Regionalized Life Cycle Analysis of Ecosystem External Cost Associated with Land-Use Change in Photovoltaic Systems

Abstract

This article presents a methodology for assessing the ecosystem external costs linked to land-use changes caused by utility-scale photovoltaic systems using a regionalized life cycle approach. The core scientific challenge is to integrate a typically non-site-specific method—life cycle assessment—with a site-specific evaluation of ecosystem services affected by land-use changes. The methodology does not model specific agricultural practices. The approach is applied to three configurations of solar-tracking photovoltaic plants installed on arable land: ground-mounted photovoltaics, elevated agrivoltaics, and spaced agrivoltaics. For each configuration, the external costs or benefits per megawatt-hour (MWh) produced are estimated, allowing a comparative life cycle analysis. The findings show that the elevated agrivoltaic system is the only configuration resulting in a net loss of ecosystem service value, albeit marginal (−0.2 EUR/MWh). In contrast, the ground-mounted system yields a net benefit (approximately 1 EUR/MWh), followed by spaced agrivoltaics (0.1 EUR/MWh). These outcomes are mainly driven by the construction and operational phases, while the impacts from component production, transport, and end-of-life stages are significantly lower. The methodology offers a replicable framework for integrating the monetary evaluation of ecosystem services into life cycle assessments of land-intensive renewable energy systems.

1. Introduction

From a life cycle assessment (LCA) perspective, photovoltaic (PV) plants are generally associated with significant benefits in terms of carbon emissions and air pollutant savings compared to fossil fuel technologies [1,2,3]. However, utility-scale PV plants in rural areas can have territorial, landscape, and ecological impacts due to their size and site characteristics [4,5]. Consequently, under European law, large-scale ground-mounted PV plants must undergo an environmental impact assessment (EIA) for authorisation (Directive 2011/92/EU [6]). The EU Taxonomy Regulation [7] and associated Delegated Acts [8,9], which define criteria for establishing environmentally sustainable economic activities, assume PV benefits for emissions reduction without imposing specific requirements on potential negative environmental effects. The EU Taxonomy Delegated Act for climate [8] recognizes the eligibility of PV systems without requiring compliance with the life cycle emissions limit of gCO₂/kWh < 100 applicable to fossil fuel plants. Regarding compatibility with biodiversity and ecosystems (Do No Significant Harm —DNSH criteria), the Taxonomy does not require adherence to quantitative technical requirements (key performance indicators), but only qualitative checks. While Taxonomy promotes PV plants, in Italy, recent legislation [10] prohibits the construction of conventional ground-mounted PV systems in agricultural areas. There appears to be no regulatory consensus on the impacts of photovoltaic plants on ecosystems. The key questions motivating this article are: Do large-scale PV systems in rural areas cause significant harm from the perspective of ecosystem services? Could PV impact due to land-use changes be significant if a life cycle approach is assumed, including land occupation by mining sites, production plants, and waste disposal facilities? Using the common metric of monetary externalities, what is the order of magnitude of external costs for producing the same amount of electricity with different PV configurations?

1.1. Aim of the Article

Based on these motivations, the aim of this article is to develop a pioneering methodology for estimating the ecosystem-related external costs or benefits associated with land-use changes caused by utility-scale photovoltaic systems built in agricultural areas, adopting a life cycle perspective. The methodology is applied to three alternative configurations of a PV system installed in an arable land context: a conventional ground-mounted plant, an ‘elevated’ agrivoltaic plant, and a ‘spaced’ agrivoltaic plant. To clarify the scope of this study, it is important to note that we do not aim to apply conventional life cycle assessment (LCA) methods to evaluate all ecosystem-related impacts of PV systems. Instead, our approach focuses exclusively on the economic value of ecosystem services associated with land-use changes. It is grounded in a distinct, economics-based methodology commonly referred to in the literature as monetary LCA [11,12]. This method quantifies the economic value of environmental impacts—so-called externalities—by replacing the characterization phase of the Life Cycle Impact Assessment (LCIA) with a unified monetary metric for the impact indicators. Monetary LCA differs from Life Cycle Cost Analysis (LCCA) in that the former aims to calculate the economic value of environmental externalities generated by a technology [12]. In contrast, LCCA analyses the costs of a technology over its economic lifetime from the perspective of an economic decision-maker [13]. These costs include acquisition, installation, operation, maintenance, refurbishment, discarding, and disposal costs [14].

To illustrate our approach, we apply the monetary life cycle method to Ecoinvent’s inventory data on land use (i.e., land transformation and occupation [15,16]). This allows us to assess a specific category of environmental externalities: those related to variations in ecosystem services resulting from land-use changes throughout the life cycle of a PV system, from asset production and operation to end of life. Addressing this research gap will enable future studies to aggregate the results obtained for ecosystem services related to the land-use-change inventory data with those associated with other external cost categories of renewable energy plant that can be assessed using the monetary LCA approach (e.g., damage costs of greenhouse gas emissions and health impacts of air pollutants, as in [17]). It will also allow researchers to compare the environmental external costs of different energy sources in a more complete way.

1.2. Definitions and Conceptual Framework

Some key definitions are necessary to frame the scope of the work and to highlight the novelty of the methodology. Ecosystem services (ES) are the contributions ecosystems make to human well-being [18]. The benefit of ES represents the monetary value of these contributions. Measuring the monetary value attributable to the flow of services generated by ecosystems over time represents a way to indirectly measure the economic value of ecosystems and biodiversity, analogous to the concept of rent associated with an asset (e.g., the rent of land or financial capital) [19]. The monetary valuation approach of ES is closely linked to that of natural capital valuation [20]; Comitato per il capitale naturale [21,22]). The premise of this approach is that, like other types of capital (fixed capital, financial capital, etc.), the stock of natural capital also produces a flow of services, benefiting human activities, called ecosystem services.

The benefit of ES associated with a specific ecosystem (or the external costs resulting from ES value reduction over time due to the installation and operation of a ground-mounted utility-scale photovoltaic system in a specific land context) can be valued using a mix of environmental and welfare economics techniques depending on the type of ES [23,24]. In some cases (e.g., ‘habitat quality–biodiversity’, ‘air pollution mitigation’), non-market value techniques are used to elicit people’s willingness to pay for the ES [24,25]. In other cases, proximity markets are used to estimate the value of ES. For example, ‘outdoor recreation’ can be valued by analyzing the transportation costs incurred by visitors to reach a natural site [26]. When ES are directly linked to the exploitation of natural resources that have a reference market where prices can be observed (so-called “provisioning” ES, with agriculture being a typical example), specific valuation techniques should be employed to distinguish the portion of economic benefit attributable to human inputs (labour, machinery, fertilizers, irrigation) from the specific contribution of natural factors (a combination of type of crop, soil, irradiation, and other climate features). Only the latter reflects the true value of the ecosystem service “crop provisioning” [27,28].

Ecosystem external costs arise when a human activity (in this article, the construction of a photovoltaic system) results in an economic well-being loss due to its impacts on ES, and when this loss is not accounted for or compensated by the responsible party [17,29]. In other words, external costs, together with the internal costs borne by the company managing the photovoltaic power plant project, constitute a component of the social costs considered by welfare economics when applying cost–benefit analysis or other economic well-being optimization techniques [30,31]. While internal costs of technologies are generally known or easier to evaluate, external costs require specific modelling based on the ‘impact-pathway approach’ and economic evaluation methodologies [32].

1.3. Literature Review

The origins of a life cycle approach to assessing the external costs of renewable energy plants can be traced back to research conducted in the 1990s on the evaluation of fuel cycle externalities associated with power generation technologies, fuel transport, and refining plants [33,34,35]. The main outcomes of this body of literature relate to the development of impact-pathway modelling for quantifying the health-related effects of noxious emissions from fossil-fuel-based power plants [36]. More recent studies employing the impact-pathway approach in the electricity sector enlarged their coverage to a broader set of air emission types, including greenhouse gases, some heavy metals, and organic pollutants, enabling country-wide sectoral assessments. Examples are the works promoted by the European Environment Agency [17,37,38,39,40,41,42], for Italy. Although the ExternE project included analysis of the external costs of renewable energy plants [35], the ecosystem-related external costs of these technologies remained only partially assessed. This limitation was primarily due to the challenges of modelling and evaluating their ecological and territorial impacts within a comprehensive analytical framework.

In recent decades, such a framework is increasingly offered by the diffusion within the scientific community of the ES approach [43,44], its classifications [18,20,45], economic assessment methods [23,24], and assessment results [46,47,48]. Particularly, the economic valuation approach of ES gained prominence in Europe with the experimentation of national natural capital accounting systems related to land use and cover mapping in the European Union [25,26,27,49] and in specific countries, such as Italy [21,22,50]. The development of georeferenced assessments of ecosystem service indicators and economic values at the national scale (INCA platform, European Commission—JRC, 2024 [51]) is contributing to the consolidation and harmonization of valuation methodologies, opening new research avenues across various domains, including the assessment of environmental and territorial impacts of renewable energy plants. The ES framework approach coupled with land-use mapping is revealing new opportunities in the fields of environmental management and social acceptability. For instance, Semeraro and colleagues have explored strategies and bio-engineering techniques to enhance ES in the environmental management of photovoltaic and agrivoltaic systems [52,53,54]. Moreover, the ES framework is becoming increasingly prominent in the analysis of social acceptability and conflict prevention related to renewable energy developments, both through qualitative methods [55,56] and the use of biophysical indicators [57,58,59].

