E-Mail/Andersson: aqua@jafet.org).

Abstract

Andersson, Jafet C.M., Alexander J.B. Zehnder, Bernhard Wehrli, and Hong Yang, 2012. Improved SWAT Model Performance with Time-Dynamic Voronoi Tessellation of Climatic Input Data in Southern Africa. Journal of the American Water Resources Association (JAWRA) 48(3): 480-493. DOI: 10.1111/j.1752-1688.2011.00627.x

Abstract:  In this study, we compared two approaches to obtain climatic time series for the Soil and Water Assessment Tool (SWAT), namely the conventional centroid method and time-dynamic Voronoi tessellation, and assessed the performance of SWAT in simulating discharge and smallholder maize yields in Southern Africa. Climatic time series were estimated with each method. The Voronoi method utilized all available precipitation and temperature data, but the centroid method used only 14.5 and 82.5%, respectively. After centroid processing, sub-basin time series were on average 42 and 63% incomplete, respectively. After Voronoi processing, all time series were complete. SWAT was fed with each climate dataset. Each model setup was independently calibrated and validated against discharge and maize yield. Similar model performance was obtained with both methods for yield. The root mean squared error during calibration was 0.26 and 0.27 t ha−1 for the centroid and Voronoi methods, respectively (p-value: 0.80). However, daily discharge simulations improved significantly with the Voronoi method. The coefficient of determination increased from 0.24 to 0.39 in the calibration period (p-value: 9.6 × 10−13) and from 0.41 to 0.48 in the validation period (p-value: 3.1 × 10−3). The Voronoi method improved the simulation of the river flow regime. The largest improvements were obtained in data scarce situations, at high spatial and temporal resolution, and where the centroid method performed the worst.

Ancillary