1 Introduction

Renewable energy consumption has grown substantially in recent years owing to their effectiveness in climate change mitigation. Renewable energy sources have higher efficiencies and lower greenhouse gas emissions per energy unit contrary to conventional fossil fuels [1, 2]. The International Energy Agency (IEA) has predicted that renewable energy will constitute 35% of the global energy supply by 2026 [3]. Among several sources of renewable energy, solar power generation has undergone rapid advancement [2]. This trajectory is anticipated to intensify, with renewable energy encompassing nearly 95% of the world’s additional power generation capacity from 2020 to 2026 and solar power generation constituting more than half of this increment [3].

Solar photovoltaic (PV) systems have a distinctive characteristic of relatively low energy density; therefore, they require larger land areas compared with fossil fuel and nuclear power plants (solar PVs: 36.9 km2 TW h1 year1, coal: 9.7 km2 TW h1 year1, nuclear power: 2.4 km2 TW h1 year1; [4,5,6,7]). Thus, the use of solar PVs and the associated transformations in land use have proliferated, leading to concerns regarding their impact on local ecosystems. Solar PVs can also modify the foundational aspects of ecosystems such as microclimatic conditions [8,9,10], sediment transport, rainfall runoff [11], and seed bank survival [12]. Furthermore, PV equipment can pose a direct threat to local biodiversity [9]; they act as barriers to the movement of wildlife [13] and facilitate the invasion of non-native plants [14].

To avoid the conflict between solar PV use and biodiversity conservation, recent efforts have been made to exclude these conservation areas as the construction sites for solar PVs [15, 16] and to include areas where such facilities already exist [17]. Unfortunately, approximately 17.4% of solar PVs worldwide have already been constructed in globally important biodiversity conservation areas, which can increase to 42% by 2028, including facilities that are currently under construction [18]. Furthermore, seminatural environments and fragmented small-scale habitats impacted by human activities have weaker regulatory restrictions [19]; thus, small- to medium-sized solar PVs are constructed therein [20].

Seminatural grasslands harbor unique ecosystems with species that have adapted to periodic disturbances resulting from human activity [21,22,23]. However, decrease in human activities, such as land abandonment, have threatened such ecosystems [24, 25]. Kim et al. [20] demonstrated that PV installations are being prioritized for seminatural grasslands that are often abandoned in Japan. Although this fosters efficient use of the abandoned land, it can pose critical threat to the biodiversity in seminatural grasslands.

Therefore, this study proposes a compatible approach to strike a balance between solar PV installations and seminatural grassland conservation. Seminatural grassland conservation requires maintenance such as regular mowing and controlled burning [22, 26]. Managing the undergrowth to prevent the reduction in power generation efficiency of the photovoltaic system due to shading of the panels also contributes to maintaining a seminatural grassland. On the other hand, the installation of solar panels can disturb the topsoil and cause microclimatic modifications. If these impacts destroy the vegetation in seminatural grasslands, their coexistence with solar PV systems becomes unattainable.

The aforementioned issues can be addressed using stilt-mounted agrivoltaic systems, which are designed to balance solar power generation and vegetation growth beneath the solar panels. These agrivoltaics systems are installed using screw piles instead of a concrete foundation, thereby minimizing topsoil disturbance and reducing the impact of existing vegetation [27, 28]. Moreover, the system is designed with less shading than traditional solar power systems (30–70% depending on the crop), ensuring sufficient solar radiation for crops with smaller height and minimizing microclimatic modifications. Notably, this technology has already been successfully implemented on farms, wherein crops flourished under the panels [29,30,31]. By repurposing this technology, the coexistence of grassland vegetation and solar PV systems can be possibly ensured. This idea is referred to as “Ecovoltaics” approach [44].

