The Potential of Variable Renewable Energy Sources in Mexico: A Temporally Evaluated and Geospatially Constrained Techno-Economical Assessment

: Due to the increasing global importance of decarbonizing human activities, especially the production of electricity, the optimal deployment of renewable energy technologies will play a crucial role in future energy systems. To accomplish this, particular attention must be accorded to the geospatial and temporal distribution of variable renewable energy sources (VRES), such as wind and solar radiation, in order to match electricity supply and demand. This study presents a techno-economical assessment of four energy technologies in the hypothetical context of Mexico in 2050, namely: onshore and offshore wind turbines and open-ﬁeld and rooftop photovoltaics. A land eligibility analysis incorporating physical, environmental, and sociopolitical eligibility constraints and individual turbine and photovoltaic park simulations, drawing on 39 years of climate data, is performed for individual sites across the country in an effort to determine the installable potential and the associated levelized costs of electricity. The results reveal that up to 54 PWh of renewable electricity can be produced at a levelized cost of electricity of less than 70 EUR · MWh − 1 . Around 91% (49 PWh) of this electricity would originate from 23 TW of open-ﬁeld photovoltaic parks that could occupy up to 578,000 km 2 of eligible land across the country. The remaining 9% (4.8 PWh) could be produced by 1.9 TW of onshore wind installations allocated to approximately 68,500 km 2 of eligible land that is almost fully adjacent to three mountainous zones. The combination of rooftop photovoltaic and offshore wind turbines accounts for a very small share of less than 0.03% of the overall techno-economical potential.


Introduction
The last decade has seen a significant increase in the contribution of renewable energy technologies to the reduction of anthropogenic CO 2 emissions. Currently, renewables account for around 34% of global installed power capacity [1], and investments in them have grown steadily by at least 10% per year over the last decade [2], spurred by cost reductions and technological improvements [3]. Moreover, international commitments in the form of environmental policies and market integration mechanisms, such as those outlined in the 2015 Paris Climate Agreement [4], underpin the trend towards increasing renewable power generation. Hence, in order to aid policy initiatives and the other decisionmaking processes necessary for an effective energy transition, the generation potential of renewable energy technologies must be assessed with consideration given to their intrinsic geospatial and temporal heterogeneity for optimizing their use.

