Increased terrestrial ecosystem carbon storage associated with global utility-scale photovoltaic installation

Wang, Qingrui; Wang, Kai; Shao, Lintao; Tang, Xinyi; Tang, Shuchang; Mašek, Ondřej; Meyer, Gesa; Kleissl, Jan; Zhang, Liwei; Wang, Mudan; Wang, Weisheng; Yang, Qing; Sitch, Stephen

doi:10.1038/s41561-025-01715-2

Article
Published: 02 June 2025

Increased terrestrial ecosystem carbon storage associated with global utility-scale photovoltaic installation
与全球大型光伏电站建设相关的陆地生态系统碳储存增加SCI升级版地球科学1区

Nature Geoscience volume 18, pages 607–614 (2025)Cite this article

2495 Accesses
2 Citations
20 Altmetric
Metrics details

Abstract 摘要

Utility-scale photovoltaic (USPV) stands out as one of the foremost renewable energy technologies crucial for achieving global climate targets, owing to its low carbon footprint. While individual case studies exist, a comprehensive global analysis of the impacts of USPV deployment on land-cover changes and subsequent carbon pool dynamics across diverse ecosystems remains lacking. Here we show that worldwide deployment of USPV plants between 2000 and 2018 would increase the carbon pool of the hosting ecosystem by a total of 2.1 TgC over their lifespans, as revealed by the ensemble mean of multiple datasets. Although the carbon pool changes associated with global USPV deployment currently contribute approximately ${15.9}_{- 5.8}^{+ 1.0} %$ ( ${e n s e m b l e m e a n}_{- d i f f e r e n c e t o p e r c e n t i l e 25}^{+ d i f f e r e n c e t o p e r c e n t i l e 75}$ ) (or an average absolute carbon footprint of approximately ${10.5}_{- 3.8}^{+ 0.5} g$ CO₂-equivalent per kilowatt-hour) of the carbon footprint of USPV plants, this share is projected to increase by around 7-fold by 2050, driven primarily by decreasing photovoltaic manufacturing emissions. Notably, optimizing land management strategies can potentially enhance carbon density in the hosting ecosystem of existing USPV plants by approximately ${3.0}_{- 0.4}^{+ 3.7} k g C m^{- 2}$ , thereby facilitating an average reduction of ${4.3}_{- 0.2}^{+ 9.3} %$ in the carbon footprint of these USPV plants.
大型光伏（USPV）作为一种具有低碳足迹的可再生能源技术，在实现全球气候目标方面具有重要意义，因此备受关注。尽管存在个别案例研究，但关于 USPV 部署对不同生态系统土地利用变化及随后碳库动态影响的全局性综合分析仍然不足。在此，我们表明，2000 年至 2018 年间全球 USPV 电站的部署将使宿主生态系统的碳库在其生命周期内总共增加 2.1 TgC，这一结果由多个数据集的集合平均数揭示。尽管与全球 USPV 部署相关的碳库变化目前约占 USPV 电站碳足迹的 ${15.9}_{- 5.8}^{+ 1.0} %$ ( ${e n s e m b l e m e a n}_{- d i f f e r e n c e t o p e r c e n t i l e 25}^{+ d i f f e r e n c e t o p e r c e n t i l e 75}$ )（或平均每千瓦时约 ${10.5}_{- 3.8}^{+ 0.5} g$ CO ₂ -当量碳足迹），但预计到 2050 年，这一比例将主要由光伏制造排放减少的驱动下增加约 7 倍。值得注意的是，优化土地利用策略可以潜在地提升现有美国光伏电站承载生态系统的碳密度，增幅约为 ${3.0}_{- 0.4}^{+ 3.7} k g C m^{- 2}$ ，从而促进这些美国光伏电站碳足迹平均减少 ${4.3}_{- 0.2}^{+ 9.3} %$ 。

You have full access to this article via University of Toronto Collections Development Department.
您可以通过多伦多大学馆藏发展部完全访问本文。

Download PDF 下载 PDF

A global inventory of photovoltaic solar energy generating units
光伏太阳能发电单元的全球清单

Article 27 October 2021
2021 年 10 月 27 日

Global terrestrial carbon fluxes of 1999–2019 estimated by upscaling eddy covariance data with a random forest
1999-2019 年全球陆地碳通量通过升级涡度协方差数据与随机森林估算

Article Open access 开放获取文章 24 September 2020 2020 年 9 月 24 日

A countrywide analysis of 27 solar power plants installed at different climates
一项全国范围内的分析，涉及在不同气候条件下安装的 27 座太阳能电站

Article Open access 开放获取文章 14 January 2022 2022 年 1 月 14 日

Main 主要

Solar photovoltaic (PV) technology is one of the world’s most promising renewable energy generation technologies¹. Global PV installed capacity increased from 0.8 GW in 2000 to 891.3 GW by 2021², and approximately 13,345 GW are projected to be in operation by 2050^3,4,5,6,7,8. Nevertheless, the booming PV technology^9,10, particularly in the domain of utility-scale installations, has concurrently imposed substantial pressures on land resources^9,11. Furthermore, the deployment of utility-scale photovoltaic (USPV) facilities alters the preexisting land-cover types, impacting the carbon (C) pool (including vegetation C and soil C) of the hosting ecosystem. These changes can subsequently affect the C footprint of the USPV plants^12,13.
太阳能光伏（PV）技术是世界上最具前景的可再生能源发电技术之一。全球光伏装机容量从 2000 年的 0.8 吉瓦增长到 2021 年的 891.3 吉瓦，预计到 2050 年将运营约 13345 吉瓦。然而，蓬勃发展的光伏技术，特别是在大型电站领域，同时对土地资源施加了巨大压力。此外，大型光伏（USPV）设施的建设改变了原有的土地覆盖类型，影响了宿主生态系统的碳（C）库（包括植被碳和土壤碳）。这些变化随后可能影响 USPV 植物的碳足迹。

Although recent years have witnessed an emerging focus on this issue, the predominant emphasis has been on the assessment of individual USPV plants with specific land-cover types¹⁴, or on a regional scenario analysis using integrated assessment model¹⁵. To our knowledge, a comprehensive evaluation covering global USPV plants sited on different land-cover types and with diverse land management strategies remains absent, which has resulted in unknown global patterns and massive uncertainty about the magnitude of the USPV effect on the C pool of the hosting ecosystem, as well as its role in the C footprint of USPV plants. Further complications arise from the fact that highly accurate worldwide USPV time series data are lacking, particularly concerning the installation dates of these facilities^{16,17,18,19,20,21}. This is because the key factors used to assess changes of the C pool and C footprint, including the PV panel efficiency, the emission intensity associated with PV panel manufacturing and the land-cover type both before and after the construction of USPV facilities, exhibit variations over time. Resolving these major gaps should improve our understanding of how the C pool is altered by USPV deployment, what role it plays in the C footprint of USPV plants and how the land of USPV plants should be managed to achieve mutual benefits in USPV development and ecosystem C pool enhancements.
尽管近年来人们开始关注这一问题，但主要精力集中于评估具有特定土地覆盖类型的单个美国光伏电站 ¹⁴ ，或使用综合评估模型进行区域情景分析 ¹⁵ 。据我们所知，目前尚缺乏对全球不同土地覆盖类型、采用多样化土地管理策略的美国光伏电站进行全面评估的研究，这导致全球模式未知，美国光伏电站对宿主生态系统碳库的影响程度及其在美国光伏电站碳足迹中的作用存在巨大不确定性。进一步复杂化的是，由于缺乏高精度的全球美国光伏时间序列数据，特别是这些设施的建设日期数据 ^{16,17,18,19,20,21} ，使得评估碳库和碳足迹变化的关键因素——包括光伏板效率、光伏板制造相关的排放强度以及美国光伏设施建设前后的土地覆盖类型——随时间变化而难以确定。解决这些主要差距将有助于我们更好地理解碳库如何因大型光伏系统部署而改变，它在大型光伏系统碳足迹中扮演什么角色，以及如何管理大型光伏系统的土地以实现大型光伏系统发展与生态系统碳库增强之间的互利共赢。