Existing literature reviews on ES and renewable energy plants indicate that despite the broad research potential of the ES framework, its application to assess and mitigate the ecosystem-related external costs of photovoltaic plants is a relatively recent area of research [60,61]. The monetary ES approach has been applied primarily to hydropower and wind energy plants, while solar photovoltaic systems have thus far rarely been considered [60]. Available studies covering the external costs of renewable energy plants typically focus on a specific impact type, e.g., loss of landscape value [62,63,64], without adopting a systematic approach based on ES classifications such as The Economics of Ecosystems and Biodiversity (TEEB [20,23,45]) or Common International Classification of Ecosystem Services (CICES [18]). The impacts of renewable energy plant expansion—specifically wind and ground-mounted PV systems—in the Basilicata Region on ecosystem-related dimensions such as soil carbon sequestration, groundwater availability, and biodiversity have been assessed by [65], using a variety of methodological approaches and biophysical indicators. However, interpreting and generalizing the results specifically for ground-mounted PV systems remains challenging due to insufficient detail in PV plant modelling, including technological configurations, ecosystem characterization beneath and between modules, and the design of plant foundations. A recent literature review on the (non-monetary) impacts of photovoltaic installations on biodiversity concludes that the current body of evidence is still too scarce to support generalizations about systematic negative or positive effects, such as those related to habitat creation through shading and nesting [66]. Studies on ecosystem-related external costs of photovoltaic plants are rare, limited in coverage, and often adopt non-conventional approaches. [67] develop a method for evaluating the environmental cost/benefits of PV solar plant installations in high-value natural areas (rich in forests, vegetation, and moorlands) versus conserving natural ecosystems, which focuses on the loss of carbon sinks and biodiversity as well as disaster risk. Ref. [68] propose three distinct approaches to quantify the impacts of ground-mounted photovoltaic (PV) plants, incorporating ES in accordance with the established classification—namely, provisioning, regulating and maintenance, and cultural services. The three approaches are: cost–benefit analysis (CBA), multi-criteria decision analysis (MCDA), and composite modelling assessment (COSIMA). Compared to our economic evaluation approach, which is specific and limited to the value of ES impacted by PV systems, CBA is an economic assessment method for investment projects that adopts a broader and more comprehensive perspective. It focuses on balancing capital and operational expenditures with the economic benefits of projects—for example, revenues from energy and crop production in the case of agrivoltaic systems—and could benefit from the inclusion of externalities in the assessment framework [69,70]. MCDA is a multi-indicator framework that evaluates impacts expressed on heterogeneous measurement scales, producing a composite index that can incorporate decision-makers’ preferences. COSIMA integrates the outputs of both CBA and MCDA into a single performance index for each alternative. In contrast to the present study, their work does not focus on the value of ES, nor does it apply the proposed methodologies to empirical case studies. Consequently, it does not define a counterfactual scenario for analyzing changes in ES associated with the construction and operation of PV plants in a rural context. Furthermore, it does not adopt a life cycle perspective.

A study that presents some similarities with ours is that of [71], who compare three types of power plants—natural gas, solar, and wind—constructed in the U.S. Chihuahuan Desert. Their analysis focuses on land occupation and the loss of ecosystem service value, applying the TEEB framework [23]. However, their scope is limited to a desert ecosystem characterized by grassland, shrubland, and woodland, excluding cropland. Despite methodological parallels, their life cycle perspective is confined to the operational phase of the energy plants and the natural gas extraction process, without incorporating a comprehensive life cycle assessment of the energy plant components or their end-of-life phases.

The integration of the ES approach in life cycle assessment (LCA) of renewable energy plants remains in an early stage. A review by [72], which employed a bibliometric analysis of three decades of literature on ES and LCA, revealed that these two academic fields have largely evolved in parallel, with relatively few interactions (particularly outside the context of renewable energy studies). According to these authors, although LCA seeks to assess the impacts of technologies on terrestrial and marine ecosystems, as well as on natural resource use and human health, it has struggled to fully account for the negative impacts of technologies or products on ES. This is partly due to the conceptualisation of ES as positive contributions provided by ecosystems to human well-being. One of the main challenges of applying a life cycle perspective is the variability of ES across different ecosystems and geographical areas, coupled with the absence of geographically differentiated established indicators for the value of ES [72]. This lack makes it difficult to integrate ES into a standardized methodology for environmental impact assessment, such as LCA.

This study contributes to the above-mentioned existing research strands with two main objectives:

• To develop a regionalized methodology for analyzing ecosystem-related external costs associated with land-use changes induced by renewable energy plants. The methodology is designed to be applicable across the main life cycle phases of a PV system: production, construction, operation, and end of life.
• To apply the proposed methodology to a utility-scale PV system, considering three alternative configurations: a 56 MW ground-mounted plant, an elevated agrivoltaic plant of the same capacity, and a ‘spaced’ agrivoltaic plant with a capacity of 39 MW, occupying the same total area as the other two. The goal is to identify the configurations that minimize the ecosystem external costs in relation to the kWh produced, calculated with a life cycle approach. The technical details of the three configurations were developed based on a project for a new ground-mounted photovoltaic plant in a rural area of central Italy. The exact geographical location of the facility has been anonymized as the project is currently undergoing authorization.

Building on our literature review and the studies surveyed by VanderWilde and Newell [72], our research is among the first to investigate photovoltaic systems using a methodology that integrates the economic valuation of ES within a life cycle perspective. Our regionalized life cycle approach is based on geographical localisation assumptions for life cycle processes, which reflect the site chosen for the PV system (Italy) and the prevailing market origins of the main components of the reference photovoltaic system. In our monetary-based life cycle analysis, we used ‘benefit transfer’ techniques at a regional (supranational) scale. Benefit transfer is an environmental economics technique used to estimate the monetary values of environmental goods in specific geographical areas by utilizing existing data and results from studies conducted in similar but different contexts [73]. This method involves transferring economic benefits measured in one location to another where specific studies have not been conducted, thereby saving time and resources by avoiding the need for new detailed surveys. Our methodology employs benefit transfer techniques to encompass various geographical areas globally, such as Italy and China. This is achieved by using available data sources on standardized monetary values of ES across different regions ([48], Supplementary Materials S1.6.1), supplemented by the authors’ original review work [74] on ecosystem service evaluation within the Italian context (see Section 2 and Supplementary Materials S1.6.2).

The article is structured into the following sections. Section 2 provides a general description of the methodology for assessing ecosystem external costs of large-scale PV systems with a regionalized monetary life cycle approach. Section 3 describes the case study used to implement the methodology, consisting of three alternative PV system configurations installed at the same site within an agricultural context. Section 4 describes the life cycle monetary results in terms of net losses (external costs of PV systems) or net gains (external benefits). It also includes a subsection on intermediate (non-economic) results—related to land use and ecosystem changes—to facilitate the interpretation of the monetary outcome. Section 5 discusses the assumptions and limitations of the proposed methodology and highlights its potential applications as a tool to support energy policy assessments, land-use planning, and the environmental practices of energy companies. Finally, Section 6 summarizes the findings of the regionalized life cycle methodology for assessing ecosystem-related externalities of PV systems, highlighting its novelty, limitations, and directions for future research on ecosystem external costs.

2. Materials and Methods

This section outlines the general methodology and the main sources used for evaluating ES related to land-use changes of a ground-based photovoltaic plant. Due to space constraints, the detailed description of all steps of the methodology, including assumptions, data sources, and methods used in each step is provided in Supplementary Materials S1. The R code used for the calculations, along with the underlying datasets, is publicly available on GitHub at the following link: https://github.com/giuliomela/land_use_change_pv_systems (accessed on 29 December 2025, commit hash: 1f9a085a57994a16ef6490d9a048621ef1df7bbc).

The goal of the methodology is to calculate the variation in the monetary value of ES resulting from land-use changes associated with the life cycle of utility-scale photovoltaic plants built in an agricultural context. The methodological goal is achieved through a regionalized life cycle approach [75], covering the production, construction, operation, and end-of-life phases of the plant’s key components. In case of agrivoltaics configurations of PV systems (“a photovoltaic installation that adopts solutions aimed at preserving the continuity of agricultural and pastoral cultivation activities at the installation site” (MITE [76], p. 4)), the methodology refers only to the technological side of the system; the extension of the LCA from energy to crop production, related to the agricultural side of agrivoltaic configurations, is out of scope.

The methodology differs from standardized LCA methods [77,78,79] in three main aspects:

• It focuses exclusively on ES associated with land-use-change data resulting from the main phases of the PV system life cycle, with information obtained through direct spatial mapping or LCA inventories. Due to the investigative assumptions and the pioneering nature of the methodology, it does not use environmental footprint categories and related impact indicators such as climate change, ecotoxicity, or acidification [79].
• It applies an economic valuation approach to environmental externalities (commonly referred to as ‘monetary LCA’ [11], and sometimes described in the literature as ‘socio-environmental externalities LCA’ [80], to estimate changes in economic well-being associated with specific environmental impact categories. We quantify the negative or positive variation in the economic value of ES resulting from land-use changes across the life cycle phases of PV systems. A reduction in value is classified as an ‘ecosystem external cost,’ while an increase is considered an ‘ecosystem external benefit’. The economic approach is characterized by the aggregation (sum) of the annual value of specific ES (e.g., CO₂ removal, erosion mitigation, agricultural pollination) per hectare of ecosystem type, using a monetary unit of measure (e.g., EUR₂₀₂₂/ha-y) that reflects environmental economics methods to elicit people preferences for ES [23,24]. A summary of the main sources of data and economic valuation methods used for each ecosystem service in Italy is provided in Table S22. To obtain differentiated values on a regional scale we used ‘benefit transfer’ techniques, elaborating results of literature studies that applied economic valuation techniques to the benefits provided by different categories of ecosystems to humans in a specific geographical and socio-economic context (e.g., national territory, a region, a rural area, a natural park).
• It refers to the so-called regionalised LCA approach [75], which models the main life cycle phases based on their actual geographic location. In this framework, different ES values are applied to the same amount of land-use change depending on where it occurs, in contrast to traditional LCA methods that use uniform characterization factors globally.

Figure 1 illustrates the general structure of the proposed methodology.

The methodology employs a general approach common to the various life cycle phases of the reference system, as exemplified by the procedural steps of the central horizontal process illustrated in Figure 1 (Steps 1–8). This process is then articulated into two parallel approaches, differing in terms of data and sources used, depending on the life cycle phases (i.e., production and end-of-life phases of the main plant components, illustrated in the lower horizontal process of Figure 1; PV plant construction and operation phases, shown in the upper horizontal process). More specifically, the common approach across the different life cycle phases consists of the following procedural steps:

• Definition of the goal of the study and preliminary delimitation of the scope of the life cycle investigation of land-use change with reference to components, processes, useful life, and energy producibility of the PV plant (Step 1 in Figure 1).
• Geographical localisation of life cycle components and processes (Step 2).
• Modelling of the plant, its components, and processes (Step 3).
• Identification of land-use-change classes, estimation of the affected areas by land-use class in each region, and duration of land occupation (Step 4).
• Application of correspondence tables between land-use classes and biome/ecosystem types for calculating the areas of the biome/ecosystem type affected by land-use changes (Step 5).
• Calculation of ecosystem service monetary values per hectare-year using benefit transfer method, differentiated by biome/ecosystem and geographical region (Step 6).
• Use of ecosystem service unit values for the monetary valuation of ES net losses or benefits associated with areas affected by biome/ecosystem changes (by phase, component/process, and geographical area, Step 7).
• Summation of results in EUR/GWh across components/processes and life cycle phases of the photovoltaic plant (Step 8).