Literature Review
Past studies have employed a mixture of methodologies and generally lack the comparable and reproducible models and data inputs that are necessary to assess renewable energy in Mexico, which hinder a direct comparison between assessments. In 1995, the first national onshore wind assessment was conducted by Schwartz et al. [15], in which regional wind zones with both high and poor potentials were identified and formed the basis of most of the subsequent regional wind energy assessments. For instance, Jaramillo et al. [16] concluded that in the federal state of Baja California, a region with high wind potential identified by Schwartz et al., the average levelized cost of electricity (LCOE) from onshore wind would be around 50 EUR·MWh −1 . Later, in 2010, Hernández Escobedo et al. [17] assessed the wind potential of the entire country, estimating that a modeled 750 kW turbine could operate with 1700 full load hours (FLH) on average. However, the corresponding LCOE was not provided, and the model used was not open source, which makes the findings difficult to compare. Subsequently, Hernández Escobedo et al. [18] performed a similar analysis with a 1.5 MW wind turbine in the state of Veracruz. The result was an FLH estimation of around 1750-2630 h per year. This second study did not include an LCOE estimation either. Likewise, the works of Figueroa Espinoza et al. [19], Carrasco Diaz et al. [20], Carreon Sierra et al. [21], and Hernández Escobedo [22] focused on different regions in Mexico but again did not include detailed techno-economical parameters in their results. In 2019, Rodriguez Hernández et al. [23] employed yet another methodology to assess the economic feasibility of small wind turbine designs in the Mexico City region and reported that only turbines with capacities of 0.5 and 0.8 kW, which are not in line with recent and future trends in utility-scale onshore wind turbines [24], could be economically viable in the region. Most assessments hitherto have only focused on onshore wind potential and, to the authors' knowledge, only one study, the "National inventory of zones with high potential for clean energy" (AZEL) [25], estimates open-field photovoltaics (PV) potential nationwide. Furthermore, again to the best of the authors' knowledge, offshore wind energy and rooftop PV have not yet been assessed on a national level. The AZEL assessment carried out by the National Ministry of Energy (SENER) consists of estimations of renewable energy potentials based on an energy density analysis presented via a geographic information system (GIS) map of onshore wind and open-field solar PV, among others. Although the AZEL report is the first assessment to offer some land exclusion constraints for the placement of PV parks and onshore turbines, such as protected areas, human settlements, transportation roads, transmission lines, minimum continuous areas (ignores the sparse land available for installations smaller than up to 12.5 km 2 ), terrain elevation, airports, coastal lines, and minimum wind speed, it does not provide techno-economical details on the costs of electricity production. The AZEL report also fails to provide time series generation profiles, account for efficiency gains over time due to technology advancement, and offer the optimal arrangement of VRES installations or the cost functions that are necessary to more precisely estimate green potential.
Internationally, relevant renewable energy assessments have incorporated complementary considerations and progressively improved their accuracy and standardization.
McKenna et al. [26] modeled onshore wind potential for Germany using wind speed data with 1 km 2 spatial resolution, multiple wind turbine designs, and a detailed land eligibility analysis (LEA). Building on McKenna's approach, Jäger et al. [27] assessed the onshore wind potential of Baden-Württemberg, Germany, positioning wind turbines according to the wind direction to maximize the utilization of the available land and accounting for wind blocking from neighboring turbines. Staffell and Pfenninger [28] used NASA's MERRA-2 database [29] over 20 years to re-analyze previous wind power assessments and found that, depending on the weather of a particular year, wind assessments could differ by up to 50% in terms of capacity factors and total generation output. The researchers concluded that taking larger weather timespans could help reducing weather variability across different years. Robinius et al. [30,31] also applied LEA and turbine positioning algorithms in a wind assessment for Germany for an advanced onshore wind turbine as part of a 2050 energy system simulation of Germany. Later, Ryberg et al. [24,32] assessed the onshore wind, open-field PV, and rooftop PV potential in Europe to investigate the occurrence of lulls in electricity production, incorporating and improving upon the cumulative progress of past approaches. Ryberg's study included an extended LEA [33,34] and the modeling of futuristic onshore wind turbines [24] and advanced PV panels to account for technological progress in the foreseeable future [24], as well as a scalable VRES placement algorithm and a quasi-physics-based algorithm to calculate the generation profiles and apply some air density and sunlight irradiation corrections. Finally, Caglayan et al. [35] adopted Ryberg's methodology to calculate the potential of offshore wind energy for Europe, also adding the investment cost estimation function described by Fingersh et al. [36] and Maness et al. [37] as a function of the foundation type, distance to shore, and ocean depth.

Contribution of the Paper
Due to the recent progress in the field not being applied to Mexico and given the large methodological divergences in previous assessments, this study aims to assess the techno-economical potential in the entire Mexican territory (Approximately 2 million km 2 of land and 3.2 million km 2 of sea (assigned to the nearest state)) of four VRES technologies: onshore wind, offshore wind, open-field PV, and rooftop PV for the future 2050 context using a state-of-the-art approach with tailor-made geospatial considerations. The ultimate objective of this work is that the results of this assessment to provide practical information to facilitate decision making for key stakeholders with respect to where renewable energy can be optimally exploited. The results can also serve as inputs for an optimal and sustainable energy system design at a national level.

Organization of the Paper
The study is structured as follows: Section 1 serves to introduce the relevance of the topic and to declare the main contribution of the paper. Next, the methodology is described in Section 2. The results and discussion are presented in Section 3 and Section 4, respectively. Finally, relevant conclusions are presented in Section 5.

Methodology
The selected assessment approach comprised three main steps. First, an extensive LEA was applied to Mexico, maintaining a particular emphasis on country-specific factors. The total eligible area per technology and distribution were obtained here. Secondly, the maximum number of VRES installments with their techno-economic parameters were individually derived by applying technology design correlations according to the local terrain and weather characteristics, as well as costing algorithms as utilized in the selected literature. Finally, quasi-physics-based renewable energy production simulations were carried out to obtain time series generation profiles for each installment. The average electricity generation, FLH, and LCOE were also calculated, and the VRES techno-economic potential was obtained by curtailing electricity production with an LCOE over 70 EUR·MWh −1 that corresponds to historic average electricity costs in Mexico according to data reported by the National Electricity Control Authority [38]. A more detailed description of the three main steps in the methodology is presented in the subsequent subchapters. The results, discussion, and conclusions are structured in accordance with the administrative state division, as well as geographical zones (shown in Figure 1) to help the reader follow the comparisons more easily.