In this study, the suitability of installing solar PV in seminatural grasslands was first examined to assess the feasibility of the proposed approach. If the location is deemed suitable, the installations could either pose a risk to the grasslands or their compatibility with the vegetation for their coexistence can be explored. The primary focus of this study is on the latter perspective. For analysis, seminatural grassland areas were identified using location information and satellite imagery. Then, six factors that are frequently used in the suitability assessment of solar PV systems were compared, including sunlight exposure, hours of sunshine, average temperature in August, i.e., the month with the highest temperature in the study region, slope, distance from roads, and distance from transmission lines [15], between seminatural grasslands, existing solar PVs and random points.

2 Methods

2.1 Study area

The study area was the capital city of Japan comprising Tokyo and seven surrounding prefectures (Fig. 1). Tokyo is the most populated city in the world (World Population Review https://worldpopulationreview.com/world-cities#top, accessed at 2024/2/21) and is expected to have high electricity demand. Furthermore, this area has a high probability of housing medium-sized solar power facilities [20], necessitating the coexistence of increased electricity generation and seminatural grassland conservation.

Fig. 1
figure 1

Geographical location and distributions of topography of the capital city (Tokyo and Saitama, Kanagawa, Chiba, Ibaraki, Tochigi, Gunma, and Yamanashi Prefectures)

Full size image

2.2 Identification of grasslands

Grassland locations were identified using two data sources. (1) The grassland database [32], containing location information, was created via literature survey; the registered locations are considered to have been grasslands at the time of survey. Of the 287 registered grasslands in the database, 33 were in the capital city area of Japan. Each location was visually confirmed on aerial photographs (Geospatial Information Authority of Japan GSI Maps https://cyberjapandata.gsi.go.jp/xyz/seamlessphoto/{z}/{x}/{y}.jpg (in Japanese); final confirmation on 2023/11/23), and polygons were created based on the outline of each grassland. If the registered location was not a grassland, but a grassland was clearly identified in proximity (~ 500 m) due to overlaid aerial photography, the closest grassland layer was used for analysis. Additionally, four sites located within a university or a research facility were excluded from the analysis. (2) The National Standard Land-Use Mesh Dataset [33, 34] contained data of the vegetation survey conducted by the Ministry of the Environment, wherein areas of land-use categories (such as natural forests and secondary grasslands) were recorded in 1 km × 1 km mesh (hereafter 3rd mesh). Herein, 3rd meshes dominated by secondary grasslands in the capital area were extracted and analyzed. The land use mesh data classified as “secondary grassland (grassland excluding natural grassland and cultivated meadow)” was identified as secondary grassland.

2.3 Identification of solar PV systems

The suitability of installing solar PV systems in grasslands was evaluated in comparison to the locations of existing solar PV installations. The boundaries of solar PV systems were identified based on the data published by Kim et al. [20] (Table 1) that contained location information provided by a point distribution map of solar power plants published in “Electricity Japan.” (http://agora.ex.nii.ac.jp/earthquake/201103-eastjapan/energy/electrical-japan/; in Japanese Last accessed: 2024/2/26). Polygon data were created by overlaying location information with aerial photographs and visually confirming the boundaries of solar PV systems. Thus, 8725 solar PV locations were identified in Japan and used for analysis [20].We focused on solar power plants located within the capital area of Japan—both large-scale (> 10MW) (PV_L) and small- to medium-sized (PV_S).

Table 1 List of identified objects for analysis, original data extension, and original data source
Full size table

Agrivoltaic systems (PV_A) were identified based on the names of farms listed in the Ministry of Agriculture, Forestry and Fisheries’ case collection [35]. These farms were then searched on Google Maps (https://www.google.com/maps/ Final confirmation: 2024/2/21), and locations that could be visually confirmed on aerial photographs were included in the analysis. Five such locations within the capital area of Japan were identified and included in the analysis. Two of these locations overlapped with the data of small- to medium-sized solar power plants provided by Kim et al. [20] and were therefore excluded from the related dataset.

