Main

Global warming, driven by carbon emissions mainly from fossil fuels, has caused a 1.07 °C rise in global surface temperature during the 2010s compared with the pre-industrial era1. This warming poses substantial climate challenges to natural systems and human societies2. The 2015 Paris Agreement underscores the importance of limiting global temperature rise to below 1.5 °C or 2 °C (ref. 3). However, without intervention, the global temperature is projected to rise by 2.8 °C by the end of the century, pushing the Earth system closer to or surpassing the climate tipping points4.

Solar energy is an essential technology for climate change mitigation, showing impressive technological learning curves and large future potential5,6. While often underestimated in climate stabilization models, it could provide the single largest contribution to mitigation in 20507. Rooftop photovoltaic (RPV), initially a niche solution8, may also offer a global-scale opportunity to reduce fossil fuel reliance9. Previous studies have shown that the carbon mitigation potential of RPVs in China is up to 4 gigatonnes (Gt), accounting for 70% of the country’s emissions from the electricity and heat sector10. However, at the global level, studies have mainly focused on the technical benefits and economic feasibility of RPVs9,11. Analyses of the impact of RPVs on carbon mitigation and, consequently, their benefits to global climate, particularly in regions with diverse natural and societal conditions, are notably absent. Bridging this gap is vital for understanding the role of RPVs in climate change mitigation and guiding effective energy policy-making.

Evaluating the potential of RPVs over large geographic areas faces a considerable challenge in obtaining a comprehensive, high-quality global distribution of the rooftop area. Existing datasets, derived from on-site surveys or high-resolution Earth observations, are region-specific and have heterogeneous spatiotemporal references, leading to inconsistencies and variabilities in coverage and quality when applied globally12,13,14. Global rooftop area estimates also suffer from uncertainties due to the limitations of conventional remote sensing technologies with moderate resolutions in the tens to hundreds of metres. This could lead to misidentification of dispersed and fine-scale building rooftops, introducing cumulative errors in rooftop area quantifications15,16. The limited understanding of fine-grained global rooftop area hinders accurate assessments of RPV potential and their spatial variations17,18.

In this study we integrate geospatial data mining and artificial intelligence techniques to estimate the global rooftop area at a resolution of 1 km, addressing the lack of detailed geographic distributions. We evaluate the technical potential of RPVs, focusing on regional differences. Technical potential quantifies electricity generation based on resource availability, system performance and physical availability for RPV development19. This metric provides a baseline for understanding maximum feasible deployment under theoretical conditions, serving as a foundation for further studies on economic or market viability. For the associated environmental and climate benefits, this study estimates carbon mitigation from substituting grid electricity with RPVs using a baseline emission methodology. Then, by using nine advanced Earth system models (ESMs) from coupled model intercomparison project phase 6 (CMIP6), we explore the potential warming mitigation achievable by mid-century under various future climate policy scenarios. We also examine the spatial distribution of this potential, particularly in climate-sensitive regions and its role in mitigating climate tipping risks. These analyses enhance understanding of impact of regional RPVs on global energy landscapes and highlight how resource endowments shape their climate mitigation contributions.

Global rooftop area estimation

We developed a two-stage approach, combining both top-down and bottom-up methods, to estimate the global rooftop area (Extended Data Fig. 1). In the top-down stage, a deep learning model based on the Vision Transformer architecture19 (Methods) was trained and applied to high-resolution (1.2 m) remote sensing imagery to quantify rooftop area in selected representative regions. In the bottom-up stage, we modelled the nonlinear relationship between the rooftop area determined in the top-down stage and various geospatial variables using an ensemble of random forests. This allowed for the global extrapolation of rooftop area estimates.

The deep learning model was initially pretrained on publicly available benchmark datasets covering ~2,500 km2 (Methods and Supplementary Table 1). This pretraining phase enabled the model to recognize common building patterns across imagery with varying spatial resolutions, architectural styles and geographic contexts20. The model was then fine-tuned using manually labelled high-resolution remote sensing imagery, which covered 1,600 km2 of diverse landscapes worldwide (Methods and Extended Data Fig. 2a), including 835,253 labelled buildings. The deep learning model was evaluated on a separate set of high-resolution remote sensing images covering 386 km² from various locations, ensuring that these images were not part of the training process or model optimization (Methods and Extended Data Fig. 2b). The validation results confirmed that the model can accurately quantify rooftop area across a variety of landscapes globally (Methods and Extended Data Fig. 3).

In the bottom-up stage, we partitioned the global land surface into discrete spatial grids, each consisting of 1 km2 units. Geospatial datasets were aggregated into these grids, including population statistics, night-time light imagery, road networks, digital elevation models and land use and land cover data (Methods and Supplementary Table 2). An ensemble of random forest models was then used to estimate the global rooftop area from these datasets. To develop the predictive models, 1,724 cities were strategically selected as regions of interest (here the term ‘city’ refers to administrative divisions that may include both rural and urban areas; Extended Data Fig. 4). The trained deep learning model was applied to quantify rooftop area in these regions. Subsequently, 8.5 million spatial grid samples were generated, with rooftop area as the dependent variable and the aforementioned geospatial datasets as independent variables, which were then used to train the predictive models (Methods).

