Aquafarm Use and Energy Transition of the Aquavoltaics Policy on Small-Scale Aquaculture in Taiwan

Abstract

Aquavoltaics policy has been introduced in Taiwan to promote the integration of solar photovoltaic facilities on aquafarms. To explore the effects of the aquavoltaics policy on aquafarm price and small-scale aquaculture, we collected data on aquaculture and renewable energy materials. Subsequently, three groups of factors that influence the use of aquafarms (land, aquaculture, and renewable energy attributes) were analyzed using the hedonic price model to examine the effects of the aquavoltaics policy on aquafarm prices. We employed spatial econometrics models to estimate each variable’s influence and analyze the factors that affect aquafarm prices, as well as the possible effects of implementing an aquavoltaics policy. The empirical results indicate that the implementation of the Two-Year Solar Promotion Plan has led to an approximately 10% increase in aquafarm prices, reflecting the policy’s influence on land valuation and market expectations. Variables such as distance to urban areas, proximity to feeder lines, shellfish farming and empty ponds were found to significantly affect aquafarm prices. These findings suggest that when aquavoltaics policies are implemented in regions dominated by small-scale aquaculture, a systematic approach to aquafarm use and pricing is required. Moreover, developing integrated energy blueprints and aquavoltaic plans that balance economic, environmental, and fishery objectives is essential for achieving synergy between the fishery and renewable energy sectors.

1. Introduction

Climate change and global warming have intensified the global call for renewable energy transition. Solar power, which surpassed fossil fuels in new capacity additions in 2016, now plays a central role in global decarbonization strategies [1]. At COP28, 130 countries including the European Union pledged to work together to triple the world’s installed renewable energy generation capacity to at least 11,000 GW by 2030 [2]. Solar power will have a pivotal role to play in this global transition toward clean energy and in achieving global energy and climate targets [3]. In Taiwan, the government has actively pursued renewable energy development since the promulgation of the Renewable Energy Development Act in 2009. This was followed by a series of legislative and policy measures, including the Greenhouse Gas Reduction and Management Act (2015), the National Climate Change Action Guidelines (2017), the Pathway to Net-Zero Emissions in 2050 (2022), and the Climate Change Response Act (2023) [4]. Taiwan’s total GHG emissions have totaled 285.1 MtCO₂e in 2020, ranking 33rd compared to the world’s total annual GHG emissions of 47,513 MtCO₂e [5]. To achieve carbon reduction targets and strive for net-zero emissions, Taiwan government has committed to sourcing 20% of its total electricity from renewable sources by 2025. They aim to achieve this by building 20 GW of solar photovoltaic (PV) capacity and more than 5.7 GW of offshore wind generation capacity. The goal is to achieve solar PV capacity target of 31 GW by 2030 and 40–80 GW by 2050. However, the installed solar PV capacity in Taiwan only reached 13.81 GW by the end of October 2024 [6]. Due to many difficulties in ground-based photovoltaic preparations, the current 20GW target has been postponed to November 2026.

Solar energy systems can effectively reduce electricity bills. By installing solar PV capacity, any individual or organization can independently generate some of the needed electricity and reduce their dependency on traditional grids while achieving energy conservation and reduced consumption. Studies worldwide have extensively explored aquaculture and renewable energy issues [7,8,9]. In Taiwan, the government launched a Two-Year Solar Promotion Plan in 2016 to reach 20 GW of installed solar PV capacity by 2025, with 3 GW from roof-mounted and 17 GW from ground-mounted systems. However, land resources in Taiwan are limited, and some of the land is used by the agricultural sector, which is comprised of a large number of small-scale farmers and fishers, each operating an average farm size of about 1.1 hectares [10]. Nevertheless, factors such as having limited land space in a densely populated environment and socioeconomic development have been a constant source of pressure compelling the land-use conversion of farmlands [11]. Moreover, the specific spatial requirements of renewable power generation facilities have created a coopetition relationship between renewable energy, farmland usage, and food security [12]. In Taiwan, the coopetition between the agricultural and renewable energy sectors has posed a significant problem for Tainan City: too many farmlands have converted to solar energy production, exerting a significant impact on the living environment of the entire farming and fishing communities [13]. In addition, the scarcity of land resources and frequent reports of competition for farmlands have heightened concerns about the various ecosystems in nature [14]. To mitigate these conflicts, the government revised the Solar PV Development Plan in 2021—setting a target of 8 GW for rooftop systems and 12 GW for ground-mounted installations—and initiated a nationwide environmental assessment to ensure a balance between energy development, agriculture, and ecology.

Another development focus of Taiwan is aquavoltaics (solar panels installed on fish farms) for which an aquavoltaics policy has been formulated. The policy describes two pathways of designating zones for aquavoltaic installations: one by the central or regional government and state-owned enterprises using the original model and the other by the Ministry of Agriculture. In the second pathway, the Ministry of Agriculture provides a range of zones that are given priority to integrate renewable energy in aquaculture operations. Excluding ecologically sensitive areas and birding hotspots, 7985 hectares of zones have been prioritized for aquavoltaic installations. According to the Regulations for Examining the Application of Structuring Farming Facilities on Agricultural Land, a renewable energy facility may not occupy more than 40% of the area of farmland on which it is to be located (which includes other facilities under management). Outdoor aquaculture farms designated for aquavoltaic installation shall adopt either one of the following three types of ground-mounted solar PV systems: solar arrays installed above the fish pond, floating solar, or solar panels installed on the embankment of the fish pond. To facilitate the adoption of the aquavoltaics policy, the Taiwan government lowered the installation threshold for aquavoltaic zones in 2020, from 25 ha to 10 ha, and exempted aquaculture farms with a capacity below 2 MW (i.e., a 5-ha farm based on 40% of farmland area) from the environmental and social inspection mechanism for solar PV systems. In spite of the government’s renewable energy policies, aquafarm prices have been severely affected by the exploits of landowners and developers. While the policy is intended to introduce aquavoltaic systems as a means of increasing non-fisheries income for aquaculture fishers, its undesirable effects—the different aquaculture benefits derived from different fish species and the cognitive discrepancies of landowners and tenants—may adversely affect the development of fishing communities, especially disadvantaged fishers, and the natural ecosystem [15]. Since aquavoltaics policy is for low-density development, in order to ensure coexistence and prosperity with society and the environment and to eliminate controversial locations, the environmental and social inspection mechanism for solar PV systems will be promoted in 2021 to confirm the possible impact of future aquavoltaics construction, and fully communicate with stakeholders before developing a site to accelerate and efficiently build it, thereby enhancing the development of the overall photovoltaic industry and jointly achieving energy transformation goals.