The components and processes considered in the land-use-based life cycle analysis of the reference photovoltaic plant and its alternative configurations are as follows:

• Modules (56 MW for the reference plant and the elevated agrivoltaics plant, 39.2 MW for the spaced agrivoltaics plant): production and transport.
• Support structure—solar trackers (a total weight of 623 kg is required to support 26 modules in both the reference and spaced agrivoltaic systems, whereas the elevated agrivoltaic system requires 886 kg for every 26 modules): production and transport.
• Inverters (13 inverters of 3437 kW each for the reference plant and the elevated agrivoltaics plant, 9 inverters for the spaced agrivoltaics plant): production and transport.
• Materials for the PV plant’s roads (11,844 m of dirt roads, of which 8568 m for perimeter roads—4 m wide). The modelling of road materials includes the extraction processes and any necessary processing to obtain the final material.
• End of life of the PV plant for the modules, inverters, and module support structure components. For each component, treatment and recycling processes were identified in accordance with available information in the existing literature (IEA Task 12 on PV sustainability [81] and EPD Italy [82]).

The transformation cabins and cables used for transporting the produced energy are excluded from the study scope.

The analysis adopts the cut-off allocation approach [83]. The cut-off approach assigns the environmental impacts associated with the primary production of materials to the initial user without granting any credit to the primary producer for the recyclable materials provided. The environmental burdens of recycling processes related to PV components’ end of life remain outside of the system boundaries, while secondary (recycled) materials used as input in components’ production processes only present the impacts resulting from the recycling processes.

The lifetime of the PV system is an important parameter for the analysis as it concerns the temporal extension of the effects of changes in ES during the life cycle’s operational phase (starting from the installation of the plant until its dismantling). The standard useful life of 30 years, recommended by the JRC in its report on photovoltaic systems and their components [84], is used as a reference for all three configurations of the PV system. Given that the construction period of the plant is less than one year, including the restoration of temporary works such as material and component storage areas, it is deemed unnecessary or negligible to model this phase separately from the operational phase. Similarly, for simplicity in modelling, it is assumed that the decommissioning phase of the plant at the end of its life, along with the restoration of the site to its original state, will also be of short duration (less than one year). Consequently, we implicitly assume that the plant’s useful life of 30 years encompasses the construction, operation, decommissioning, and site-restoration phases.

Figure 2 presents an overview of the modelled components and processes of the photovoltaic system from a life cycle perspective (Step 1—Scope Definition), along with the assumptions regarding their geographical localization, which reflect the siting of the case study in Italy (Step 2—Localization of Processes).

To establish the geographical areas to which the components and processes refer, we suggest using, where possible, the criterion of representativeness of the supply markets for a photovoltaic (PV) system built in the chosen country. In our case study, we assume that the photovoltaic plant (three alternative configurations) is built in an agricultural area of central Italy, while the geographical areas for the production of the main components of the plant are defined with a market analysis (see Supplementary Materials S1.2).

The main differences between the two parallel approaches we adopted for, respectively, the construction and operation phases of the PV system and for the production, transport, and end of life of its components (upper and bottom horizontal processes in Figure 1), are related to methods and data sources used in Step 4 (land-use change and calculation of affected areas) and Step 6 (calculation of regional ecosystem service values per hectare-year).

In Step 4, characterization of land-use changes for the construction and operation phase employed spatially explicit mapping of the PV system in its territorial context (analysis of land cover in the fenced area hosting the system before and after the building phase). Conversely, to model land-use changes associated with the remaining life cycle phases (component production, transportation of components from the import area to the PV building site, and end-of-life processes), we utilize an LCA inventory of land-use data (obtained through LCI modelling of each PV component, as described in Step 3), which serves as a secondary information source compared to direct mapping. More specifically, we relied on land-use-change data provided by the Ecoinvent inventory dataset version 3.9 (Wernet et al., 2016 [15]). Using the LCA software SimaPro, version 9.5 [85] and Ecoinvent’s elementary flows for Land Transformation and Land Occupation, it is possible to calculate the total extension (area) of the various land-use classes subject to change due to elementary activities, and the average duration of land occupation associated to these transformations. The dataset separately quantifies the areas (measured in m² of affected surface) of the land-use classes as they appear in the ex-post phase (indicator ‘Transformation to [land-use class i]’) and those characterizing the ex-ante situation (indicator ‘Transformation from [land-use class j]’). ‘Land Occupation, in a given land-use class i’ (measured in m²/year) provides the time-related information on the surfaces occupied by a land use class. Further information on the Ecoinvent land-use inventory data is provided in Supplementary Materials S1.4.1.

Similarly, in Step 6, the calculation of regional per-hectare ES values adopted two different sources of data. For the production, transport, and disposal phases (which processes can be located globally, outside Italy), we used the Ecosystem Services Valuation Database—ESVD [48,86], a well-known source of ecosystem service ‘standardized’ values, derived from studies conducted globally. Established in 2010, ESVD aims to be the reference point for environmental economists and all those working in the field of ecosystem and service valuation. Over the years, thanks to the growing literature on the subject, the database has grown from about 1200 to over 9000 values, referring to a wide range of ES and biomes/ecosystems worldwide. For each literature estimate, the database provides a series of descriptive information such as biomes types and ES subject to valuation; monetary valuation results expressed in the original units of measure; area of the valuation; geographical coordinates; country and continent; study scale; valuation method; protection status and ecosystem conditions; bibliographic references; review status of the study; and the standardized value (the original estimate is reworked in international USD at 2020 prices per hectare-year). The standardized values of the ESVD database are related to the Common International Classification of Ecosystem Service (CICES), which distinguished three main categories of ES: provisioning (for example, crop production), regulation and maintenance (e.g., carbon removal by ecosystem) and cultural (for example, landscape-related ES). To use the database values for this work, we processed the standardized values for each combination of biome and ES class (provisioning, regulation, and cultural) in each region and converted results into EUR at 2022 prices. We calculated the average values at these geographical areas following a procedure like that proposed by the ESVD database developers for calculating summary values, which involves excluding outliers. Outliers were identified using the IQR method (excluding values that overcome the interquartile range multiplied by 1.5). The geographical areas that we selected for the PV components and processes related to production, transport, and end-of-life phases are three:

• China, for module production.
• The entire world, for maritime transport services from China to Italy (the route involves three continents and international cargo ships are built in many different world areas).
• The EU for all other components (trackers, inverters, gravel roads) and processes (transport, disposal, and incineration of the PV system components).

Further information on how we elaborated ESVD values and the results obtained are available in Supplementary Materials S1.6.1 and Table S19—Average values of ES by biome type, study geographical area (China, European Union, and World), and source (geographical area covered by the original studies).

The geographical coverage of the ESVD database is satisfactory for large regions—comprising over 2300 data points for the EU, more than 650 for the USA, and approximately 450 for China. However, coverage becomes critically limited for smaller countries such as Italy, which is represented by only 64 data points. Given that the ESVD database includes 15 biome types and classifies ES into three macrocategories—provisioning, regulation, and cultural—the number of estimates available for Italy is insufficient to derive statistically significant average values for specific combinations of ES and biomes.

For the construction and operation phases of the PV plant (located in Italy), given the limited representation of Italian ecosystems in the ESVD database, we used a specific set of national values for ES developed by Molocchi and Mela in a previous work published in Italian [74]. These values represents national-level averages for eleven ES associated with nine ecosystem types (and further subtypes of land-use classes) selected from a literature review of the main empirical works focused on developing a national mapping and accounting of ES in Italy, differentiated per ecosystem type. The dataset is shown in Table S21 and commented on in Supplementary Materials S1.6.2. The main sources of this dataset are:

• The reports developed by the JRC [25,26,27] with the aim of establishing a system of experimental ecosystem economic accounts at the European Union level, consistent with the United Nations SEEA-EEA methodological standard, which have led to a biophysical and monetary mapping for various types of ES in many member states of the EU (including Italy).
• The Annual Reports on the State of Natural Capital in Italy, also aimed at developing national accounting, whose four editions considered here [21,22,87,88] present empirical evaluations for the national territory of specific ES.
• The methodology developed by ISPRA/SNPA for the monetary evaluation of ecosystem service losses due to land consumption, illustrated in a specific methodological annex of the 2018 report [89].

The main characteristic of the ES values dataset developed for Italy [74] is that it provides average Italian values for 11 categories of ES, which can be traced back to the same three main macrocategories of the CICES classification used by ESVD (see

Table 1. Categories of ecosystem services (ES) evaluated in the life cycle phases of the study.

Construction and Operation Phase	Production, Transport, and End of Life of PV Plant Components
Ecosystem services (TEEB [45])	Ecosystem services (CICES macrocategories [21])
ES1 Agricultural supply	Provisioning
ES2 Timber supply
ES3 Water supply
ES4 Net carbon removal	Regulation and maintenance
ES5 Habitat quality (biodiversity)
ES6 Agricultural pollination
ES7 Air pollution mitigation
ES8 Freshwater purification
ES9 Erosion mitigation
ES10 Flood mitigation
ES11 Nature-based recreation	Cultural

).

For calculating the variation in the value of ES with the proposed life cycle approach, the following formula is used:

$$\mathsf{\Delta }V=T\sum _{i=1}^n\sum _{\mathrm{j}=1}^c{V}_{ij}\mathsf{\Delta }Are{a}_{ij}$$

(1)

where

• $\mathsf{\Delta }V$ is the differential value (net monetary loss or net benefit) of ES associated with land-use changes in the production, transport, and end-of-life phases of the reference PV system.
• $T$ is the average land occupation duration, expressed in years, of the production, transport, and end-of-life phases.
• $V_{ij}$ is the value, expressed in EUR/ha-year, of the ES annually provided by land-use class $i$ in a given geographical area j.
• $\mathsf{\Delta }Are{a}_{ij}$ is the net balance, for a given land-use class i and a given geographical area j, between the summed area of ‘Transformation to [land-use class i]’ and ‘Transformation from [land-use class i]’ across all processes related to a given geographical area j.