In this study, we leveraged a time series dataset of USPV plants to quantify the alterations in land cover and the subsequent changes in the C pool of the hosting ecosystem induced by USPV deployment and elucidate its role in the C footprint of USPV plants. Our dataset encompasses USPV plants worldwide and includes diverse historical data, particularly the installation date, thus allowing us to discern a global pattern in the response of C pool to the rapid expansion of USPV and its contribution to the C footprint of USPV plants, examine how it reacts to changes in climate, microclimate and technology levels, and offer insights for land management strategies of USPV plants.
在本研究中，我们利用了美国光伏电站的时间序列数据集，以量化土地覆盖的变化以及由美国光伏电站部署引起的宿主生态系统碳库的后续变化，并阐明其在美国光伏电站碳足迹中的作用。我们的数据集涵盖了全球的美国光伏电站，并包括多样化的历史数据，特别是安装日期，从而使我们能够揭示碳库对美国光伏电站快速扩张的响应模式及其对美国光伏电站碳足迹的贡献，考察其如何对气候变化、微气候和技术水平的变化做出反应，并为美国光伏电站的土地管理策略提供见解。

Land-cover changes after USPV deployment
美国光伏电站部署后的土地利用变化

Based on the global inventory of PV solar energy generating units proposed previously¹⁶, we developed a time series dataset for global USPV installations constructed from 2000 to 2018, leveraging remote sensing data and machine learning algorithms (Supplementary Discussion, section A). On the basis of the dataset, we found that cropland, particularly that growing C3 crops, emerged as the preferred site for the deployment of USPV plants with the exception of Africa (Fig. 1a,c,d), which aligns with the outcomes of prior studies¹⁶. However, a reduced use of cropland for USPV deployment was witnessed, particularly for USPV plants constructed after 2013. This reduction in cropland use can be attributed to the geographical shift in USPV deployment focus from Europe towards Asia and North America²² (Supplementary Discussion, section B), where croplands were utilized less frequently for USPV construction compared with Europe. In addition, food security considerations may also have a role, as the increased size of USPV plants (Supplementary Fig. 1) necessitates expansive land areas for their construction. By contrast, since 2013, the use of grassland, dominantly C3 grassland, for USPV construction has increased by approximately 5% compared with prior years.
基于先前提出的全球光伏发电单元清单 ¹⁶ ，我们利用遥感数据和机器学习算法，构建了一个 2000 年至 2018 年全球美国光伏电站部署的时间序列数据集（补充讨论，A 部分）。基于该数据集，我们发现耕地，尤其是种植 C3 作物的耕地，成为美国光伏电站部署的首选地点，除非洲外（图 1a,c,d），这与先前研究的结果 ¹⁶ 一致。然而，我们观察到耕地在美国光伏电站部署中的使用减少，尤其是在 2013 年后建设的美式光伏电站。耕地使用的减少可归因于美国光伏电站部署重点从欧洲转向亚洲和北美的地理变化 ²² （补充讨论，B 部分），与欧洲相比，亚洲和北美的耕地在使用频率上较少用于美国光伏电站建设。此外，粮食安全问题也可能发挥作用，因为美国光伏电站规模的增加（补充图 1）需要更广阔的土地面积进行建设。相比之下，自 2013 年以来，用于美国光伏电站建设的草地（主要是 C3 草地）使用量比前几年增加了约 5%。

Fig. 1: Land-cover patterns before and after USPV deployment.
图 1：美国光伏电站部署前后的土地利用模式。

Several land management decisions may be implemented after the construction of USPV plants, including land clearing²³, which removes the existing vegetation and covers the ground with gravel, potentially leading to a gradual loss of belowground C; maintaining the original land cover¹⁵; and converting to another land-cover type, which is usually grassland²⁴. We found that, on average, over 77% of the land encompassing USPV plants was predominantly managed as grassland after construction, with a continuously increased proportion of approximately 15% from 2000 to 2018 (Fig. 1b,e). The adoption of land clearing regimes showed a decreasing trend overall but at high variability between 6% and 25%. Spatially, Europe exhibited the highest percentage of USPV plants managing their grounds as grassland at approximately 92%, followed by North America, Asia and Australia (Fig. 1f). However, over 45% and 40% of the grounds of USPV plants installed in Africa and South America were managed as bare land, respectively.
在美国家用光伏电站建设完成后，可能会实施几种土地管理决策，包括土地清理 ²³ ，这会移除现有植被并用砾石覆盖地面，可能导致地下碳逐渐流失；维持原始土地覆盖 ¹⁵ ；以及转换为另一种土地覆盖类型，通常是草地 ²⁴ 。我们发现，平均而言，超过 77%的美国家用光伏电站所涵盖的土地在建设后主要以草地为主管理，从 2000 年到 2018 年，其比例持续增加，约增加了 15%（图 1b,e）。采用土地清理制度总体呈下降趋势，但在 6%到 25%之间具有高度变异性。从空间上看，欧洲的美国家用光伏电站以约 92%的比例管理其土地为草地，其次是北美、亚洲和澳大利亚（图 1f）。然而，非洲和南美洲安装的美国家用光伏电站的土地中，分别有超过 45%和 40%被管理为裸地。

These stark regional differences in management practices reflect varying approaches to USPV land stewardship. The high prevalence of grassland management in Europe demonstrates successful integration of ecosystem preservation with solar deployment, while the predominance of bare land management in Africa and South America suggests opportunities for improved land management strategies. Moreover, there was notable spatiotemporal heterogeneity of land-cover changes (Supplementary Fig. 2). For example, more than 90% of the croplands encompassing USPV plants were converted to grasslands in Europe, while approximately 20% of the croplands were changed to bare land in Asia. Potential explanations of this regional discrepancy include differences in climatic conditions, the availability of land types and local policies^{16,25,26,27,28}.
这些在管理实践上的显著区域差异反映了美国光伏土地管理的不同方法。欧洲草地管理的普遍性表明了生态系统保护与太阳能部署的成功整合，而非洲和南美洲裸地管理的优势则暗示了改进土地管理策略的机会。此外，土地覆盖变化存在明显的时空异质性（补充图 2）。例如，在欧洲，包含美国光伏植物的大部分耕地（超过 90%）被转化为草地，而在亚洲，约 20%的耕地被改造成了裸地。这种区域差异的潜在解释包括气候条件差异、土地类型可用性和地方政策的差异。

C pool alterations from land-cover changes
陆地覆盖变化引起的碳库变化

The land-cover changes ensuing from the deployment of USPV plants induced alterations in the C pool of the hosting ecosystems. We estimated a global C pool change of ${2.1}_{- 2.3}^{+ 0.6} T g C$ ( $e n s e m b l e$ ${m e a n}_{- d i f f e r e n c e t o p e r c e n t i l e 25}^{+ d i f f e r e n c e t o p e r c e n t i l e 75}$ ) from the hosting ecosystems induced by USPV installations constructed between 2000 and 2018, with apparent individual differences among USPV plants (Fig. 2a). Spatial analysis revealed that USPV plants sited in Europe were the predominant contributors to the C gain, amounting to ${2.1}_{- 1.5}^{+ 0.3} T g C$ , followed by the installations in Asia. By contrast, a loss in the C pool was observed for USPV plants in the other continents, particularly North America by ${- 0.6}_{- 0.3}^{+ 0.1} T g C$ . The C gain arose mainly from USPV plants that were sited on C3 cropland, which were subsequently converted to C3 grassland after construction (Fig. 2b), owing to the relatively large difference of C density between C3 croplands and C3 grasslands hosting USPV plants (Supplementary Fig. 3). By contrast, USPV plants sited on C3 grassland and C3 cropland whose grounds were managed as bare land dominated the C loss, attributed to the relatively large plant area and C density difference, respectively. Notably, the source of the C loss represented apparent variability at continent scale. USPV plants sited on C3 grassland resulted in the majority of C loss in Asia, while those located on forest area and shrubland dominated the C loss in Europe and North America, respectively (Supplementary Fig. 4).
由美国光伏电站的部署所引发的土地利用变化导致宿主生态系统的碳库发生改变。我们估计了 2000 年至 2018 年间由美国光伏电站建设所引起的宿主生态系统的全球碳库变化为 ${2.1}_{- 2.3}^{+ 0.6} T g C$ ( $e n s e m b l e$ ${m e a n}_{- d i f f e r e n c e t o p e r c e n t i l e 25}^{+ d i f f e r e n c e t o p e r c e n t i l e 75}$ )，其中美国光伏电站之间存在明显的个体差异（图 2a）。空间分析表明，位于欧洲的美国光伏电站是碳增加的主要贡献者，达到 ${2.1}_{- 1.5}^{+ 0.3} T g C$ ，其次是亚洲的电站。相比之下，其他大洲的美国光伏电站碳库出现了减少，特别是在北美地区减少了 ${- 0.6}_{- 0.3}^{+ 0.1} T g C$ 。碳增加主要来自位于 C3 农田上的美国光伏电站，这些电站建设后转变为 C3 草地（图 2b），这是由于 C3 农田和 C3 草地承载的美国光伏电站之间的碳密度差异相对较大（补充图 3）。相比之下，位于 C3 草地和 C3 农田且地面管理为裸地的美国光伏电站主导了碳的减少，这归因于相对较大的植物面积和碳密度差异。值得注意的是，C 损失的来源表现出大陆尺度的明显变化。位于 C3 草地上的美国光伏电站导致了亚洲大部分的 C 损失，而位于森林区域和灌丛地带的光伏电站则分别主导了欧洲和北美的 C 损失（补充图 4）。