Earlier studies have investigated agricultural land value or change of its use [16,17,18]. Most of these studies have focused on the rent-seeking behavior caused by the urbanization of agricultural lands, and a few have explored renewable energy issues in the agricultural sector. For example, Lehn and Bahrs [19] have identified urban sprawl and livestock production as the factors driving up the prices of farmlands in Germany. Chang et al. [20] have observed that the conventional land use policy tends to release farmland to achieve economic development and satisfy land use demand and has resulted in the conversion of farmland into nonfarm uses. They have also found that developers are often willing to pay a premium and further encourage farmland transactions and the frequent change on farmland ownerships. Lee et al. [21] have reported a considerable effect on farmland use and agricultural development in Taiwan, with even a negative effect on farmland prices, when the government has encouraged aquavoltaic practices through solar farm policies to promote solar power generation. Franziska et al. [22] and Yang et al. [23] have indicated that the agricultural sector in Germany has been affected by biogas policies such as the German Renewable Energy Act in recent years, which has not only given rise to environmental issues but also caused distortions within the agricultural sector, including agricultural production, farms, and land markets. Their research results reveal that this policy has also increased the rental prices of farmlands. Using cases of renewable energy development in Norway and Japan, Valeria et al. [24] have examined the relationship among renewable energy development, food supply, water, and farmlands and found that a scale up of renewable power generation contradicts the goal of local sustainable development. The aquaculture industry is one of the main producers of aquatic products in Taiwan. It supplies 250,000 to 300,000 metric tons of quality aquatic products each year, including tilapia, milkfish, bass, groupers, Asian hard clam, and oysters. Presently, aquavoltaic practices are primarily adopted in major aquaculture areas of Taiwan, which are largely located along the coast where space is limited for large-scale fish farms. Given the current changes to policy goals, the aquavoltaics policy has driven rent-seeking and opportunistic behaviors in the aquaculture sector, which not only influences aquafarm prices in aquaculture production areas but also poses a risk of escalating costs for symbiotic projects, thus affecting the socioeconomic development of the fisheries industry [25].

The objective of Taiwan’s aquavoltaics policy is to develop renewable energy facilities without compromising aquaculture production and to optimize the aquaculture production environment for industrial transformation, generating additional income (feed-in tariff) for aquaculture farmers. However, the feed-in tariff incentive has also created rent-seeking opportunities for coastal aquafarms on state-owned lands, leading to aquafarm price fluctuations and potential impacts on the aquaculture industry, fisheries economy, and solar PV investment. This policy context underscores a growing tension between energy transition and aquaculture sustainability, highlighting the need to empirically assess how aquavoltaics policy affect aquafarm prices and land-use dynamics. The present study is significant in that it is the first to explore the effects of the aquavoltaics policy on the aquaculture industry, aquafarm prices, and land use. Whereas previous studies have mainly addressed agriculture or renewable energy issues using random samples [26,27,28], this study analyzes comprehensive data on aquafarm transactions in relation to aquaculture production and renewable energy development. Specifically, it investigates the linkage between the aquavoltaics policy and the socioeconomic development of the fisheries sector through three groups of determinants: land, aquaculture, and renewable energy attributes.

The hedonic price theory provides the analytical foundation for this study. According to Rosen [29], a hedonic price function—determined by the supply and demand for product attributes—reflects the implicit price that consumers are willing to pay for each characteristic. Following Goodman and Kawai [30], who applied this approach to estimate agricultural land and housing prices, numerous studies [30,31,32,33,34,35,36,37] have extended hedonic pricing methods to assess property and farmland values based on key physical and locational attributes. In this study, aquafarm price characteristics were defined through a literature review and consultations with five expert scholars specializing in fisheries economics and aquaculture management. The Hedonic Price Model (HPM) is employed to identify aquavoltaic factors influencing aquafarm prices, while Spatial Econometric Models (SEM) are used to capture spatial dependencies and regional effects. Together, these analytical approaches provide a comprehensive understanding of how the aquavoltaics policy influences land-use patterns, aquafarm valuation, and the adaptive behavior of aquaculture stakeholders. The findings are expected to offer valuable insights for policymakers and industry participants seeking to promote energy transition and sustainable development within small-scale aquaculture systems.

2. Research Background and Scope

Aquaculture has been practiced in Taiwan for more than 300 years. The industry experienced unbridled expansion during the 1960s, 1970s, and 1980s—the so-called golden years of Taiwan’s aquaculture industry—fueled by the continuous aquaculture breakthroughs, the crowning of Taiwan as the Kingdom of Giant Tiger Prawns, and the thriving eel industry attributed to exports to Japan [38]. Following a period of prosperity, the aquaculture industry has entered a declining phase due to land subsidence and soil salinization problems in the southwest coast of Taiwan as well as a bout of shrimp disease outbreaks. Since then, the industry has never managed to recover fully, despite its success in exporting tilapia to the United States and despite the increased groupers production. In addition, global issues such as global competition, free trade, disease outbreaks in aquaculture, climate change, and aquavoltaic issues are all challenges hampering the development of the entire aquaculture industry.

Based on the development of the aquaculture industry in Taiwan, aquaculture output has remained at 250,000 metric tons in recent years, accounting for 25% to 30% of the total fisheries output. According to Taiwan Fisheries Agency, the country’s annual aquaculture output in 2022 was 264,000 metric tons, totaling NT$33.5 billion. Onshore aquaculture output was 245,000 metric tons, totaling NT$28.9 billion, with higher output observed in Tainan City, Pingtung County, Kaohsiung City, Yunlin County, Chiayi County, and Changhua County. In other words, the southwest coast of Taiwan plays host to onshore aquaculture activities. Of these regions, Tainan City has the highest aquaculture output (74,219 metric tons), highest production value (NT$7.81 billion), largest aquaculture area (11,619 ha), and the highest number of aquaculture farmers (15,000 people).

