Soil Geochemistry Predictive Maps for Western Kenya

British Geological Survey produced predictive raster maps for 56 chemical elements, pH, and organic matter content across western Kenya. The maps are based on 452 soil samples collected between 2015 and 2020 and were generated using random forest machine learning algorithms at a 500m spatial resolution.

Use Cases

• Train models to predict concentrations of 56 chemical elements from environmental covariates using the random forest methodology described.
• Analyze spatial correlations between pH or organic matter content and specific elemental distributions for agricultural planning.
• Use the 500m resolution raster data to investigate potential links between soil geochemistry and public health concerns like iodine or zinc deficiency.
• Create baseline maps for specific elements to inform soil management and micronutrient sourcing strategies.

Strengths

• Predictive maps cover 56 chemical elements plus pH and organic matter, providing a wide geochemical profile.
• Maps derived from 452 physically collected and laboratory-analyzed soil samples.
• Random forest algorithms applied to create high-resolution (500m) predictive raster outputs.
• Data collection spanned a 5-year period (2015-2020) for temporal consistency.

Limitations

• Data consists of predictive model outputs (raster maps), not the raw 452 sample measurements.
• Geographic scope is limited to western Kenya, not nationally representative.
• Model accuracy and uncertainty metrics for the predictions are not described in the input.

Provenance

Source

British Geological Survey (BGS)

Collection Method

Machine learning analysis (random forest) applied to 452 soil samples collected during field surveys; samples analyzed by BGS inorganic chemistry laboratories.

Time Range

2015-2020

Freshness

null

Geography

Western Kenya