**Fig. 2: C pool changes induced by USPV deployment.**

Temporally, the C gain peaked at 0.42 TgC for USPV plants constructed in 2011, and showed a sudden decrease for those constructed in 2013, subsequently experiencing a continuous increase thereafter (Fig. 2c). The sudden decrease in the C gain was attributable to fewer USPV plants constructed in 2013, and most importantly the geographical shift in USPV deployment focus from Europe towards Asia and North America²² (Supplementary Discussion, section B). The land of the majority of USPV plants sited in Europe was maintained as grassland after construction, while in Asia and North America, more than 20% of the USPV plants led to the conversion of their grounds into bare land (Supplementary Fig. 2). Fortunately, the proportion of USPV plants managing their grounds as bare land showed a decreasing trend since 2014 (Fig. 1e), which can explain the continuous increase in the C gain from 2014 to 2017.

Although the overall impact of USPV deployments on the ecosystem C pool, quantified herein as a C gain of 2.1 TgC, may currently seem modest owing to the relatively limited spatial footprint of USPV plants constructed between 2000 and 2018 (<4,000 km²), it was projected that approximately 54,737 km² of land will be occupied for USPV deployment by 2050 (Supplementary Discussion, section C), potentially leading to a considerable alteration of the terrestrial C pool¹³. Thus, we suggest that the impact of USPV deployment on land cover and the subsequent C pool of the hosting ecosystem should be carefully considered in PV planning and management^25,29,30.

Effect of C pool change on C footprint of USPV

The C footprint of global USPV plants, defined as the greenhouse gas emissions per unit of power generation from a life cycle perspective, considering the contribution from C pool changes (ranging from ${- 76.3}_{- 26.5}^{+ 7.5}$ to ${94.0}_{- 30.0}^{+ 12.3}$ grams of CO₂ equivalent (CO₂e) per kilowatt-hour, with an average absolute value of ${10.5}_{- 3.8}^{+ 0.5} g$ CO₂e kWh⁻¹) (Supplementary Fig. 5a), spanned from ${- 7.8}_{- 7.4}^{+ 25.6}$ to ${182.7}_{- 29.6}^{+ 31.9} g$ CO₂e kWh⁻¹, with a mean value of approximately ${56.3}_{- 0.1}^{+ 3.4} g$ CO₂e kWh⁻¹ (Supplementary Fig. 5b). The impact of changes in the C pool on the C footprint of USPV plants exhibited notable variability across individual plants, with a range extending from ${- 112.9}_{- 4.5}^{+ 48.2} %$ to ${106.0}_{- 19.9}^{+ 42.4} %$ (Fig. 3a), and an average absolute impact of ${15.9}_{- 5.8}^{+ 1.0} %$ (Fig. 3b). In particular, USPV plants in Europe and North America were more sensitive to C pool alterations, with an average absolute impact (median value) reaching up to ${14.3}_{- 0.8}^{+ 6.0} %$ and ${11.6}_{- 1.5}^{+ 6.6} %$ , respectively. Therefore, ignoring the contributions from changes in the C pool can result in substantial misestimates of the C footprint of USPV plants.

**Fig. 3: Effect of C pool changes on C footprint of USPV plants.**

Moreover, we performed a scenario analysis to reveal the impacts of technology innovation, which would lead to a reduction in the emission intensity of USPV construction, and the C pool alteration caused by climate change³¹ on the role of PV deployment induced C pool changes in the USPV C footprint among all the current USPV plants. Predicated on the anticipated reduction in emission intensity of USPV installations themselves (Supplementary Discussion, section D)³² and the predicted C density in the future (Supplementary Methods, section A), we estimated that the absolute impact of changes in the C pool on the C footprint of USPV plants will be amplified by around 1-, 2- and 7-fold by 2030, 2040 and 2050, respectively (Fig. 3c). We further investigated the contributions of technology innovation and climate-change-induced C pool alteration to the overall impact and found that the amplified impact was predominantly attributed to technology innovation rather than climate-change-induced C pool changes (Supplementary Fig. 6).

In addition, we compared the emissions or emission reductions induced by C pool changes due to USPV deployment with the emissions avoided by replacing fossil energy with PV power generation. Although the emissions or emission reductions induced by C pool changes were currently modest compared with the avoided emissions (an absolute ratio of approximately 1.9%), this figure is expected to increase continuously and rapidly as the electricity mix continues to decarbonize (Supplementary Discussion, section E), indicating that nature-based solution would become imperative to achieve climate targets³³.

Implication for USPV land management

Given that C pool changes exert a critical role in the C footprint of existing USPV plants and they are induced by land-cover changes, we next conducted a scenario analysis to investigate the potential impacts of different land management strategies on the C pool of the hosting ecosystem and, subsequently, on the C footprint of these existing USPV plants. To do that, four hypothetical scenarios were designed to alter the current land management strategies of these global USPV plants (Methods). The findings revealed a substantial C gain totalling ${4.6}_{- 4.3}^{+ 1.8}$ and ${10.6}_{- 1.3}^{+ 13.1} T g C$ (or ${1.3}_{- 1.2}^{+ 0.5}$ and ${3.0}_{- 0.4}^{+ 3.7} k g C m^{- 2}$ ) under S2 and S3, which were 5-fold and twice that under the current baseline, respectively (Fig. 4a). By contrast, under S1 and S4, a potential C loss of ${2.8}_{- 1.1}^{+ 0.9}$ and ${20.7}_{- 0.8}^{+ 5.5} T g C$ (or ${0.8}_{- 0.3}^{+ 0.2}$ and ${5.8}_{- 0.2}^{+ 1.6} k g C m^{- 2}$ ) was estimated, and it was estimated that 53% of global USPV plants could achieve C pool enhancements by optimizing land management strategies according to the implication from the S3 scenario. Spatially, C loss predominantly originated from USPV plants sited in North America, Europe and Asia under S1 (Supplementary Fig. 7a), while approximately 80% of the C loss was induced by USPV plants located in Asia and Europe under S4 (Supplementary Fig. 7d). By contrast, USPV plants sited in Asia and Europe dominated the C gain by more than 97% and 88% under S2 and S3, respectively (Supplementary Fig. 7b,c).

**Fig. 4: C pool and C footprint changes under different land management scenarios.**

An estimated average C footprint of ${62.7}_{- 0.9}^{+ 0.3}$ and ${73.1}_{- 0.6}^{+ 0.4} g$ CO₂e kWh⁻¹ was obtained for global USPV plants under S1 and S4, respectively (Supplementary Fig. 8a,d), which were ${11.5}_{- 7.4}^{+ 8.7} %$ and ${30.0}_{- 13.7}^{+ 11.7} %$ higher than that of the current baseline (Fig. 4b). Notably, USPV plants in Europe were apparently more sensitive to C pool alterations under S1 and S4, with the C footprint increased by more than 20% and 50%, respectively. The potential explanation included the high C density of the hosting ecosystem in the USPV plants sited in Europe owing to managing the grounds as grassland after construction, and the relatively low C emission intensity associated with the manufacturing of USPV plant components in Europe (Supplementary Fig. 9). This observation further underscores the imperative of integrating C pool considerations into the comprehensive assessment of the C footprints of USPV plants, in light of the anticipated continual reduction in C emissions from USPV installations themselves^32,34. By contrast, under S2 and S3, the estimated average C footprints of global USPV plants were ${55.1}_{- 0.6}^{+ 4.3}$ and ${53.6}_{- 5.3}^{+ 1.0} g$ CO₂eq kWh⁻¹, respectively (Supplementary Fig. 8b,c), a reduction of ${2.0}_{- 0.9}^{+ 0.1} %$ and ${4.3}_{- 0.2}^{+ 9.3} %$ from the current baseline, with the most pronounced plant-scale decrease exceeding 100%. USPV plants sited in Asia, South America and Africa exhibited the greatest potential for C footprint mitigation, attributed primarily to the relatively high proportion of USPV plants currently managing their grounds as bare land in these continents (Fig. 1f).