Table 1. A comparison of 2012–2022 aquaculture statistics between Tainan City and nationwide.

	Output (Metric Tons)		Output Value (NT$ Million)		Area (Hectare)		Number of Aquaculture Farmers
Year	Tainan City	Nationwide (%)	Tainan City	Nationwide (%)	Tainan City	Nationwide (%)	Tainan City	Nationwide (%)
2012	87,426	27.6%	6989	20.8%	13,733	34.5%	19,482	29.4%
2013	83,233	26.2%	7503	21.9%	12,672	33.1%	19,278	28.3%
2014	84,785	27.4%	8450	23.4%	11,526	31.7%	18,650	27.6%
2015	77,476	26.8%	6740	20.2%	12,233	34.9%	18,338	26.6%
2016	73,365	31.9%	6828	24.0%	11,728	34.6%	17,118	25.6%
2017	75,558	29.5%	7580	24.6%	12,180	35.5%	16,740	23.9%
2018	77,486	29.9%	7122	22.4%	11,753	34.9%	15,083	21.7%
2019	84,476	31.3%	7169	22.9%	13,355	37.6%	14,872	23.2%
2020	77,476	30.3%	6043	22.3%	13,666	38.5%	14,737	21.7%
2021	75,951	29.9%	7072	26.0%	13,000	38.2%	14,898	22.9%
2022	74,219	30.3%	7810	27.0%	11,619	36.0%	14,896	22.9%

summarizes the onshore aquaculture statistics of Tainan City (annual output, value, farm area, and number of fishers). These data show that the onshore aquaculture activities in Tainan City play a crucial role in Taiwan [39]. Given its economic importance and high concentration of aquafarms, Tainan City serves as an ideal study area to investigate how Taiwan’s aquavoltaics policy influences aquafarm prices, land-use patterns, and the socioeconomic dynamics of small-scale aquaculture. This study focuses on six major aquaculture districts in Tainan City which represent the core areas of onshore aquaculture and aquavoltaic development in southern Taiwan.

In 2022, Tainan City had a total aquaculture area of 11,619 ha, of which 5684.3 ha, 3281.84 ha, and 1713.88 ha were used to cultivate fishes, shellfish, and shrimps, respectively—mainly tilapia, milk fish, Asian hard clam, and king prawn. In particular, 27,275 metric tons of tilapia (accounting for 47.15% of the total output) were harvested from an aquaculture area of 1059.24 ha; 23,629 metric tons (49.23%) of milk fish were harvested from an area of 4111.04 ha; 13,143 metric tons (24.28%) of Asian hard clam were harvested from an area of 3281.84 ha; and 2979 metric tons (31.39%) of king prawn were harvested from an area of 3564.93 ha. In terms of administrative regions (Figure 1), aquafarms in Tainan City are mostly located along the coast, with Qigu District having the largest output and aquaculture area, followed by Annan and Beimen Districts, then Jiali District, while both Madou and Jiangjun Districts have the smallest output and aquaculture area.

3. Methods

3.1. The Hedonic Price Model (HPM)

A hedonic price model is subsequently constructed to estimate and predict the implicit prices of an aquavoltaic farm. To explore the effects of aquavoltaic systems on aquafarm lands, the characteristics that influence land transaction price P are grouped into three categories: general land attribute matrix N, aquaculture production characteristic matrix F, and renewable energy-related variables G. The log–log model is used in estimations to reduce heteroscedasticity in data. The hedonic price model is expressed as Equation (2).

$${\mathit{P}}_i=a_0+a_{1}N+a_{2}F+a_{3}G+\epsilon$$

(1)

$${ln\mathit{P}}_i={\beta }_0+{\beta }_{1}lnN+{\beta }_{2}lnF+{\beta }_{3}lnG+\epsilon$$

(2)

3.2. Spatial Econometrics Models

Spatial econometrics constructs regression models to examine relationships between dependent and independent variables while accounting for spatial dependence. Unlike conventional regression, spatial data are not randomly sampled, and their observations may violate assumptions of independence and homoscedasticity. Because land prices are influenced by geographic proximity and spatial autocorrelation, Moran’s I test is applied to assess whether aquafarm prices exhibit spatial clustering or dispersion. If the null hypothesis of spatial randomness is rejected, spatial regression models are employed to analyze the spatial effects on aquafarm transaction prices [40,41]. Anselin [42] has described two types of spatial autocorrelation models: spatial lag model (SLM) and spatial error model (SEM). The SLM considers the characteristics of spatial autocorrelation to be determined by spatial proximity, and the performance of neighboring locations affects variable value. The SEM attributes the characteristics of spatial autocorrelation to the lack of certain independent variables in the model that are spatially structured in relation to error terms [43]; it estimates model parameters by using the maximum likelihood method.

3.2.1. Spatial Lag Model (SLM)

The SLM can be used when an error is caused by neighborhood effects. Given the “lag” effect of a dependent variable, a matrix neighboring the sample is incorporated in the model and is expressed as Equation (3):

$${\mathrm{Y}}_i=\mathsf{\alpha }+\mathsf{\rho }\mathrm{W}{\mathrm{Y}}_{\mathrm{i}}+\mathsf{\beta }\mathrm{X}+\mathsf{\epsilon };\mathsf{\epsilon }~\mathrm{N}(0,{\sigma }^2)$$

(3)

where Y is a dependent variable of land price; α is a constant; ρ is a spatial autoregressive coefficient; W is a spatial weight matrix; β is a regression coefficient; X is a dependent variable: general land attribute N, aquaculture attribute F, and renewable energy attribute G; and $\mathsf{\epsilon }$ is an error term (random error). In (3), $\mathsf{\rho }$ ≠ 0 means that land price is affected by spatial contiguity.

