Reproducibility package for Scanning For Soil Health: Assessing The Accuracy And Scalability Of Innovations For In-Situ Soil Measurement

Cost-effective and scalable soil testing tools can play an instrumental role in guiding agricultural decisions and policies, particularly in Sub-Saharan Africa, where declining soil fertility poses a serious threat to productivity. Laboratory tests of soil samples, which provide the most accurate measures of soil properties, remain costly and challenging to scale in the context of national surveys or programs. To test and validate alternative approaches to soil measurement, the World Bank launched the Uganda CLASS study. This paper evaluates the accuracy and scalability of three soil testing tools (the AgroCares Scanner, the Palintest kit, and the Sonkir pH meter) as well as two publicly available soil maps (iSDA and SoilGrids), against laboratory-based soil measurements conducted by CIFOR-ICRAF. While differences in soil parameter estimates may be expected to vary across testing methods, understanding the nature and magnitude of the divergence in estimates from those measured using established laboratory approaches is critical for practitioners and analysts seeking to leverage soil analyses to inform policy and program design. We find that the soil tools – except for the Sonkir pH meter, which performs poorly – provide only moderately adequate predictions of soil properties, and that both soil maps fail to capture sample variation. These findings highlight key trade-offs in accuracy, cost, and scalability, and offer practical guidance for scaling soil monitoring strategies in Uganda and similar contexts.