A dataset of the global potential for onshore field-scale solar photovoltaic systems
The deployment of solar photovoltaics (PV) can contribute to global decarbonization and mitigation of climate change[1,2]. A spatially explicit potential map of solar PV worldwide can facilitate decision-makers to plan PV projects in a sustainable way and provide important input for investigating the long-term impacts of PV expansion under different scenarios. Here we assess the land suitability for PV development using machine learning and estimate the potential of generating capacity for onshore field-scale solar PV (including commercial, industrial, and utility scales) worldwide. Our mapping is focused on the Earth’s surface between 65° N
Resource information
- Data purpose
- Research dataset
- Format
- tif
- Original repository
- 15239019
- DOI
- 10.5281/zenodo.15239019
- Credit basis
- Creators in Zenodo record metadata
- Access
- Open files available from the original Zenodo deposit
- Last checked
- Nov 5, 2026
- License
- cc-by-4.0
- Last record update
- Nov 5, 2026
About this resource
The deployment of solar photovoltaics (PV) can contribute to global decarbonization and mitigation of climate change[1,2]. A spatially explicit potential map of solar PV worldwide can facilitate decision-makers to plan PV projects in a sustainable way and provide important input for investigating the long-term impacts of PV expansion under different scenarios. Here we assess the land suitability for PV development using machine learning and estimate the potential of generating capacity for onshore field-scale solar PV (including commercial, industrial, and utility scales) worldwide. Our mapping is focused on the Earth’s surface between 65° N and 60° S, using the World Geodetic System 1984 coordinate system with the spatial resolution of 0.0083°. The global field-scale PV inventory data from ref. [3] were used as labeled positive data, and 68,661 random sites extracted globally were used as unlabeled data. Nineteen predictors were selected to model the potential distribution of PV systems, accounting for solar resource, climate, terrain, environment, hazard, and socioeconomic factors. The positive and unlabeled learning with constraints (PBLC) algorithm was used to train a one-dimensional convolutional neural network using positive-unlabeled samples[4]. The trained model can predict where PV developments are likely to occur, but actually only a small proportion of the suitable land will be converted to PV installations. Using the global PV inventory from ref. 2, we derived an empirical area factor for each land type. If a pixel is predicted to be suitable, the actual area for PV installation is equal to the area of the pixel multiplied by the area factor. Following the approach in ref. 2, we estimated the nominal peak alternating current generating capacity based on the potential installation area. We estimate that the area of suitable land for PV development is about 21,459,552 km2 globally, and the potential generating capacity is 169.52 (±3.21) TW. Rangeland contributes the largest proportion of estimated capacity (57.3%), followed by cropland (28.3%). File descriptions: pv_probability.tif: the global probability (suitability) map of solar PV development. pv_cap_ac_mw.tif: the global map of nominal peak alternating current (AC) generating capacity of solar PV (unit: mW). NoData Value: -1