The impact of off-site land use energy intensity on the overall life cycle land use energy intensity for utility-scale solar electricity generation technologies

Concerns about the potential environmental impacts of energy acquisition can present a barrier to development, and for solar energy one such concern is the amount of land used per megawatt-hour (MW h) of electricity generated, which is referred to as the land-use energy intensity (LUEI). The LUEI presents some challenges for solar energy because solar energy tends to have a larger power plant footprint per MW h when compared to electricity from natural gas or coal. For example, the LUEI for utility-scale solar energy facilities ranges from 0.1 to 10 m² per megawatt-hour.¹ Although Ong et al.² provide a narrower range, it is one that still falls within these bounds. By comparison, Fthenakis and Kim³ estimate that the LUEI for coal generation power plants ranges from 0.006 to 0.032 m² per megawatt-hour.

However, these LUEI estimates are incomplete from a life cycle perspective as they represent only the land area occupied by the infrastructure associated with the production of electricity, that is, the “on-site” LUEI. All electricity generation uses land “upstream” of electricity generation for manufacturing of the generating equipment and, in many cases, for fuel extraction. Land is used “downstream” as well—for handling of fuel wastes and byproducts and for equipment disposal. The extraction of coal, for example, can require mining across vast areas, resulting in upstream LUEIs that are an order of magnitude or more larger than the on-site LUEIs for the power plants. Off-site LUEIs for coal surface-mining in the U.S. range from 0.043 to 1.45 m² per megawatt-hour, with an average of 0.4 m² per megawatt-hour.³ It should be noted that analyses of LUEI can use plant capacity instead of generation in the denominator (m² per megawatt instead of m² per megawatt-hour). We use a generation-based LUEI in this analysis because a capacity-based LUEI is not suited to account for an ongoing stream of fuel inputs for electricity generation, but rather one-time material inputs for construction. While solar electricity fuel (solar radiation) does not use land, we use a generation-based LUEI so that it can be compared to similar values for fossil fuel generation.

Much of the LUEI literature shares common methods and data; it uses Life-Cycle Analysis (LCA) software in combination with Life-Cycle Inventory (LCI) databases. Most researchers use these tools because measuring inputs from all upstream and downstream components in a process can quickly lead to lists of hundreds of materials. For example, Burkhardt et al.⁴ published an LCI database for concentrated solar thermal (CSP) technologies that identified more than 200 material and energy requirements for constructing, operating, and decommissioning a CSP facility. Furthermore, each of these 200 inputs has its own production pathway, increasing the scope of the analysis exponentially. This complexity is where the LCI databases make performing LCAs manageable; indeed, it is for these reasons that much of the literature has relied on LCA software that leverages the vast LCI databases to perform upstream and downstream analyses.

The limited literature that reports solar LUEI calculations is no exception. Fthenakis and Kim,³ for example, calculate both the on-site and off-site LUEI for solar energy “from the life-cycle inventories of the PV (photovoltaic) modules and BOS (balance of system) components given in recent publications, along with the land-use factors from the Ecoinvent Database.” In addition, Mason et al.,⁵ Alsema et al.,⁶ and de Wild-Sholten et al.⁷ use the Ecoinvent LCI database to calculate LUEIs. Unfortunately, neither the land use factors used for each technology nor the mathematics used to calculate the LUEI are included in any of these publications, and, as a result, the reader must assume that the land-use factors for each technology were in fact accurate and that the mathematical way in which the land-use factors were calculated was correct. Last, because seemingly all publications on LUEI calculations seem to use the same methodology, there is no way to externally validate the results. Klein and Rubin⁸ provide a more transparent LCI, but only for on-site land use.

Given these concerns, we developed a methodology for analyzing off-site land use impacts within the solar industry, using the LUEI (m²y/MWh) as the functional unit. For a full explanation of this unit, see Horner and Clark¹ and the explanation of the LUEI algorithm below. Because this analysis differs with regard to the data and methods used in Fthenakis and Kim,³ it serves as a useful point of validation of existing LUEI methods. Furthermore, the off-site solar LUEI values provided by this study could be used for an equitable comparison with those of coal and natural gas, the main fossil fuels used to produce electricity in the United States.⁹ Our specific objectives in this analysis were to: (1) develop a new method for calculating the off-site LUEI for both upstream and downstream production pathways for silicon PV, thermal CSP, and thin-film PV; and (2) use the results to validate existing calculations for off-site LUEIs.

