Inter-Research > CR > v59 > n1 > p39-60  
Climate Research

via Mailchimp

CR 59:39-60 (2014)  -  DOI:

Spatial interpolation of daily rainfall stochastic generation parameters over East Africa

Pierre Camberlin1,*, Wilson Gitau2, Pascal Oettli3, Laban Ogallo4, Benjamin Bois1

1Centre de Recherches de Climatologie (CRC), Biogéosciences, UMR 6282 CNRS/Université de Bourgogne, 6 Bd Gabriel,
21000 Dijon, France
2Department of Meteorology, University of Nairobi, PO Box 30197, 00100 Nairobi, Kenya
3Department of Earth and Planetary Science, Graduate School of Science, The University of Tokyo, Science Building #1, Room 810, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan
4IGAD Climate Prediction and Applications Centre (ICPAC), PO Box 10304, 00100 Nairobi, Kenya
*Corresponding author:

ABSTRACT: Downscaling seasonal rainfall predictions to daily time-scale, for crop yield simulation for instance, can be performed using stochastic generators (SGs). The spatial interpolation of the SG parameters is required to generate rainfall time-series at ungauged places. A methodology is defined which makes use of topography to interpolate these parameters, in a region with a rugged terrain covering Kenya and north-eastern Tanzania. A first-order Markov chain was used to model rainfall occurrence, and a gamma distribution was used to model amounts. The 2 parameters of the Markov models, p01 and p11, and the 2 parameters of the gamma distribution are computed at 121 stations. The Kolmogorov-Smirnov test for goodness-of-fit shows that 88% (99%) of the stations and months have their dry (wet) spell frequencies successfully reproduced by first-order Markov chains, and two-third of the stations have their daily amounts satisfactorily fitted by the gamma distribution. Local regression, using elevation as the predictor and weighting stations according to distance from the target pixel and to environmental variables, is used to interpolate the 4 SG parameters. Cross-validation indicates that distance-weighted regression provides good estimates, but the inclusion of topographical variables (aspect in particular) improves the results further. The final maps show a strong orographic control of both the Markov and gamma parameters. However, while elevation has an effect on rainfall occurrence, rainfall intensity is more strongly related to local slope aspect, with eastward to southeastward oriented foothills and coastlines displaying the highest gamma scale values. These results suggest that a statistical disaggregation of daily rainfall is improved by taking explicitly into account topography through its effect on the spatial distribution of SG parameters.

KEY WORDS: Kenya · Tanzania · Disaggregation · Interpolation · Topography · Markov chains · Gamma distribution

Full text in pdf format 
Cite this article as: Camberlin P, Gitau W, Oettli P, Ogallo L, Bois B (2014) Spatial interpolation of daily rainfall stochastic generation parameters over East Africa. Clim Res 59:39-60.

Export citation
RSS - Facebook - - linkedIn