Climate change may disturb the terrestrial C pool as well as affect the power generation potential of PV systems. Although we did not investigate this aspect in detail in our C pool impact estimation and C footprint accounting because it remains outside the scope of this study, it deserves attention and further investigation because climate change can exert a substantial influence on the C pool^35,36,37,38 and can moderately affect PV power outputs³⁹. In addition, the potential indirect land-use changes induced by USPV management decisions and their subsequent impact on the C pool warrant further investigation¹⁵.

Sensitivity to microclimate impact

Given that the existence of PV panels can alter the local albedo, affect the energy balance and further result in microclimate changes⁴⁰, we carried out a sensitivity analysis to preliminarily reveal the impact of microclimate changes on our results. The sensitivity analysis comprised two components: an albedo-based analysis and a ‘whole scene’ assessment (Methods). The results presented here were estimated using the observation-based dataset. However, the results calculated utilizing the model-derived C density datasets can also be found in Supplementary Fig. 10 and the ‘Data availability’ section. For the albedo-based analysis, we found that, on average, the C pool of the hosting ecosystem after construction would be enhanced by ~16% to ~34%, primarily due to temperature decreases induced by the effective albedo increase estimated for most USPV plants (Supplementary Fig. 10a). An estimated average increase in C pool changes of ~30% to ~101% was then obtained (Supplementary Fig. 10b), which subsequently led to a reduction in the C footprint of USPV plants, spanning from approximately 4% to 12% (Supplementary Fig. 10c). Regarding the whole scene assessment, we found that, in general, lower temperatures and higher precipitation would lead to an increase in C pool of the hosting ecosystem after construction, with the highest C gain surpassing 60% (Supplementary Fig. 11a). By contrast, higher temperatures and lower precipitation would result in a C loss of up to approximately −40%. As a result, USPV-deployment-induced C pool changes would vary widely, ranging from −141.2% to 208.1% under different scenarios (Supplementary Fig. 11b). Correspondingly, the C footprints of USPV plants were projected to change by −26.2% to 12.7% (Supplementary Fig. 11c).

The results indicated that the C gain of the hosting ecosystem and the C footprint reduction of the USPV plants may be more substantial if microclimate impacts are considered. However, given the complexity of microclimate impacts exerted by PV panels^24,41,42,43, which cannot be fully captured by the sensitivity analysis conducted in this study^44,45,46, it is suggested that more studies on microclimate changes in response to PV deployment should be carried out across PV plants located in diverse ecosystems and climate zones to bridge the existing knowledge gap⁴⁷.

Our study pioneers the evaluation of the impact that global USPV deployment has on the C pool of the hosting ecosystem, revealing its importance in the C footprint of USPV plants, although uncertainties remain (Supplementary Discussion, section F). Stakeholders, developers and policymakers can utilize our results to achieve a mutually beneficial outcome, simultaneously realizing C pool enhancements and reducing the C footprint of USPV plants.

Methods

Developing a time series dataset

We developed a time series dataset of global USPV installations leveraging remote sensing data and machine learning algorithms (see the overall flow chart in Supplementary Fig. 12). This dataset encapsulates key details such as the construction date of each USPV plant, along with installed capacity, panel efficiency and ground coverage ratio.

A state-of-the-art global inventory of solar PV energy generating units that contains all the commercial, industrial and USPV installations around the world deployed before 2018 was used as the basic data¹⁶. We selected 37,106 out of 68,661 generating units from the inventory according to the definition of USPV from International Energy Agency (IEA), that is, greater than 1 MW nameplate capacity⁴⁸. Then, spatial clustering was performed to amalgamate the generating units referring to the same plant using a neighbourhood search distance of 400 m (ref. ²¹). This process resulted in 28,769 USPV plants used for installation date identification.

Landsat 5 and Landsat 8 atmospherically corrected surface reflectance data with 30-m resolution^49,50 served as the primary inputs for implementing the temporal cluster matching (TCM) algorithm^20,51 coupled with the Pettitt change-point detection algorithm⁵² to identify the construction date of each USPV plant. The temporal coverage of Landsat 5 and Landsat 8 remote sensing data spans from 1984 to the present, with a gap between May 2012 and February 2013, when Landsat 5 was decommissioned and Landsat 8 was not yet in operation. The decision against utilizing Landsat 7 data to bridge this gap stems from the scan line corrector malfunction in May 2003, which substantially compromised the utility of the data post-failure⁵³. Consequently, the absence of remote sensing data during this interim period may influence the identification accuracy of the construction date from May 2012 to February 2013, leading to the exclusion of this period from our comprehensive analysis.

The TCM algorithm, initially developed by the Microsoft AI for Good Research Lab in collaboration with Stanford RegLab^20,51, is predicated on the observation that the construction of a PV plant alters the colour and texture characteristics of the land, making the PV plant’s footprint distinct from its surroundings. Utilizing k-means clustering, the algorithm partitions pixels within the PV plant area and its surroundings into k clusters, subsequently calculating the discrete distribution of cluster indices for both the PV plant and its surrounding area separately. Kullback–Leibler (KL) divergence is used to determine the extent to which the colour and texture characteristics within the PV plant area diverge from those of the surrounding landscape. A higher KL-divergence value suggests a greater likelihood that a PV plant has been constructed. The integration of the Pettitt change-point detection algorithm with the TCM algorithm was undertaken to identify the construction dates of USPV plants instead of using the decision threshold method originally used in the TCM algorithm to identify PV construction based on whether the KL-divergence value exceeds the threshold. The rationale behind this modification lies in the observation that the decision threshold exhibits variability across diverse PV plants⁵², making it challenging to accurately ascertain the construction date of USPV plants worldwide using a single threshold. The Pettitt change-point detection algorithm, by contrast, offers a more generalized approach that accommodates the heterogeneity of PV plants, circumventing the subjective determination of a decision threshold, thereby enhancing the accuracy of construction date identification across the global spectrum of USPV plants.

The installed capacity of each USPV plant was estimated as¹⁶

C_{A C} = A \times G T I \times G C R \times η \div P V O U T \div I L R,

(1)

where C_AC is the installed capacity of a USPV plant, A is the area of the USPV plant, GTI is the incident irradiance at the optimal tilt angle of the USPV plant location, which is obtained from the Global Solar Atlas⁵⁴, GCR is the ground coverage ratio, $η$ is the panel efficiency, PVOUT is the PV solar energy production intensity, which is also obtained from the Global Solar Atlas, and ILR is the inverter loading ratio.

By utilizing our identification of the construction dates for USPV plants, we incorporated time-varying values of GCR, $η$ and ILR into equation (1), as opposed to using constant values. GCR was calculated on the basis of the latitude and the mounting types used⁵⁵. Given the unavailability of specific mounting type data for each USPV plant, we relied on the global market distribution of PV mounting types, which encompasses both fixed-tilt and tracking axis⁵⁶. The annual variation in $η$ for various panel types was obtained from the literature^57,58,59. Similar to the mounting type, the exact panel types utilized in each USPV plant were not directly accessible. Thus, we applied the global market shares of different PV panel types⁶⁰, including monocrystalline silicon (Mono-Si), polycrystalline silicon (Poly-Si), amorphous silicon (a-Si), cadmium telluride (CdTe) and copper indium (gallium) selenide (CI(G)S), to compute a weighted average $η$ for USPV plants constructed in different years. The data for the annual ILR were sourced from the Energy Information Administration⁶¹.

The estimation of land-cover changes resulting from the deployment of USPV plants was conducted leveraging the identified construction dates of these plants. The European Space Agency (ESA) Climate Change Initiative 300-m-resolution land-cover data covering 1992–2020 were used to determine the preconstruction land-cover types⁶². To assess the dominant land-cover types present at the USPV plant locations after construction, the ESA WorldCover dataset for the year 2021, featuring a higher resolution of 10 m, was used⁶³. The enhanced resolution of the ESA WorldCover data facilitates the accurate extraction of the predominant land-cover type within the boundaries of the USPV plants.