Land Eligibility Analysis
The Geospatial Land Availability for Energy System (GLAES) model [39] developed by Ryberg et al. [34] was employed to determine the geospatially constrained eligible land for the investigated VRES technologies. In total, 34 land constraints and their technologyspecific exclusion zones were considered. This large number of constraints was intended to underpin a conservative approach and avoid "environmental problem shifting" [40]. Table A1 in the Appendix A lists the land constraints and exclusion zones in detail for onshore wind, offshore wind, and open-field PV. The LEA for rooftop PV was only constrained by human settlements (house rooftops) in accordance with Ryberg's approach [32]. For each technology, the spatial reference system EPGS 6263 [41] was used with a spatial resolution of 100 m 2 , as it represents a good trade-off between spatial fidelity, data availability, and computational intensity.
Of a total of 34 constraints, 27 were previously employed by either Ryberg et al. [32], Caglayan et al. [35], or Heuser et al. [42] and were kept unchanged because they do not conflict with the particularities of the Mexican context. Official datasets published by Mexican authorities were preferred when possible, as detailed in Table A1 in the Appendix A. Detailed justifications for the selection of these constraints, filtering procedures, and the buffer zones used can be found in the preceding works of Ryberg et al. [32], Caglayan et al. [35], and Heuser et al. [42], whereas the following subchapter describes in more detail the adjusted and additional considerations for the Mexican context.

Specific LEA Constraints for Mexico
Seven primary constraints specific to the context of Mexico are taken into account, which constitute the novel contributions of this work, namely: military areas, harbors, LNG terminals, geothermal sites, primary jungles, active volcanoes, and hurricanes. The reasons for their inclusion and the corresponding buffer zones assumed are also outlined. Details on the data sources and assumed buffer distances are included in Table A1 in the Appendix. In general, international data sources were used, following Ryberg et al.'s [43] methodology for the most part. Nevertheless, data from national sources were preferred with respect to constraints relating to demographics, energy infrastructure, and land use, as they were generally more up to date than international data sources.
All VRES placements within 1 km distance from military areas, harbors, and LNG terminals were excluded due to the security issues entailed and the assumption that sig-

Specific LEA Constraints for Mexico
Seven primary constraints specific to the context of Mexico are taken into account, which constitute the novel contributions of this work, namely: military areas, harbors, LNG terminals, geothermal sites, primary jungles, active volcanoes, and hurricanes. The reasons for their inclusion and the corresponding buffer zones assumed are also outlined. Details on the data sources and assumed buffer distances are included in Table A1 in the Appendix A. In general, international data sources were used, following Ryberg et al.'s [43] methodology for the most part. Nevertheless, data from national sources were preferred with respect to constraints relating to demographics, energy infrastructure, and land use, as they were generally more up to date than international data sources.
All VRES placements within 1 km distance from military areas, harbors, and LNG terminals were excluded due to the security issues entailed and the assumption that significant interference between VRES production and the operations conducted in these Energies 2021, 14, 5779 6 of 24 areas could occur. In a 200 m radius from geothermal sites, where geothermal energy systems can potentially be installed, placements were restricted to maintain the potential of this renewable energy source. Placements in primary jungles and within 1 km of their surrounding areas were excluded to protect biodiversity, as recommended by the SENER's environmental evaluation [44]. Based on the conclusions of Capra et al. [45], circular exclusion areas with a radius of 2 km from active volcanoes' vents were excluded for onshore wind turbines and open-field PV parks to avoid the risks related to volcanic activity. Offshore wind turbines were considered by the authors not to be subject to these risks, as they must be placed in water at least 15 km from coastlines [35], and no records of marine volcanic eruptions in those areas of Mexico could be found. Similarly, areas hit by hurricanes of category 3 or above, as per data from 1980 to 2019 by the US National Oceanic and Atmospheric Administration (NOOA) [43], were excluded for both wind turbine installation variants. Category 3 hurricanes have a maximum sustained wind speed above 60 m/s, in accordance with the limit of most commercial wind turbines' resistance (40-80 m/s) [46]. The exclusion zone, in this case, is taken as 30 km, as this is the maximum radius at which any hurricane has shown sustainable wind speeds above 60 m/s [47]. This security buffer is increased to 50 km for PV installations where not only wind speeds but also flying objects can cause damage.
Land constraints such as radio towers, land instability (i.e., the likelihood of landslides), and ground composition considered by previous authors [48][49][50] were not included herein due to a lack of these data existing for Mexico. A minimum distance to a transmission line is not considered applicable in the case of Mexico in 2050, as the country's electricity demand is expected to grow considerably [11], adding the need for extensive grid expansion and, therefore, making the current status quo of connection not applicable to a 2050 context. Arguably, other LEA constraints such as earthquakes and tsunamis can potentially be included in an LEA for a country such as Mexico. Nevertheless, this study does not include these due to empirical evidence found in the literature presented in the Appendix A.

