Disrupted montane forest recovery hinders biodiversity conservation in the tropical Andes
Description
This dataset provides the tables, R scripts and Google Earth Engine scripts for the paper 'Disrupted montane forest recovery hinders biodiversity conservation in the tropical Andes'. The general workflow is: 1) Selection, clipping and filtering of Landsat data and GFCC data in Earth Engine 2) Export of data into tabular format (in latitudinal bands as very large files) 3) Calculation of z-scores using the latitudinal bands tables 4) Andes-wide Analysis of tree cover change, time series and recovery trajectories in separate scripts 5) Landscape-scale Analysis of tree cover change and time series in separate scripts. 6) Validation with Sentinel and Classification data
Files
Steps to reproduce
GEE analysis: ----------------- Definition of driest month per pixel: https://code.earthengine.google.com/a69a5f133b669507ba5de2d3458aaa1f Extraction of Landsat data from Google Earth Engine (this filters the Landsat data to the potential recovery areas (6 different types of potential recovery areas) and lastly exports a table with a row per pixel with data for various bands and products for the years 2000-2020). The script was manually run for individual latitudinal bands to produce the separate tables found in the attached folder Latitudes_Small which was then used for inputs in R https://code.earthengine.google.com/7579756964287caf65dbbad5468ebc09 R analysis -------------- The script Analysis_timeseries_Andes uses these latitudinal tables to generate NDWI time series (note - only the potential recovery class '5' = abandoned land is used for analysis). The R script Z-scores calculations uses the latitudinal tables to calculate the z-scores for the recovery trajectories and creates the table Andes_z_scores recalculated, which is used as input in the R scripts Analysis_recovery_classification and Analysis_treecover. (note - only the potential recovery class '5' = abandoned land is used for analysis). The R script Analysis_recovery_classification analyses recovery trajectories and displays figures across the Andes and by country and elevation. (note - only the potential recovery class '5' = abandoned land is used for analysis). The R script Analysis_treecover analyses tree cover change and displays figures across the Andes and by country. The scripts Analysis_timeseries_casestudies.R and Analysis_treecover_casestudies.R are the landscape scale analysis - They uses the three tables Table_indices_allyears_Cusco.csv, Table_indices_allyears_Intag.csv, Table_indices_allyears_Challabamba.csv which are all subsets of the Andean-wide analysis. For validation with supervised classification and Sentinel data in the case study landscape of Intag this GEE script was used to create and export validation data: https://code.earthengine.google.com/f30affc6c8fb06b8c94e6b3c487fd4ef. Validation analysis was done in R using these scripts: ANalysis_validation_Intag_Landsat_resampled_successional_classes.R and Analysis_validation_Intag.R. These scripts use the tables validation_Intag_Sentinel_improvedaccuracy.csv and validation_Intag_Landsat_improvedaccuracy.csv.