The Google Earth Engine (GEE) platform was used to acquire and preprocess remote sensing data for our analysis, which included tasks such as image clipping, cloud cover filtering and image band extraction. GEE is a comprehensive platform designed for Earth science data and analysis at a planetary scale, which houses a vast catalogue of satellite imagery and geospatial datasets including the Landsat series datasets. More details can be found at https://earthengine.google.com. The calculation of the KL divergence for the USPV plants was also performed utilizing GEE, leveraging a corresponding Python package^64,65. The Pettitt change-point detection and subsequent postprocessing operations were conducted offline using Python, ArcGIS and other auxiliary software and tools.

Quantifying impacts on C pool

We leveraged several C density datasets to investigate C pool changes induced by worldwide USPV deployment, encompassing one observation-based and ten model-derived datasets (Supplementary Methods, section B, and Supplementary Table 1). In instances where some grid cells lack C density data for specific land-cover types present before or after USPV construction, owing to the relatively coarse spatial resolution of the C density datasets when compared to the scale of an individual USPV plant, an extrapolation method was used. This method estimates the C density for each land-cover type on the basis of the similarity of temperature and precipitation in the cell to those in the other cells with the same land-cover type containing C density data. A threshold of 1 °C for temperature and 50 mm for precipitation was utilized to identify cells with similar climatic conditions. The average C density value from these similar cells was then allocated to the cells lacking C density information. Data on the annual average temperature and total precipitation for each cell were sourced from the ERA5-land hourly dataset with a spatial resolution of approximately 0.1° (ref. ⁶⁶).

We calculated C pool changes induced by a USPV plant, denoted as $I_{e c o}$ , by multiplying the difference in C density before and after construction by the plant’s area (equation (2)). The C density difference was estimated by comparing the mean C density (averaged over multiple decades) of land-cover types before and after construction. This mean C density data were calculated individually on the basis of each of the 11 C density datasets.

I_{e c o} = A \times ({C D A}_{m e a n} - {C D B}_{m e a n}),

(2)

where CDA_mean is the mean C density of the hosting ecosystem within the boundary of the USPV plant after construction, CDB_mean is the mean C density of the hosting ecosystem within the boundary of the USPV plant before construction. Both CDA_mean and CDB_mean encompass the C density of both vegetation and soil for all land-cover types, except for cropland. For cropland, only the C density of soil is considered, predicated on the understanding that the majority of vegetation C in cropland returns to the atmosphere in the form of CO₂ within a short time period⁶⁷. CDA_mean and CDB_mean were derived from the previously mentioned C density datasets, based on the land-cover types identified before and after the construction of the USPV plant.

Accounting for USPV C footprint

The C footprint of each USPV plant was assessed leveraging the developed time series dataset. The system boundaries encompass emissions or emission reductions from PV panel manufacturing, the balance of system (BOS) and the hosting ecosystem. Emissions arising from transportation, operation and maintenance, and the end-of-life phase were excluded, because the contribution from these phases is comparatively minimal^68,69, and the majority of USPV plants are still in operation, leading to a lack of data regarding the end-of-life stage.

The emission from PV panel manufacturing for a specific USPV plant, denoted as E_p, is calculated on the basis of the emission intensity of PV panel manufacturing, the nominal capacity of PV modules, trade data of PV panels and the market share distribution of various PV panel types

E_{p} = C_{D C} \times \sum_{j = 1}^{m} (S_{{p a n e l}_{j}} \times \sum_{i = 1}^{n} (S_{{t r a d e}_{i}} \times {E F}_{i j}))

(3)

C_{D C} = C_{A C} \times I L R,

(4)

where C_DC is the nameplate capacity of the PV modules installed within the PV plant, $S_{{p a n e l}_{j}}$ is the market share of the panel type j, $S_{{t r a d e}_{i}}$ is the market share of PV panels exported from country or region i to the nation hosting the PV plant, inclusive of domestic supply, EF_ij is the emission intensity of panel type j produced in country or region i, m = 1, …, 5 denotes the enumeration of PV panel types, specifically Mono-Si, Poly-Si, a-Si, CdTe and CI(G)S, and n denotes the enumeration of countries or regions producing PV panels.

We compiled a year-specific country-to-country trade look-up table for PV panels using data from the International Trade Centre Trade Map⁷⁰ to obtain country-specific market share information for PV panels. This approach accounted for the trade matrix among countries, rather than relying on an overall market share that overlooked country-specific differences. In addition, the proportion of domestic PV panel supply was derived from a report by the IEA Photovoltaic Power Systems Programme⁷¹. These datasets were amalgamated to calculate the year- and country-specific $S_{{t r a d e}_{i}}$ . The emission intensities of the various types of PV panel manufactured in distinct countries or regions were taken from the existing literature and the ecoinvent database (Supplementary Table 2). For panel types supported by ample data, regression analyses were conducted to obtain the temporal evolution of emission intensity for our analysis. By contrast, for panel types without sufficient data (for example, only 1 or 2 years of data available), a default reduction rate in emission intensity was used (a 45% decrease from 2011 to 2020), as documented in a special report by the IEA³².

The emissions from the BOS, denoted as E_BOS, encompass emissions from the mounting system, emissions related to electric installations (excluding inverters) and emissions from inverters.

E_{B O S} = E_{m} + E_{e i} + E_{i n v e r t e r}

(5)

E_{m} = A \times G C R \times {E F}_{m}

(6)

E_{e i} = C_{D C} \div 570 \times {E F}_{e i}

(7)

E_{i n v e r t e r} = C_{A C} \div 500 \times 2 \times {E F}_{i n v e r t e r},

(8)

where E_m represents the emissions from the mounting system, E_ei denotes the emissions from the electric installation and E_inverter represents the emissions from the inverters utilized within the PV plant. Correspondingly, EF_m is the emission intensity for the mounting system, EF_ei denotes the emissions from the electric installation component for a module with a capacity of 570 kWp and EF_inverter denotes the emissions from an inverter with a capacity of 500 kW. The factor of 2 is introduced to account for the 15-year lifespan of an inverter, which is half of the lifespan of a PV plant. All emission intensity figures used in the calculation of the BOS emissions were sourced from the ecoinvent database.

The power output of a specific PV plant at time t, denoted as $P_{o u t} (t)$ , was calculated by the PV power generation potential (PV_pot) and the installed capacity of the plant⁷² as

P_{o u t} (t) = {P V}_{p o t} (t) \times C_{A C}

(9)

{P V}_{p o t} (t) = P_{R} (t) \frac{R S D S (t)}{{R S D S}_{S T C}}

(10)

P_{R} (t) = 1 + γ [T_{c e l l} (t) - T_{S T C}]

(11)

T_{c e l l} (t) = c_{1} + c_{2} T A S (t) + c_{3} R S D S (t) + c_{4} V W S (t)

(12)

where ${P V}_{p o t} (t)$ is the PV power generation potential of the PV plant at time t, $P_{R} (t)$ is the performance ratio, encapsulating the effects of temperature on the efficiency of PV cells, $R S D S (t)$ is the surface-downwelling shortwave irradiance at time t, ${R S D S}_{S T C}$ is the irradiance under standard test conditions (1,000 W m⁻²), $T_{c e l l} (t)$ is the temperature of the PV cell at time t, $T_{S T C}$ is the temperature under standard test conditions (25 °C), $T A S (t)$ is the surface air temperature at time t and $V W S (t)$ is the surface wind velocity at time t. The variables $γ$ , $c_{1}$ , $c_{2}$ , $c_{3}$ and $c_{4}$ are PV panel type-specific factors sourced from the existing literature^57,72.

The hourly data for TAS, RSDS and VWS were acquired from the ERA5-Land dataset, which is a widely used reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts with a spatial resolution of approximately 11 km (ref. ⁷³). The hourly power output for each USPV plant, spanning from 1991 to 2020 for each type of PV panel, was calculated utilizing equation (9), and the market shares of the different PV panel types were used to compute a weighted average power output. This output was then aggregated to reflect the total power generated over the plant’s operational lifespan. Moreover, a degradation rate of 0.7% per year was incorporated into the power output calculations across the 30-year lifespan of the USPV plant⁷⁴.

