Spatial prediction of dog population distribution in Kenya.

Abstract

Free-roaming dogs pose a significant public health concern due to their role in disease transmission. Rabies has been endemic in Kenya for over a century, yet a sustainable and standardized method for estimating dog populations remains unestablished. To address this gap, we applied kriging and co-kriging spatial interpolation techniques to predict the distribution of free-roaming domestic dogs across different Kenyan counties. To improve the model's accuracy, we incorporated environmental and demographic predictors such as daily temperature, the Normalized Difference Vegetation Index (NDVI), and human density. Dog population data at the village level were collected from 34 counties through an online survey of veterinary professionals in both the public and private sectors. A spherical model was used to construct the semivariogram, integrating temperature, NDVI, and human density to refine spatial predictions. Kenya's total free-roaming dog population was estimated to be 7.46 million. The density of dogs per square kilometer varied across counties, corresponding to a median national density of 12.13 dogs per square kilometer. Compared to models incorporating no or only single or multiple covariates, the co-kriging model incorporating human density provided the best fit, with the minimum estimated difference between observed and predicted values. The spatial distribution map highlights arid and sparsely populated pastoral counties having lower dog densities, whereas peri-urban, densely populated, and agricultural counties exhibit higher dog numbers. This study provides a spatial framework for estimating free-roaming dog populations, which can inform the design and implementation of rabies control programs and public health interventions in Kenya and other infected countries.