The extrapolated global rooftop area was assessed using 16,000 grids across macroregions (Methods and Extended Data Figs. 5, 6 and 7 and Supplementary Fig. 3). The grid-level analysis showed high consistency across regions (r2 = 0.89; bias error = 2%). Additionally, the global estimates demonstrated high consistency with other reference datasets, including the Google building footprints (GBF)21, the Microsoft building footprints (MBF)22, the global human settlement layer (GHSL)23 and the world settlement footprint 3D (WSF3D)24, with r2 values ranging from 0.75 to 0.85 (Methods and Extended Data Fig. 8). The global rooftop area was estimated to be 286,393 km2, with 30% of the total area contributed by East Asia (E-ASI) and 12% by North America (N-AME). At the national level, China (74,426 km2) and the United States (30,928 km2) ranked first and second, respectively, followed by India (23,087 km2) in third place (Fig. 1a). These global rooftop area estimates provide comprehensive coverage, accurately capturing the morphology of urban agglomerations in both developing and developed regions (Fig. 1b–d).

Fig. 1: Global rooftop area distribution.
figure 1

a, Rooftop area (in km2) at the macroregional and national scales. b, Global spatial distribution on a 1-km grid. c,d, Close-up of specific regions (indicated in b), covering approximately 1 and 0.7 million km2 in Nigeria (c) and China (d). Supplementary Fig. 3 contains the definitions and abbreviations for the macro-regions and countries.

Full size image

Electricity generation and carbon mitigation

Owing to geographical constraints such as rooftop orientation, slope and shading between buildings, only a fraction of rooftops is suitable for RPV installation. We calculated available rooftop area using a scaling factor of 30% based on general cases reported (Supplementary Table. 3). Solar energy input was derived from a high-resolution global surface solar radiation dataset (3-h temporal, 10-km spatial resolution)25. The dataset, derived from ISCCP and ERA5 datasets using improved physical inversion algorithms, has demonstrated better accuracy than GEWEX SRC, CERES and ISCCP datasets25. For regional studies, higher-quality datasets such as NRSDB, CAMS or CM-SAF are recommended, although they currently lack global coverage26. The electricity generation potential of an RPV system was defined as the maximum achievable electricity production for a given rooftop area. By considering a 20% panel conversion efficiency and an 80% overall efficiency, the conversion from solar energy to electricity generation can be accomplished (Methods). The carbon mitigation potential was defined as the CO2 reduction from replacing grid electricity with RPV electricity and thus calculated using baseline grid emission factors developed by the technical working group of the international financial institutions (Methods). To explain the variations in the main results caused by different panel efficiencies and rooftop availability, we conducted sensitivity evaluations (Supplementary Tables 4–6).

Solar radiation intensity largely determines the electricity generation of RPVs. The global average surface solar radiation is ~1,500 kWh m−2 yr−1, with hotspots concentrated in Africa (2,000–2,200 kWh m−2 yr−1). However, owing to limited building stock, the installation potential in Africa (1,188 GW) is only 7% of the global total. In contrast, the macroregions (N-AME, Northern America; N-EUR, Northern Europe; W-EUR, Western Europe and E-EUR, Eastern Europe) with lower solar radiation (900–1,200 kWh m−2) have an installation potential of 4,339 GW, which constitutes 25% of the world’s total (Extended Data Fig. 9a,b). National disparities between installed capacity and solar resources are evidenced by the volume and intensity of electricity generation (Fig. 2a). More countries (62%) have advantages (in the top 50%) in only one aspect of volume or intensity.

Fig. 2: Global distribution of RPV electricity generation and carbon mitigation potential.
figure 2

a,b, Electricity generation (a) and carbon emission reduction (b). Intensity and volume combinations are divided by quartiles, with the percentages of countries in each combination summarized. While the volume index refers to the benefits provided by all potential RPV installations, the intensity index indicates the benefits provided per unit of potential RPV installation.

Full size image

The carbon mitigation benefits of RPVs depend on grid baseline carbon intensity. While the global average carbon intensity stands at 529 gCO2 kWh−1, national values range from 0 to 1,330 gCO2 kWh−1 (Extended Data Fig. 9c). Macroregions with higher clean energy shares, such as W-EUR, generally exhibit carbon intensities below the average (206 gCO2 kWh−1), while fossil fuel-reliant regions such as E-ASI exceed the average (700 gCO2 kWh−1). Spatial disparities in installed capacity and grid intensity are reflected in mitigation volume and intensity (Fig. 2b). Countries with advantages (in the top 50%) in both mitigation volume and intensity accounted for 22% of the total. Notably, 9% of the countries, including Egypt, Sudan, South Africa and many other African, Asian and American countries, achieved dual volume-intensity benefits in both electricity generation and carbon mitigation.

