Global Renewables Watch: A Temporal Dataset of Solar and Wind Energy Derived from Satellite Imagery
Caleb Robinson, Anthony Ortiz ∗, Allen Kim, Rahul Dodhia, Andrew Zolli, Shivaprakash K Nagaraju, James Oakleaf, Joe Kiesecker, Juan M. Lavista Ferres · 2025
Read Global Renewables Watch: A Temporal Dataset of Solar and Wind Energy Derived from Satellite Imagery by Caleb Robinson, Anthony Ortiz ∗, Allen Kim, Rahul Dodhia, Andrew Zolli, Shivaprakash K Nagaraju, James Oakleaf, Joe Kiesecker, Juan M. Lavista Ferres (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-14860.txt.
Introduction
Caleb Robinson and co-authors built, in a 2025 paper, a world map of solar farms and wind turbines read from satellite photos over time. Tracking clean energy growth usually leans on national reports that lag and clash. Their Global Renewables Watch instead watches the hardware itself, quarter by quarter, from late 2017 to mid-2024.
Deep learning models scanned over 13 trillion pixels to spot panels and turbines. Each find carries a build date and the land use that came before. The final set holds 375,197 turbines and 86,410 solar sites. Country totals matched agency capacity figures closely. Planners gain a neutral, current ledger of the energy transition, useful for checking pledges against steel in the ground.
Catalog record
| CREATOR | Caleb Robinson, Anthony Ortiz ∗, Allen Kim, Rahul Dodhia, Andrew Zolli, Shivaprakash K Nagaraju, James Oakleaf, Joe Kiesecker, Juan M. Lavista Ferres |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-14860
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Global Renewables Watch: A Temporal Dataset of Solar and Wind Energy Derived from Satellite Imagery Caleb Robinson, Anthony Ortiz ∗, Allen Kim, Rahul Dodhia, Andrew Zolli, Shivaprakash K Nagaraju, James Oakleaf, Joe Kiesecker, Juan M. Lavista Ferres Abstract We present a comprehensive global temporal dataset of commercial solar photovoltaic (PV) farms and onshore wind turbines, derived from high-resolution satellite imagery analyzed quarterly from the fourth quarter of 2017 to the second quarter of 2024. We create this dataset by training deep learning-based segmentation models to identify these renewable energy installations from satellite imagery, then deploy them on over 13 trillion pixels covering the world. For each detected feature, we estimate the construction date and the preceding land use type. This dataset offers crucial insights into progress toward sustainable development goals and serves as a valuable resource for policymakers, researchers, and stakeholders aiming to assess and promote effective strategies for renewable energy deployment. Our final spatial dataset includes 375,197 individual wind turbines and 86,410 solar PV installations. We aggregate our predictions to the country level — estimating total power capacity based on construction date, solar PV area, and number of windmills — and find an r 2 r^{2} value of 0.96 0.96 and 0.93 0.93 for solar PV and onshore wind respectively compared to IRENA’s most recent 2023 country-level capacity estimates. Figure 1: ( Left ) Distribution of 86,410 solar PV installations and 375,197 onshore windmills detected by our models in 2024 Q2. ( Right ) Close-up visualizations of solar and wind installations in the village of Farmsum in the Dutch province of Groningen. Climate change poses significant challenges globally, manifesting through shifting weather patterns, threats to food production, and rising sea levels that increase the risk of catastrophic flooding [ 31 ] . The energy sector, responsible for approximately three-quarters of global greenhouse gas emissions, plays a pivotal role in these environmental changes [ 4 ] . To mitigate the most severe impacts of climate change, a transformation toward renewable energy sources is essential. Notably, 87% of Nationally Determined Contributions (NDCs) under the Paris Climate Accord aim to increase renewable energy usage, with over half specifying quantifiable targets 1 1 1 https://www.climatewatchdata.org/net-zero-tracker . Monitoring the expansion of renewable energy, particularly wind and solar installations — which constitute over 80% of newly installed renewable capacity — is crucial for meeting climate-related commitments [ 44 ] . Further, the rapid development of renewable energy infrastructure necessitates careful consideration of land use, as land is a limited resource in many regions [ 30 ] . The physical footprint and siting of these installations can impact ecosystems, cultural and historical resources, scenic landscapes, and agricultural production [ 19 ] . Proactive spatial planning can guide renewable energy development toward low-conflict areas, balancing