3.2.2. Spatial Error Model (SEM)

If spatial autocorrelation exists in an error term, the spatial autocorrelation problem must be corrected and included in the error term to correct for spatial error. In this case, the error term is not a random error and expresses spatial dependence. The SEM is expressed as Equation (4):

$${\mathrm{Y}}_{\mathrm{i}}=\mathsf{\alpha }+\mathsf{\beta }\mathrm{X}+\mathsf{\epsilon };\mathsf{\epsilon }=\mathsf{\lambda }\mathrm{W}\mathsf{\epsilon }+\mathsf{\eta };\mathsf{\eta }~\mathrm{N}(0,{\mathsf{\sigma }}^2)$$

(4)

where Y is a dependent variable of land price; α is a constant; β is a regression coefficient; X is a dependent variable: general land attribute N, aquaculture attribute F, and renewable energy attribute G; ε is an error term; λ is a spatial error coefficient; W is a spatial error matrix based on contiguity; and η is a random error. If λ ≠ 0, it means a possibility of interfering factors in the SEM causing spatial correlation. The Breusch–Pagan test is then used to test for spatial heteroskedasticity; the Lagrange Multiplier (LM) test and Robust LM are used to test the SLM and SEM for heterogeneity, and the loglikelihood ratio (LIK), the Akaike information criterion (AIC), and the Schwarz criterion (SBC) are used to test for model fitness to determine which model—the SLM or the SEM—is a better fit [44].

4. Data Description

4.1. Data Sources

In this study, a real estate transaction inquiry system (https://lvr.land.moi.gov.tw, accessed on 16 November 2025) developed by the Ministry of Interior has been used to collect data on agricultural land transactions between October 2012 and December 2021 over a period of nine years and two months. The scope of the data includes main aquaculture areas in Beimen, Xuejia, Madou, Jiangjun, Qigu, and Annan Districts of Tainan City. The data documents the transactions involving lands in Tainan City that are used for agriculture, livestock farming, and aquaculture. In total, 3078 sets of data have been collected, each containing the following information: township/city/district, land section location or building section number, urban/non-urban land-use zoning, total land transfer area, date of transaction (year/month/day), total land price, and the coordinate of the transacted land. Next, this study has used a map information system created by the Ministry of Interior and a fishery management system developed by the Fisheries Agency to filter and select aquafarm data for variable verification. In total, 1024 sets of data on aquafarm transactions have been collected. Figure 2 shows the distribution of transactions involving aquafarm lands.

This study has adopted the hedonic price model to analyze the effects of the aquavoltaics policy on aquafarm prices and employed spatial regression models to estimate the influence of each variable. The objective is to analyze the aspects that affect aquafarm prices and the possible effects of implementing an aquavoltaics policy.

4.2. Factors Influencing Aquafarm Prices

Farmland prices are affected by many factors, some of which also influence aquafarm land prices. Therefore, factors that influence aquafarm lands are grouped into three categories (

Table 2. Variables representing the characteristics of aquafarm prices.

Category	Variable	Definition	Expected Relationship
Dependent Variable	Price	Total price ÷ Land transfer area (m2)	−
Land Attributes	Tarea	Transacted land surface area (m2)	−
	Tarea2	Square of transacted land transfer surface area (m2)	−
	Dcity	Linear distance from transacted land to neighboring city (m)	−
	Droad	Linear distance from transacted land to neighboring road (m)	−
	AreaQ, B, G, M, S	Administrative regions in which transacted land is located, including Qigu, Beimen, Jiangjun, Madou, Xuejia, and Annan (control group) Districts	−
Aquaculture Attributes	Fish	Cost of cultivating fishes in the aquafarm	+
	Shrimp	Cost of cultivating shrimps in the aquafarm	+
	Shellfish	Cost of cultivating shellfish in the aquafarm	+
	YD	Holder or not a holder of aquaculture permit or license	+
	Empty	Presence or absence of empty pond in the aquafarm	+
Renewable Energy Attributes	Line	Presence or absence of feeder lines nearby to power the land	+
	SQ	Presence or absence of suitable landform/shape	+
	After	Before or after the launch of the Two-Year Solar Promotion Plan	+
	Green1	The aquafarm is in an existing aquaculture production area, has an area of more than 10 ha used for farming, and has a fish pond occupying more than 60% of the land	+
	Green2	The aquafarm has an area of more than 25 ha used for farming, and has a fish pond occupying more than 60% of the land	+
	Dsolar	Distance from transacted land to nearest solar farm (m)	−

). Category 1 consists of variables representing the land attributes that influence the value of aquafarm and includes the size, shape, and location of the piece of land subject to transfer. Category 2 comprises variables related to the aquaculture attributes, such as the productivity and surface area of an aquafarm. Category 3 consists of variables related to renewable energy, including the distance from transacted land to renewable energy facility, the location prioritized for renewable energy installation, the time at which renewable energy policy is implemented, and the distance from the transacted aquafarm to a solar farm.

The empirical analysis in this study was conducted in two stages to examine the spatial effects of regional heterogeneity on aquafarm land prices. In Stage 1, the model excluded the regional variable to establish a baseline estimation focusing on general land, aquaculture, and renewable energy attributes. In Stage 2, the regional variable (i.e., area) was incorporated into the model as an additional explanatory attribute to capture spatial differentiation across districts. By introducing this stepwise modeling approach, we aimed to distinguish whether price variations arise primarily from intrinsic property characteristics or from locational factors associated with specific regions. Using Annan District, which is closest to the city center, as the control group, the model further assessed how different areas within Tainan City influence aquafarm land transactions and the extent to which location-based disparities affect the implementation of the aquavoltaics policy.

4.2.1. Aquafarm Prices

To explore the effects of each attribute variable on aquafarm transaction prices and to ensure consistent unit of measurement, we estimate the effects of each factor on aquafarm transaction prices by using unit price per m² of transacted land as the dependent variable. To reduce heteroscedasticity, the effect of extreme values has been controlled for by using the log of the dependent variable. Therefore, the log of per m² land price is the dependent variable in the empirical model [40].

4.2.2. Land Attributes

Aquafarm transactions are influenced by land-related factors such as land size, distance to roads or cities, and location. Since land size exhibits diminishing returns, a quadratic term is used to capture its negative effects. Distance to roads and cities is determined using GIS by obtaining coordinates from map data. We anticipate that aquafarm prices to decrease with larger land size (Tarea), greater distance to cities (Dcity), and farther from roads (Droad). Prices are also affected by potential land value increases after land-use changes and regional characteristics, such as proximity to urban areas and public transportation [41]. In Stage 2, the region variable is included to compare different areas, using Annan District as the control. Results are anticipated to show a negative correlation between price and distance from Annan District, which diminishes as distance increases.