Techno-Economic Parameters
Once the available areas were obtained, GLAES was again used to place minimally separated elliptical or polygonal geometries representing either single turbines or solar PV parks, respectively. Around 1.6 million elliptical shapes representing onshore turbine rotors of 136 m in diameter oriented towards the 39-years average wind direction were incorporated in the available land with an inter-turbine separation of 8 × 4 times the rotor diameter (transversal x longitudinal axis) in accordance with Ryberg et al. [24]. Similarly, nearly 600,000 offshore turbines were modeled with a fixed rotor diameter of 210 m and separated by 10 × 4 times the rotor diameter, as per Caglayan et al. [35]. For PV technologies, polygonal shapes representing about 660,000 PV farms with several PV strings and 20 km 2 of PV-covered house roofs were assigned following Ryberg's [32] approach. Next, the technology's design parameters were obtained. Futuristic onshore wind turbine design incorporated in the Renewable Energy Simulation toolkit (RESKit) [51] model for Python by Ryberg et al. [24] for a 2050 context were used to account for technological development. Utilizing this algorithm, design parameters such as hub height and specific capacity were inferred for each turbine from local wind speed and terrain characteristics for a given rotor diameter. When applied to the previously obtained placements, onshore turbine designs resulted in nominal capacities and hub heights ranging from 3.2 to 5.9 MW and 88 to 185 m, respectively. Individual onshore turbines' total capital costs (CAPEX), ranging from 1090 to 1350 EUR·kW −1 , were obtained according to the costing procedure used by the same author. A future 1100 EUR·kW −1 baseline cost design was assumed and assumed to cover the cost of capital, balancing of the system, and financial costs according to their CAPEX shares, suggested by NREL's turbine cost breakdown [52] and costing relationships noted by Fingersh et al. [36] and Maples et al. [53]. With respect to offshore wind, a fixed turbine design with a hub height of 135 m, rotor diameter of 210 m, and capacity of 9.4 MW was selected following the approach of Caglayan et al. [35]. The only design parameter used to select offshore turbines was the foundation type that yielded the lowest LCOE of the four investigated options, i.e., monopole, jacket, semisubmersible, and floating spar according to the placement criteria used by Maness et al. [37]. Similar to their onshore counterparts, the CAPEX cost function of offshore turbines uses a 2300 EUR·kW −1 baseline design, resulting in a range of 1570-10,000 EUR·kW −1 due to cost additions arising from foundation type, sea floor depth, and distance to shore [35].
For PV technologies, the models Winaico WSx-240P6 (fixed tilt) for open-field and LG 360Q1C-A5 for rooftops were selected, in accordance with Ryberg's methodology [32]. The tilt and azimuth angles were obtained on the basis of the Global Solar Atlas (GSA) [54] database for each PV park location, whereas the CAPEX assumptions correspond to 2050 projections by the Fraunhofer Institute [55] of 500 EUR·kW −1 for open-field PV and 800 EUR·kW −1 for rooftop PV modules. As in the case of wind turbines, the PV CAPEX is assumed to cover all costs associated with the installation. Furthermore, the technical parameters of the PV models were derived from the Go Solar California database [56], with a projected module efficiency of 24% and 30% for open-field and rooftop PV, respectively, as per the conservative projection for 2050 in the same Fraunhofer report.
The technologies' annual operational costs (OPEX), as percentages of the PV CAPEX, and economic service lives were kept the same as the ones suggested by the corresponding costing methodology. The weighted average cost of capital (WACC) was fixed at 8% for all technologies to maintain consistency with similar assessments [24,32,42,57]. Regional differences in technology costs and access were not considered in the costing procedure. Nevertheless, the technology costs cited here are around 20% higher than those utilized by Sarmiento [58] for Mexico in the same 2050 context and, hence, represent a conservative approach. Table 1 summarizes the VRES techno-economic parameters utilized.