Relating Equation (1) to the functional unit E (electricity production over the life span of the reference PV plant configuration), we obtain the following formula:

$${\displaystyle \frac{\mathsf{\Delta }V}{E}}={\displaystyle \frac{T}{E}} \sum _{i=1}^n\sum _{\mathrm{j}=1}^c{V}_{ij}\mathsf{\Delta }Are{a}_{ij}$$

(2)

where ${\displaystyle \frac{\mathsf{\Delta }V}{E}}$ is the variation in the monetary value of ES associated with land-use changes in the life cycle of the energy plant per MWh of energy produced by the plant.

The same formulas, in a simplified version, are also used for the building and operation phases, given that the evaluation refers to the ES values V_ij of the only region where the plant is built (Italy).

A final comment to compare the proposed methodology with standard LCA [77,79]. The present study does not aim to be a fully LCA-compliant assessment. However, for the production and end-of-life phases of the main components of the photovoltaic plant (illustrated in the lower horizontal procedure in Figure 1), it draws on LCA inventory data and follows the logical sequence of steps defined in the LCA framework:

(a)

Goal and Scope Definition.

(b)

Life Cycle Inventory (LCI).

(c)

Life Cycle Impact Assessment (LCIA).

(d)

Interpretation.

Within this framework, Steps 2 to 5 of our methodology—ranging from the localization of life cycle processes to the calculation of areas affected by ecosystem changes across regions—correspond to Step (b): LCI. In the monetary-based approach adopted here, Steps 6 to 8 correspond to Step (c) of the LCA framework—LCIA—as the economic valuation of ES, based on per hectare-year values across different regions and ecosystem types, serves as a means of deriving characterization factors applied to land-use inventory data. The final Step, (d) Interpretation, is simplified in this study, as no multiple-impact indicators are used. Instead, a single monetary-based impact indicator is applied: the variation in ecosystem service value associated with land-use changes.

3. Description of the Case Study

The Reference Plant and Alternative Configurations

The proposed methodology is applied to a reference PV system and two alternative configurations—elevated and spaced agrivoltaics—using technical data from an anonymized project located in a rural area of central Italy currently undergoing authorization by the Ministry for the Environment and Energy Security (MEES). The selection of this case study from among new projects under authorization is motivated by the opportunity to work with primary data that would otherwise be difficult to access. This is made possible through the MEES portal on environmental assessments and authorizations [90], which publicly provides extensive project documentation—both technical and for dissemination—prepared by project proponents to meet the requirements for authorization, public information, and stakeholder consultation, as stipulated by environmental impact assessment (EIA) regulations. The online platform allows users to filter new projects by type and access documentation on photovoltaic and agrivoltaic projects (more than 1500 as of the end of 2024) currently undergoing authorization [90].

The reference project envisages the construction of a 56 MW ground-mounted photovoltaic plant in an agricultural area characterized by arable farming (durum wheat). The total area affected by the plant is 79.6 hectares (79.2 ha fenced area hosting the modules, inverters, and service roads; 0.4 ha for the distribution cabin; see

Table 2. Technical data of the three photovoltaic (PV) system configurations.

Parameter	Reference Plant (Conventional Ground-Mounted)	Alternative Configuration 1 (Elevated Agrivoltaics)	Alternative Configuration 2 (Spaced Agrivoltaics)
Fenced Area Hosting the Plant	79.2 ha	79.2 ha	79.2 ha
Total Power	56.02 MWp	56.02 MWp	39.21 MWp
Power Density (on fenced area)	0.71 MWp/ha	0.71 MWp/ha	0.52 MWp/ha
Number of Modules	98,280	98,280	68,796
Module Area	2.73 m2	2.73 m2	2.73 m2
Power/Module	570 Wp	570 Wp	570 Wp
Modules per Tracker	78	78	78
Tracker Length (39 modules)	45 m	45 m	45 m
Tracker Width (2 modules)	4.82 m	4.82 m	4.82 m
Number of Trackers	1260	1260	882
Height of the Tracker Support Structure	2.59 m	4.59 m	2.59 m
Distance Between Parallel Rows (between support structure posts)	9.5 m	9.5 m	13.5 m
Minimum Distance Between Parallel Rows (between edges of horizontal/zenith panels)	4.7 m	4.7 m	8.7 m
Estimated Annual Production	94.96 GWh/year	94.96 GWh/year	66.47 GWh/year

). The site is predominantly flat and situated in a hilly area. Regarding land-use type in the ex-ante situation, the project documentation of the reference plant indicates agricultural use with durum wheat and maize crops, uniformly classified in the cartography as ‘irrigated arable land’ Corine Land Cover (CLC) 2.1.2 [91] in all portions of land affected by the plant. We assume any pre-existing agricultural activity within the fenced area is interrupted by the building and operation of the PV system.

The technology consists of bifacial monocrystalline silicon modules mounted on ground structures with a single-axis N–S solar tracker (rotation axis parallel to the ground-plane-oriented N–S). The support structure is typical of ground-mounted photovoltaics and consists of posts directly driven into the ground, without artificial foundations in concrete or other materials. In the reference photovoltaic plant, the minimum operational height of the modules at maximum inclination (60°) is 0.5 m, while the maximum height is 4.68 m. When the modules reach the horizontal position, their height from the ground is about 2.6 m.

Assuming a useful life of the PV plant of 30 years [87], the estimated energy producibility of the base case at the chosen site is 2849 GWh (about 95 GWh/year), equivalent to 1695 h at the nominal power of the plant (Supplementary Materials S1.1.3).

Regarding alternative configurations, it was considered appropriate to compare a conventional ground-mounted photovoltaic plant with an agrivoltaics plant in its two main configurations [76,92]: elevated and spaced.

Alternative configuration 1: Elevated agrivoltaics (with continuation of the pre-existing agricultural activity). We assume an agrivoltaics system with the same technological characteristics as the reference plant (bifacial monocrystalline silicon modules with tracker), but with trackers mounted at 4.6 metres above ground. When inclined at 60°, the minimum height of modules is 2.5 metres, allowing the passage of machinery and the continuation of agricultural activity (the requirement of the Italian Guidelines on Agrivoltaic is 2.1 m [76]). Even if some combine harvesters can overcome this height, the tractor can pass between module rows while implements (which are much shorter and wider) pass below elevated AV panels. The elevated agrivoltaic configuration is characterized by the same area as the reference photovoltaic plant (fenced area 79.2 ha); to ease the comparison with other configurations, we also assume the same power (56 MWp) and power density (0.71 MWp/ha). To avoid crop assumptions influencing the LCA comparison results between technological configurations, the maintenance of the pre-existing crop, i.e., durum wheat, is assumed. This implies that the scope of the LCA investigation in this work is limited to the technological plant and its energy production, excluding analyses related to agricultural production. The estimated energy producibility of the elevated agrivoltaics plant at the chosen site is the same as the reference plant (95 GWh/year, 2849 GWh over 30 years), given that they share the same power and technology.

Alternative configuration 2: Spaced agrivoltaics (with continuation of the pre-existing agricultural activity). A ground-mounted agrivoltaics system with spaced structures, built at the same height from the ground as the reference plant (min 0.5–max 4.7 m at 60°). It maintains the same technological characteristics and covers the same area (79.2 ha) but features wider spacing between rows. As a result, the total installed power is 30% lower than that of the reference plant (39.2 MWp total power and 0.50 MWp/ha power density, respectively). This configuration allows the resumption of pre-existing agricultural activity, but to a lesser extent than elevated agrivoltaics, limited to the open spaces between the module rows. Again, we assume the continuation of the pre-existing crop (durum wheat). The estimated energy producibility of the spaced agrivoltaics configuration is scaled down by 30% compared with the reference plant: 1994 GWh over the useful life of 30 years (66.5 GWh/year).

The same areas for ancillary works are assumed in the three configurations: cabins for inverters and other equipment (0.08 ha), gravel roads surrounding the plant inside the fenced area (4.74 ha), and one transformation substation outside the fenced area (0.40 ha). Table 2 summarizes the main data of the PV reference plant and of the two alternative configurations, while

Table 3. Components and materials used to model the three system configurations under study. The reported values represent the quantities needed to produce a single plant.

Components	Materials	Reference Plant	Alternative Configuration 1	Alternative Configuration 2
Module	Number of modules	98,280	98,280	68,796
	Tab ribbon (kg)	1.37 × 104	1.37 × 104	9.56 × 103
	Glass-fibre-reinforced plastic (kg)	1.43 × 104	1.43 × 104	1.00 × 104
	Photovoltaic cell, HJT (kg)	1.97 × 105	1.97 × 105	1.38 × 105
	Solar glass (kg)	2.58 × 106	2.58 × 106	1.81 × 106
	Bus bar + ribbon (kg)	1.67 × 103	1.67 × 103	1.17 × 103
	N-olefins (kg)	7.18 × 104	7.18 × 104	5.3 × 104
	Anti-reflex coating (kg)	1.34 × 102	1.34 × 102	9.4 × 101
	Aluminum alloy (kg)	3.22 × 105	3.22 × 105	2.26 × 105
Tracker	Number of trackers	1260	1260	882
	Steel, low-alloyed (kg)	7.75 × 105	1.11 × 106	5.42 × 105
	Electric motor (kg)	9.83 × 103	9.83 × 103	6.88 × 103
	Printed wiring board for tracker (kg)	7.6 × 101	7.6 × 101	4.94 × 101
	Cables (kg)	2.52 × 103	2.52 × 103	1.76 × 103
Inverter	Number of inverters	13	13	9
	Aluminum, cast alloy (kg)	8.32 × 103	8.32 × 103	5.76 × 103
	Epoxy resin (kg)	3.33 × 103	3.33 × 103	2.30 × 103
	Heat sink in aluminum (kg)	6.24 × 103	6.24 × 103	4.32 × 103
	Printed wiring board (kg)	3.74 × 103	3.74 × 103	2.59 × 103
	Inductors (kg)	1.66 × 104	1.66 × 104	1.15 × 104
	Switches (kg)	8.32 × 102	8.32 × 102	5.76 × 102
	Electric connector, peripheral-type buses (kg)	8.32 × 102	8.32 × 102	5.76 × 102
	Cables (kg)	1.66 × 103	1.66 × 103	1.15 × 103
TOTAL	(kg)	4.3 × 106	4.36 × 106	2.82 × 106

shows the mass balance of components and materials required by each PV configuration. The detailed life cycle inventory is provided in Supplementary Materials S1.3.

