Title: Application of non-parametric regression in estimating missing daily rainfall data

Authors: Rachel Makungo; John Ogony Odiyo

Addresses: Department of Hydrology and Water Resources, School of Environmental Sciences, University of Venda, Private Bag X 5050, Thohoyandou 0950, South Africa ' Department of Hydrology and Water Resources, School of Environmental Sciences, University of Venda, Private Bag X 5050, Thohoyandou 0950, South Africa

Abstract: Most daily rainfall time series data are too short and/or possess missing records hindering them to perform reliable and meaningful analyses. Robust locally weighted scatter smoother non-parametric regression approach (NPR), with tricube weighting function was used to estimate missing daily rainfall data in the upper reaches of Nzhelele and Luvuvhu River Catchments in Limpopo Province of South Africa. The approach proposed in this study has not yet been widely applied for estimating missing rainfall data. Model performance ranged from acceptable to excellent. Graphical fits of observed and estimated rainfall data showed a general agreement. Scatter plots indicated that there was no definite pattern of underestimation and overestimation of peak rainfall events. Scatter points for low rainfall values were closer to the best fit line showing good agreement between observed and simulated rainfall values for most of the stations. The study showed that NPR effectively estimated missing rainfall data.

Keywords: catchment; missing rainfall data; model performance; non-parametric; regression; time series; tricube weighting function.

DOI: 10.1504/IJHST.2019.102321

International Journal of Hydrology Science and Technology, 2019 Vol.9 No.3, pp.236 - 250

Accepted: 13 Nov 2017
Published online: 13 Sep 2019 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article