VRES Simulations and Techno-Economic Potential
The computational quasi-physics-based model Renewable Energy Source toolkit (RESKit) [51] was employed to simulate VRES production according to localized climate inputs and geospatial characteristics, as per the methodology of Ryberg et al. [24] for onshore wind and PV potentials. The methodology of Caglayan et al. [35], which also uses the RESKit model, was followed for offshore wind energy. A general description of the simulation procedure is offered here, but the reader is highly encouraged to refer to the original sources for more detailed information on the mathematical and physical foundations. For onshore and offshore wind simulations, 39 years of available MERRA-2 weather data (1980-2019) from NASA [29] was processed to obtain long-term 50 m-height wind speed values at a 50 km 2 spatial resolution. Onshore turbines were then adjusted to local contexts by using the global wind atlas (GWA) [59] to further improve the spatial resolution to 1 km 2 . Then, local wind speeds were projected to the turbines' hub height using the roughness length factors estimated by Silva et al. [60] for the terrain type specified in the Corina Land Cover (CLC) dataset [61]. Offshore wind simulations utilize a constant 0.0002 m roughness lengths factor with bilinear interpolation, according to Caglayan et al. [35]. Subsequently, ideal gas air density was adjusted at the hub height by extracting the surface pressure and air temperature values from the MERRA-2 dataset, as suggested in IEC 61400-12 [62]. Finally, to account for stochastic wind energy losses, local wind speed values at the hub height were convoluted into synthetic power curves that were derived for each turbine as a function of its specific power. In this manner, the corresponding capacity factor at each hour in the weather data was obtained. The total annual generation and FLH were obtained by applying the simulations across the 39 years of data. The mean annual production and LCOE obtained represent historical averages over the evaluated time period.
For PV simulations, RESKit utilizes the already defined PV system location, module characteristics, and tilt and azimuth angles alongside weather data parameters such as air and dew point temperature near the surface, surface pressure, and near-surface wind speed. The model then outputs the capacity factors at each time step as a result of a multi-step process. First, the solar position is computed for each time step at each location using the NREL Solar Position Algorithm [63], for which the terrain elevation is obtained from the Global Multi-resolution Terrain Elevation Data [64] and pressure and temperature values from MERRA-2 are required. Then, extraterrestrial irradiance, relative air mass, and global horizontal irradiance (GHI) values at each time step were estimated as proposed by Spencer [65], Kasten and Young [66], and Perez et al. [67], respectively. Next, due to the low spatial resolution of the MERRA-2 dataset, the adjustment procedure used for the wind simulations was applied using the global solar atlas (GSA) [54] to the obtained GHI values. Subsequently, the plane of array (POA) as a sum of its components' contributions (beam, ground reflected, and sky-diffuse) was computed in accordance with the methodology of Myers [68] and Perez et al. [69,70] and modified according to their respective angles of incidence according to De Soto [71] and Brandemuehl and Beckman [72]. Next, the cell temperature of the module was determined considering POA irradiance, wind speed, and air temperature with the method outlined by King et al. [73]. Afterwards, the DC electricity generation time series by the module was estimated as per the Single Diode model [71]. Finally, an 18% holistic loss was applied to account for inversion, wiring, soiling, and other losses according to values reported in the literature [54,74]. For more information about the PV simulation procedure, the reader is encouraged to refer to the original source [32].
Like any other resource, VRES generation potential is constrained by technological development (technical potential) but also by its cost of production (economic potential) [75]. Based on the relationship between VRES market value (refers to the amount of money that electricity generators can sell their electricity for in the wholesale electricity market without subsidies [76]) and historical electricity prices (refers to the time-weighted average wholesale electricity price [76]) proposed by Hirt [76], an LCOE threshold of 70 EUR·MWh −1 based on past electricity prices in Mexico [38] was applied to distinguish techno-economic VRES potential. The logic behind this assumption was that electricity production with costs higher than the electricity price would not be economically viable. The authors are of the view that comparing the electricity price to the LCOE is adequate as an initial approach to determining economic VRES potential.

