ABSTRACT Access to water, sanitation, and hygiene (WASH) services in Zambia remains low and inadequately monitored, hindering effective evaluation and resource allocation. Traditional methods offer limited, generalised data, impeding the assessment of program impact. This study employs model-based geostatistical techniques to generate continuous, georeferenced predictions from the limited observed data. By using Pearson and linear regression analyses to select relevant covariates, a Bayesian model implemented via R-integrated nested Laplace approximation was developed, while utilising the Gaussian statistical family. Validation through the k-fold analysis and ground truthing demonstrated high accuracy, with mean absolute errors of 0.146 (water) and 0.119 (sanitation), and validation accuracies of 90% for water and 88% for sanitation. Results indicated that in Kanyama, sanitation access was classified as ‘basic’ at 82.61%, while water access was in the ‘limited’ category at 69%. The findings confirm that geostatistical methods can effectively improve data granularity and accuracy for WASH monitoring. The study recommends integrating geospatial approaches into existing systems to enhance resource allocation and project implementation in Zambia's WASH sector.
Malama et al. (Thu,) studied this question.