While global RPV currently offers substantial carbon mitigation potential, ongoing energy transitions are gradually lowering grid carbon intensity. Scenarios—stated policies (STEPS), sustainable development (SDS) and net-zero emission (NZE)—represent different visions for energy transition and climate goals (Supplementary Note 1). Each outlines varying levels of constraints, where the increasing penetration of clean energy and optimization of power technologies progressively reduce the carbon mitigation potential of RPVs. From 2020 to 2050, the grid carbon intensity is expected to decrease by 48%, 96% and 101%, respectively, under the three scenarios (Supplementary Note 2 and Supplementary Fig. 4). Correspondingly, RPV electricity generation contributes to cumulative global carbon mitigation of 268, 144 and 102 Gt over three decades, respectively (Supplementary Fig. 5). At the macroregional level, different resource endowments in terms of installed potential, solar radiation and grid carbon intensity remain key factors driving differences in RPV electricity generation and carbon mitigation potential (Fig. 3).

Fig. 3: Cumulative potential from 2020 to 2050 in the STEPS scenario.
figure 3

a,b, Electricity generation (a) and carbon mitigation (b) for different macroregions are shown as bubbles of different sizes. The indicators of installed capacity, solar radiation and grid carbon intensity explain the reasons behind regional differences.

Full size image

Mitigations of global warming and climate tipping point risk

The conversion from RPV carbon mitigation to global warming mitigation potential was performed using the transient climate response to cumulative emissions of CO2 (TCRE)27, a well-established metric that quantifies the linear relationship between cumulative carbon emissions and global temperature. Regional temperatures also exhibit a linear response to cumulative carbon emissions27. We calculated regional TCRE (RTCRE) by aggregating data from nine ESMs (Methods). By applying this method, we were able to translate the carbon mitigation potential of RPVs into an estimated global and regional warming mitigation. The spatial pattern of RTCRE values varied across different latitudes and elevations (Extended Data Fig. 10a). The RTCRE values on land range from 0.7 to 6.9 °C per trillion tonnes of carbon (TtC), with an average of 2.4 ± 0.8 °C per TtC. The highest values were observed for high-latitude E-EUR and N-AME (>3°C TtC−1). Additionally, high-altitude regions such as the Qinghai-Tibet Plateau and the Western Australian Plateau in mid-to-low latitudes also have relatively high RTCRE (2.5–3.5 °C TtC−1). The spatial pattern of the intermodel distribution of RTCRE (Extended Data Fig. 10b, calculated as the standard deviation among the models) indicated that RTCRE uncertainty was generally higher in high-latitude regions than in mid- to low-latitude regions in the Northern Hemisphere.

In total, GHG emission reduction through RPVs from 2020 to 2050 under the STEPS, SDS and NZE scenarios are projected to mitigate global warming by 0.13–0.05 °C (Fig. 4). Since the global average temperature has already risen by 1.2 °C above pre-industrial levels by 20203, meeting the goals of the 1.5 or 2 °C warming limit requires an additional margin of only 0.3 or 0.8 °C. Under the STEPS scenario, we find that if RPV potential is not tapped, the additional carbon emissions from the power grid from 2020 to 2050 could lead to an approximate increase of 0.13 °C in global temperature. This equals 43% and 16% of the remaining allowable warming below the 1.5 and 2.0 °C thresholds (0.3 and 0.8 °C), respectively. However, more stringent climate policies (SDS and NZE scenarios) will optimize the power mix and further mitigate its baseline carbon emissions, resulting in greater mitigation of the relative avoidable global warming (0.07 and 0.05 °C). These contribute an additional 17–23% and 6–9% to meet the 1.5 or 2.0 °C thresholds, respectively.

Fig. 4: Cumulative global warming mitigation potential from 2020 to 2050 under different scenarios.
figure 4

a–c, STEPS (a), SDS (b) and NZE (c). Macroregional statistics show that regions (E-EUR and N-AME) close to the Arctic are more sensitive to global climate change and these regions are expected to benefit more from global RPV carbon mitigation efforts.

Full size image

The contributions of RPVs to global warming mitigation may be particularly substantial in regions with high climate sensitivity, where irreversible impacts on society and ecosystems are likely to occur once tipping points are reached28. Currently, at the warming level of ~1.2 °C, five tipping points have already been approached, with an additional five expected to be reached when the 1.5 °C warming is achieved29. In this context, efforts of RPV carbon mitigation are expected to reduce the likelihood of triggering these climate-critical thresholds. For example, RPVs may potentially provide a 25–66% additional warming margin to reach the lower warming threshold for the Atlantic meridional overturning circulation collapse (Fig. 5). RPVs could also potentially contribute an additional 17–44% margin for reaching the lower threshold for Barents Sea ice abrupt loss and mountain glaciers loss.

Fig. 5: Potential climate impacts of RPVs on climate-sensitive areas.
figure 5

The estimation of temperature thresholds for tipping points of climate-sensitive areas is revised from ref. 29. The 1.5–2 °C temperature target is indicated by the blue range. The potential warming mitigation derived from RPV relative to the current warming level (green dashed line) is shown in the yellow range.