We used data gathered from mineral extraction, solar manufacturing, and disposal companies combined with geographic information systems (GISs) to calculate the off-site LUEI for three different utility-scale solar energy technologies: silicon-PV (Si-PV), cadmium-telluride (CdTe) thin-film PV, and parabolic trough CSP. First, we used LCI data published in the literature for each of these technologies to identify the materials used in the finished product. The LCI sources, including the bases of their analyses, are given in Table I. One of the sources used an existing power plant as its basis. This power plant, Solar Electric Generating Systems (SEGS) VIII, includes an auxiliary natural gas system that provides backup power during low and non-solar hours and accounts for approximately 25% of annual production. Because we were only interested in the solar production from this facility, we assumed a solar-only capacity factor (see notes in Table I) but conservatively included all material inputs for the plant.

The other two used design specifications at the module level. Next, we performed an internet search to identify upstream facilities that produce those materials, and for which location data and production/disposal rate data were given. Finding adequate information to properly allocate land use proved difficult, so all facilities we found that provided this information were used in this analysis. The area of these upstream and downstream facilities was calculated using GIS software and aggregated to form the off-site LUEI that includes both upstream and downstream components. A basic description of the methods used in this analysis follows. Detailed tables containing the facilities found and analyzed in the upstream and downstream pathways for each technology, including all key parameters, can be found in the online supplementary material.¹⁰ 

We limited the scope of this analysis to the major material inputs (by mass) to each technology, and as a result, this analysis does not include all upstream and downstream materials. Among these major inputs, only those for which we could identify a facility, calculate its area, and acquire data regarding its annual production capacity were used in the LUEI estimation. For example, data for the entire silicon production pathway was found for PV, from poly-Si to wafers to cells to modules. For CdTe thin film, values were found for the production of cadmium (Cd) and tellurium (Te), starting from the refinery/smelter and ending with the construction of the panel. A list of the stages from each production pathway that is included in this study is provided in Table II. We performed searches for extraction data for the parabolic trough CSP technology but did not find any that was useful for this analysis. In fact, it was difficult to find the necessary extraction data for all of the technologies because for the data to be useful, it needed to include allocation factors, mine areas, and extraction totals or rates, which are often unpublished by private operators. As a result, most data were available for the processing stage and above (see Table II). For the disposal/recycling pathways, it was assumed that the mass of a material used to produce a technology, and included in the upstream analysis, equals the mass of that material needing disposal/recycling at the end of its life, and thus is included in the downstream analysis.

This study defines off-site LUEI as the aggregate land area (in square meters or m²) occupied by the upstream and downstream stages of the solar industry per unit of energy production (megawatt-hours, MW h) per year (yr) for the components of a given utility-scale solar facility. As such, the off-site LUEI is in units of m²y/MWh (which is read as meters squared years per megawatt-hour). This definition of LUEI is consistent with that used in Horner and Clark.¹ The off-site LUEI was calculated for each of the three technologies studied by summing the individual LUEI values for each stage of a technology's upstream and downstream pathway. The following algorithm gives that summation:

where α (unitless, with a value between 0 and 1) is the land-use allocation factor for a production or disposal facility i; A_i (m²) is the land area occupied by that facility i; C_i (MW/yr) gives the annual production or disposal capacity of facility i in units that reflect the amount of electricity generating capacity embedded in that produced or disposed material (relative to the amount of that material required for a full plant of a certain capacity); t (yr) is the assumed lifetime of the production or disposal facility i in years; 8760 (h/yr) is the number of hours in a year; and C_f (unitless, with a value between 0 and 1) is the capacity factor of the utility-scale solar plant. The variable C_i represents the allocation of the power generating capacity of the final, fully constructed solar technology to the production of raw materials like sand, steel, and concrete (which do not generate electricity themselves) and intermediate components such as PV cells, in order to properly scale the land used by the material or component production facilities. The capacity factor of the solar plant must be included even though this algorithm does not calculate on-site LUEI because the material inputs to the plant, and thus the facilities that create them, must be scaled for the actual plant electricity output just as the on-site LUEI is scaled. Thus, the on-site and off-site LUEI are consistent and can be compared. For all facilities, a 30-yr lifetime was used for the value of t.¹⁴ A capacity factor of 0.18 was used for both the Si-PV and CdTe thin film technologies, whereas a value of 0.25 was used for CSP.¹⁵ The arithmetic mean of the LUEI values for multiple facilities representing the same production or disposal/recycling stage is obtained by summing across the facilities and dividing by the number of facilities (n). Then, the LUEI values for each stage (j) are summed over the number of stages (s) to determine the cumulative off-site LUEI for that solar technology.

