Dataset for Machine Learning and AHP-Based Land Suitability Evaluation for Tea Cultivation in Ganyange Ward, Tarime District, Tanzania

Title
Dataset for Machine Learning and AHP-Based Land Suitability Evaluation for Tea Cultivation in Ganyange Ward, Tarime District, Tanzania

Description
This dataset supports the study titled “Land Suitability Evaluation for Tea Cultivation: A Machine Learning and AHP Integrated Approach” . It contains soil, environmental, climatic, land cover, and accessibility variables used to evaluate land suitability for tea cultivation in Ganyange Ward, Tarime District, Tanzania.

The dataset integrates field-based soil observations, laboratory analyses, remote sensing data, and derived geospatial covariates within a multi-criteria decision analysis (MCDA) framework combining Random Forest (RF) modelling and the Analytical Hierarchy Process (AHP).

A total of 64 topsoil samples (0–30 cm depth), including 16 soil profile-derived surface samples and 48 auger samples, were collected using a conditioned Latin Hypercube Sampling (cLHS) approach to capture spatial variability. Laboratory analyses include soil pH, electrical conductivity (EC), soil organic carbon (SOC), total nitrogen (N), available phosphorus (P), exchangeable base cations (K, Ca, Mg, Na), cation exchange capacity (CEC), and soil texture fractions (sand, silt, clay).

Environmental covariates used for Random Forest prediction include Sentinel-2A spectral bands and indices (NDVI, SAVI, OSAVI, BSI, CMR, CI, FII, GSI, SER), and terrain attributes derived from Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) such as elevation, slope, aspect, and topographic wetness index (TWI). Climatic variables (rainfall and temperature) were derived from CHIRPS datasets, while accessibility variables (distance to roads and market centres) were obtained from OpenStreetMap.

Random Forest models were applied to generate spatially continuous soil property layers, which were subsequently integrated with other criteria using AHP-derived weights in a GIS-based weighted overlay framework. The final output includes land suitability classes for tea cultivation categorized into highly suitable (S1), moderately suitable (S2), marginally suitable (S3), and not suitable (N).

This dataset is intended to support reproducibility of GIS-based land suitability modelling, machine learning applications in soil science, and decision-support tools for sustainable tea cultivation in tropical agroecosystems.

Geographical coverage
Ganyange Ward, Tarime District, Mara Region, Tanzania

Coordinate system
WGS84 / UTM Zone 36S (EPSG:32736)

Keywords
Land Suitability Evaluation; Tea Cultivation; Random Forest; Analytical Hierarchy Process; GIS; Digital Soil Mapping; Environmental Covariates; Tanzania