Global 10 arc-seconds land suitability maps for projecting future agricultural expansion

This repository contains data from “Global 10 arc-seconds land suitability maps for projecting future agricultural expansion” study. Abstract: Conversion of nature into agricultural land is a major driver of environmental change globally. However, the spatial resolution of current global land change models is relatively coarse and does not capture fine-grained patterns relevant for quantifying local impacts. To fill this gap, we created global suitability maps for conversion into agricultural land at a resolution of 10 arc-seconds (~300 meters at the equator). To that end, we developed Artificial Neural Network models relating locations of conversions into agricultural land, as recorded from 2003 to 2013, to various explanatory variables reflecting topography, climate, soil and accessibility. Cross-validation of the models indicated good performance with a mean Area Under the Curve (AUC) value of 0.95. Hindcasting from 1992 to 2002 revealed a similar performance (AUC = 0.94), indicating that our models are robust when applied to a different temporal context. Our high-resolution land suitability maps can be used to project future expansion of agricultural land worldwide, which is an asset for global environmental assessment studies. Data description: We provide here raster maps of suitability for agricultural expansion into three categories of agriculture - (i) homogeneous cropland, (ii) mosaic cropland with >50% crops, and (iii) mosaic cropland with >50% vegetation. The source for delineating categories was the ESA CCI land cover data. ESA CCI land cover data recognizes additional categories of agricultural land, however some of them have limited spatial coverage. For that reason, we merged the rainfed cropland and irrigated cropland categories into a single category - homogeneous cropland, where a grid cell is largely dominated by crops. Rainfed croplands account for 87% of the homogeneous croplands, while irrigated croplands account for the remaining 13%. Mosaic categories were defined in the same way as in the ESA CCI land cover dataset. Numerical designations of these categories in the ESA CCI land cover dataset are 10, 20, 30, and 40 for rainfed, irrigated, mosaic with >50% crops, and mosaic with <50% crops, respectively. Global suitability maps are provided at the spatial resolution of 10 arc-seconds (~300 meters at the equator). These files are available for three categories in the main folder with the filename prefix "Agri_suitability_mosaic_*". The numerical value in the file name refers to the agricultural category type (10 - homogeneous cropland, 30 - mosaic cropland with >50% crops, and 40 - mosaic cropland with >50% vegetation). In addition to the 10 arc-second layers, we provide aggregated layers with the spatial resolution of 30 arc-seconds, 5 and 10 arc-minutes, for more coarse-grained applications and less computationally-intensive analyses. We provide the aggregated layer maps for the minimum, mean, and maximum values of the aggregated 10 arc-seconds values within the coarser cells. There are in total 9 files provided for each of the aggregated spatial resolutions. Repository content: Main folder: - “Agri_suitability_mosaic_10.tif” is the global raster map for homogeneous cropland category at the spatial resolution of 10 arc-seconds. - “Agri_suitability_mosaic_30.tif” is the global raster map for mosaic cropland with >50% crops category at the spatial resolution of 10 arc-seconds. - “Agri_suitability_mosaic_40.tif” is the global raster map for mosaic cropland with >50% vegetation category at the spatial resolution of 10 arc-seconds. - "readme.txt" is the text file with the basic description and the metadata for the repository. Aggregated_layers folder: This folder contains three subfolders for files with a different spatial resolution (30s, 5m, 10m; see argument "RESL" below), with 9 files in each folder. File names for the aggregated maps of global land suitability contain the following information: “Agri_suitability_aggregated_RESL_TYPE_CATG.tif” - "RESL" is the spatial resolution of the layer. Value is either "30s", "5m", or "10m", corresponding to spatial resolution of 30 arc-second, 5 arc-minutes, and 10 arc-minutes. - "TYPE" is the type of aggregated values. Value is either "min", "avg", or "max", corresponding to the minimum, mean, and maximum values of the aggregated 10 arc-seconds values within the coarser cells. - "CATG" is the category of agricultural land. Value is either "10", "30", or "40", where category 10 is homogeneous cropland, category 30 is mosaic cropland with >50% crops, and category 40 is mosaic cropland with >50% vegetation. Raster metadata: Driver: GTiff Projection proj4string: +proj=longlat +ellps=WGS84 +no_def