���ճԹ�

This ���ճԹ� ArcGIS web application was created to develop a predictive soil geochemistry map of western Kenya. The interactive app provides to the agri-community using ���ճԹ�’s measured data combined with machine learning.

The original data, relating to land use, crops grown, drinking water source/usage and any local health problems, was generated from field collections between 2016 and 2019 during the ���ճԹ� ODA-I programme, as part of a geochemistry and health project to investigate the spatial incidences of diseases within the Rift Valley (e.g. oesophageal cancer; micronutrient deficiencies). You can read more about the project and our time in western Kenya in our previous blogs on and .

On completing our fieldwork trips, we brought the collected samples back to the ���ճԹ� headquarters at Keyworth, Nottingham, and analysed them at the Inorganic Geochemistry Laboratories. We created a dataset that compiled soil prediction maps for 56 chemical elements (mg/kg), pH and organic matter content (per cent) using machine learning (Random Forest) analysis. The predictive maps, displayed as raster files with a spatial resolution of 500 m, were based on the 452 soil samples collected from discrete sampling locations across western Kenya and relevant environmental covariate data, such as elevation and rainfall.

Once all of the predictive layers were created, they were entered into an , accessible ‘free’ online via a PC or mobile device. Stakeholders in Kenya from the academic and outreach sectors were consulted during development and tested the web tool, making useful suggestions for refinements to better communicate both the tool and the data itself directly to farmers. Future developments will enable us to continue to add more data and expand the area, subject to funding.

The ArcGIS map will have other research uses for ���ճԹ� grant-funded projects, in particular a Royal Society International Collaboration grant to study dynamics for land-to-lake transfers within the Winam Gulf catchment of Lake Victoria resulting from soil degradation. This tool can also provide valuable information to agricultural extension services regarding areas where fertility may be at greater risk due to soil degradation and contribute to source apportionment models for transfers into the lake basin.

The publication of our web tool will enable ���ճԹ� to seek wider stakeholder input to the tool to stimulate new developments, ideally with new funding. For example, combining this data with health incidence statistics could provide investigation assistance regarding the spatial influence of geochemistry on health conditions (e.g. oesophageal cancer and micronutrient deficiency) and exposure to geogenic potentially harmful metals with chronic health implications.

Acknowledgements

This research was supported by the ���ճԹ�-NERC grant  ‘Geoscience for sustainable futures’ and ���ճԹ�/ programmes for financial support. It was delivered via the ���ճԹ� Eastern Africa Official Development Assistance (ODA) Research Platform and the activities were coordinated by Michael Watts (���ճԹ�), Odipo Osano (University of Eldoret) and Diana Menya (Moi University).

We would like to thank the many people who assisted in the collection of samples, including the public health officers from each county administrative area and, in particular, the field and laboratory staff from the University of Eldoret (UoE), Moi University (Moi U) and ���ճԹ�.

University of Eldoret

David Samoie, Doreen Meso, Charles Owano, Melvine Anyango and Job Isaboke.

Moi University

Esilaba Anabwani and Amimo Anabwani.

Queen’s University Belfast

Daniel Middleton.

���ճԹ�

Andrew Marriott, Amanda Gardner, Elliott Hamilton, Nicholas Porter and Sophia Dowell.

About the author

Dr Olivier Humphrey

Environmental chemist

���ճԹ� Keyworth

Find out more