economic viability with conservation priorities. Identifying “Renewables Acceleration Areas” through robust spatial planning — including ecosystem and wildlife sensitivity mapping, ecological priorities, and social value assessments — can facilitate sustainable development [ 34 ] . Analyzing past patterns of renewable energy expansion can also inform predictive models that optimize future siting, minimizing environmental conflicts [ 14 , 29 , 21 ] . Therefore, tracking the spatial patterns of renewable energy development is vital to ensure that such projects do not compromise biodiversity and other essential ecosystem services [ 20 , 22 ] . However, existing data about the location and history of renewable energy installations is limited and existing datasets become quickly outdated with the exponential growth of solar and wind power. Renewable Energy Capacity grew by more than 80% in the last 5 years with solar PV and wind accounting for 95% of the expansion [ 2 ] . Figure 2: Inference workflow . We run separate solar PV and windmill semantic segmentation models over PlanetScope quarterly satellite imagery from 2017 Q4 through 2024 Q2 to create a solar PV polygon dataset and windmill point dataset with estimated construction dates from each feature. Global satellite imagery layers combined with machine learning methods for extracting renewable energy locations offer the promise of creating continuously updated datasets of renewable energy installations along with their history and other metadata identifiable from space. Deep learning, in general, is a powerful tool for extracting information from satellite imagery, and Convolutional Neural Networks (CNNs) have been widely used for detecting roads [ 28 ] , buildings [ 7 ] , agricultural fields [ 42 ] , and performing building damage assessment [ 38 ] from satellite imagery. However, such approaches bring novel challenges and have required extensive resources to execute at global scales in the past — for example, a previous effort to create a dataset of global solar photovoltaic installations from satellite imagery required processing 550 TB of imagery, used 1 million CPU-hours, and 20,000 GPU-hours [ 23 ] . Training models that generalize over the variety of landscapes on Earth requires large representative labeled datasets, running such models over terrabytes of imagery requires concentrated engineering efforts, and filtering false positives in the resulting output to produce usable data products requires non-trivial labor investments. In this work we aim to address these challenges and present a new global map of commercial solar photovoltaic (PV) installations and onshore wind turbines based on high-resolution satellite imagery from 2024 Q2 (Figure 1 ) as well as per feature construction dates back through October 2017. Specifically, we propose a data-centric machine learning [ 41 ] approach to modeling solar PV and windmills from annotations sourced from OpenStreetMap along with a simple false positive filtering model based on MOSAIKS features [ 39 ] . We test our methods on public benchmark datasets and use them to train models that we apply globally and temporally (Figure 2 ). We validate our global map by comparing aggregate power capacity estimates at a country level to IRENA’s latest 2023 per-country estimates of renewable energy (Figure 6 ). Finally, we tie each feature to the land cover that preceded it using global 300-m resolution land cover data from the European Space Agency [ 9 ] . This, for the first time, provides a detailed understanding of the land use change associated with wind and solar development. 1 Materials and Methods We produce global, temporal datasets of solar photovoltaic (solar PV) power plants and onshore wind turbines by training and applying semantic segmentation models on PlanetScope basemap satellite imagery every three months from October, 2017 (2017 Q4) through June, 2024 (2024 Q2). We use these models to generate two global datasets: geospatial polygons that represent individual commercial, industrial, or utility scale solar PV installations and geospatial point data that represent individual wind turbines. Specifically, we follow the convention set in Kruitwagen et al. [ 23 ] for solar – targeting installations in excess of 10,000 square meters — and for wind we target any onshore turbine visible at a 4.7\text{px/m} image resolution (which we estimate to be a windmill with a rotor diameter of 47 meters or larger). This would include most wind projects built in the last 40 years [ 13 ] . We estimate each feature’s date of construction based on when it first appeared in the satellite imagery – already present by 2017 Q4, or quarterly from 2018 Q1 onward – and, finally,