GISs were used to determine the area of the facilities located as part of the upstream/downstream LUEI analysis for each technology. To perform this step, we first acquired the address of each facility by visiting each facility's website. Using the address, the facility was located in Google Earth. Google Maps and its Street View component were used to verify sites if there was any uncertainty about boundaries in Google Earth. Google Earth's “Add Polygon” function was then used to trace the outline of a facility. The entire property occupied by the business, rather than just the building(s), was outlined and considered part of the land area occupied by the facility. The approach of using the property boundary instead of the building boundary was adopted mainly to account for parking lots or other features that were not buildings per se but nonetheless impact land-use (Fig. 1).

To determine the area of polygons drawn around production facilities, the polygons were first saved as.kmz files in Google Earth (Fig. 1). These files were then converted from.kmz to GIS layer files in Esri's ArcMap 10.0. The Feature function of the Measure tool was then used to calculate the area of the facility polygons. These land areas were recorded in square meters for use in Eq. (1).

Facility production and disposal capacity data must be presented as MW of solar capacity produced per year for use in Eq. (1). In other words, the material flows going into building a power plant and the material flows going out of disposing one need to be quantified relative to the power capacity of the plant. Thus, each amount of material can be represented as a proportion of the plant's capacity to produce energy. Most upstream firms, however, record output in terms of mass (i.e., kilograms [kg] or metric tons) of a material, which then must be converted into MW/yr. To convert mass to MW, the LCI databases associated with each technology (listed in Table I) were used. These LCI databases list the materials and energy consumed or emplaced during the construction of a specific amount of a technology (measured in capacity). For example, for a PV panel manufacturer, the LCI database lists the quantity (mass) of material and energy consumed to construct what is termed a “1-silicon module piece” (pc), which has a power rating of 165 watts (W). Because all of the data listed in the LCI database are relative to this 165-W power rating, data supplied by manufacturers (in kg) can be converted into units of power capacity (MW). In this manner, the output of upstream firms, reported in kg, were converted to MW by dividing by the kg/MW listed in the LCI database. All such conversion ratios are listed in tables in the online supplementary material (Tables S.IV, S.VII, and S.X).¹⁰ 

A complicating factor in this analysis is the fact that a material of interest may come from a manufacturing facility or refinery that produces more than one product. Because the solar plant is consuming only the one output, the area allocated to the solar plant should be smaller than the total area of the entire facility; that is, the allocation factor should be proportionate to the fraction of output from the entire facility that is used by the solar plant. However, if a facility only produces one product, it does not matter how much of it goes to a solar power plant or to some other unrelated industry, because the ratio of facility land area to material produced is what matters for LUEI. Some of the major facilities used for this report are single-product, and thus have an allocation factor of 1.

To place the results of this study in context, we compared our calculated off-site LUEI results with on-site LUEIs from a previous study.¹ The on-site LUEI values are the arithmetic mean of a range of literature values for each technology, and as such provide a basis for a general order-of-magnitude comparison between the two values.

Our results are presented in Table III. Comparison with on-site LUEI values from the literature indicate that the off-site solar LUEI is probably two or three orders of magnitude smaller than the on-site LUEI for each of the technologies examined. We summarize the results by technology below. The online supplementary material includes tables that detail the facility-specific results.¹⁰ 

The cumulative “off-site” LUEI for the Si-PV technology is 0.16 m²y/MWh, which amounts to approximately 1.1% of the on-site LUEI for this technology (15.2 m²y/MWh).¹ Most of the upstream LUEI (0.16 m²y/MWh) derives from the module pathway, and only a very small portion (0.01 m²y/MWh) can be assigned to the BOS pathway. Although individual LUEI values ranged widely for different materials in the BOS pathway (Table S.I, online supplementary material¹⁰), the LUEI estimates were typically much smaller than those in the solar module pathway. This result occurs because a much smaller mass of components is used per megawatt of capacity for the BOS production than for the module production, which is evident in the conversion ratios listed in Table S.III in the online supplementary material.¹⁰ 

The downstream pathway for Si-PV has an LUEI of 0.0023 m²y/MWh, which amounts to 0.015% of the generation-site LUEI for Si-PV. Only scrap metal for BOS components is included in this pathway. There is no downstream LUEI for PV modules given that recycling of silicon PV modules occurs mainly in the same facilities that produce the panels, meaning that the land-use component accounting for recycling of the panels is embedded in the upstream LUEI in this analysis.