Full size image

Discussion

Although previous studies demonstrated substantial electricity generation and carbon mitigation benefits of RPVs in energy transition, further investigation is required to understand their role in addressing climate change challenges. Here we estimated global rooftop areas at 1-km resolution by using geospatial data mining and artificial intelligence techniques, addressing the critical bottleneck in detailed energy assessments. Furthermore, we systematically assessed the climatic benefits of global RPV electricity generation, highlighting its potential contribution to climate change mitigation, particularly towards achieving global temperature targets.

Using nine advanced ESMs from CMIP6, we fill a critical gap in evaluating the climatic benefits of RPVs. From 2020 to 2050, potential carbon mitigation from RPVs corresponds to a global warming mitigation of 0.05–0.13 °C. This provides an additional 17–43% margin towards the 1.5 °C warming threshold and reduces risks of triggering the tipping points in climate-sensitive regions. High-resolution global rooftop area estimation further supports detailed energy assessments. Policy-makers often face trade-offs between coverage and resolution, but our robust and scalable method enables accurate large-scale assessments. Specifically, the top-down approach, which applies a deep learning model on high-resolution remote sensing imagery to quantify rooftop areas in selected regions, reduces bias errors and cumulative inaccuracies from medium-to-low resolution imagery or crowdsourced geographic information9,30. The bottom-up method enables efficient extrapolation of global rooftop area, ensuring accurate and seamless predictions. While this study successfully estimated the global rooftop area distribution at high resolution, accuracy was constrained by uncertainties in explanatory variables and geographical heterogeneity, causing regional variations in our data quality. Future work could refine region-specific variables, introduce spatiotemporal causal reasoning and integrate domain-specific knowledge to improve the estimations. The detailed global baseline data and potential distribution simulations provided may offer valuable micro-level support for future macro-level climate change integrated assessment models or energy system models31,32. This could enhance their decision-making capabilities at regional and city levels, promoting more precise energy planning and climate policy evaluation.

Encouraging citizen participation in energy production and turning them into consumers of the energy they produce are crucial steps in tackling global climate challenges33. Our results show considerable global potential for widespread RPV deployment. However, future sustainable development should not rely solely on universal solutions. Our results highlight the necessity for regional planning that adapts to local conditions to enhance the energy and climate benefits of RPVs. Considering volume and intensity metrics and perspectives of electricity generation and carbon mitigation, we delineate advantageous combinations for developing RPVs in different regions, advocating for the rational use of resource endowments. For instance, despite holding 60% of the best solar resources globally, Africa accounts for only 1% of photovoltaic (PV) installations. Poor energy services and reliance on polluting energy sources further impede the progress of Africa in addressing environmental degradation and climate change34. Our results suggest that RPV development could provide substantial energy and environmental benefits in most African countries, serving as a key driver of sustainable development. Additionally, our results indicate the most influential regions and those most likely to benefit from the global effort of broad deployment of RPVs. For example, in developing countries with large populations such as China and India, the widespread adoption of RPVs not only facilitates their own energy transition, but also plays a pivotal role in assisting other regions around the globe to combat climate change, especially in regions with high latitudes and altitudes or those approaching their tipping points.

The main results of this study focus on the technical potential of RPVs and their associated environmental and climate benefits. We adopted theoretical assumptions to provide clear and quantifiable references. We also documented comparisons of our main results with selected studies at global, regional, national and municipal levels, with details provided in Supplementary Table 7. Under the same assumptions, at the global level, our estimated annual electricity generation potential of 19,483 TWh is comparable to the 18,040 TWh estimated by ref. 35. For Switzerland, our estimated rooftop area of 320 km2 is similar to the 267 ± 71 km2 estimated by ref. 36. For states and counties in the United States, our results on rooftop area, electricity generation and carbon mitigation are comparable to those reported by Google Project Sunroof37. These comparisons with high-resolution, top-down studies demonstrate the high accuracy of our main results, falling within the range of estimates reported in the current literature.

For global-scale assessments, we used simplified methods with globally uniform assumptions to account for complex external factors. The impacts of varying rooftop availability and PV panel efficiency on the main results are presented in Supplementary Tables 4–6. We suggest that future research investigate local benchmarks when targeting various deployment scenarios. Specifically, rising temperatures affect the actual efficiency of PV panels and this effect is material-dependent (Supplementary Table 8)38. Panel materials with smaller band gaps are generally more sensitive to temperature and moisture in the atmosphere39. Furthermore, based on IPCC global warming scenarios, by 2100, the energy output of silicon PV could decrease by 15 kWh kWp−1, with some regions experiencing losses of up to 50 kWh kWp−1 (ref. 40). Additionally, compared with horizontal installation assumed in this study, installations with optimal tilt angles, single-axis vertical tracking, single-axis horizontal tracking and dual-axis tracking could increase the final radiation received by approximately 1.19, 1.22, 1.35 and 1.39 times, respectively41. Globally, as latitude increases, the advantages of tilted and tracking systems over horizontal installations also become more pronounced41. Our study provides a starting point for large-scale assessments of RPV potential, which can be expanded in future studies by incorporating meteorological simulation and urban morphology. This would enhance regional applicability and support the development of optimal PV deployment strategies in complex urban environments.