2.4 Environmental conditions for site suitability analysis

The suitability of solar PV installations in an area was assessed based on six evaluation factors (Table 2). These factors were selected from among those used frequently for evaluation and directly affect the financial performance of solar power projects. Factors related to electricity generation include total solar radiation, sunshine duration, and average summer temperatures. Solar radiation represents the total amount of solar energy per unit time and unit area (kWh/m2); it directly impacts the assumed annual electricity generation [16]. Sunshine duration represents the duration (in hours) during which direct normal irradiance exceeds 120 W/m2. This factor also directly influences electricity generation and is frequently used for site evaluation [15, 36, 37]. High solar panel temperatures reduce the efficiency of solar systems and impacts electricity generation [38]. Therefore, temperature was also considered an evaluation factor, and the average temperature in August, when temperatures are the highest in Japan, was used. Although slope orientation is a commonly used factor [15], it was excluded herein because for stilt-mounted agrivoltaic systems, the slope orientation of ground and solar panels do not necessarily align. Factors related to facility construction costs often include slope, distance from roads, distance from transmission lines, and distance from urban areas [15]. Herein, distance from urban areas was excluded because we focused on the capital area and the distance from the city to the PV system location was not large. Raster format layers in 1-km mesh were created for all these factors using Q-GIS from source data and were used in subsequent analyses. Although the data on distance to roads were outdated, the tendency for road density to be higher in central Tokyo and lower in areas farther from the city was indicative of the present-day scenario.

Table 2 List of factor variables, original data extension, and original source
Full size table

2.5 Conversion of factor variables using the fuzzy membership function for PV site suitability analysis

The fuzzy method was employed to analyze PV site suitability. This method uses the fuzzy membership function (FMF) to quantify the degree to which a factor belongs to a specific set, defined as the fuzzy set. This value ranges from 0 to 1 and is called fuzzy membership. Herein, PV site suitability is considered a fuzzy set, and fuzzy membership (FM) of 0 is assigned to PV sites with low suitability for and 1 to those with high suitability. The FMF applied to each evaluation factor was selected by assessing the impact of the factor value on PV site suitability (Table 3).

Table 3 Applied fuzzy membership functions and the parameters for each factor variable
Full size table

Because of the wide variety of relationships between input values and membership, numerous FMFs have been developed [39]. For factors related to sunlight (solar radiation and solar hours), the higher observed value is a higher suitability for solar PV installations. Therefore, consistent with previous studies [37, 40], we applied a fuzzy large function (Fig. 2a) with parameters such as the midpoint (m) and spread (s). The spread value affects the width of the graph, we set a value of 5 that matches the data range of solar radiation and sunshine duration. And the midpoint was determined to be a value between the maximum and minimum values observed across Japan. For example, the observed annual mean solar radiation in Japan over the past three decades was 3.06–4.42 kWh/m2. Consequently, the midpoint for solar radiation was set at 3.70, representing the midpoint of these two values. Similarly, the midpoint for solar hours was set at 4.90, approximately midway between the highest and lowest observed values. This was based on the average annual hours of sunlight at the location with the longest duration of sunlight in Japan (6.31 h/day) and the location with the shortest duration (3.52 h/day).

Fig. 2
figure 2

Fuzzy membership functions used in this study: a fuzzy large function and b fuzzy linear decreasing function [x: factor value, μ: fuzzy membership, m: midpoint, s: spread, min: threshold of factor value at which FM becomes 0 (lowest), and max: threshold of factor value at which FM becomes 1 (highest)]