4. Results

This chapter is divided into two parts. In Section 4.1, we present the monetary results obtained by applying the methodology to the case studies described in the previous section—namely, the ecosystem-related external costs per GWh due to land-use changes associated with the three photovoltaic plants assessed from a life cycle perspective. To enhance the understanding of these results, in Section 4.2, we present selected intermediate results related to land use obtained through the partial application of the proposed methodology to the case study (up to Step 5, before the final economic valuation).

4.1. Monetary Results

The section is organized into three parts: Section 4.1.1 first presents the results obtained for the production, transport, and end-of-life phases of the PV components; Section 4.1.2 reports the results for the construction and operation phases of the PV plants; and Section 4.1.3 summarizes the aggregated results for the entire life cycle.

4.1.1. Results for Production, Transport, and End-of-Life Phases

Figure 3 illustrates the external costs associated with the production, transport, and end-of-life phases of three photovoltaic (PV) plant configurations, expressed in EUR per GWh of electricity generated over a 30-year operational lifetime.

Panel (a) presents the net variation in ecosystem service (ES) value for each configuration. Across all three cases—base case (ground-mounted PV), Alternative 1 (elevated agrivoltaics), and Alternative 2 (spaced agrivoltaics)—the net variation is negative, indicating external costs. The average loss is approximately –3 EUR/GWh, reflecting a small net degradation of ES. Among the configurations, Alternative 1 shows a slightly higher external cost, primarily due to the increased use of steel in the elevated support structures, which raises the material intensity per unit of energy produced. However, the differences in net losses across configurations are marginal, suggesting that the choice of configuration has limited influence on the overall ES cost profile for these phases of the life cycle.

Panel (b) provides a breakdown of ES impacts by category—provisioning, regulation and maintenance, and cultural services—for the base case. The results are not uniform across categories. Both provisioning and regulation and maintenance services exhibit clear net losses, largely driven by the impacts of module production. In contrast, cultural services show a slight net gain, attributed to positive contributions from module and inverter production. These gains in cultural services are partially offset by negative impacts from module transport (notably from China to the EU) and tracker production. This heterogeneity highlights the importance of disaggregating ES impacts to capture biome-specific dynamics.

For a deeper understanding of the results, Figure 4 provides a breakdown of the net variations in the specific values of ES by biome classes and ecosystem service categories.

Almost all biome categories—particularly ‘marine ecosystems’, indicated in blue in Figure 4—experience a net loss in the value of ES (values with a negative sign). This outcome is primarily attributed to the significant loss in marine areas, measured in hectares per kilowatt-hour (ha/kWh), due to land-use transformations driven by PV processes. Marine ecosystems also exhibit relatively high per-hectare ecosystem service values in the ESVD (1139–1206 EUR/ha-year, see Table S19), amplifying the impact of their loss. A representative example of such transformations is the installation of drilling rigs on the seabed for fossil fuel extraction. This activity involves reclassifying land use from ‘seabed, unclassified’ to ‘seabed, drilling and mining’, thereby diminishing the associated ecosystem service value.

The only biome category showing a net increase in the value of ES is ‘anthropogenic systems’ (highlighted in red in Figure 4). This is consistent with findings in Section 4.1, which show that this is the only biome category that expands its net surface area (ha/kWh) due to PV-related land-use transformations. Although ‘urban areas with artificial cover’ also increase in surface area, they are not associated with ES in the ESVD [48,86]. The ‘anthropogenic systems’ biome, as defined by the ESVD classification [93], encompasses natural aquatic ecosystems (both terrestrial and marine) that have been heavily modified by human infrastructure. In our methodology, due to the limited availability of comprehensive ESVD estimates across all ecozones within this biome, we monetized only land-use changes related to a specific subclass—‘anthropogenic freshwater systems’. This subclass includes artificial reservoirs used for hydroelectric power generation. According to studies included in the ESVD database, artificial reservoirs are associated with a high average cultural ecosystem service value (1530 EUR/ha-year), primarily due to recreation and tourism opportunities. In summary, the observed increase in the value of ES for anthropogenic systems in Figure 4 is driven by land-use changes linked to hydroelectric power use in PV component production.

4.1.2. Results for the Building and Operation Phase

Unlike the production, transport, and end-of-life phases, the operation phase (specifically, the phases of construction, operation, and dismantling of the photovoltaic plant) yields markedly different results depending on the configurations (Figure 5).

The base configuration (ground-mounted PV) results in a net increase in the value of ES, averaging EUR 929 per GWh produced over the 30 years of plant operation. The a2 configuration (spaced agrivoltaics) maintains a net benefit but to a lesser extent (80 EUR/GWh), while the a1 configuration (elevated agrivoltaics) results in a net loss in the value of ES, amounting to −155 EUR/GWh. The positive outcome in the two ground-mounted configurations (base case and a2) is fundamentally due to the net increase in ES value that occurs when land-use class ‘permanently irrigated arable land’ (CLC code 2.1.2) and its associated ecosystem type ‘cropland’ changes into land-use class ‘natural grassland’ (CLC code 3.2.1) and its associated ecosystem type ‘grassland’. This transformation involves the spaces beneath the modules in the two ground-mounted configurations (base case and a2—spaced agrivoltaics) and the open spaces between PV module rows in the base case only, since spaced agrivoltaics maintains the agricultural activity between PV rows.

Figure 6 offers a breakdown of results obtained for the operation phase, highlighting the contribution of specific ES (panel a) and the per-hectare ES values of the ecosystem types affected by the installation of the PV system, based on a literature review of studies supporting the mapping and accounting of ES in Italy (panel b).

In panel (a), the net positive effect occurring to the two ground-mounted configurations is primarily driven by ‘ES5—habitat and biodiversity’ and ‘ES9—erosion mitigation’. In panel (b), these ES values are significantly higher for the ‘grassland’ ecosystem compared to ‘cropland’, reflecting the relative loss of habitat and higher erosion rates associated with agriculture activities (particularly those involving ploughing, as in arable land) as compared to spontaneous grassland. Regarding provisioning services (ES1—crops and ES3—water), the per-hectare value of 107 EUR/ha-year for ES1 related to cropland reflects the contribution of nature to the market value of durum wheat production—the most common arable crop on Italy. This estimate is based on the $\mathrm{E}\mathrm{c}\mathrm{o}\mathrm{C}\mathrm{o}{\mathrm{n}}_{\mathrm{c}\mathrm{r}\mathrm{o}\mathrm{p}\mathrm{s}}$ coefficient of 0.11, calculated by [28] using the “emergy” method, as suggested by the JRC in the pilot European ecosystem accounting study [27]. The per-hectare values for ES3—water provisioning (37 EUR/ha-year for CLC 2.1.2 ‘permanently irrigated arable land’ and 140 EUR/ha-year for CLC 3.2.1 ‘natural grassland’), calculated using the rent of resource method (System of Environmental Economic Accounting, chapter 5 [94]) applied to Italian data on groundwater provisioning by ecosystem categories [87], reflect differences in water balance across ecosystems (e.g., the lower value for cropland as compared to natural grassland reflects the lower groundwater recharge due to water consumptions in agriculture). The net result for these two provisioning services is slightly positive for both ground-mounted plant configurations, primarily due to the improvement in the water provisioning service (ES3) resulting from the conversion from cropland to grassland cover. As for the cultural ecosystem service ES11—natural recreation, our methodology for Italy considers the mapped recreational value (cost of intraday trips) of different ecosystem categories for both residents [26] and non-residents [50,87]. All three configurations experience a loss of cultural ES value as compared to the ex-ante situation, due to the fencing of the area hosting the PV plants and the resulting inaccessibility of potential visitors. The loss per GWh is relatively greater for spaced agrivoltaics due to its lower production for the same fenced area.

The results on the operational phase of PV plants highlight the importance of analyzing land use and ecosystem characteristics when planning utility-scale PV development. Minimizing ecosystem-related external costs requires prioritizing areas with negligible ecosystem service values—such as urban and industrial zones, mineral extraction sites, or landfills—or selecting agricultural areas with relatively low ecosystem service values, such as arable land in non-protected zones (characterized by low habitat quality for wildlife species) used for low-value crops, such as durum wheat, that are intensive in water use and chemical inputs (e.g., pesticides and fertilizers).

The findings of our case studies also emphasize the relevance of implementing basic ecological best practices in the management of ground-mounted PV systems. These practices include periodic vegetation cutting beneath the modules through animal grazing within the facility area (thus utilizing the ecosystem service of forage provision), avoiding the use of herbicides and pesticides, and promoting beekeeping to support the pollination ecosystem service. In this regard, the contributions of Semeraro et al. [55,56,57] are noteworthy for their proposals to enhance the ES of photovoltaic plants through ecological, botanical, and bio-engineering measures applicable to both existing and newly designed installations.

4.1.3. Aggregation of Life Cycle Phases

The final results of our monetary LCA exercise on different PV system configurations are obtained by aggregating partial results for the life cycle phases of the plants (Figure 7). The overall net variation in the specific value of ES, associated with land-use changes along the life cycle, is positive for the base case and a2 (spaced agrivoltaics), with 925 EUR/GWh and 77 EUR/GWh, respectively. Conversely, the net variation is negative for the a1 configuration (elevated agrivoltaics), with −159 EUR/GWh. To provide a comparison with the average wholesale reference price (so-called PUN) of electricity purchased on the Italian Power Exchange market (0.143 EUR/kWh in January 2025, GME, 2025 [95]), the external benefit of the base case is 0.65% of PUN, while the external cost of the elevated agrivoltaics is 0.11%.