Our assessment of the theoretical potential of RPVs highlights their substantial climate mitigation benefits. However, realizing this potential within the existing energy infrastructure involves complex factors. Achieving high PV scenarios requires policies supporting PV technological innovation and coordinated plans to accommodate large-scale PV power generation. This involves modernizing power market regulations, expanding transmission grids and scaling up storage technologies, all while addressing issues related to supply and demand balance42,43. In addition, peer effects have been recognized as important motivators for installations, as they are critical in influencing climate-related decision-making through geographic effects and social network effects44. To most effectively expand the use of solar panels, policy-makers can increase cumulative use by deploying them in areas where they are sparse, thus taking advantage of the proximity effect45. Moreover, through a combination of government interventions and adaptive business models, racial inclusiveness of PV can be effectively realized and the penetration of PV in all communities may be facilitated46.

In summary, our study not only proves the immense potential of RPVs in promoting sustainable energy solutions but also suggests that fully exploiting their potential can hasten the transition to a more climate-resilient world. Although unlocking this potential still requires policy-makers and society to overcome future organizational and financial challenges, it is expected that RPVs will provide effective clean energy solutions in many regions afterwards. Our quantification of the substantial climate benefits of RPVs will motivate energy planners to capitalize on this opportunity, while also encouraging researchers to further explore the potential of other forms of building-integrated PVs that seamlessly integrate energy production with urban infrastructure, thereby expanding the scope and impact of renewable energy applications.

Methods

Two-stage process for global rooftop area estimation

The global rooftop area estimation followed a two-stage process (Extended Data Fig. 1): (1) a top-down approach using deep learning to quantify the rooftop area in selected representative regions from high-resolution remote sensing imagery and (2) a bottom-up approach using random forest ensembles and various geospatial datasets to extrapolate rooftop area estimates at a global scale.

In the top-down stage, we used the SegFormer model, a state-of-the-art Vision Transformer-based deep learning model in computer vision47. The choice of this architecture was motivated by its capacity to process large datasets and its ability to capture global context while handling complex patterns and long-range dependencies—crucial for large-scale building identification48. For pretraining, we used publicly available benchmark datasets focused on building identification13,49 (Supplementary Table 1). These datasets, generated by various institutions between 2013 and 2020, cover ~2,500 km2 across diverse regions with varying spatial resolutions, ranging from very fine (0.1 m per pixel) to coarser (3 m per pixel). The selected datasets represent a wide range of global regions and building types, including urban, suburban and rural areas, as well as a mix of built-up and natural landscapes. They also cover various urban layouts and densities, offering high-resolution imagery that captures different architectural styles and building sizes. This broad representation ensures that the model is trained across diverse environments, enhancing its ability to generalize to different geographic contexts.

For fine-tuning the deep learning model, we used high-resolution Google Earth imagery with a consistent spatial resolution of 1.2 m. Google Earth imagery is obtained through the Google Static Map API, which applies rigorous sifting and preprocessing to produce cloud-free, high-quality images50. The imagery is sourced from a combination of airborne and satellite platforms, including WorldView, QuickBird, IKONOS, GeoEye-1, Pleiades, SuperView-1 and Kompsat-3A, resulting in harmonized imagery51. Further details on the SegFormer architecture, along with the pretraining and fine-tuning processes, are provided in Supplementary Note 3.

We applied the trained deep learning model to quantify rooftop area in 1,724 cities, where ‘city’ refers to administrative divisions sourced from the database of global administrative areas (Extended Data Fig. 4). These regions were strategically selected using a spatial sampling scheme, with cities treated as sample units. The objective was to maximize the distance between samples to ensure representativeness, optimized using a simulated annealing algorithm52. To further enhance the geographic and environmental diversity of the selected regions, we first stratified the global grids using K-means clustering, based on natural and human characteristics of each grid (Supplementary Fig. 6 and Supplementary Table 2). The optimal number of clusters for the K-means algorithm was determined using silhouette coefficients. After stratification, spatial sampling was performed within each stratum.

In the bottom-up stage, we collected multi-source geospatial datasets and aggregated them as statistical variables—such as built-up proportion, night-time light intensity, road length, population, tree-cover proportion, terrain elevation, terrain slope and geographical coordinates (Supplementary Fig. 2 and Supplementary Table 2)—into global discrete spatial grids. These variables are strong predictors of rooftop areas because they capture both human activity and the physical characteristics of the environment9,10. For instance, night-time light intensity is often correlated with urbanization and building density, while terrain elevation and slope can influence building placement and rooftop orientation. Each spatial grid covered an area of 1 km2 and was based on the world Mollweide equal area projection (EPSG: 6933). To model the nonlinear relationship between the rooftop area and these geospatial variables, we aggregated the rooftop area determined in the selected regions into the corresponding grids. Grids lacking high-resolution remote sensing imagery were excluded, resulting in 8.5 million grid samples. The rooftop area was treated as the dependent variable in these samples, while the other statistical variables were used as independent variables.