Full size image

The efficiency of solar panels is influenced by temperature, leading to variations in electricity generation. In general, the power generation efficiency of a solar system decreases linearly with increasing cell temperature [38]. Therefore, the fuzzy linear decreasing function (Fig. 2b), which is effective when the degree of adaptation decreases linearly with a specific factor, was applied for evaluating the influence of temperature based on the average temperature in August. In the fuzzy linear function, parameters such as min and max are specified. The temperature deemed most suitable for PV power generation (max value: FM = 1) was 20 ℃, a value suitable for normal terrestrial environment; it was considered a standard condition for measuring the PV system efficiency [38]. The min value was set at 28.8 °C, representing the highest temperature in Japan, excluding the Southwest Islands with a subtropical climate. If the ground slope is steep, construction costs will be high because soil runoff will have to be prevented and appropriate drainage channels will have to be constructed [41]. Therefore, a flat ground was assumed to be optimal herein. A fuzzy liner decreasing function was applied, where the membership decreased as the slope became steeper. The lowest value (FM = 1) was set to 0°, and the highest value (FM = 0) was set to 30°, which is recommended in Japan for special consideration during installation [41]. The distance to infrastructure such as roads and power lines can considerably impact the cost of solar panel installation. When solar facilities are located far from roads and power lines, additional land development may be necessary to reach such facilities. To evaluate these factors, a fuzzy linear decreasing function was applied, where membership decreased linearly with increasing distance. For distance from roads, the most suitable condition was considered when the solar facility was directly adjacent to the road. The minimum value (FM = 1) was set at 2 m based on the standard road width of 4 m specified by the Building Standards Law; the centerline was the reference point. The maximum value (FM = 0) was set at 13,530 m, representing the furthest distance from a road within the study area. For distance from power lines, the most suitable condition was considered when the solar facility was directly adjacent to the power lines. The minimum value (FM = 1) was set to 0 m, whereas the maximum value (FM = 0) was set to 11,700 m, representing the furthest distance from power lines within the study area.

2.6 Statistical analysis

To evaluate the suitability of grasslands for solar PV installations, the FM of each assessment factor was compared between grasslands, nongrassland locations, and existing solar PVs. First, two grassland layers (GL1: extracted from the grassland database) [32], GL2: extracted from the National Standard Land-Use Mesh Dataset covering Japan [33, 34] were generated. Then, each grassland layer and evaluation factor (F1–F6) was overlaid, and the FM of each grassland was extracted. For F1–F4, the mean value within each grassland polygon was employed as the FM. For F5 and F6, the maximum value in the polygon (the point with the lowest distance to the infrastructure) was used as the FM for each grassland. The extracted FMs for each grassland were averaged to obtain the respective FMs for GL1 and GL2. Similarly, for PV plants, the FMs of F1–F6 for each polygon were calculated for large PV systems (PV_L), small-to medium-sized PV systems (PV_S), and agrivoltaic systems (PV_A); then, their average value was computed.

For locations other than grasslands, random points were generated within the evaluation area and 95% confidence intervals were computed from the FM of each point. Initially, 29,000 random points were generated, 10,000 times the count of GL1, across the entire metropolitan area, excluding grasslands, water areas, and isolated islands. A circle with a radius of 86.93 m was generated around each random point and used as the FM evaluation range for each point. The value of radius was derived from the average area of small- and medium-sized solar PV systems, serving as an assumed value for the scale of solar sharing in grasslands. The FM for each evaluation factor at each random point was calculated using the same methodology as that used for grasslands (F1–F4: average of the values within the circle; F5 and F6: maximum of the values within the circle). For each evaluation factor, a 1,000 bootstrap average was computed from the generated points with a sample size of 29 (from GL_1) to determine the top 2.5% (R_high) and bottom 2.5% (R_low) values.

3 Results

3.1 Identification of grasslands and solar PV systems

For grasslands, 29 sites (GL_1) were extracted from the grassland database, and 273 meshes (GL_2) were identified using the National Standard Land-Use Mesh Dataset (Fig. 3a). The number of solar PV systems included 50 large power plants exceeding 10 MW and 2397 small- and medium-sized power plants with a capacity of < 10 MW (Fig. 3b).

Fig. 3
figure 3

Identified objects for analysis: a grasslands and b PVs in study area

Full size image

3.2 Generation of fuzzy index layers for factor variables

Six FM layers were generated, each corresponding to a specific factor variable. The scale ranged from 0 to 1, with a higher FM indicating a more favorable location for a PV plant. Regions with a high FM in Fig. 4 align with areas characterized by high solar irradiation, longer sunshine hours, moderate air temperature, flat terrain, proximity to transmission lines, and closeness to roads.