Our results indicate that the cumulative off-site LUEI for CdTe Thin Film technologies is 0.028 m² yr/MW h, which is only 0.19% of the on-site LUEI. Similarly to the Si-PV results, the upstream LUEI (0.025 m²y/MWh) is an order of magnitude larger than the downstream LUEI (0.0026 m² yr/MW h) for this technology. Between 90% and 95% of the total mass of the panel is recycled, leaving only a small fraction for disposal, which is accounted for here. Because recycling occurs on the site of the manufacturing facility and is not accounted for separately in the data used for this study, this land use is accounted for in the upstream LUEI calculations.

Our analysis of the CSP pathway resulted in a cumulative LUEI value of 0.011 m²y/MWh, which is roughly 0.11% of the generation-site LUEI. The cumulative LUEI is derived from about two-thirds of upstream components (0.0076 m²y/MWh) and one-third downstream (0.0033 m²y/MWh). Some of the inputs to CSP production, such as aluminum and copper, have a very small LUEI, mainly because of the low quantity of material that is needed per megawatt of installed capacity. The largest downstream material, by mass, is concrete, which is recycled on-site, and therefore does not have a downstream LUEI associated with it (see Tables S.VIII–S.X in the online supplementary material¹⁰).

The study by Fthenakis and Kim³ is one of the few existing studies to report estimates of upstream LUEIs. Our estimate for Si-PV (0.16 m²y/MWh) is smaller than theirs (0.77 m²y/MWh), but on the same order of magnitude. The methods used by Fthenakis and Kim³ are clearly distinct from the methods used in this study (which may account for the difference in results), and as a result this study serves as external validation of their similar conclusion that off-site LUEI is significantly smaller than on-site LUEI for solar electricity generation.

It should be mentioned that this analysis does not examine all materials contained within the upstream and downstream pathways for each technology. For CSP, Burkhardt et al.⁴ list 41 materials in their LCI database; however, for the process-based lifecycle approach used here, we were only able to acquire the necessary data for five materials. That said, we were able to acquire comprehensive data for the downstream CdTe pathway from an industry source, and even with this comprehensive data, the results of the analysis are similar to those for the other pathways—that is, the off-site LUEI is a small fraction of the on-site LUEI. It seems, therefore, that even if we were able to acquire data for the remaining upstream and downstream materials, the final off-site LUEI value would remain small compared to the on-site value.

Any upward revisions that might occur by including more upstream and downstream stages would also be offset to some degree by utilizing less conservative assumptions. For example, because of a paucity of data, we were only able to calculate land-use allocation factors for two facilities in this analysis, and all other facilities assumed an allocation factor of 1. This treatment implies that 100% of the output of the facility is the material used by the solar industry. Although this may be the case for some facilities, such as a Si-wafer facility, in most cases the true value of the allocation factor would be a small fraction of 1. For example, the calculated allocation factors for the smelters that produced cadmium and tellurium were 0.0027 and 0.000015, respectively, because those elements represented such a small proportion of the total output (by mass) of the smelters. Similar allocation factors would be expected for many other facilities (e.g., steel and aluminum facilities) and we acquired the appropriate data.

The main conclusion from our analysis is that the upstream and downstream LUEIs for solar technologies are <1% (CdTe and CSP) or approximately 1% (Si-PV) of on-site LUEI. Therefore, this study confirms what is general practice—using on-site LUEI estimates is sufficient for estimating lifecycle LUEI for utility-scale solar technologies. Notably, this assumption does not apply to other electricity generation technologies, such as coal, which studies of Fthenakis and Kim³ show a much higher level of off-site land use compared to on-site levels.

This analysis serves as external validation for current LCI-database methods for calculating off-site solar LUEI. By focusing solely on off-site LUEI, and by not relying on proprietary databases, it is able to provide complete and accessible calculations for the reference of other researchers (available in the online supplementary material¹⁰). While the study could benefit from a greater availability of production data at various upstream and downstream stages, it does not appear likely that the provision of such data would substantially change its overall conclusions.

The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory (“Argonne”). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract No. DE-AC02-06CH11357. Argonne's work was supported by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, Office of Solar Energy Technology. The U.S. Government retains for itself, and others acting on its behalf, a paid-up nonexclusive, irrevocable worldwide license in said article to reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display publicly, by or on behalf of the Government.