Finally, the C footprint of a USPV plant, denoted as F_c, was calculated as

F_{c} = \frac{E_{p} + E_{B O S} + E_{e c o}}{P_{o u t}}

(13)

E_{e c o} = - 3.67 \times I_{e c o},

(14)

where E_eco is the emission from alterations in C pool induced by USPV deployment. The coefficient of −3.67 converts changes in C pool into equivalent releases or sequestration of CO₂.

Designing land management strategies

Four hypothetical scenarios were designed to alter the current land management strategies of the existing global USPV plants. These scenarios aimed to explore the potential impacts of different land management strategies on the C pool of the hosting ecosystem and, subsequently, on the C footprint of these USPV plants. The first (all croplands postconstruction) scenario (S1) assumes that the grounds of all these USPV plants are managed as cropland. The second (all grasslands postconstruction) scenario (S2) assumes that all these USPV plants manage their grounds as grassland. The third (optimal) scenario (S3) assumes that all these USPV plants manage their grounds with land-cover types optimized for the highest C density potential based on local environmental conditions. The fourth (worst case) scenario (S4) assumes that the grounds of all these USPV plants are managed as bare land.

Analysing the impact of microclimate changes

Given that USPV deployment can result in microclimate changes but with large uncertainties^24,41,42,43, we performed a sensitivity analysis to reveal the potential impact of USPV-induced microclimate changes on our results. The sensitivity analysis comprised two components: an albedo-based analysis and a whole scene assessment. For the albedo-based analysis, we selected surface air temperature as a proxy for microclimate impact estimation, because temperature was proved to be most affected by USPV construction attributed to effective albedo change (reflectivity of PV panel plus solar conversion efficiency⁷⁵)^45,46. In addition, other microclimate impacts, such as reduced evaporation caused by the PV panels, could also be partially reflected through changes in temperature. A total of four scenarios were set, including ±0.5, ±1, ±1.5 and ±2 °C after construction, of which the direction of temperature change was determined by the direction of change in the effective albedo. The effective albedo of the USPV area was deemed to decrease for USPV plants sited on bare land, while the opposite is true for those sited on the area with the other land-cover types⁷⁶. The decrease in effective albedo will lead to an increase in surface temperature because more energy will be absorbed by the ground, while the increase in effective albedo will result in a cooling effect. Then, the C density of the hosting ecosystem of each USPV plant after construction was estimated leveraging the extrapolation method mentioned before based on the four scenarios, and subsequently, C pool changes, the C footprint and the contribution of C pool change to the C footprint were calculated. Given the complexity of the microclimate impacts of USPV plants, which affect not only temperature but also other meteorological variables, particularly precipitation, we conducted a whole scene assessment. This assessment considered the effects of changes in both temperature and precipitation, as well as their combined impacts. Specifically, we set up four temperature change scenarios (±0.5, ±1, ±1.5 and ±2 °C) and four precipitation change scenarios (±20%, ±30%, ±40% and ±50%). Unlike the albedo-based analysis, the whole scene assessment did not constrain the direction of temperature and precipitation changes, aiming to provide a comprehensive understanding of the potential influences of these changes on our results, even though some scenarios may not occur in reality.

Data availability

The global inventory of solar PV energy generating units utilized as basic data in this study is available via Zenodo at https://doi.org/10.5281/ZENODO.5005868 (ref. ⁷⁷). The Landsat 5 and Landsat 8 atmospherically corrected surface reflectance data are available at https://www.usgs.gov/landsat-missions or in the Google Earth Engine (GEE) repository. The 300-m-resolution land-cover data are available at http://maps.elie.ucl.ac.be/CCI/viewer/download.php. The 10-m-resolution land-cover data are available at https://esa-worldcover.org/en or in the GEE repository. The observation-based C density data of soil are available via Zenodo at https://doi.org/10.5281/ZENODO.2536040 (ref. ⁷⁸). The observation-based C density data of vegetation are available at https://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=1763 (ref. ⁷⁹). The C density data derived from models included in the Trendy dataset are available at https://mdosullivan.github.io/GCB. The ERA5-Land hourly reanalysis data are available at https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land?tab=overview (ref. ⁸⁰) or in the GEE repository. The time series dataset developed in this study and the relevant data are available via Figshare at https://doi.org/10.6084/m9.figshare.28328765 (ref. ⁸¹). Source data are provided with this paper.

Code availability

The source code used in this study is available via Figshare at https://doi.org/10.6084/m9.figshare.28328810 (ref. ⁸²).

