Pervasive gaps in Amazonian ecological research

Raw data and code for the analyses on research probability across the Brazilian Amazonia. Briefly, the code reproduces the Random Forest models and the intersection of research probability with susceptibility to current and future anthropogenic disturbances, performs a site-level analysis of the resulting outputs; and then illustrates the building of figures used in the main manuscript.

The modelling framework was built using the directory structure informed in the README.pdf file. The R-code provided has steps designed to replicate the directory structure as reported, but understanding it is a good starting point to navigate the output produced. There are five zip files whose content we described below:

AvgOutputs.zip: a folder containing four files with the results layer from the research probability model. The files represent the final result for each habitat type and the average across habitat types.

Datasets.zip: represents the "Datasets" folder directory, as illustrated in the README.pdf.  This zip file contains the files ‘SamplingSites.csv’, which informs metadata for the entire list of occurrence data of community data (organism group, coordinates (long and lat) and habitat); and the ‘VarImportance.csv’, which was produced in the Script2_RCode_RandomForestModels.R, and used to build Figure 3 in the manuscript.

GlobalChangeLayers.zip: folder with input and intermediary layers used for generating Figure 4. Degradation files are from Lapola et al. (2023, DOI: 10.1126/science.abp8622), while climate layers are from IPCC interactive atlas (Gutiérrez et al. 2021, DOI: 10.1017/9781009157896.021.).

Predictors.zip: includes the raster files produced for the "Predictors" folder. There are five rasters representing the layers: DryMonths, LandTenure, NearbtDND, ResearchEduc, and TravelTime. All rasters were generated at the spatial resolution of 1 km spatial resolution using the files in Script1_GEECode_PredictorLayerPreparation.zip. 

Projections.zip: the projections of research probability for each habitat type and organism are contained in this folder. There are 11  files, named as aquatic_benthos.tif, aquatic_fishes.tif, aquatic_heteropterans.tif, aquatic_macrophytes.tif, aquatic_odonates.ti, upland_ants.tif, upland_beetles.tif, upland_birds.tif, upland_trees.tif, wetland_birds.tif, wetland_trees.tif.

RasterMasks.zip: contains the masks created for each habitat to remove areas outside each habitat and areas without forest. The files are aquatic.tif, upland.tif and wetland.tif. They were created in the Google Earth Engine platform

RData.zip: represents the "RData" folder directory, which is designed to store RData files produced while running the R-scripts. It currently includes three files to facilitate the plotting of main figures.

Scripts: a set of seven files, described in extra details in the following.

• Script1_GEECode_PredictorLayerPreparation.zip: a set of three GEE-scripts developed to prepare raster masks and predictor layers in the Google Earth Engine platform.

• Script2_RCode_RandomForestModels.R: R-script to perform data partition, Random Forest model training, model validation, and extraction of variable importance and partial effects per predictor.

• Script3_RCode_DeltaClimate.R: R-script to compute the metric of absolute climate change.

• Script4_RCode_Figure1.R: R-script to build graphical pieces composing the Figure 1.

• Script5_RCode_Figure2.R: R-script to build the Figure 2.

• Script6_RCode_Figure3.R: R-script to build graphical pieces composing the Figure 3.

• Script7_RCode_Figure 4.R: R-script to build graphical pieces composing the Figure 4.

Shapefiles.zip: represents the "Shapefiles" folder directory, as illustrated in the README.pdf. There is one shapefile in this zip, which corresponds to the study area limits according to Bullock et al. (2020, DOI: 10.1111/gcb.15029).