4.2.3. Aquaculture Attributes

Aquaculture attributes were derived from the Fisheries Agency’s Aquaculture Fishery Management System, including fish fry costs, aquaculture area, species, permits or licenses, and empty ponds. By overlaying transaction land maps with onshore aquaculture pond maps, the serial numbers of ponds within each transacted parcel were identified, linking to detailed aquafarm and activity data. This study focuses on aquaculture costs by species and the presence of permits or licenses as key variables. The aquaculture cost per m² was calculated by multiplying the quantity of fish fry released by their price at the time of release, representing aquaculture input. Species were grouped into fish, shrimp, and shellfish, as each type has distinct effects. Theoretically, higher aquaculture investment implies greater economic potential and, thus, higher land value. Farms with land use permits (YD) likely have stronger infrastructure, increasing attractiveness to buyers, while those with empty ponds are more suitable for aquavoltaic development. Overall, aquaculture cost, permit possession, and empty ponds are expected to positively affect aquafarm prices.

4.2.4. Renewable Energy Attributes

Renewable energy facility attributes include Taipower line capacity and distribution, landform, policy implementation time, prioritized renewable energy zones, and distance to solar farms. The suitability of a fish pond for aquavoltaic installation depends mainly on its proximity to feeder or transmission lines, which connect solar systems to Taipower’s grid. Installation near feeder lines reduces setup costs, as grid reinforcement expenses are shared between the transmission and electricity enterprises, while new line installation costs are borne by the latter. Thus, a dummy variable represents whether a feeder line within 300 m of the aquafarm can supply sufficient capacity to power 40% of the transacted land [40]. Flat landforms favor solar system installation, whereas irregular terrain increases costs and reduces feasibility. Policy implementation time is also relevant, as the Ministry of Economic Affairs launched the Two-Year Solar Promotion Plan in July 2016 to encourage ground-mounted solar systems on aquaculture lands. Overall, feeder line proximity, favorable landforms, and policy timing are expected to positively affect aquafarm prices.

Locations prioritized for solar energy installation follow the Ministry of Agriculture’s Guidelines on the Review of Aquavoltaic Facilities (Amendments to the guidelines on 31 July 2020 have given priority for solar energy installation to aquafarms with more than 10 ha of land used for farming and a fish pond occupying 60% of the land), which regulate agricultural land and aquaculture production areas. An aquafarm is considered suitable for aquavoltaic development if its surface area meets the GIS-based criteria for installation. Large, homogenous land parcels are preferred for such projects. Two dummy variables (Green1 and Green2) represent priority locations: Green1 refers to aquafarms within existing production areas exceeding 10 ha, with ponds covering over 60% of the land; Green2 applies to aquafarms larger than 25 ha with similar pond coverage that meet solar promotion plan criteria. Both are expected to correlate positively with aquafarm prices. The variable distance to solar farm (Dsolar), representing the distance to the nearest solar farm, captures potential spatial competition and is expected to have a negative effect. Table 2 summarizes all variables used in this study.

The ordinary least squares (OLS) method, SLM, and SEM are used to estimate the research variables. The models are defined as Equations (5)–(7):

$$\begin{array}{ll}ln{Price}_i={\alpha }_0+\hfill & {\alpha }_1{Tarea}_i+{\alpha }_2{Tarea}^2+{\alpha }_3{Dcity}_i+{\alpha }_4{Droad}_i+{\alpha }_5{AreaQ}_i\\ & +{\alpha }_6{AreaB}_i+{\alpha }_7{AreaG}_i+{\alpha }_8{AreaM}_i+{\alpha }_9{AreaS}_i\\ & +{\alpha }_{10}{Fish}_i +{\alpha }_{11}{Shrimp}_i+{{\alpha }_{12}Shellfish}_i+{\alpha }_{13}{YD}_i\\ & +{\alpha }_{14}{Empty}_i+{\alpha }_{15}{Line}_i+{\alpha }_{16}{SQ}_i+{\alpha }_{17}{After}_i\\ & +{\alpha }_{18}{Green1}_i+{\alpha }_{19}{Green2}_i+{\alpha }_{20}{Dsolar}_i+{\epsilon }_i i =1,\dots ,n\end{array}$$

(5)

$$\begin{array}{ll}\hfill ln{Price}_i={\alpha }_0+& {\rho WlnPrice+\beta }_1{Tarea}_i+{\beta }_2{Tarea}^2+{\beta }_3{Dcity}_i+{\beta }_4{Droad}_i\\ & +{\beta }_5{AreaQ}_i+{\beta }_6{AreaB}_i+{\beta }_7{AreaG}_i+{\beta }_8{AreaM}_i\\ & +{\beta }_9{AreaS}_i+{\beta }_{10}{Fish}_i+{\beta }_{11}{Shrimp}_i+{{\beta }_{12}Shellfish}_i\\ & +{\beta }_{13}{YD}_i+{\beta }_{14}{Empty}_i+{\beta }_{15}{Line}_i+{\beta }_{16}{SQ}_i+{\beta }_{17}{After}_i\\ & +{\beta }_{18}{Green1}_i+{\beta }_{19}{Green2}_i+{\beta }_{20}{Dsolar}_i+{\epsilon }_i i=1,\dots ,n \end{array}$$

(6)

$$\begin{array}{ll}\hfill ln{Price}_i={\alpha }_0+& {\beta }_1{Tarea}_i+{\beta }_2{Tarea}^2+{\beta }_3{Dcity}_i+{\beta }_4{Droad}_i+{\beta }_5{AreaQ}_i\\ & +{\beta }_6{AreaB}_i+{\beta }_7{AreaG}_i+{\beta }_8{AreaM}_i+{\beta }_9{AreaS}_i\\ & +{\beta }_{10}{Fish}_i+{\beta }_{11}{Shrimp}_i+{{\beta }_{12}Shellfish}_i+{\beta }_{13}{YD}_i\\ & +{\beta }_{14}{Empty}_i+{\beta }_{15}{Line}_i+{\beta }_{16}{SQ}_i+{\beta }_{17}{After}_i\\ & +{\beta }_{18}{Green1}_i+{\beta }_{19}{Green2}_i+{\beta }_{20}{Dsolar}_i+{\eta }_i\\ & \hfill {\eta }_i=\lambda W+u_i i=1,\dots ,n\hfill \end{array}$$

(7)

5. Results and Discussion

5.1. Descriptive Statistics

Table 3. Descriptive statistics of aquafarm price characteristic variables.