References

Renewables 2023 – Analysis (IEA, 2024); https://www.iea.org/reports/renewables-2023
Renewables 2022 (IEA, 2022); https://www.iea.org/reports/renewables-2022
Chen, S. et al. Deploying solar photovoltaic energy first in carbon-intensive regions brings gigatons more carbon mitigations to 2060. Commun. Earth Environ. 4, 369 (2023).
Google Scholar
Wang, S. et al. Future demand for electricity generation materials under different climate mitigation scenarios. Joule 7, 309–332 (2023).
CAS Google Scholar
Future of Solar Photovoltaic: Deployment, Investment, Technology, Grid Integration and Socio-Economic Aspects (International Renewable Energy Agency, 2019).
Global Renewables Outlook: Energy Transformation 2050 (International Renewable Energy Agency, 2020).
Net Zero by 2050 – Analysis (IEA, 2021); https://www.iea.org/reports/net-zero-by-2050
World Energy Outlook 2022 – Analysis (IEA, 2022); https://www.iea.org/reports/world-energy-outlook-2022
Zhang, N. et al. Booming solar energy is encroaching on cropland. Nat. Geosci. 16, 932–934 (2023).
CAS Google Scholar
Snapshot 2023 (IEA-PVPS, 2023); https://iea-pvps.org/snapshot-reports/snapshot-2023/
Grodsky, S. M. & Hernandez, R. R. Reduced ecosystem services of desert plants from ground-mounted solar energy development. Nat. Sustain. 3, 1036–1043 (2020).
Google Scholar
Hernandez, R. R., Hoffacker, M. K. & Field, C. B. Efficient use of land to meet sustainable energy needs. Nat. Clim. Change 5, 353–358 (2015).
Google Scholar
Hernandez, R. R., Hoffacker, M. K., Murphy-Mariscal, M. L., Wu, G. C. & Allen, M. F. Solar energy development impacts on land cover change and protected areas. Proc. Natl Acad. Sci. USA 112, 13579–13584 (2015).
CAS Google Scholar
Zhang, B., Zhang, R., Li, Y., Wang, S. & Xing, F. Ignoring the effects of photovoltaic array deployment on greenhouse gas emissions may lead to overestimation of the contribution of photovoltaic power generation to greenhouse gas reduction. Environ. Sci. Technol. 57, 4241–4252 (2023).
CAS Google Scholar
van de Ven, D. et al. The potential land requirements and related land use change emissions of solar energy. Sci. Rep. 11, 2907 (2021).
Google Scholar
Kruitwagen, L. et al. A global inventory of photovoltaic solar energy generating units. Nature 598, 604–610 (2021).
CAS Google Scholar
Yu, J., Wang, Z., Majumdar, A. & Rajagopal, R. DeepSolar: a machine learning framework to efficiently construct a solar deployment database in the United States. Joule 2, 2605–2617 (2018).
Google Scholar
Zhang, X., Xu, M., Wang, S., Huang, Y. & Xie, Z. Mapping photovoltaic power plants in China using Landsat, random forest, and Google Earth Engine. Earth Syst. Sci. Data 14, 3743–3755 (2022).
Google Scholar
Stowell, D. et al. A harmonised, high-coverage, open dataset of solar photovoltaic installations in the UK. Sci. Data 7, 394 (2020).
Google Scholar
Ortiz, A. et al. An artificial intelligence dataset for solar energy locations in India. Sci. Data 9, 497 (2022).
Google Scholar
Dunnett, S., Sorichetta, A., Taylor, G. & Eigenbrod, F. Harmonised global datasets of wind and solar farm locations and power. Sci. Data 7, 130 (2020).
Google Scholar
Snapshot 2022 (IEA-PVPS, 2022); https://iea-pvps.org/snapshot-reports/snapshot-2022/
Beatty, B., Macknick, J., McCall, J., Braus, G. & Buckner, D. Native Vegetation Performance under a Solar PV Array at the National Wind Technology Center (NREL, 2017).
Armstrong, A., Ostle, N. J. & Whitaker, J. Solar park microclimate and vegetation management effects on grassland carbon cycling. Environ. Res. Lett. 11, 074016 (2016).
Google Scholar
Dunnett, S., Holland, R. A., Taylor, G. & Eigenbrod, F. Predicted wind and solar energy expansion has minimal overlap with multiple conservation priorities across global regions. Proc. Natl Acad. Sci. USA 119, e2104764119 (2022).
CAS Google Scholar
Bosmans, J. et al. Determinants of the distribution of utility-scale photovoltaic power facilities across the globe. Environ. Res. Lett. 17, 114006 (2022).
Google Scholar
Balta-Ozkan, N., Yildirim, J., Connor, P. M., Truckell, I. & Hart, P. Energy transition at local level: analyzing the role of peer effects and socio-economic factors on UK solar photovoltaic deployment. Energy Policy 148, 112004 (2021).
Google Scholar
Thormeyer, C., Sasse, J.-P. & Trutnevyte, E. Spatially-explicit models should consider real-world diffusion of renewable electricity: solar PV example in Switzerland. Renew. Energy 145, 363–374 (2020).
Google Scholar
Hernandez, R. R. et al. Techno–ecological synergies of solar energy for global sustainability. Nat. Sustain. 2, 560–568 (2019).
Google Scholar
Wang, Y. et al. Accelerating the energy transition towards photovoltaic and wind in China. Nature 619, 761–767 (2023).
CAS Google Scholar
Sitch, S. et al. Evaluation of the terrestrial carbon cycle, future plant geography and climate‐carbon cycle feedbacks using five dynamic global vegetation models (DGVMs). Glob. Change Biol. 14, 2015–2039 (2008).
Google Scholar
International Energy Agency. Special Report on Solar PV Global Supply Chains (OECD, 2022).
Stern, R. et al. Photovoltaic fields largely outperform afforestation efficiency in global climate change mitigation strategies. PNAS Nexus 2, pgad352 (2023).
Google Scholar
Global Energy and Climate Model – Analysis (IEA, 2024); https://www.iea.org/reports/global-energy-and-climate-model
Ren, S. et al. Projected soil carbon loss with warming in constrained Earth system models. Nat. Commun. 15, 102 (2024).
CAS Google Scholar
Piao, S. et al. Characteristics, drivers and feedbacks of global greening. Nat. Rev. Earth Environ. 1, 14–27 (2019).
Google Scholar
Ruehr, S. et al. Evidence and attribution of the enhanced land carbon sink. Nat. Rev. Earth Environ. 4, 518–534 (2023).
CAS Google Scholar
Wu, D. et al. Accelerated terrestrial ecosystem carbon turnover and its drivers. Glob. Change Biol. 26, 5052–5062 (2020).
Google Scholar
Feron, S., Cordero, R. R., Damiani, A. & Jackson, R. B. Climate change extremes and photovoltaic power output. Nat. Sustain. 4, 270–276 (2020).
Google Scholar
Long, J. et al. Large-scale photovoltaic solar farms in the Sahara affect solar power generation potential globally. Commun. Earth Environ. 5, 11 (2024).
Google Scholar
Kannenberg, S. A., Sturchio, M. A., Venturas, M. D. & Knapp, A. K. Grassland carbon–water cycling is minimally impacted by a photovoltaic array. Commun. Earth Environ. 4, 238 (2023).
Google Scholar
Lambert, Q., Bischoff, A., Cueff, S., Cluchier, A. & Gros, R. Effects of solar park construction and solar panels on soil quality, microclimate, CO₂ effluxes, and vegetation under a Mediterranean climate. Land Degrad. Dev. 32, 5190–5202 (2021).
Google Scholar
Armstrong, A., Waldron, S., Whitaker, J. & Ostle, N. J. Wind farm and solar park effects on plant–soil carbon cycling: uncertain impacts of changes in ground-level microclimate. Glob. Change Biol. 20, 1699–1706 (2014).
Google Scholar
Hu, A. et al. Impact of solar panels on global climate. Nat. Clim. Change 6, 290–294 (2016).
Google Scholar
Li, Y. et al. Climate model shows large-scale wind and solar farms in the Sahara increase rain and vegetation. Science 361, 1019–1022 (2018).
CAS Google Scholar
Lu, Z. et al. Impacts of large‐scale Sahara solar farms on global climate and vegetation cover. Geophys. Res. Lett. 48, e2020GL090789 (2021).
Google Scholar
Krasner, N. Z. et al. Impacts of photovoltaic solar energy on soil carbon: a global systematic review and framework. Renew. Sustain. Energy Rev. 208, 115032 (2025).
CAS Google Scholar
Renewables 2019 (IEA, 2019); https://www.iea.org/reports/renewables-2019
USGS Landsat 5 Level 2, Collection 2, Tier 1. Earth Engine Data Catalog https://developers.google.com/earth-engine/datasets/catalog/LANDSAT_LT05_C02_T1_L2 (accessed 2024).
USGS Landsat 8 Level 2, Collection 2, Tier 1. Earth Engine Data Catalog https://developers.google.com/earth-engine/datasets/catalog/LANDSAT_LC08_C02_T1_L2 (accessed 2024).
Robinson, C., Ortiz, A., Ferres, J. M. L., Anderson, B. & Ho, D. E. Temporal cluster matching for change detection of structures from satellite imagery. In ACM SIGCAS Conference on Computing and Sustainable Societies (COMPASS) 138–146 (ACM, 2021).
Pettitt, A. N. A non-parametric approach to the change-point problem. Appl. Stat. 28, 126–135 (1979).
Google Scholar
Chen, J., Zhu, X., Vogelmann, J. E., Gao, F. & Jin, S. A simple and effective method for filling gaps in Landsat ETM + SLC-off images. Remote Sens. Environ. 115, 1053–1064 (2011).
Google Scholar
Global Solar Atlas https://globalsolaratlas.info/map (accessed 2024).
Tonita, E. M., Russell, A. C. J., Valdivia, C. E. & Hinzer, K. Optimal ground coverage ratios for tracked, fixed-tilt, and vertical photovoltaic systems for latitudes up to 75°N. Sol. Energy 258, 8–15 (2023).
Google Scholar
International Technology Roadmap for Photovoltaic (ITRPV) 9th edn (VDMA, 2018); https://www.vdma.org/international-technology-roadmap-photovoltaic
Bosmans, J. H. C., Dammeier, L. C. & Huijbregts, M. A. J. Greenhouse gas footprints of utility-scale photovoltaic facilities at the global scale. Environ. Res. Lett. 16, 094056 (2021).
CAS Google Scholar
Chen, Y. et al. From laboratory to production: learning models of efficiency and manufacturing cost of industrial crystalline silicon and thin-film photovoltaic technologies. IEEE J. Photovolt. 8, 1531–1538 (2018).
Google Scholar
Bhandari, K. P., Collier, J. M., Ellingson, R. J. & Apul, D. S. Energy payback time (EPBT) and energy return on energy invested (EROI) of solar photovoltaic systems: a systematic review and meta-analysis. Renew. Sustain. Energy Rev. 47, 133–141 (2015).
Google Scholar
Philipps, S. & Warmuth, W. Photovoltaics Report Fraunhofer Institute for Solar Energy Systems (Fraunhofer ISE, 2019).
Solar plants typically install more panel capacity relative to their inverter capacity. EIA https://www.eia.gov/todayinenergy/detail.php?id=35372 (2018).
Land Cover CCI Product User Guide Version 2. ESA maps.elie.ucl.ac.be/CCI/viewer/download/ESACCI-LC-Ph2-PUGv2_2.0.pdf (2017).
Zanaga, D. et al. ESA WorldCover 10 m 2021 v200. Zenodo https://doi.org/10.5281/zenodo.7254221 (2022).
Wu, Q. geemap: a Python package for interactive mapping with Google Earth Engine. JOSS 5, 2305 (2020).
Google Scholar
Wu, Q. et al. Integrating LiDAR data and multi-temporal aerial imagery to map wetland inundation dynamics using Google Earth Engine. Remote Sens. Environ. 228, 1–13 (2019).
Google Scholar
ERA5-Land Hourly Data from 2001 to Present (ECMWF, 2019).
Yang, Y. et al. Terrestrial carbon sinks in China and around the world and their contribution to carbon neutrality. Sci. China Life Sci. 65, 861–895 (2022).
CAS Google Scholar
Mehedi, T. H., Gemechu, E. & Kumar, A. Life cycle greenhouse gas emissions and energy footprints of utility-scale solar energy systems. Appl. Energy 314, 118918 (2022).
CAS Google Scholar
Nian, V. Impacts of changing design considerations on the life cycle carbon emissions of solar photovoltaic systems. Appl. Energy 183, 1471–1487 (2016).
CAS Google Scholar
Trade statistics for international business development. Trade Map https://www.trademap.org (accessed 2024).
Life cycle inventories and life cycle assessments of photovoltaic systems. IEA-PVPS https://iea-pvps.org/key-topics/life-cycle-inventories-and-life-cycle-assessments-of-photovoltaic-systems/ (2020).
Jerez, S. et al. The impact of climate change on photovoltaic power generation in Europe. Nat. Commun. 6, 10014 (2015).
CAS Google Scholar
ERA5-Land Monthly Averaged Data from 2001 to Present (ECMWF, 2019).
Frischknecht, R. et al. Methodology guidelines on life cycle assessment of photovoltaic 2020. IEA-PVPS https://iea-pvps.org/key-topics/methodology-guidelines-on-life-cycle-assessment-of-photovoltaic-2020/ (2020).
Taha, H. The potential for air-temperature impact from large-scale deployment of solar photovoltaic arrays in urban areas. Sol. Energy 91, 358–367 (2013).
CAS Google Scholar
Xu, Z., Li, Y., Qin, Y. & Bach, E. A global assessment of the effects of solar farms on albedo, vegetation, and land surface temperature using remote sensing. Sol. Energy 268, 112198 (2024).
Google Scholar
Kruitwagen, L. et al. A global inventory of solar photovoltaic generating units – dataset. Zenodo https://doi.org/10.5281/ZENODO.5005868 (2021).
Hengl, T. & Wheeler, I. Soil organic carbon stock in kg/m² for 5 standard depth intervals (0–10, 10–30, 30–60, 60–100 and 100–200 cm) at 250 m resolution. Zenodo https://doi.org/10.5281/ZENODO.2536040 (2018).
Spawn, S. A. & Gibbs, H. K. Vegetation collection global aboveground and belowground biomass carbon density maps for the year 2010. ORNL DAAC https://doi.org/10.3334/ORNLDAAC/1763 (2020).
ERA5-Land hourly data from 1950 to present. Copernicus Climate Change Service https://doi.org/10.24381/CDS.E2161BAC (2019).
Wang, Q. Increased terrestrial ecosystem carbon storage associated with global utility-scale photovoltaic installation: datasets. Figshare https://doi.org/10.6084/M9.FIGSHARE.28328765 (2025).
Wang, Q. Increased terrestrial ecosystem carbon storage associated with global utility-scale photovoltaic installation: code. Figshare https://doi.org/10.6084/M9.FIGSHARE.28328810 (2025).

