Accessibility drives research efforts on Amazonian sarcosaprophagous flies

We provide the raw data and code derived from the article “Accessibility drives research efforts on Amazonian sarcosaprophagous flies” (DOI: 10.1098/rspb.2025.2109), published in Proceedings B. These materials were used in the analyses of the probability of knowledge of sarcosaprophagous flies in the Brazilian Amazon. Using occurrence records and machine learning, we mapped knowledge distribution at three levels (two taxonomic and one comparative) for sarcosaprophagous flies in the Brazilian Amazon: by families, by the best-sampled species, and through a null model that simulates the probability of knowledge by chance. These flies are of great ecological and public health importance, and our results reveal substantial and striking knowledge gaps across the region. The observed patterns were consistent across all taxonomic levels, emphasizing that addressing knowledge perspectives requires more than simply increasing research effort. Our results highlight the importance of studying less charismatic groups, such as sarcosaprophagous flies, which play essential roles in ecosystem functioning. They also provide new insights into the value of targeted research in remote areas and collaborative engagement with local traditional communities, both crucial for building a comprehensive understanding of biodiversity and promoting effective conservation in the Amazon. The provided code uses Random Forest models to estimate the probability of knowledge for the group based on occurrence data combined with environmental and accessibility predictors. It also includes step-by-step scripts to generate the figures that illustrate the results, as presented in the main manuscript. A modeling framework was built following the workflow and predictor files from Carvalho et al. (2023, DOI: 10.1016/j.cub.2023.06.077 ), whose data are also available in a section of this repository (DOI: 10.5281/zenodo.7951033 ), with some modifications based on additional information included in the current manuscript. In summary, the R scripts and all necessary files are organized in a directory-based structure. There are seven compressed folders, each described below. 1 - Predictors.zip: Contains raster layers stored in the "Predictors" folder. There are five raster predictors: Degradation, DistanceResearch, DrySeasonLength, LandOwnership e TravelTime. hese layers were produced and made available by Carvalho et al. (2023, DOI: 10.1016/j.cub.2023.06.077 ) at a spatial resolution of 1 km (also available on Zenodo DOI: 10.5281/zenodo.7951033 ). 2- Datasets.zip: This directory corresponds to the "Datasets" folder and includes the following files: "occ.csv", "occ_AM.csv", and "occ_sp.csv", containing the occurrence data used in the manuscript, both for the null model randomization and for generating models at the family and species levels, as well as for the null model itself; "VarImportance.csv" and "VarImportance_figuras.csv", used to generate Figure 3; "figure5_A-D.csv" and "figure5_A-D_sp.csv", used to generate Figure 4; "figure5_E.csv" e "figure5_E_sp.csv", used to generate Figure 5. 3- RasterMasks.zip: Contains the file "upland.tif", which was used as a spatial mask to generate the results presented in the manuscript. This layer was created in Google Earth Engine by Carvalho et al. (2023, DOI: 10.1016/j.cub.2023.06.077 ; 10.5281/zenodo.7951033 ). 4- Shapefiles.zip: Corresponds to the "Shapefiles" directory. It contains a shapefile that defines the study area boundaries, as described by Bullock et al. (2020, DOI: 10.1111/gcb.15029 ). 5- Scripts.zip: Includes all R scripts used for model generation and figure production: "Predictors_correlation.R": generate the correlation among predictors. Additional analysis for the supplementary material; "Null_model_sort.R": generates random occurrences for the null model; "knowledge_distribution_models.R": constructs the Random Forest models; "Figures_knowledge_distribution.R": creates the figures used in the main manuscript. 6- Projections.zip: This folder includes the probability projection maps for each family and species, as well as for the null model, stored as GeoTIFF files. There are 19 .tif files in total: (1) “Calliphoridae_flies.tif”, (2) “Chloroprocta_idioidea_Calliphoridae.tif”, (3) “Chrysomya_albiceps_Calliphoridae.tif”, (4) “Cochliomyia_macellaria_Calliphoridae.tif”, (5) “Hemilucilia_semidiaphana_Calliphoridae.tif”, (6) “Lucilia_eximia_Calliphoridae.tif”, (7) “Mesembrinellidae_flies.tif”, (8) “Mesembrinella_batesi_Mesembrinellidae.tif”, (9) “Mesembrinella_bellardiana_Mesembrinellidae.tif”, (10) “Mesembrinella_bicolor_Mesembrinellidae.tif”, (11) “Mesembrinella_quadrilineata_Mesembrinellidae.tif”, (12) “Mesembrinella_randa_Mesembrinellidae.tif”, (13) “Sarcophagidae_flies.tif”, (14) “Oxysarcodexia_intona_Sarcophagidae.tif”, (15) “Oxysarcodexia_thornax_Sarcophagidae.tif”, (16) “Peckia_(Pattonella) intermutans_Sarcophagidae.tif”, (17) “Peckia (Peckia) chrysostoma_Sarcophagidae.tif”, (18) “Peckia (Sarcodexia)_lambens_Sarcophagidae.tif”, and (19) “Model_nulo_nulo.tif”. 7- Av gOutputs.zip: This folder contains three raster files representing the final output layers of the research probability models. Each file corresponds to one of the modeled levels: (i) families, (ii) well-sampled species, and (iii) the null model. The family (3) and species (15) models represent the mean probability within their respective groups. Finally, a README.pdf file is provided to explain the functionality and structure of the directories.