Type of Variable	Variable	Mean	Standard Deviation	Min.	Max.
Dependent Variable	Price	1418.175	1735.976	68	13,202
Land Attributes	Tarea	10,986.06	32,370.72	2.16	577,448
	Dcity	727.437	825.073	1	5233
	Droad	293.486	438.139	1	2897
	AreaQ	0.288	0.453	0	1
	AreaB	0.187	0.390	0	1
	AreaG	0.045	0.207	0	1
	AreaM	0.077	0.267	0	1
	AreaS	0.135	0.342	0	1
Aquaculture Attributes	Fish	5.524	14.516	0	261.409
	Shrimp	1.465	4.231	0	76.694
	Shellfish	0.908	4.551	0	116.332
	YD	0.397	0.490	0	1
	Empty	0.352	0.478	0	1
Renewable Energy Attributes	Line	0.281	0.450	0	1
	SQ	0.479	0.500	0	1
	After	0.573	0.495	0	1
	Green1	0.782	0.413	0	1
	Green2	0.102	0.302	0	1
	Dsolar	2190.834	2932.389	6.662	12,705.18

shows the descriptive statistics (mean, standard deviation, minimum, and maximum) of the 1024 sets of data on aquafarm transactions between 2002 and 2021. The average unit price of the aquafarm transactions is NT$1418 per m², and the average surface area of the land being transferred is 10,986 m². The average costs of cultivating fishes, shrimps, and shellfish are NT$5.524, NT$1.465, and NT$0.908 per m², respectively. Of the transacted aquafarms, 39.7% hold an aquaculture permit, 35.2% have an empty pond, and 28.1% are located near feeder lines. The average distance to neighboring solar PV farm is 2191 m. To reduce heteroscedasticity, regression analysis has been performed by taking the log of the ‘price’ dependent variable.

5.2. Spatial Econometrics Analysis

5.2.1. Spatial Clustering and Heterogeneity of Transacted Aquafarms

Based on the variables defined above, the analysis reveals that transacted aquafarms are spatially clustered, exhibiting both neighborhood effects and spatial heterogeneity. The implementation of aquavoltaics policy has influenced aquafarm use and transaction patterns, leading to spatial variations in aquafarm prices. These variations are primarily attributed to the “area” attribute variables, which introduce regional differences and clustering. After controlling for these variables, residuals in certain regions remain unusually large or small, suggesting omitted key factors—an indication of spatial heterogeneity. Moreover, even after accounting for all observable attributes, neighboring areas still exert mutual influence, known as the neighborhood effect.

To test for spatial dependence, this study employed GeoDA 1.22 software. In stage 1, the “area” variable was excluded from the general land attribute model. As shown in

Table 4. OLS, SLM, and SEM estimation and test results (*** represents a significant level of 1 %, ** represents a significant level of 5 %, and * represents a significant level of 10 %).

Model	OLS			SLM			SEM
	Coefficient	p-Value		Coefficient	p-Value		Coefficient	p-Value
CONSTANT	2.77754	0.000	***	0.57431	0.000	***	2.80394	0.000	***
Tarea	−5.90 × 10−7	0.239		−1.65 × 10−7	0.676		3.12 × 10−8	0.937
Tarea2	1.40 × 10−12	0.250		4.44 × 10−13	0.644		2.24 × 10−14	0.981
Dcity	−8.87 × 10−5	0.000	***	−5.14 × 10−6	0.619		−3.26 × 10−5	0.080	*
Droad	5.48 × 10−6	0.812		−2.46 × 10−5	0.175		−3.42 × 10−5	0.092	*
Fish	0.00096	0.062	*	0.00085	0.036	**	0.00106	0.007	***
Shrimp	0.00150	0.404		0.00098	0.491		0.00089	0.525
Shellfish	−0.00275	0.102		−0.00224	0.090	*	−0.00178	0.176
YD	−0.01573	0.295		0.00862	0.467		0.01031	0.390
Empty	0.03740	0.024	**	0.03001	0.022	**	0.02210	0.093	*
Line	0.10046	0.000	***	0.03788	0.017	*	0.02427	0.185
SQ	−0.00199	0.893		−0.0031	0.791		−0.01049	0.378
After	0.09626	0.000	***	0.10639	0.000	***	0.11754	0.000	***
Green1	−0.08623	0.000	***	−0.09202	0.000	***	−0.09312	0.000	***
Green2	−0.2316	0.000	***	−0.11303	0.000	***	−0.11167	0.005	***
Dsolar	9.85 × 10−5	0.000	***	1.88 × 10−5	0.000	***	6.39 × 10−5	0.000	***
Pho (ρ)				0.79213	0.000	***
LAMDA ($\mathsf{\lambda }$)							0.86688	0.000	***
R-squared	0.69610			0.80776			0.8097
Adj-R-squared	0.69158
F-statistic	153.925
BP	55.6794	0.000	***	61.4536	0.000	***	61.5041	0.000	***
LIK	69.6863			278.347			273.9041
AIC	−107.373			−524.694			−515.808
SBC	−28.4691			−438.858			−436.905
S.E	0.22605			0.17979			0.17889
Moran’s I	0.2956	0.000	***
LM (lag)	767.8211	0.000	***
Robust LM (lag)	76.2751	0.000	***
LM (error)	816.2545	0.000	***
Robust LM (error)	124.7084	0.000	***
Likelihood-Ratio (lag)				417.321	0.000	***
Likelihood-Ratio (error)							408.436	0.000	***

, both Pho (ρ) and LAMDA ($\mathsf{\lambda })$ reject the null hypothesis, confirming significant neighborhood effects—prices of aquafarms in adjacent areas continue to affect each other. The residuals also display spatial heterogeneity, implying spatial autocorrelation in the OLS results and potential missing variables. The Breusch–Pagan test confirms heteroscedasticity, while Moran’s I test (0.2956) rejects the null of spatial independence, indicating that neighboring transactions share similar attributes and that both high and low values tend to cluster geographically [40,41].