The breakdown of the final results for ecosystem service macrocategory varies across the three PV configurations. As seen in the previous section (results of the operational phase), the variability across configurations is particularly high for the regulation and maintenance ES macrocategory. The significant net positive effect of the base case is due to the abandonment of the ‘cropland’ ecosystem type (which has relatively low values for regulation-type ES, such as habitat for biodiversity and erosion mitigation) and its replacement with ‘grassland’ surfaces (below the modules and between the strings). This effect counterbalances and surpasses the negative impact caused by the partial substitution of ‘cropland’ with ‘urban’ surfaces (artificial coverage due to ancillary works such as service roads and cabins, which are assumed to cover the same surfaces in all three configurations). In the case of elevated agrivoltaics with continuation of agricultural activity (a1) the substitution of cropland ecosystem surfaces is minimized (the dual objective is to produce energy and conduct agricultural activities), and the agrivoltaic system continues to provide the ES of the pre-existing form of agriculture. A net loss in ES value is avoidable only if the agrivoltaic system implements a change in the type of crops (and cultivation practices) that enhances the quality of the regulatory ES associated with agricultural activities. However, this aspect of further improvement offered by the change in agricultural practices related to agrivoltaic systems is not within the scope of this article and should be considered as a topic for further research.

4.2. Inside the Box: Intermediate Results on Land Use and Ecosystem Changes Due to PV Plants

This section presents selected intermediate results obtained through the partial application of the proposed methodology to the case study. The focus is on land use and related ecosystem changes—specifically, the net loss or gain in terms of hectares of ecosystem types affected by land-use changes, as well as the duration of these changes, as calculated in Steps 4 and 5 of our methodology ($\Delta Are{a}_{ij}$ and T of Equation (1) of Section 2). It is important to note that these are extrapolated intermediate results, presented to support and enhance the interpretation of the monetary valuation outcomes presented in Section 4.1.

4.2.1. Intermediate Results on Land Use and Ecosystem Changes for Production, Transport, and End-of-Life Phases

We start with the most significant intermediate results for the production, transport, and end-of-life phases (the building and PV plant operation phases are addressed in Section 4.2.2). These results pertain to the analysis of ecosystems affected by land-use changes —measured in hectares per biome type—associated with each component/process of the PV system configurations.

As mentioned in methodology Section 2, the ‘land use’ flows resulting from the Ecoinvent inventories of the three PV plant configurations are grouped into three macro-classes [16]: ‘Transformation to’ (m²), ‘Transformation from’ (m²), and ‘Occupation’ (m²-year). ‘Occupation’ provides information that accounts for the period of land occupation in a specific land-use destination (the occupied areas in a given land-use state are multiplied by the years of occupation) but does not quantify land-use changes (it refers exclusively to land use in the ex-post situation—for example, after a production facility has been built). The two ‘Transformation’ indicators, on the other hand, allow for accounting of the areas affected by land-use changes occurring in the production, transport and end-of-life phases. For the calculation methodology, the average duration of land use, expressed in years (the T component of Equation (1)), is motivated by the fact that the external costs of land transformations induced by the PV system are calculated as a difference between the ex-post and ex-ante values related to the time flow of ES, which are evaluated in EUR/ha-year. The average duration of land use is not directly specified by the Ecoinvent dataset [15] but can be inferred retrospectively. To simplify the calculation procedure, this work assumes an average land occupation duration equal for all production, transport, and disposal processes of the PV system. This is obtained by dividing the sum of all ‘Occupation’ elementary flows (unit of measure m²-year) by the sum of the ‘Transformation to’ elementary flows, measured in m²). The average duration of land occupation T, expressed in years, of the production, transport, and end-of-life processes of the plant components was found to be approximately 44.3 years. Small differences are obtained depending on the plant configuration: 44.33 years for the base case, 44.30 years for elevated agrivoltaics, and 44.32 years for spaced agrivoltaics. This means that the material quantities that differentiate the design of the three PV configurations (e.g., the tracker support structure of the ‘elevated agrivoltaic’ is much heavier than the base case; the material weight of the base case is much higher than the ‘spaced’ PV system) minimally affect the average duration of land occupation processes.

In our methodology, detailed results obtained with Ecoinvent for land-use classes are aggregated into biome classes (ecosystem types) by using a correspondence table (Table S17 in Supplementary Materials S1.5.1) between the land-use classes used by the Ecoinvent database [15] and the biome classification of the ESVD ecosystem services valuation database [93]. This table’s function is to link Ecoinvent’s land-use data (used to elaborate the expression $\Delta Are{a}_{ij}$ of Equation (1)) with the monetary values of ES in a given geographical area associated with ESVD biomes ($V_{ij}$ of Equation (1)).

Table 4. Reference ground-mounted photovoltaic plant. Net variation in surface area (ha/kWh) of the transformed biome. Negative values indicate a loss of surface area for the type of biome, while positive values indicate an increase.

	Total	Tracker	Module	Inverter	Tracker Transport	Module Transport	Inverter Transport	Gravel	Module End of Life	Inverter End of Life	Tracker End of Life
Components/Region (Columns) Biome Types (Rows)		EU	China	EU	EU	China- Italy	EU	EU	EU	UE	EU
(Aquatic) anthropogenic systems	4.79 × 10−11	4.90 × 10−12	3.28 × 10−11	9.02 × 10−12	1.20 × 10−13	2.49 × 10−13	4.25 × 10−15	5.87 × 10−13	2.00 × 10−13	5.69 × 10−15	2.47 × 10−14
Forests	−3.65 × 10−11	−4.27 × 10−12	−1.99 × 10−11	−8.00 × 10−12	−5.00 × 10−13	−3.05 × 10−12	−1.77 × 10−14	7.87 × 10−14	−5.82 × 10−13	1.26 × 10−15	−2.59 × 10−13
Inland wetlands	−3.25 × 10−12	−7.15 × 10−14	−3.08 × 10−12	−7.26 × 10−14	−7.60 × 10−15	−1.26 × 10−14	−2.69 × 10−16	−4.23 × 10−16	−6.58 × 10−15	−1.11 × 10−16	−6.17 × 10−16
Intensive land use (cropland)	−2.50 × 10−12	−4.70 × 10−12	−3.73 × 10−12	6.87 × 10−12	−2.34 × 10−13	−1.82 × 10−13	−8.28 × 10−15	−1.64 × 10−14	−3.25 × 10−13	−8.33 × 10−15	−1.59 × 10−13
Marine	−1.45 × 10−10	−2.01 × 10−11	−8.99 × 10−11	−1.51 × 10−11	−2.68 × 10−12	−1.15 × 10−11	−9.46 × 10−14	−1.64 × 10−13	−3.66 × 10−12	−1.32 × 10−14	−1.37 × 10−12
Rangelands and natural grasslands	−1.77 × 10−12	−4.30 × 10−13	−1.04 × 10−12	−2.51 × 10−13	−8.20 × 10−15	−2.08 × 10−14	−2.90 × 10−16	−4.70 × 10−16	−1.16 × 10−14	−1.78 × 10−16	−8.69 × 10−16
Rivers and lakes	−1.60 × 10−12	−3.28 × 10−14	−1.51 × 10−12	−2.93 × 10−14	−5.55 × 10−15	−1.32 × 10−14	−1.96 × 10−16	−2.68 × 10−16	−5.99 × 10−15	−8.99 × 10−17	−1.77 × 10−16
Shrubland and shrubby woodland	−5.55 × 10−12	−6.91 × 10−13	−4.22 × 10−12	−5.88 × 10−13	−8.52 × 10−15	−1.30 × 10−14	−3.01 × 10−16	−5.19 × 10−15	−2.87 × 10−14	−4.27 × 10−16	−1.47 × 10−15
Urban areas with artificial cover	3.77 × 10−12	5.30 × 10−12	1.20 × 10−12	−6.95 × 10−12	6.45 × 10−13	3.04 × 10−12	2.28 × 10−14	−6.45 × 10−13	7.61 × 10−13	2.19 × 10−15	3.96 × 10−13

reports the area variation in terms of ha/kWh, for each biome type (ESVD classification) and component/process involved in the construction of the photovoltaic base case (ground-mounted photovoltaics). Negative values represent a net loss of area for a given biome type, while positive values indicate a net gain. To enhance the interpretability of the results, Table 4 includes data for the biome category ‘urban areas with artificial cover’, even though this category has recently been excluded from the ESVD classification system [86]. This exclusion implies an assumed ecosystem service value of zero for urban and industrial land covers, which are characterized by a high degree of soil sealing.

The total values in Table 4 show that the production, transport, and disposal processes of photovoltaic (PV) components result in a net loss of area for pristine marine ecosystems. This is due to their transformation into ‘aquatic anthropogenic systems’ (which show a positive net change), primarily driven by the creation of offshore fossil fuel extraction sites. Besides the net loss of pristine marine ecosystems, there is a total net loss of area for the following biomes: ‘forests’, ‘shrubland and shrubby woodland’, ‘inland wetlands’, and ‘intensive agricultural land’. Conversely, in addition to the net increase in ‘aquatic anthropogenic systems’, the PV system also leads—as expected—to a total net increase in ‘urban areas with artificial cover’. Since this biome category includes industrial zones, the increase is attributable to the land requirements of extraction and production sites for PV components, transport infrastructure, and waste management facilities.

Similar results are obtained for Alternative 1 (elevated agrivoltaics configuration does not result in significant land-use variations for the production, transport, and end-of-life phases), while higher net ha/kWh values are obtained for Alternative 2 (spaced agrivoltaics) due to the lower lifetime production of the latter. Further analysis of intermediate results (areas) is offered in Supplementary Materials S2.2 Analysis of Contributions to Land-use Change by Component and Process.

4.2.2. Intermediate Results on Land-Use and Ecosystem Changes—Construction and Operation Phase

For the construction and operation phase, we directly employ spatially explicit mapping by comparing the land-cover situation before the plant’s construction (ex-ante) with the post-construction cover. Given the importance of the construction and operation phases in driving land-use changes associated with photovoltaic (PV) installations, we opted for a direct investigation methodology—spatial analysis of land-use changes—rather than relying on standard data from LCA inventory databases, which represent unverifiable secondary sources. We assumed that the new land-cover classes, identified through direct observation of the land-use types characterizing photovoltaic systems in their various configurations, persist throughout the operation phase until the plant is decommissioned. At that point, the land-use class is assumed to revert to its pre-construction state following site restoration. Also, for the construction and operation phase, we use a correspondence table (Table S18 in Supplementary Materials S1.5.2) between land-use classes (Corine Land Cover, the European system of land-cover detection and classification based on the visual interpretation of high-resolution satellite images [91]) and the MAES ecosystem classes adopted by our ES values dataset [74]. This correspondence table was originally elaborated by JRC within the KIP-INCA project [25,26,27,49], aimed at experimenting with both physical and monetary accounts of ES in the EU.

