This repository contains R scripts and data used in the analysis of vegetation productivity patterns across different governance types in Kenya's rangelands, addressing the research question: "What insights can patterns of relative vegetation productivity as quantified by the RPI over space and time provide regarding the relative effectiveness of various governance strategies for rangeland management employed across different tenure and protection regimes?" This project is presented by Isaac Nduta Mureithi as part of a masters dissertation in Conservation and Biodiversity at the University of Exeter, as part of the Oppenheimer Programme in African Landscape Systems (OPALS) https://opals-exeter.org/.
The folder data contains study area shapefiles for Kenya's administrative boundaries, protected areas from WDPA, conservancy boundaries, and Relative Productivity Index (RPI) raster datasets. The Kenya's Rangeland Protected Areas.Rmd is the main R Markdown document containing the complete analysis workflow including data import, spatial processing, trend analysis using Theil-Sen slope estimation, and visualisation of RPI patterns across Amboseli National Park, Tsavo East and West National Parks, AET conservancies, and surrounding buffer areas. The analysis utilises libraries including sf, terra, tidyverse, tmap, and trend for spatial data processing, statistical analysis, and mapping. The document includes custom themes for publication-quality visualizations, temporal performance metrics (MAE, R², RMSE, Rppt), and comparative analysis of vegetation productivity across different governance types.
Spatial datasets for conservancies (AET_Conservancies), and community lands (AET_Other_Conserved_Areas) found in the Amboseli_Conservation_Areas folder were acquired from Sustain East Africa Ltd. with support from Dr. Peter Tyrrell. Spatial datasets for ranches (RanchesTTWCA), and settled areas (SettledAreas) within the Tsavo Conservation Area, found in the Tsavo_Conservation_Areas folder, were obtained from Amos Chege Muthiuru, a PhD student in the Geography Department at King’s College London. Border demarcations of National Park spatial maps are available from the World Database on Protected Areas (WDPA), and is organised in the WDPA_Amboseli_Tsavo folder. UNEP-WCMC and IUCN (2025), Protected Planet: The World Database on Protected Areas (WDPA) and World Database on Other Effective Area-based Conservation Measures (WD-OECM) [Online], April 2025, Cambridge, UK: UNEP-WCMC and IUCN. Available at: www.protectedplanet.net.
The reproducible code is openly available on GitHub under a GNU GPL V3 licence (https://github.com/TESS-Laboratory/Mureithi_RPI_Conservation_Governance), and the RPI datasets required to run the scripts are available at https://zenodo.org/records/14843888 , including:
- actual_gpp_ea_v2.tif - Integrated annual GPP estimates for 22 hydrological years (September-August) from September 2000 to August 2023, based on the PML_V2 ET and GPP dataset.
- potential_gpp_ea_v2.tif - The estimated potential GPP for each grid cell and hydrological year based on a quantile regression forest model with climatic, topographic and soil predictors (using the 90th percentile of predicted GPP as the potential).
- rpi_ea_v2.tif - The relative productivity index values, defined as the ratio between actual and potential GPP in a given grid cell and hydrological year.
- rpi_ea_temporal_performance.tif - A three-band raster reporting the pixel-level mean absolute error (mae), root mean squared error (rmse) and r-squared (rsq) for the quantile regression forest model. These layers can be used as quality control to identify pixels where the model is performing well or poorly at modelling inter-annual variability in GPP as a function of covariates.