Considering the spatial heterogeneity of grid cell variables and addressing the long-tail regression modelling issue, we developed both regression- and classification-oriented random forest models at both continental and global scales. The trained random forest ensembles were then applied to estimate the global rooftop area using a set of global independent variables. Further details on the random forest ensemble can be found in Supplementary Note 4. In the postprocessing stage, a water map was used to refine the initial estimates by assigning a rooftop area of 0 m2 to grids that were 100% water-covered based on land use and land cover data.

Evaluation of estimated global rooftop area

The evaluation involved assessing the accuracy of the deep learning model in the top-down stage and the random forest ensemble in the bottom-up stage using independent test datasets. Additionally, the quality of the global rooftop area estimates was evaluated through qualitative and quantitative comparisons with reference datasets.

To evaluate the performance of the deep learning model, we created 386 plots, each covering an area of 1 km2 (Extended Data Fig. 2b). These plots were selected to ensure global representation, with two plots randomly distributed within the built-up areas of each country. High-resolution remote sensing images were obtained for all 386 plots and each was manually labelled to identify rooftop areas. The average rooftop area across the plots was 0.14 ± 0.10 km2 (mean ± s.d.), with a maximum of 0.53 km2. These plots span different countries and geographic regions, encompassing a wide range of building distributions and densities, providing a comprehensive evaluation. The plots were further divided into 2,951 image patches, each processed to match the size and colour bands of the training images used for the deep learning model, ensuring consistency in the validation process (examples in Supplementary Fig. 1). This test dataset was kept independent from the training data to prevent bias in the performance assessment of the model.

Against the labelled image patches, the deep learning model successfully identified 76% of the rooftops (true-positive rate), with a 2.7% false-positive rate where non-rooftop objects were mistakenly identified as rooftops. This level of performance is acceptable for building identification at a global scale: we compared the MBF and GBF—currently state-of-the-art building footprint datasets—against our test dataset. MBF had a true-positive identification rate of 61.6% and a false-positive rate of 4%, while GBF had a true-positive rate of 66.5% and a false-positive rate of 3.8%.

Additionally, we compared the rooftop area quantified by the deep learning model with the manually labelled rooftop area for each image patch. Overall, the predictions of the model showed a strong correlation with the actual data, with an r2 value of 0.93 and a slope of 1.04 across all test samples (Extended Data Fig. 3a). However, we observed variations in prediction accuracy across different macroregions (Extended Data Fig. 3b–u): the model performed with greater accuracy in economically developed regions, where r2 values exceeded 0.95. In contrast, accuracy decreased in less economically developed regions, with r2 values around 0.9. These discrepancies could be due to the varying ability of the model to recognize landscapes that differ notably across regions.

To evaluate the accuracy of the global rooftop area estimation by the random forest ensemble, we randomly selected 16,000 grids across various macroregions (800 grid samples per macroregion). These grids were not part of the training process of the random forest model. We used deep learning and high-resolution remote sensing imagery to quantify the rooftop area within these grids. The rooftop area determined from the images was then compared with the estimates generated by the random forest ensemble. Our results demonstrated a high level of accuracy, with an r2 value of 0.89 and a slope of 0.87 (Extended Data Fig. 5), indicating strong consistency between the estimated and observed rooftop area, although with a slight underestimation.

However, the results also indicated a drop in accuracy for the Pacific Islands (r2 = 0.61, bias error = 55%; Extended Data Fig. 5c) and Western Asia (r2 = 0.67, bias error = 24%; Extended Data Fig. 5k). This suggests potential limitations in capturing relevant factors in these regions, probably due to a lack of local data in the training set, especially in areas where high-resolution remote sensing imagery is scarce. The bias error here is defined as the relative bias, normalized by the absolute value of the sum of the observed rooftop areas and is calculated using the following equation53:

$$\text{bias}=\frac{\left|\sum _{N}\left({Y}_{\text{obs}}-{Y}_{\text{pred}}\right)\right|}{\left|\sum _{N}{Y}_{\text{obs}}\right|}$$
(1)

We also compared the estimated and observed rooftop area by identifying deviations (residuals) across the 16,000 test grids. The results revealed a correlation between rooftop area and residuals: grids with larger rooftop area tended to exhibit greater residuals. Specifically, for grids with rooftop area ranging from 0.01 to 1 km2, the variation in residuals was substantial, showing a slight underestimation (Extended Data Fig. 6a). Residuals demonstrated a bell-shaped normal distribution, with most clustering between ±5,000 m2 (Extended Data Fig. 6b). Further examination of the global distribution of estimated rooftop area revealed that most 1-km2 grids had rooftop area <0.4 km2 (Extended Data Fig. 7a). This consistent pattern of rooftop area distribution was evident across various scales (Extended Data Fig. 7b–u).

To assess the quality of our global rooftop area estimates, we compared them with several reference datasets, including GBF, MBF, GHSL and WSF3D. The qualitative characteristics of these reference datasets are described in Supplementary Table 10. While GBF and MBF represent advanced building footprint products, their coverage is incomplete, omitting extensive areas across multiple continents (Supplementary Fig. 7). For example, aggregated GBF lacks large portions of North America, Europe, East Asia and Australia, while MBF lacks coverage over China, Russia and parts of Europe. Compared with these datasets, our estimates provide globally seamless coverage, produced through the two-stage method. By incorporating natural and human-related datasets from 2020, our estimates capture current urban development and building areas, with the exception of the digital elevation model from SRTM v.4 (collected in the 2000s), which remains appropriate given the relatively stable nature of global terrain over time.