For the reference photovoltaic plant, the fenced area within which the main structures and almost all ancillary works are constructed (excluding the user substation of 0.4 hectare) covers an area of 79.2 hectares, of which 26.9 hectares pertain to the ground projection of the modules, 4.7 hectares to the perimeter and internal service roads, and 0.08 hectares to various cabins.

Regarding land-use type in the ex-ante situation, the reference PV project documentation indicates agricultural use with durum wheat and maize crops, uniformly classified in the cartography as ‘irrigated arable land’ (CLC 2.1.2) in all portions of land affected by the plant (including the user substation). Additionally, the same project documentation assumes that the construction of the ground-mounted photovoltaic plant will lead to the abandonment of pre-existing agricultural activity and that, during the plant’s management phase, periodic cutting of spontaneous vegetation will be carried out for maintenance purposes. Ground-mounted PV systems do not typically require costly concrete foundations or other structural platforms. The support poles are directly driven into the soil. Rainwater can infiltrate the ground and contribute to groundwater recharge. Beneath the modules, there is an open space—whose height varies depending on the tilt of the modules, ranging from 0.5 to 4.7 m—that allows vegetation to grow, and no soil sealing occurs. Among land-use classes, ‘natural grassland’ is the most appropriate for representing the spontaneous vegetation that develops beneath and between the panels and is periodically mowed for maintenance purposes. Therefore, for the evaluation of the artificial land cover of the reference plant, only the ancillary works should be accounted for (5.2 ha out of 79.6 hectares, equivalent to 6.5%), while the remaining part of the area (including the spaces under the panels) can be classified as the CLC 3.2.1 ‘natural grassland’ sub-class (‘grassland’ ecosystem type in MAES classification).

Figure 8 summarizes the land-use changes of the reference photovoltaic plant with the respective affected surfaces.

For elevated agrivoltaics (Figure 9), it is assumed that in the ex-post situation, agricultural activity can be carried out not only in the open spaces between the module rows but also in 80% of the ground projection area of the modules (a non-cultivable strip of land approximately one metre wide running along the tracker support poles was assumed). Overall, the system’s surface area designated for agricultural activity amounts to 87%.

In the spaced agrivoltaics (Figure 10), agricultural activity can only continue between the PV module strings. Since the plant is characterized by a lower power density (fewer module rows, more spaced out), compared to the reference plant, the surface area under the modules (spontaneous vegetation) is reduced, while it is possible to use the larger spaces between the rows for agricultural purposes, so the agricultural use surface reaches 70% of the system’s total surface area.

Table 5. Building and operation phases. Net ex-post–ex-ante change in surface area (in ha and ha/kWh) of the transformed ecosystem type. Negative values indicate a loss of surface area for the ecosystem type, while positive values indicate an increase.

Land Cover (Ecosystem Type)	Base Case Ground-Mounted Conventional		Alternative 1 Elevated Agrivoltaics		Alternative 2 Spaced Agrivoltaics
	ha	ha/kWh	ha	ha/kWh	ha	ha/kWh
Cropland	−79.6	−2.79 × 10−8	−10.6	−3.72 × 10−9	−24.0	−1.20 × 10−8
Grassland	74.4	2.61 × 10−8	5.4	1.89 × 10−9	18.8	9.43 × 10−9
Urban cover	5.2	1.83 × 10−9	5.2	1.83 × 10−9	5.2	2.61 × 10−9
Total variations	0	0	0	0	0	0

summarizes the net variation in surface area of the affected ecosystem types—cropland, grassland, and urban cover—across the three photovoltaic (PV) system configurations, expressed in both absolute terms (ha) and relative terms (ha/kWh). In the base case, cropland loss is total (−79.6 ha), but it is almost entirely offset by an increase in grassland cover (+74.4 ha), with the remaining difference attributed to artificial surfaces associated with ancillary infrastructure (+5.2 ha). In the elevated agrivoltaics configuration, cropland loss is limited to 10.6 ha (13% of the available area) and is only partially compensated by grassland cover established beneath the supporting structures. In the spaced agrivoltaics configuration, cropland loss is greater (−24 ha) than in the elevated system, but approximately 78% of this loss is compensated by an increase in grassland cover. All PV configurations require the same amount of artificial land cover in absolute terms (5.2 ha, that is, about 7% of the affected surfaces).

The land-use impacts per kWh in this phase are at least three orders of magnitude higher than those observed in the other life cycle stages (4.2.1). Intermediate results highlight how system design choices directly influence the land-use efficiency of PV installations, with elevated agrivoltaics offering the lowest land transformation per kWh (as compared to the ex-ante agricultural land use), and the base case showing the highest cropland loss.

5. Discussion

The proposed methodology represents an original approach to assess ecosystem-related external costs of ground-mounted PV systems, while acknowledging certain simplifications and limitations. Our methodology estimates the loss or gain in ES value associated with land-use change; it does not independently assess the direct effects of the renewable energy plant on ES separate from land-use change. This scope limitation excludes landscape effects over a broader geographical area surrounding the plant, particularly those related to its visual impact, where land-use remains unchanged following the plant’s installation. This type of impact is typically assessed when evaluating the external costs of renewable energy plants [61,62]: the scope of our methodology allows for the monetary assessment of ecosystem effects at a local scale (i.e., land surfaces undergoing land-use change), which—if a comprehensive evaluation is to be achieved—should be considered alongside landscape external costs related to visual intrusion occurring at a broader geographical scale. Another boundary of the methodological scope is that the LCA of crop production related to the agricultural side of agrivoltaic configurations is excluded. We assess different technological configurations of PV plants, assuming—in the case of agrivoltaic plants—the maintenance of the previous type of agricultural activities. The assessment of different agrivoltaic configurations using the same PV technology but different crop choices is out of scope.

Further limitations can be identified in relation to specific steps of the methodology.

LCI Modelling Phase: The proposed methodology relies on the availability of inventory databases detailing land use and land-use changes at the level of elementary activities, such as raw material extraction and primary transformation processes. The Ecoinvent database [15], commonly used in LCA, provides inventory data on land-use transformation and occupation, enabling aggregated accounting of land-use changes. However, it lacks granularity in tracing elementary land-use transformations from one state to another, presenting an obstacle in verifying elementary data and interpreting results at the component or process level. Future developments should aim to create a life cycle inventory-level indicator that fully characterizes land-use change with detailed temporal and spatial information on occupation and transformation. Additionally, aligning this indicator with recognized ecosystem classifications, such as the CLC—Corine Land Cover [91] and MAES—Mapping and Assessment of Ecosystems and their Services [49,96], would enhance its utility.

Monetary Evaluation Phase: The methodology’s monetary evaluation phase involves applying benefit transfer techniques to calculate regionalized unit values for ES based on available literature-based datasets. Two datasets were used: a global data set [48] and a country-specific one for Italy [74]. The first one (ESVD) analyzes site-specific studies, standardizing values in relation to surface area and currency. To obtain representative regional values, outliers must be excluded and large samples used, which requires expanding the geographic area for averages. However, ESVD’s limited observations for Italy necessitated the use of a separate source of values for ES based on experimental national accounting studies for ES in Italy and in the EU. These studies calculate average national ES values per hectare, though they are limited by the number of ES categories assessed and the heterogeneity of statistical sources on ES across Member States. The constraints in monetary evaluation data sources can be alleviated through continued empirical research and efforts to catalogue more studies and experimental reports on national accounting of ecosystem service value in the ESVD database. Finally, the scope of analysis in this article is limited to ES, excluding human contributions to economic value formation. This article explicitly excludes the contribution of labour and machinery to the formation of agricultural added value. However, this value component—preserved over time by agrivoltaic systems and lost by conventional ground-mounted systems—should be accounted for separately within cost–benefit analyses of such projects.

In summary, while the methodology for valuing ES related to land-use changes of utility-scale photovoltaic plants from an LCA perspective is innovative, it faces challenges in data granularity, regional representativeness, and scope limitations. Future research and database enhancements are essential for improving the accuracy and applicability of this approach.

Despite its limitations, this study offers valuable insights for energy policy, land-use planning, and the environmental practices of energy companies—particularly in the design and siting of photovoltaic (PV) installations. The proposed methodology is not intended to replace existing tools for assessing the environmental and land-use impacts of renewable energy systems [4] during the environmental impact assessment (EIA) procedure [6]. Rather, it aims to complement them by introducing an additional monetary evaluation grounded in the ES theoretical framework.

Monetizing impacts of land-use change due to large-scale PV systems through the ES approach presents several advantages. It provides a concise and quantitative basis for analyzing trade-offs among the complex environmental dimensions affected by the construction of the system in agricultural areas. For instance, it helps balance external benefits such as higher water availabilty against external costs like the loss of recreational opportunities. Furthermore, it allows practitioners of cost–benefit analysis (CBA) of renewable energy projects to enhance the robustness of their exercises by incorporating ecosystem-related external costs into the broader accounting framework of CBAs, alongside technology costs and renewable-energy-related benefits such as CO₂ emissions savings [70]. Similarly, incorporating estimation of ecosystem-related external costs with a life cycle perspective—alongside the health-related external cost [17]—enriches energy scenario evaluations. This enables more accurate comparisons between the externalities of renewable and fossil-fuel-based energy technologies. As noted in the Introduction, LCA is increasingly required in public policy and private investment contexts, such as the EU Taxonomy for Environmentally Sustainable Finance [7]. Although this study does not directly address the technical screening criteria for PV systems under the Taxonomy [8,9], the case studies presented here offer preliminary evidence of negligible ecosystem-related external costs. This supports the current absence of quantitative thresholds for the “Do No Significant Harm” (DNSH) criteria concerning ecosystem impacts of PV installations. Qualitative guidelines could instead be introduced in the Taxonomy to ensure that ground-mounted photovoltaic installations do not result in land take, such as the use of structures mounted on poles directly driven into the ground, avoiding cement platforms, or artificial foundations.

