Global nitrogen use efficiency for wheat, maize, and rice at 5-arc-minute resolution from 1995 to 2020

OverviewHere, we present a global gridded dataset of NUE for wheat, maize, and rice at a 5-arc-minute resolution from 1995 to 2020, reconstructed using a multi-source ensemble framework based on the mass-balance approach. The dataset integrates major N input pathways, including synthetic fertilizer, manure, atmospheric deposition, biological N fixation, and seed N, together with crop-specific grain N output. We evaluated the consistency and agronomic plausibility of our estimates through independent validation against national-scale data, farmer surveys, and existing literature, while also outlining potential pathways for future refinement. This dataset provides a consistent basis for analysing spatiotemporal patterns of crop NUE and supports global research on nutrient management, food security, and environmental sustainability. Dataset Description The dataset provides global coverage (180°W to 180°E, 90°S to 90°N) at a spatial resolution of 5-arc-minute for the period 1995–2020, using the standard WGS84 coordinate reference system. All spatial data are provided in GeoTIFF (.tif) format, ensuring high compatibility with standard geographic information systems and programmable data analysis tools (e.g., Python, R). A detailed description of the file organization and naming conventions within the data repository is provided below: - File Organization: The repository contains three separate ZIP archives, corresponding to the three major crops: wheat, maize, and rice. Each archive contains both the annual NUE and the normalized range spatial layers for that specific crop. - File Naming Conventions: § NUE Data: Files are named following the format NUE_{Crop name}_Median_xxxx.tif. § Normalized Range Data: Files are named following the format NUE_{Crop name}_Nor_xxxx.tif. § (Note: {Crop name} represents the specific crop, i.e., Wheat, Maize, or Rice, and xxxx denotes the four-digit year, such as 1995 or 2020). - Units: The NUE datasets use percentage (%) as the unit, whereas the normalized range is a dimensionless metric. Code availability All data processing was conducted using Python (version 3.13.5). The code supporting this study is publicly available at: https://github.com/TUMQiang/TUM_PA_NUE.git