Modelling extreme discharge response to several geostatistically interpolated rainfall using very sparse raingage data
Sarann Ly1, Catherine Sohier2, Catherine Charles3, and Aurore Degré21Institute of Technology of Cambodia, Department of Rural Engineering, Russian Federation Boulevards, PO Box 86, Phnom Penh, Cambodia 2Univ. Liège, Gembloux Agro-Bio Tech, Soil Water Plant Exchange, Passages des Déportés, 2, 5030 Gembloux, Belgium 3Univ. Liège, Gembloux Agro-Bio Tech, Ecosystems-Atmosphere Exchanges, avenue de la Faculté d'Agronomie 8, 5030 Gembloux, Belgium
Received: 11 Jan 2016 – Accepted for review: 10 Feb 2016 – Discussion started: 16 Feb 2016
Abstract. This study presents modelling work of extreme discharge response to rainfall inputs interpolated by geostatistical approaches. Multivariate geostatistics are used by incorporating elevation as external data to improve the rainfall prediction. Thirty year daily rainfall in the Ourthe and Ambleve nested catchments, located in the Ardennes hilly landscape in the Walloon region, Belgium are interpolated and then used as inputs for a distributed physically-based hydrological model (EPIC-GRID). The effect of different raingage densities and particularly the effect of the raingage positions for very sparse raingage data used for rainfall interpolation, on extreme flow is analysed. We propose an index that can illustrate the quality of the raingage distribution with respect to the calculation of extreme discharge. In high elevation sub-catchment, we found that the multivariate geostatistics can significantly improve the rainfall prediction to produce very good simulated peak discharge. In the low elevation sub-catchment and the low raingage density, our results indicated that the Universal Kriging (UNK) is not appropriate. The IDW, Ordinary Kriging (ORK) and Ordinary Cokriging (OCK) methods provide generally good performance. The Thiessen polygon (THI) and Kriging with External Drift (KED) provide good performance for the whole catchment but less good for sub-catchments. The position of the raingages is the key factor for rainfall interpolation, particularly in the data-scarce region. UNK and KED methods are the most sensitive.
Ly, S., Sohier, C., Charles, C., and Degré, A.: Modelling extreme discharge response to several geostatistically interpolated rainfall using very sparse raingage data, Nat. Hazards Earth Syst. Sci. Discuss., doi:10.5194/nhess-2016-16, 2016.