Results
The results are grouped by their renewable source and are presented in the following subchapters. In the first place, the available area resulting from the LEA is expressed as Energies 2021, 14, 5779 9 of 24 a percentage of available land. Afterwards, the VRES simulations are presented using potential maps, graphs, and tables. Figure 2 displays the area eligible for the placement of wind turbines. Around 754,000 km 2 of land (~39% of the total) is eligible for onshore installations. Most of it is in the country's northern states, where there are vast areas of uninhabited land and, to a lesser extent, in the southeastern states. The center of the country appears to have the least area eligible due to settlement-related constraints such as human habitation areas and roads. For offshore installations, a total area of around 183,000 km 2 (~6% of the territorial sea) was found to be eligible.  Open-field PV parks account for a total area of around 578,000 km 2 (~30% of the territorial extension), which is around one-quarter smaller than the eligible area for onshore wind. There are three main reasons for this. The first is that, given the mountainous geography of Mexico, tougher slope-related constraints (with a maximum inclination of 17° for wind compared to 3° for PV) have a greater impact on the available land. Second, the exclusion zone for hurricane areas is larger compared to that for wind turbines, which considerably reduces the eligible area in the Yucatan peninsula. Third, agricultural areas, which account for around 17% of the total territorial land, are excluded for PV installations but not for wind. The total eligible rooftop area is around 60,000 km 2 and has a distribution focused in major urban areas across the country, especially in the central region.

VRES Techno-Economic Generation Potential
The overall annual techno-economic VRES generation potential of Mexico was found to be 54 PWh. Of this, 49.2 PWh derives from open-field PV (91%) and 4.8 PWh from onshore wind (9%). Offshore wind (38 TWh) and rooftop PV (1 TWh) generation represent less than 0.03% of the combined total. Despite having the largest available land share, the wind resources in Mexico are regionally concentrated and less favorable by comparison to the solar ones. Larger eligible onshore wind areas do not translate into higher generation potential. The following subchapters present the wind and PV potential distribution across the country, subdivided into the zones presented in Figure 1. By comparison, the eligible area for offshore turbines is four times less than that for onshore ones. Most of this is allocated along the Gulf of Mexico and Yucatan peninsula, where shallow waters qualify as eligible despite being affected by hurricane-prone and conservation constraints. The Pacific coast features little eligible sea area due to the local arrangement of the tectonic plates that results in abrupt ocean depth drop-offs. The western side of the Baja California peninsula and the Isthmus coast partially escape this effect, as they do not lie along the inter-tectonic line. Figure 3 displays two maps corresponding to the land eligible for the placement of PV technologies: (A) for open-field PV; and (B) for rooftop PV. Open-field PV parks account for a total area of around 578,000 km 2 (~30% of the territorial extension), which is around one-quarter smaller than the eligible area for onshore wind. There are three main reasons for this. The first is that, given the mountainous geography of Mexico, tougher slope-related constraints (with a maximum inclination of 17 • for wind compared to 3 • for PV) have a greater impact on the available land. Second, the exclusion zone for hurricane areas is larger compared to that for wind turbines, which considerably reduces the eligible area in the Yucatan peninsula. Third, agricultural areas, which account for around 17% of the total territorial land, are excluded for PV installations but not for wind. The total eligible rooftop area is around 60,000 km 2 and has a distribution focused in major urban areas across the country, especially in the central region.   Figure 4 shows the 39-year average LCOE from 20 (the lowest LCOE yielded) to 70 EUR‧MWh -1 of onshore and offshore wind turbines. A white background is assigned to ineligible land. Four notable zones with good wind resources, all of which are mountainous, can be identified, namely: (1) The Isthmus (dotted in dark blue), where two different mountain systems create a gap that funnels and intensifies onshore wind resources towards the Pacific Ocean. This zone also covers some offshore wind potential. (2) The Western Sierra (dotted in purple), a chain of mountains that runs alongside the pacific coast through several northwestern states; (3) the Eastern Sierra (dotted in red), which is the eastern parallel with similar wind resources. Additionally, (4) a smaller area of offshore potential within the shallow waters of the Yucatan peninsula's eastern coastal line (dotted in light blue). Due to the geospatial dependence on mountains, slope-related constraints directly affect the wind potential adjacent to these regions. Some other good wind resources coincide with the coastal line of the Gulf of Mexico, and scattered locations in both peninsulas can also be highlighted. As some of these areas are protected for environmental reasons, turbine placement is bypassed in them, and the remaining installations do not exhibit apparent clusters. Moreover, the turbine and grid infrastructure deployment needed to explore wind potential in those regions conflicts with environmentally protected areas. Figure 4 also confirms that offshore wind zones not only offer less eligible areas but also less techno-economic potential than onshore sites. Relatively higher wind speeds in far-to-shore waters do not translate into cheaper LCOEs due to higher installation caused by increased sea floor depths and distance to the coast. Consequently, the most cost-attractive offshore zone is found near the coast of the Isthmus, where it partially benefits from the isthmus' wind funnel.

