Potential of deterministic and geostatistical rainfall interpolation under high rainfall variability and dry spells: case of Kenya’s Central Highlands

Abstract

Drier parts of Kenya’s Central Highlands endure persistent crop failure and declining agricultural productivity. These have, in part, attributed to high temperatures, prolonged dry spells and erratic rainfall. Understanding spatial-temporal variability of climatic indices such as rainfall at seasonal level is critical for optimal rain-fed agricultural productivity and natural resource management in the study area. However, the predominant setbacks in analysing hydro-meteorological events are occasioned by either lack, inadequate, or inconsistent meteorological data. Like in most other places, the sole sources of climatic data in the study region are scarce and only limited to single stations, yet with persistent missing/unrecorded data making their utilization a challenge. This study examined seasonal anomalies and variability in rainfall, drought occurrence and the efficacy of interpolation techniques in the drier regions of eastern Kenyan. Rainfall data from five stations (Machang’a, Kiritiri, Kiambere and Kindaruma and Embu) were sourced from both the Kenya Meteorology Department and on-site primary recording. Owing to some experimental work ongoing, automated recording for primary dailies in Machang’a have been ongoing since the year 2000 to date; thus, Machang’a was treated as reference (for period of record) station for selection of other stations in the region. The other stations had data sets of over 15 years with missing data of less than 10 % as required by the world meteorological organization whose quality check is subject to the Centre for Climate Systems Modeling (C2SM) through MeteoSwiss and EMPA bodies. The dailies were also subjected to homogeneity testing to evaluate whether they came from the same population. Rainfall anomaly index, coefficients of variance and probability were utilized in the analyses of rainfall variability. Spline, kriging and inverse distance weighting interpolation techniques were assessed using daily rainfall data and digital elevation model in ArcGIS environment. Validation of the selected interpolation methods were based on goodness of fit between gauged (observed) and generated rainfall derived from residual errors statistics, coefficient of determination (R ²), mean absolute errors (MAE) and root mean square error (RMSE) statistics. Analyses showed 90 % chance of below cropping-threshold rainfall (500 mm) exceeding 258.1 mm during short rains in Embu for 1 year return period. Rainfall variability was found to be high in seasonal amounts (e.g. coefficient of variation (CV) = 0.56, 0.47, 0.59) and in number of rainy days (e.g. CV = 0.88, 0.53) in Machang’a and Kiritiri, respectively. Monthly rainfall variability was found to be equally high during April and November (e.g. CV = 0.48, 0.49 and 0.76) with high probabilities (0.67) of droughts exceeding 15 days in Machang’a. Dry spell probabilities within growing months were high, e.g. 81 and 60 % in Machang’a and Embu, respectively. Kriging interpolation method emerged as the most appropriate geostatistical interpolation technique suitable for spatial rainfall maps generation for the study region.