Fig. 4
figure 4

Fuzzy scaled inputs for each factor variable: a solar radiation; b equivalent sunshine hours; c average temperature in August; d slope; e distance from nearest transmission line; and f distance from nearest road

Full size image

3.3 Comparison of FMs of grasslands, PVs and other environments

The calculated FM values are shown in Table 4. Additionally, comparisons of FMs between grasslands and other locations are illustrated in Fig. 5. For environmental factors related to electricity generation (F1–F3), GL_1 and GL_2 exceeded the R_low values. For F3 (temperature), GL_1 and GL_2 surpassed the R_high values, and for F1 (solar radiation), GL_2 exceeded the R_high value, indicating that grasslands exhibit a higher suitability for solar PV installations compared to other environments. Furthermore, for F1 and F3, the FMs were higher than those of existing solar PV systems. In contrast, for evaluation factors related to construction costs (F4–F6), GL1 was within the R_low–R_high range for all factors, indicating standard evaluations. However, in the case of GL2, F5 and F6 remained within the R_low-R_high range, whereas F4 (slope) was below the R_low value. In contrast, existing solar power plants exhibited high values surpassing R_high for F4–F6. Notably, for F5 and F6, the high FM values may be influenced by infrastructure development due to power plant construction, as these values pertain to already operational facilities.

Table 4 FMs of grasslands, PVs, and upper and lower bound of 95% confidence interval of random sites in the study area
Full size table
Fig. 5
figure 5

Comparison of fuzzy membership: a between grasslands and 95% confidence intervals for other random areas and b between PVs and 95% confidence intervals for other random areas. GL_1: grassland areas from the grassland database. GL_2: grassland areas from the National Standard Land-Use Mesh Dataset. PV_L: large solar PV systems. PV_S: small- to medium-sized solar PV systems. R_high: upper bound of 95% confidence intervals for random sites, except grasslands. R_low: lower bound of 95% confidence intervals for random sites, except grasslands

Full size image

4 Discussion

We evaluated the suitability of existing seminatural grasslands as potential sites for constructing solar PV plants to explore the potential for coexistence. Results revealed that seminatural grasslands in the metropolitan area in Japan exhibit promising electricity generation potential, which may even surpass that of existing solar power plants. Previous studies have shown that natural grassland environments are highly suitable for photovoltaic power generation [42], and a similar trend was observed in seminatural grasslands. Additionally, the construction costs for these grasslands were generally comparable to those for randomly selected sites. These findings indicate that the suitability of existing seminatural grasslands for solar PV plant installations is not considerably lower than that of other land types. Particularly, seminatural grasslands located on flat terrains are relatively conducive for coexistence with solar power facilities, as they are less likely to pose significant disadvantages in terms of construction costs.

Factors affecting the electricity generation, including solar irradiation, sunshine hours, and temperature were assessed. Both GL1 and GL2 exhibited suitability equal to or exceeding that of random points in terms of these factors. In contrast, the existing solar PVs, whether large or small- to medium-sized, demonstrated lower evaluations for solar irradiation and temperature compared to seminatural grasslands. Their suitability was also similar in terms of sunshine hours. Even when compared to random points, solar PVs fell short in terms of solar irradiation and temperature evaluations. These findings suggest that the weather conditions of remaining seminatural grasslands are conducive for electricity generation, making them suitable for solar power facilities. Seminatural and artificial grasslands have been highly impacted by solar PV development in Japan, following bare lands [20]. Unfortunately, the current development methods often involve disturbing the topsoil and destroying vegetation when placing solar panels [43]. Seminatural grasslands are valuable environments that host a diverse range of rare species [21,22,23]; therefore, intensive development in such habitats should be avoided. In recent years, biodiversity conservation in renewable energy development has garnered considerable attention [16] In response to this situation, the concept of “Ecovoltaics” approach, has been proposed to reconcile power generation with biodiversity conservation [44]. Indeed, there have been several reports of successful coexistence between solar panels and grasslands [45, 46]. Furthermore, these results suggest guidelines for designing solar PV systems to promote their coexistence with vegetation, such as wider panel spacing and elevated mounts [47]. By incorporating these technologies, solar PV systems in seminatural grasslands could help achieve biodiversity conservation while ensuring sufficient electricity generation.