VRES Techno-Economic Generation Potential
The overall annual techno-economic VRES generation potential of Mexico was found to be 54 PWh. Of this, 49.2 PWh derives from open-field PV (91%) and 4.8 PWh from onshore wind (9%). Offshore wind (38 TWh) and rooftop PV (1 TWh) generation represent less than 0.03% of the combined total. Despite having the largest available land share, the wind resources in Mexico are regionally concentrated and less favorable by comparison to the solar ones. Larger eligible onshore wind areas do not translate into higher generation potential. The following subchapters present the wind and PV potential distribution across the country, subdivided into the zones presented in Figure 1. Figure 4 shows the 39-year average LCOE from 20 (the lowest LCOE yielded) to 70 EUR·MWh −1 of onshore and offshore wind turbines. A white background is assigned to ineligible land. Four notable zones with good wind resources, all of which are mountainous, can be identified, namely: (1) The Isthmus (dotted in dark blue), where two different mountain systems create a gap that funnels and intensifies onshore wind resources towards the Pacific Ocean. This zone also covers some offshore wind potential. (2) The Western Sierra (dotted in purple), a chain of mountains that runs alongside the pacific coast through several northwestern states; (3) the Eastern Sierra (dotted in red), which is the eastern parallel with similar wind resources. Additionally, (4) a smaller area of offshore potential within the shallow waters of the Yucatan peninsula's eastern coastal line (dotted in light blue). Due to the geospatial dependence on mountains, slope-related constraints directly affect the wind potential adjacent to these regions. Some other good wind resources coincide with the coastal line of the Gulf of Mexico, and scattered locations in both peninsulas can also be highlighted. As some of these areas are protected for environmental reasons, turbine placement is bypassed in them, and the remaining installations do not exhibit apparent clusters. Moreover, the turbine and grid infrastructure deployment needed to explore wind potential in those regions conflicts with environmentally protected areas. Figure 4 also confirms that offshore wind zones not only offer less eligible areas but also less techno-economic potential than onshore sites. Relatively higher wind speeds in farto-shore waters do not translate into cheaper LCOEs due to higher installation caused by increased sea floor depths and distance to the coast. Consequently, the most cost-attractive offshore zone is found near the coast of the Isthmus, where it partially benefits from the isthmus' wind funnel.  Figure 5 shows the wind capacity potential versus an LCOE of up to 70 EUR‧MWh -1 grouped by zones. With the curtailment applied at 70 EUR‧MWh -1 , 15% of the VRES generation's installable capacity, pertaining almost exclusively to wind energy, was filtered out, but only 8% of the total VRES power output was reduced, which confirms that the filtered-out wind turbines contribute only a small fraction of the generation potential.  Figure 5 shows the wind capacity potential versus an LCOE of up to 70 EUR·MWh −1 grouped by zones. With the curtailment applied at 70 EUR·MWh −1 , 15% of the VRES generation's installable capacity, pertaining almost exclusively to wind energy, was filtered out, but only 8% of the total VRES power output was reduced, which confirms that the filtered-out wind turbines contribute only a small fraction of the generation potential.