Moreover, the component of our methodology addressing the operational phase of PV plants can inform land-use planning by guiding the selection of suitable areas for large-scale PV deployment. In Italy, recent legislation [10] prohibits the construction of conventional ground-mounted PV systems in agricultural areas, while permitting innovative agrivoltaic systems with a minimum clearance of 2.1 m to accommodate agricultural machinery ([76] and Decreto-Legge 24 gennaio 2012, n. 1 [97]). This policy choice appears to be driven more by socio-economic considerations related to agriculture than by the protection or enhancement of ES.

Our case studies—based on three configurations located in agricultural areas characterized by low-value intensive farming (i.e., no crop rotation, use of synthetic fertilizers and pesticides, and mechanized tillage)—suggest that siting PV systems in arable, non-protected areas does not necessarily imply a net loss of ecosystem service value related to land-use changes. This is particularly true when these areas do not host high-value ecosystems such as forests, woodlands, heathlands, natural grasslands, and wetlands. Accelerating investments in solar PV systems in arable areas could even yield net ecosystem service benefits, provided that energy companies adopt simple, practical management measures. Such measures could also benefit local communities and ecosystem-related professions [52]. For existing ground-mounted PV systems (excluding agrivoltaic configurations), transitioning from a purely energy-focused management approach to one that incorporates ES should be encouraged. This includes prohibiting the use of pesticides and herbicides for grassland cover maintenance (this would reduce the assumed value for ES5 “habitat quality and biodiversity”), promoting beekeeping and pollination services (ES6 “pollination”), and planting native botanical species (ES5) [53,59].

In new agrivoltaic systems, integrated management of energy, agriculture, and ecosystem components should be pursued [54]. Despite our methodology not being conceived to specifically compare agrivoltaic systems differentiated on the basis of new types of crops, it allows for preliminary recommendations based on the technological side of such systems. For example, adoption of water-saving irrigation systems can contribute to increase the assumed cropland-related average value of ES3 “water provisioning”, and enabling access to the PV plant’s service roads for recreational use could contribute to maintain the recreational value of agricultural ecosystems (ES11), further enhancing the multifunctionality of these installations.

6. Conclusions

This study presents a methodology for analyzing ecosystem external costs associated with land-use changes of PV systems through a ‘regionalized’ life cycle approach. This approach assumes that key components and processes of photovoltaic (PV) plants, along with associated land-use-change patterns, occur in specific geographical areas representative of the supply chain markets of the host nation. The methodology aim is to assess the net ecosystem service value change related to life cycle land-use change of PV plants at a regional scale, covering wide regions such as China and Europe. In the case of agrivoltaics configurations of PV systems, the methodology refers only to the technological side of the system, excluding changes in crop types related to the agricultural side of agrivoltaic configurations. The methodology was applied to three utility-scale PV plant configurations with solar tracking built in an arable agricultural area: ground-mounted PV, elevated agrivoltaics, and spaced agrivoltaics. The study estimated the ecosystem service losses (external costs) or gains (external benefits) generated by each configuration in relation to the GWh produced, enabling a comparison in a life cycle perspective.

The results of the chosen case study indicate that the elevated agrivoltaics configuration, which allows for the continuation of pre-existing agricultural activities (e.g., durum wheat), is the only one that generates ecosystem external costs as compared to the pre-contruction situation, albeit small, at −159 EUR/GWh. This cost is much less than a thousandth of a EUR per kWh, equivalent to 0.1% of the electricity price on the Italian wholesale market. The other two configurations—ground-mounted PV and spaced agrivoltaics—generate net ecosystem-related benefits for society as compared to the ex-ante assumed rural context, characterized by arable land used for cereal cropping. The ground-mounted PV configuration generates a net ecosystem benefit of 925 EUR/GWh, nearly a thousandth of a EUR per kWh, equivalent to 0.65% of the Italian wholesale electricity price. This benefit arises because solar trackers are mounted on poles driven into the ground without foundations, allowing for grass cover regeneration, which combats erosion, restores soil structure and fertility, supports biodiversity, and contributes to carbon sequestration. The spaced agrivoltaic configuration generates a net benefit of 77 EUR/GWh, as cropland transformation into grassland is limited to areas beneath the modules, while agricultural activities continue between the strings. The results for agrivoltaic configurations are closely linked to the assumption of maintaining the same agricultural activity after the construction phase. The proposed methodology should be further developed to allow a more granular modelling of ES valuation of the agricultural side of agrivoltaic systems, but this outcome needs further research. The adoption of different crops in the agrivoltaic system could potentially increase the economic value of agriculture-related provisioning ES (ES1 crop production and ES3 water provision) but effects on the economic value of regulation and maintainance ecosystem services (ES4–ES10) should be specifically further investigated in relation to main agricultural practices.

The life cycle perspective adopted by our methodology also reveals that all three PV configurations produce ecosystem external costs during the production, transport, and end-of-life phases of their components. However, these costs are minimal, just over 3 EUR/GWh, with small differences between configurations, and significantly lower than those due to the construction and operation phase. The substantial difference between the two phases can be explained by the duration of land-use changes. The land-use changes due to the construction of the PV plant persist for its 30-year operational life (10,950 days). In contrast, the production of the photovoltaic modules required for a 56 MW plant—assuming a manufacturing facility in China with an annual capacity of 1 GW—can be completed in approximately 20 days. This duration represents only 0.13% of the total operational phase of the plant.

The main practical implications of the proposed methodology, taking into account the preliminary results obtained from our case study, can be listed as follows:

• Not all cropland has equal ES value and well-designed ground-mounted PV on intensive arable land can deliver a net ecosystem service benefit. Our results for the ground-mounted conventional PV plant confirm the current approach of the EU Taxonomy (no quantitative DNSH criteria for ecosystem services) while at the same time refuting the ecosystem-based justifications for the recent Italian restrictions on ground-mounted PV in agricultural areas.
• Supply chain ES impacts are comparatively smaller than those related to construction-site land-cover changes; therefore, siting, PV plant design, and Operation and Maintenance (O&M) are decisive for achieving a net positive outcome.
• During the siting process, low ES value areas should be prioritized (e.g., arable land with intensive cereal cropping in non-protected zones or already urbanized/industrial land) to avoid high-value ecosystems (forests, wetlands, natural grasslands).
• Design and vegetation management matter: where trackers are mounted on posts without concrete slabs and spontaneous herbaceous cover is maintained during operation (no herbicides), ES value related to habitat quality, erosion mitigation, and water provisioning increases compared to the ex-ante agricultural situation.

Based on our analysis, the following stakeholder-specific recommendations can be formulated:

• For policymakers, regulators, and regional planners:
  Land mapping of ES benefits by land-use class can be used during EIA to prioritize PV plant siting; ES externalities can be integrated into CBA as a decision-making tool; low-impact construction standards can be developed in line with ES (pile-driven posts, minimal road footprints, permeable surfaces, and vegetated inter-row areas); good environmental practices should be encouraged (prohibition of herbicide, managed grazing, and pollinator habitat). Given that ES value for nature-based recreation (ES11) decreases due to fencing, perimeter trails or guided access to preserve some recreational value is recommended. If elevated agrivoltaics is supported to maintain agricultural income, ecological management requirements (e.g., crop rotations, low-input practices, water-saving irrigation, pollinator strips) can be linked to incentives so that regulation and maintenance ES also improve in portions of land where agricultural activities are maintained.
• For developers and engineering companies:
  They can choose and design configurations to fit the ES context: on low-value arable land, conventional ground-mounted PV with vegetation-positive O&M tends to maximize ES net benefits; where arable crops must be partially maintained, spaced agrivoltaics can balance energy production with ES value better than elevated agrivoltaic; if crop type can be changed to higher-value crops more suitable for the shading features of elevated agrivoltaic, the expected increase in crop provisioning value (ESV1) must compensate for possible negative variations in other ES to maintain a net positive result compared to previous agricultural context.
• For farmers:
  In ground-mounted PV plants, farmers can explore side revenues (grazing services, honey/pollination partnerships) enabled by PV vegetation management plans. Farmers play an essential role in agrivoltaics. Even though our methodology does not model agricultural practices (as they are outside the scope of the article), other sources have demonstrated that improved crop rotations, reduced use of fertilizers and pesticide (Weißhuhn et al., 2017 [98]), integration of cover crops during fallow periods (Huang et al., 2025 [99]), and precision irrigation (De Francesco et al., 2025 [100]) can positively contribute to —respectively— regulation ES, erosion mitigation, and water provisioning. However, these benefits should be still evaluated in economic terms.

While the proposed methodology offers a novel framework for assessing ecosystem external costs related to land-use changes of photovoltaic systems through a regionalized life cycle approach, several limitations warrant further investigation. First, the scope is restricted to land-use changes, excluding broader landscape effects such as visual impacts and ecosystem fragmentation in surrounding areas where land use remains unchanged. Future research should integrate spatially explicit landscape valuation methods to capture these externalities. Second, as previously said, more research is needed to enhance the granularity of the methodology to assess specific crop types and agricultural practices suitable for agrivoltaic configurations. Third, the current features of life cycle inventory data—particularly in the Ecoinvent database—limit the traceability of elementary land-use transformations. Developing enhanced inventory-level indicators with temporal and spatial resolution, aligned with standardized ecosystem classifications such as CLC and MAES, would improve analytical precision. Fourth, the monetary evaluation relies on benefit transfer techniques constrained by limited regional data. Expanding empirical studies and harmonizing national accounting efforts on ecosystem service value within databases like ESVD is essential to strengthen regional representativeness of monetary evaluations. Lastly, our assessment approach is limited to nature-related contributions to economic value (ecosystem services), excluding the human-related contribution of agriculture to economic value formation. This narrow but innovative perspective particularly penalizes the outcome regarding the wider social and economic benefits of agrivoltaic systems. Future work should also explore integrated valuation frameworks that distinguish between ecosystem-related and anthropogenic inputs in agricultural productivity. Addressing these gaps will enhance the robustness, policy relevance, and scalability of ecosystem service valuation in renewable energy planning.