5.2.2. The Spatial Econometrics Model Outperforms to Analyze the Impact Assessment of Aquavoltaics Policy

The results of the regression OLS method (Table 4) show that the variables Dcity, Fish, Empty, Line, After, Green1, Green2, and Dsolar have achieved significance level. The R-squared value was 0.69. Aquafarm prices are notably affected by proximity to cities, fish cultivation costs, the presence of empty ponds, nearby feeder lines, and the timing of aquavoltaics policy implementation—consistent with expectations. However, the variables Green1, Green2, and Dsolar have not significantly influenced aquafarm prices. This outcome reflects Taiwan’s highly fragmented aquaculture sector, composed mainly of small-scale farmers. The aquavoltaics policy requires over 70% of landowners and aquaculture operators, representing more than 70% of the eligible area, to consent to development before July 2020—a target difficult to achieve due to dispersed ownership. Although the regulation was later relaxed, market prices did not adjust accordingly.

Because OLS residuals exhibit spatial autocorrelation, the SLM was applied to capture neighborhood effects. The results confirm a significant spatial diffusion effect, where aquafarm prices in neighboring areas influence one another. Unlike the SEM, the SLM reveals that shellfish cultivation costs are significantly and negatively correlated with aquafarm prices—an unexpected finding. In Tainan, where Asian hard clam farming dominates, limited freshwater availability (e.g., in Qigu District) reduces aquaculture density and income potential. Furthermore, under the aquavoltaics policy, rent-seeking behaviors may occur; thus, lower shellfish costs could be associated with higher land prices. The SEM, in contrast, identifies Droad as significant and negatively related to aquafarm prices, consistent with expectations. Comparing model performance, higher log-likelihood (LIK) and lower AIC and SBC values indicate better fit, with the SLM outperforming others. Subsequently, the Lagrange Multiplier test was performed on the SLM and SEM, and both produced significant results. Next, the Robust LM test was conducted, and the results show that both the Robust LM (lag) and Robust LM (error) models are significant, with the SEM yielding a more significant result [43].

5.2.3. Incorporating ‘Area’ as a Variable Captures the Relationship with Aquafarm Prices

To estimate the effects of aquafarm prices on each attribute variable after the ‘area’ variable is incorporated, this study included other five districts as general land variables in the Stage-2 model, with Annan District as the control group. The results reveal that the ‘area’ variable exerted a significant influence on the dependent variable and exhibited a negative relationship as expected (

Table 5. OLS, SLM, and SEM estimation and test results (with area variables) (*** represents a significant level of 1 %, ** represents a significant level of 5 %, and * represents a significant level of 10 %).

Model	OLS			SLM			SEM
	Coefficient	p-Value		Coefficient	p-Value		Coefficient	p-Value
CONSTANT	3.2940	0.000	***	1.06 × 100	0.000	***	3.39 × 100	0.000	***
Tarea	−3.60 × 10−7	0.415		−7.56 × 10−8	0.848		2.10 × 10−8	0.957
Tarea2	6.37 × 10−13	0.552		2.09 × 10−13	0.827		9.46 × 10−15	0.991
Dcity	−4.60 × 10−5	0.000	***	−2.02 × 10−6	0.847		−2.38 × 10−5	0.154
Droad	−3.34 × 10−5	0.100		−3.26 × 10−5	0.073	*	−4.16 × 10−5	0.038	**
AreaQ	−0.5448	0.000	***	−0.18432	0.000	***	−0.6636	0.000	***
AreaB	−0.6148	0.000	***	−0.19097	0.000	***	−0.6784	0.000	***
AreaG	−0.5666	0.000	***	−0.19058	0.000	***	−0.6267	0.000	***
AreaM	−0.3711	0.000	***	−0.08401	0.034	**	−0.4737	0.000	***
AreaS	−0.5035	0.000	***	−0.18032	0.000	***	−0.6764	0.000	***
Fish	0.00084	0.063	**	0.00085	0.036	**	0.00096	0.015	**
Shrimp	0.00125	0.431		0.00072	0.612		0.00101	0.468
Shellfish	−0.00243	0.099	*	−0.00225	0.087	*	−0.00205	0.116
YD	0.00649	0.432		0.01399	0.238		0.01445	0.225
Empty	0.02485	0.091	*	0.02418	0.066	*	0.01882	0.149
Line	0.04228	0.022	**	0.03033	0.065	*	0.01973	0.275
SQ	0.00401	0.758		−0.00153	0.896		−0.00846	0.473
After	0.10698	0.000	***	0.11122	0.000	***	1.16 × 10−1	0.000	***
Green1	−0.09216	0.000	***	−9.12 × 10−2	0.000	***	−0.09090	0.000	***
Green2	−1.39 × 10−1	0.000	***	−0.10428	0.000	***	−0.10713	0.004	***
Dsolar	3.24 × 10−5	0.000	***	8.85 × 10−6	0.060	*	1.47 × 10−5	0.215
Pho (ρ)				0.67525	0.000	***
LAMBDA ($\mathsf{\lambda }$)							0.73331	0.000	***
R-squared	0.768			0.811			0.814
Adj-R-squared	0.763
F-statistic	166.086
BP	61.072	0.000	***	75.853	0.000	***	77.891	0.000	***
LIK	208.080			296.437			302.077
AIC	−374.160			−548.874			−562.155
SBC	−270.599			−440.382			−458.594
S.E	0.199			0.178			0.177
Moran’s I	0.1662	0.000	***
LM (lag)	232.146	0.000	***
Robust LM (lag)	8.425	0.004	***
LM (error)	257.994	0.000	***
Robust LM (error)	34.273	0.000	***
Likelihood-Ratio (lag)				176.715	0.000	***
Likelihood-Ratio (error)							187.995	0.000	***

), consistent with previous studies [40,41]. According to the results of the regression OLS method, the R-squared value is 0.768, in contrast to the Stage-1 model, which captured a significant negative relationship of the cost of harvesting shellfish. Because the errors in OLS regression also exhibited spatial autocorrelation, the neighborhood effect of the SLM and spatial heterogeneity of the SEM are significant. Meanwhile, the difference between the SEM and OLS model is that the SEM did not capture the relationship of Shellfish, Empty, Line, and Dsolar with aquafarm prices. In comparing the fitness of the three models, the result shows that the SEM has a better fit, which is different from the results for the Stage-1 model.

