No evidence of increased forest loss from a mining rush in a biodiversity hotspot

This repository contains the data and code used in the study entitled 'No evidence of impacts from a mining rush in a protected forest'. 2024. Authors: Katie Devenish, Simon Willcock, Kathryn M. Goodenough, Rio Heriniaina, O. Sarobidy Rakotonarivo, Julia P.G. Jones.

The main focus of this study was an evaluation of the impacts of the 2016-17 sapphire rush at Bemainty in Eastern Madagascar on deforestation and forest degradation (temporary tree cover loss). We used the Tropical Moist Forests (TMF) data (Vancutsem et al., 2021) to measure forest changes and the synthetic control method to estimate counterfactual outcomes. Supplementary data from informal interviews and lemur surveys conducted in the field October - November 2019 by Rio Heriniaina were drawn upon to explore the wider impacts and trade-offs of the mining rush, and assess the health of lemur populations 2 years after the rush. 

This repository contains the raw interview and lemur survey data collected by Rio Heriniaina at Bemainty October - November 2019. It also contains the code (entitled Script_updated_for_publication), spatial data, and input data required to reproduce the results from the main forest loss analysis. 

Data on deforestation and degradation was obtained from the Tropical Moist Forests dataset (Vancutsem et al, 2021). Annual deforestation was measured using the Deforestation year data. Annual deforestation per drainage basin was calculated in ArcGIS and is already contained within the attributes of the basins layer (Mada_basins_Lev9_edited_final).

However, annual degradation data was obtained by adapting the  TMF rawAnnual Disruptions data in Google Earth Engine. The output of this process (Annual_Disruptions_YEAR_masked) was loaded into R and annual degradation calculated per drainage basin. Annual deforestation and degradation rates were calculated as a percentage of forest cover in each drainage basin at the start of each year. To obtain annual forest cover estimates for each sub-basin we reclassified the TMF Annual Change datasets and extracted forest area per sub-basin. Both of these processes were extremely computationally expensive. The code includes these steps, but also contains an option (Option 1) to skip these steps by uploading csv files of annual forest cover (For_Area_ext2_21_04) and annual degradation per basin (Deg_Area_ext2_21_04) included in this repository. This data can be joined back to the basins layer. This allows users to skip this computationally expensive stage and progress straight to the analysis. The For90 layer (Vieilledent et al., 2018) and the Annual Disruptions data (Annual_Disruptions_YEAR_masked) are only needed if following Option 2 in the code.