Factors affecting the development costs of solar PVs, including slope, distance from power lines, and distance from roads, were also assessed. Results showed that among the remaining seminatural grasslands, only the slope of GL2 was below the random points, whereas other evaluation criteria were comparable to the random points. In contrast, existing solar PVs, both large and small- to medium-sized, exceeded random points in all three categories, i.e., slope, distance from power lines, and distance from roads. Thus, existing solar PVs prioritize cost-effective locations, and from this perspective, the remaining seminatural grasslands do not hold a significant advantage. The emphasis on cost in selecting solar PV locations has been well-documented in previous studies [48, 49]. When considering the economic balance between installation costs and electricity generation, the merits of establishing PV system using agrivoltaic type mounts in seminatural grasslands may seem limited. However, as solar PV plants require a large area [4,5,6,7], we can expect that solar PV installations will be undertaken in relative expensive areas with efficient electricity generation after they are installed in low-cost areas. Moreover, when considering the overall benefits of land-use rather than just the singular function of solar power generation, this evaluation might change. For instance, in agricultural land, ecosystem services such as habitat preservation, biodiversity conservation, carbon sequestration, water purification, and water regulation can be maintained via ecologically sensitive management, providing broader benefits despite reduced yields [50]. As suggested in Sturchio and Knapp [44], maintaining grassland ecosystems beneath solar panels can offer multiple benefits beyond electricity generation. It opens the possibility of using seminatural grasslands as multifunctional green infrastructure, wherein besides power generation, ecosystem services such as habitat preservation, biodiversity conservation, carbon sequestration, water purification, and water regulation can be exploited. The quantity and quality of ecosystem services resulting from the coexistence of solar PVs and seminatural grasslands need to be examined in further study.

5 Conclusions

In this study, we confirmed that the environmental conditions of seminatural grasslands are advantageous for electricity generation. Seminatural and artificial grasslands have frequently been chosen as locations for constructing solar PV plants, and this trend is expected to continue. This situation suggests that the installation of solar power plants could contribute to the conservation of seminatural grasslands, which are otherwise degraded due to abandonment. However, to realize this idea poses two major challenges. The first is the impact on grassland constituent species from an ecological perspective. We should examine that whether the plant community of seminatural grasslands can persist even under solar panels. However, considering that agricultural activities are already feasible under panels, it is deemed sufficiently achievable. The second challenge pertains to landscape concerns for the general public, including tourists. Installing solar PVs in rural landscapes can sometimes create negative perceptions among the public [51, 52], leading to regulations on plant construction [53]. On the other hand, landscape design of infrastructure projects, even for the same land use, increases social acceptability [54]. For instance, the use of materials and colors for mounts that blend with the landscape has been proposed [55] and commercialized in Japan (Mirai no Hatake: https://miraino-hatake.jp/ accessed at: 2024/06/05). Additionally, the concept of a “new urban photo-ecological gardens” where vegetation maintenance and solar power generation coexist and the site is open to the public, has already been implemented [56]. Such accessibility could potentially enhance public acceptance [57].

The coexistence of sustainable energy development and biodiversity conservation has gained recognition as a global challenge. In recent years, private companies have faced increasing demands for disclosing climate-related risks under the Task Force on Climate-related Financial Disclosures. In September 2023, the Taskforce on Nature-related Financial Disclosures (TNFD) released a full version of the “TNFD Recommendations” (TNFD, 2023). A similar trend is anticipated for disclosing nature-related risks (Science based targets network, https://sciencebasedtargetsnetwork.org/, accessed at: 2024/02/20). In this context, the installation of stilt-mounted agrivoltaic systems in seminatural grasslands offers the potential for environmentally conscious economic activities that can help mitigate both climate-related and nature-related risks within businesses.