For the quantitative comparison, we conducted a detailed grid-level and continental-level analysis. Owing to the limited coverage and potential spatial completeness issues in the GBF and MBF datasets, we restricted our comparison to grids that contain building areas in both datasets. For comparisons with GHSL and WSF3D, we used global grids. The results of these grid comparisons are shown in Extended Data Fig. 8, where our global rooftop area estimates exhibit high correlations with the reference datasets. However, our estimates tend to be lower than those of the reference datasets, as reflected by the slopes of the linear fits, all of which are <1 (0.71 for GBF, 0.83 for MBF, 0.58 for GHSL and 0.72 for WSF3D). This underestimation trend is consistent with our test dataset results, which also produced a slope of 0.87. Among the reference datasets, the comparison with GHSL shows the largest discrepancies, with an r2 value of 0.76 and a bias error of 37%. We attribute this to the tendency of GHSL to overestimate rooftop areas by including impervious surfaces.

We also evaluated the estimates across different macroregions. The grid-level analysis shows high correlations with both GHSL and WSF3D (Supplementary Figs. 8 and 9). However, in less economically developed regions—such as the Pacific Islands, Central Asia and various regions of Africa (Eastern, Southern, Western and Middle Africa)—we observed relatively lower r2 values and higher biases. Furthermore, the comparison of total rooftop area across macroregions, summarized in Supplementary Table 11, shows that WSF3D and our estimates are generally consistent, while the total area of GHSL is substantially higher. This trend persists in less economically developed regions, where our estimates are typically lower. The potential underestimation could be due to geographic variations in building identification accuracy, as previously discussed. At the global level, our estimates closely align with those of WSF3D, although the totals of WSF3D are slightly higher. In contrast, the global area of GHSL is substantially larger, probably indicating overestimation. Compared with a previous study that estimated a total global rooftop area of 193,875 km2 (which was noted for systematic underestimation)9, our estimates (286,393 km2) demonstrate improved alignment with reference datasets. Although slight underestimation remains in our results, we have mitigated many of the issues identified in earlier studies.

Assessment of electricity generation and carbon mitigation potential

Unified assumptions were applied to the RPV systems in this study. On the basis of the current technical level of the PV industry54, the scale and performance parameters of PV applications were determined (Supplementary Tables 12 and 13), including a PV panel conversion efficiency of 20%, an overall PV system efficiency of 0.8 and a rated power of 200 W m−2. All the PV panels in the system were assumed to be fixed horizontally. The available installed capacity of RPVs, essential to determine their potential for electricity generation and carbon mitigation, was calculated by converting the total rooftop area into the available rooftop area using a scaling factor. The factor typically considers a range of geographical constraints for RPV system installation, such as the societal function, rooftop slope, orientation, shadows and obstacles of the building. On the basis of a literature review (Supplementary Table 3), the scaling factor used in the main results was assumed to be 30%.

The potential installed capacity, \({P}_{\rm{installed}}\), was calculated using equation (2) as follows:

$${P}_{\rm{installed}}={P}_{\rm{rated}}\times {S}_{\rm{rooftop}}\times {C}_{\rm{scaling}}$$
(2)

where \({P}_{\rm{rated}}\) is the rated power of the PV panel, \({S}_{\rm{rooftop}}\) is the total rooftop area obtained from the global rooftop area estimation and \({C}_{\rm{scaling}}\) is the scaling factor for calculating the available rooftop area for RPV installation.

The potential electricity generation \({P}_{\rm{power}}\) of the RPVs was estimated using equation (3):

$${P}_{\rm{power}}={S}_{\rm{rooftop}}\times {C}_{\rm{scaling}}\times {\rm{GHI}}\times {C}_{\rm{PV}}\times K$$
(3)

where GHI is the global horizontal irradiance received by the surface. We took the average values from 2010 to 2018 to represent the general radiation conditions at different assessment times. CPV is the conversion efficiency of the PV panel and K is the overall efficiency of the RPV system.

In this study, our focus was on the electricity generation stage of RPV systems, excluding consideration of other life-cycle stages. This decision stems from the considerably lower life-cycle carbon emissions of RPV systems compared with the emissions mitigated during the operational stage55 (Supplementary Table 14). We followed the baseline methodology56 proposed by the United Nations Framework Convention on Climate Change (UNFCCC) to measure the environmental benefits of renewable energy projects. The baseline methodology provides a simplified analytical framework and facilitates a uniform comparison across different projects, thereby promoting the development of related policies. The carbon emissions mitigated by RPV power replacing grid power were calculated using the baseline emission factors of the national grid57,58. The baseline emission factors included the operating margin (OM) and build margin (BM). While OM represents the cohort of existing power plants most affected (reduced) by the project (generally high-emission or high-cost power plants), BM represents the cohort of prospective/future power plants whose construction and operation could be affected by renewable energy projects, based on an assessment of planned and expected new generation capacity. The combined margin (CM), obtained from the weighted average of the OM and BM, represents the overall impact of both aspects, and reflects the existing carbon intensity of the national grid. Because the electricity generated through RPVs is considered clean energy, the CM factor, \({\rm{EF}}_{\rm{grid,CM}}\), can be used as the carbon mitigation factor, \({\rm{EF}}_{\rm{PV,mitigation}}\), of the RPV system. The calculations were conducted using equation (4) as follows:

$${\rm{EF}}_{\rm{PV,mitigation}}={\rm{EF}}_{\rm{grid,CM}}={\rm{EF}}_{\rm{grid,OM}}\times {W}_{\rm{OM}}+{\rm{EF}}_{\rm{grid,BM}}\times {W}_{\rm{BM}}$$
(4)

where \({\rm{EF}}_{\rm{grid,OM}}\) is the OM factor and \({\rm{EF}}_{\rm{grid,BM}}\) is the BM factor. WOM is the weight of the OM factor and WBM is the weight of the BM factor. The values for PV projects were set to 0.75 and 0.25, respectively, based on the UNFCCC methodologies58.

According to the carbon mitigation factors of the RPV system for different power grids, carbon mitigation, \({P}_{\rm{carbon},{mitigation}}\), was calculated according to equation (5):

$${P}_{\rm{carbon,mitigation}}={P}_{\rm{power}}\times {\rm{EF}}_{\rm{PV,mitigation}}$$
(5)

The selected baseline emission factors reflect the average conditions of national grids in recent years and are updated at least every 2 years (ref. 58). For assessment purposes, we designated 2020 as the base year. Future baseline emission factors are subject to alterations in the power structure of each country. We assume that this change is consistent with the trend in grid carbon intensity projected by the International Energy Agency for 2020–205059. On the basis of this assumption, we obtained future changes in the \({\rm{EF}}_{\rm{PV,mitigate}}\) under three climate policy scenarios: STEPS, SDS and NZE (Supplementary Notes 1 and 2 and Supplementary Fig. 4).

In this study, the energy assessment relied on conversion factors, such as rooftop area availability and PV panel efficiency, using harmonized global reference values. We acknowledge the importance of examining these factors at the regional level. However, such a detailed analysis was beyond the scope of this study. Sensitivity analyses were included to mitigate this limitation and provide an initial understanding of how variations in these factors could affect our main results (Supplementary Tables 4–6).

Assessment of global and regional warming mitigation potential

The relationship between global warming and cumulative carbon emissions was established using the well-documented TCRE metric. TCRE is estimated as the change in global surface air temperature per cumulative CO2 emissions60. It considers both physical climate processes and the dynamics of land and ocean carbon sinks, presumed to be constant over time and independent of emission pathways61. The relationship between regional climate warming and carbon emissions was established using the RTCRE, which describe changes on a regional scale. In this study, the TCRE was scaled to a regional scale to generate an RTCRE using simple pattern scaling62. According to a previous study27, RTCRE can be calculated using equation (6):

$${\rm{RTCRE}}(m,x,t)=\frac{\varDelta T(m,x,t)}{E(m,t)}$$
(6)

where \(\varDelta T(m,x,t)\) is the temperature change simulated by the mth model in the ensemble, for the spatial domain x at time t. Parameter \(E(m,t)\) is the diagnostic value of the cumulative CO2 emissions (calculated as the sum of the annual changes in atmospheric CO2 concentration and the CO2 uptake by ocean and land carbon sinks) of the model.

We used the 1pctCO2 simulations from nine ESMs with dynamic carbon cycle from the CMIP6 archive. The ensemble comprised ACCESS-ESM1.5, BCC-CSM2-MR, CanESM5, CESM2, IPSL-CM6A-LR, MIROC-ES2L, MPI-ESM1.2-LR, NorESM2-LM and UKESM1-0-LL. We followed the standard method63 for calculating \(\varDelta T(m,x,t)\), averaging over a 20-yr window (years 60–79) centred on the year of the CO2 doubling (year 70). We applied this calculation to the spatial temperature fields from each model and then calculated the ensemble mean of the nine models. We considered the ensemble mean as the best estimate of the RTCRE values and represented the uncertainty range using one standard deviation of the model responses.

After obtaining the RTCRE, the warming mitigation potential, \({P}_{\rm{warming},{mitigate}}\) of RPVs were calculated according to equation (7):

$${P}_{\rm{warming},{mitigate}}={P}_{\rm{carbon},{mitigate}}\times \rm{RTCRE}$$
(7)

While the concept of TCRE simplifies the relationship between CO2 emissions and global warming into a linear model, this relationship is fraught with uncertainties owing to numerous complex factors, such as the effects of permafrost melting and non-CO2 GHGs64. Land surface changes caused by PV panels could also impact the regional climate to some extent by redistributing surface energy, water fluxes and even atmospheric circulation65; however, global climate change mitigation should be dominated by carbon emission reductions. In addition, future solar energy generation could be modulated by climate change to some extent66. Future climate change impacts should be considered for finer assessments.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.