5.3. The Effects of an Aquavoltaics Policy on Small Scale Aquaculture

5.3.1. Policy Implementation Increases Aquafarm Prices

Our analysis result shows that land attributes, such as distance to neighboring city and roads and land location, are factors that increase aquafarm prices. This result is consistent with previous research findings [19,41]. However, several factors must still be considered when implementing an aquavoltaics policy, such as the effects that solar panels have on the quality of life of neighboring residents and on aquaculture species, as well as the investment costs for investors (solar energy enterprises or landowners). High aquafarm prices or high aquaculture opportunity cost will discourage aquavoltaic investments. The empirical results indicate that aquafarm prices increased by approximately 10% following the implementation of the aquavoltaics policy.

5.3.2. An Aquavoltaics Policy Should Give Priority to Shellfish Farming and Empty Ponds

Aquaculture attributes reflect the current status of aquaculture production. The cost invested in aquaculture is positively correlated with aquafarm prices, while shellfish farming is negatively correlated. This shows that despite the adoption of an aquavoltaics policy, use of aquafarms remains highly correlated with aquaculture input and expected returns. Aquafarms with low aquaculture input are more susceptible to policy impact. The cost of harvesting shrimps has exhibited a non-significant effect primarily because shrimps are cultivated together with other fish species; therefore, its relationship with aquafarm prices is not pronounced. The positive effect of empty ponds on aquafarm prices reflects a high degree of correlation between land-use conversion and aquavoltaics policy. This is because a fish pond must be drained before solar PV systems can be installed.

5.3.3. Renewable Energy Facilities Influence Solar PV Investment Intentions

The availability of effective renewable energy facilities and policies is a prerequisite for solar-aquaculture integration. Our empirical results show that after the government implemented an aquavoltaics policy for energy transition, aquafarm prices rose significantly, which also affected landowners’ assessment of long-term benefits and decisions. Neighboring facilities related to renewable energy (e.g., Line) also influence aquafarm prices and the feasibility of installing solar energy systems on the farm. Policy requirements and aquafarm surface area (Green1, Green2) have not significantly influenced aquafarm prices. The aquavoltaics policy was continuously revised during its period of implementation to relax policy requirements (i.e., the aspect about more than 70% of landowners who own more than 70% of land). Nevertheless, new data are required to verify whether the policy changes offer sufficient incentives to facilitate aquavoltaic promotion. Meanwhile, increasingly more measures can be adopted to increase market transparency [14].

5.4. Toward a Systematic Approach to Aquafarm Use and Pricing in Small-Scale Aquaculture

Solar-aquaculture integration can be regulated by not only land surface area and size and the proportion of aquaculture activities, but also by a systematic approach to using and pricing aquafarms. This approach can be used to create regional energy plans and can be incorporated into policy plans, while taking into consideration factors related to small-scale aquaculture, such as stakeholders, food supply, ecosystem and landscape, and the living environment of neighboring settlements [45]. It can be used to draw up blueprints of aquavoltaic systems, including economic conditions and energy benefits, and coupled with policy incentives or community energy as policy goals [46,47], it can be used to seek consensus among local residents to create a synergy between energy, economy, and local living activities [48,49].

6. Conclusions

Along the path to energy transition, the Taiwan government hopes to source 20% of its total electricity from renewable sources by 2026 by building 20 GW of solar PV capacity. Given the small surface area of Taiwan’s land, the key to energy transition is an aquavoltaics policy that promotes solar-aquaculture integration. With its aquaculture sector composed mainly of small-scale aquafarmers, the government of Taiwan has continually revised policies in an attempt to ensure the economy of the renewable energy and aquaculture sectors, environmental sustainability, and fishery development. Although prior studies have examined farmland values, renewable energy policies, and land-use transitions, most have centered on urbanization-driven price changes or rent-seeking behaviors associated with farmland conversion. Empirical assessments of how renewable energy development affects aquafarm prices—particularly under aquavoltaics policy—remain limited. Existing research often relies on random samples or sector-specific case studies that fail to capture the complexity of aquaculture production systems, regional heterogeneity, and the spatial dependencies inherent in coastal aquaculture regions. Furthermore, few studies have systematically evaluated how policy incentives interact with aquaculture attributes, land characteristics, and renewable energy infrastructure to shape land market dynamics in small-scale aquaculture systems.

This study addresses these gaps by compiling comprehensive transaction-level data (1024 records from 2012 to 2021, rather than random samples), thereby establishing a more robust empirical foundation. An integrated analytical framework is constructed to jointly incorporate land, aquaculture, and renewable energy attributes. By applying both the Hedonic Price Model and spatial econometric techniques, the analysis accounts for spatial autocorrelation, neighborhood effects, and location-based disparities that previous studies often overlook. This approach enhances the precision of policy impact assessment and provides empirical evidence on how aquavoltaics policy influence land-use decisions, price structures, and the adaptive behavior of small-scale aquaculture stakeholders. In doing so, the study advances existing knowledge by addressing deficiencies in data completeness, spatial identification, and policy-relevant evaluation. The results of our empirical study show that transacted aquafarms are clustered in space, and aquavoltaics policy pushes up aquafarm prices. In addition to land attributes, attributes related to aquaculture (e.g., shellfish farming and empty fish ponds) should be considered in policy promotions. Regulatory requirements for aquavoltaics installations should ensure the economic aspect of aquaculture activities. Lastly, a systematic approach to using and pricing aquafarms is required when implementing an aquavoltaics policy. Meanwhile, more measures should be increasingly adopted to improve market transparency, strengthen consensus among local residents, develop energy blueprints and aquavoltaic plans that ensure the economy of the fishery sector, and create a synergy between the development and energy transition of small-scale aquaculture. This study, however, has several limitations that suggest directions for future research. The analysis is based on cross-sectional data, which may not capture long-term price dynamics; thus, future studies could apply longitudinal price modeling to assess temporal changes in aquafarm values under evolving policy conditions. Moreover, the present study focuses primarily on economic and spatial factors, while environmental and ecological impacts—such as changes in water quality, biodiversity, or land-use sustainability—remain to be systematically evaluated. Future research may also explore stakeholder behavior, social acceptance, and regional disparities in aquavoltaic adoption to provide a more comprehensive understanding of policy effectiveness and its broader implications for sustainable coastal development.