Download references

Acknowledgements

We thank N. Carvalhais, P. McGuire, X. Yue, S. Falk and Q. Sun for their assistance. This work was financially supported by the National Natural Science Foundation of China (grant nos. 72293601 (Q.Y.) and 72242104 (Q.W.)), the Joint Research Fund in Smart Grid under cooperative agreement between the National Natural Science Foundation of China and State Grid Corporation of China (grant no. U1966601) (W.W.), and the Postdoctoral Innovation Talents Support Program of China (grant no. BX20240019) (K.W.). We also thank the members of the Harvard-China Project on Energy, Economy and Environment for their valuable comments and suggestions. We are grateful to the Harvard Global Institute for providing funding support to the Harvard-China Project on Energy, Economy and Environment. The computation is completed in the HPC Platform of Huazhong University of Science and Technology.

Author information

These authors contributed equally: Qingrui Wang, Kai Wang, Lintao Shao.

Authors and Affiliations

State Key Laboratory of Coal Combustion, Huazhong University of Science and Technology, Wuhan, China
Qingrui Wang, Lintao Shao, Xinyi Tang & Qing Yang
Department of New Energy Science and Engineering, School of Energy and Power Engineering, Huazhong University of Science and Technology, Wuhan, China
Qingrui Wang, Lintao Shao, Xinyi Tang & Qing Yang
Institute of Carbon Neutrality, Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing, China
Kai Wang & Shuchang Tang
UK Biochar Research Centre, School of GeoSciences, University of Edinburgh, Edinburgh, UK
Ondřej Mašek
Climate Research Division, Environment and Climate Change Canada, Victoria, British Columbia, Canada
Gesa Meyer
Department of Mechanical and Aerospace Engineering and Center for Energy Research, University of California, San Diego, CA, USA
Jan Kleissl
State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai, China
Liwei Zhang
Research Institute for Eco-civilization, Chinese Academy of Social Sciences, Beijing, China
Mudan Wang
State Key Laboratory of the Operation and Control of Renewable Energy and Storage Systems, China Electric Power Research Institute, Beijing, China
Weisheng Wang
China-EU Institute for Clean and Renewable Energy, Huazhong University of Science and Technology, Wuhan, China
Qing Yang
John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA
Qing Yang
Faculty of Environment, Science and Economy, University of Exeter, Exeter, UK
Stephen Sitch

Authors

Qingrui Wang
View author publications
Search author on:PubMed Google Scholar
Kai Wang
View author publications
Search author on:PubMed Google Scholar
Lintao Shao
View author publications
Search author on:PubMed Google Scholar
Xinyi Tang
View author publications
Search author on:PubMed Google Scholar
Shuchang Tang
View author publications
Search author on:PubMed Google Scholar
Ondřej Mašek
View author publications
Search author on:PubMed Google Scholar
Gesa Meyer
View author publications
Search author on:PubMed Google Scholar
Jan Kleissl
View author publications
Search author on:PubMed Google Scholar
Liwei Zhang
View author publications
Search author on:PubMed Google Scholar
Mudan Wang
View author publications
Search author on:PubMed Google Scholar
Weisheng Wang
View author publications
Search author on:PubMed Google Scholar
Qing Yang
View author publications
Search author on:PubMed Google Scholar
Stephen Sitch
View author publications
Search author on:PubMed Google Scholar

Contributions

Q.Y. and Q.W. designed the study. Q.W., K.W., L.S. and G.M. collected the data. M.W. contributed to economic analysis. X.T. performed accuracy validation of the datasets. S.S. and S.T. contributed to microclimate analysis. Q.W., K.W., L.S., S.S., J.K., O.M., L.Z. and W.W. performed data analysis and wrote the manuscript with substantial contributions from all co-authors.

Corresponding author

Correspondence to Qing Yang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Geoscience thanks Zhengyao Lu and Dirk-Jan van de Ven and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Xujia Jiang, in collaboration with the Nature Geoscience team.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Methods A and B, Supplementary Figs. 1–22, Tables 1 and 2 and Supplementary Discussions A–F.

Source data

Source Data Fig. 1

Statistical source data.

Source Data Fig. 2

Statistical source data.

Source Data Fig. 3

Statistical source data.

Source Data Fig. 4

Statistical source data.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article

Wang, Q., Wang, K., Shao, L. et al. Increased terrestrial ecosystem carbon storage associated with global utility-scale photovoltaic installation. Nat. Geosci. 18, 607–614 (2025). https://doi.org/10.1038/s41561-025-01715-2

Download citation

Received: 18 April 2024
Accepted: 07 May 2025
Published: 02 June 2025
Issue Date: July 2025
DOI: https://doi.org/10.1038/s41561-025-01715-2

Subjects

This article is cited by

The extra climate benefits of solar farms
- Zhengyao Lu
Nature Geoscience (2025)

Abstract 摘要

Similar content being viewed by others其他人正在查看相似内容

Main 主要

Land-cover changes after USPV deployment美国光伏电站部署后的土地利用变化

C pool alterations from land-cover changes陆地覆盖变化引起的碳库变化

Effect of C pool change on C footprint of USPV

Implication for USPV land management

Sensitivity to microclimate impact

Methods

Developing a time series dataset

Quantifying impacts on C pool

Accounting for USPV C footprint

Designing land management strategies

Analysing the impact of microclimate changes

Data availability

Code availability

References

Acknowledgements

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Competing interests

Peer review

Peer review information

Additional information

Supplementary information

Source data

Rights and permissions

About this article

Cite this article

Share this article

Subjects

This article is cited by

Similar content being viewed by others
其他人正在查看相似内容

Land-cover changes after USPV deployment
美国光伏电站部署后的土地利用变化

C pool alterations from land-cover changes
陆地覆盖变化引起的碳库变化