Wind Energy Potential and Distribution
Despite being relatively small relative to the size of the country, the Isthmus contains around 53% (6.5 GW) of all onshore capacity with an estimated future LCOE equal to or less than 30 EUR·MWh −1 . This wind potential distribution confirms and explains why this zone alone contained around 90% (1.8 GW) of the national wind capacity as in 2019 [77]. The sunrise-facing Western Sierra (light purple) and Eastern Sierra (light red) uphills create high-speed wind corridors that can each accommodate up to 214 GW of capacity (or 37% of the total) with an LCOE of 50 EUR·MWh −1 or lower. By comparison, the smaller Isthmus zone comprises approximately 12% (63 GW) of the total capacity within the same LCOE range, whereas scattered onshore resources in the rest of the country (gray) make up the remaining 14% (83 GW). Similarly, offshore wind energy potential (in back) below 70 EUR·MWh −1 of 92 GW (97% of total) exists almost exclusively in the seas around the Isthmus and Yucatan peninsula.  Figure 5 shows the wind capacity potential versus an LCOE of up to 70 EUR‧MWh -1 grouped by zones. With the curtailment applied at 70 EUR‧MWh -1 , 15% of the VRES generation's installable capacity, pertaining almost exclusively to wind energy, was filtered out, but only 8% of the total VRES power output was reduced, which confirms that the filtered-out wind turbines contribute only a small fraction of the generation potential.   Table 2 presents the values for capacity, hub height, FLH, and mean generation, corresponding to an average turbine design per geographical zone for various LCOE batches that could serve as a design reference for comparison or as modeling parameters when an optimal wind turbine design algorithm is unavailable. For onshore installations, investments and mean generation electricity production are proportional to the turbines' baseline costs and FLH, respectively. In general, turbines in the Isthmus area have slightly larger capacities and lower hub heights due to the region's higher wind speeds compared to those in the Western and Eastern Sierras, which have similar designs because of the similar geographical features. As the LCOE threshold increases to 50 or 70 EUR·MWh −1 , average designs in the three zones tend to look similar until merging into a national average design of 4.1 MW capacity, 125 m hub height, and 126 m rotor diameter. On the other hand, the FLHs exhibit a large deviation based on the geographical zone. The FLH in the best locations, such as the Isthmus, is almost 45% bigger than the national average. Consequently, the national onshore LCOEs vary considerably, as noted in the previous figure.  Table 3 displays the average offshore turbine characteristics with an LCOE restricted to 70 EUR·MWh −1 or lower by zone. As the offshore turbines were modeled with a fixed design (see Section 2.2), investment costs varied slightly from 19 to 20.8 million EUR due to the different contributions of the distance to shore, ocean depth, and foundation type on the costing function. Nevertheless, the resulting investment costs were around four times higher than those for onshore turbines, whereas the resulting mean generation was not proportionally higher. In fact, comparing the best locations, two onshore turbines could produce around 86% of the electricity output of one offshore turbine at around half (52%) the installation cost. Therefore, the LCOE of offshore wind is considerably higher than onshore.  Figure 6 shows the corresponding LCOE map for PV technologies. Overall, Mexico has large PV energy potential throughout the country. Nevertheless, two effects slightly affect this: the effect of latitude and tropical weather. PV LCOEs tend to be slightly more expensive as latitude increases due to fewer hours of sunlight throughout the year. Northern states in the proximity of the border to the U.S. (marked with a dotted yellow line) have LCOEs of approximately 30 EUR·MWh −1 , whereas, towards the center, the LCOEs are mostly around 25 EUR·MWh −1 . Despite this latitude effect, the eligible area for PV parks in the northern states is much larger, which leads to increased solar potential. The tropical climate also affects PV panels.

PV Potential and Distribution
In contrast to the latitude effect, the climate in the Yucatan peninsula and most of the states along the coasts (marked by dotted blue lines) are subject to more rain and clouds that cause less FLH and therefore increase the LCOE. As there is no fundamental modeling difference between PV technologies, rooftop PV exhibits higher LCOEs than open-field due to the CAPEX difference and suboptimal orientations. Its total potential is negligible compared to open-field PV. Figure 7 presents the PV capacity potential by LCOE and is grouped into zones. Because of the PV potential's direct dependency on the amount of eligible land, zones with larger areas have greater PV potential. Accordingly, the northern zones, such as the Western Sierra, Eastern Sierra, and Northwest, have the greatest potential. 90% (20.8 TW) of the national capacity is within 24 and 30 EUR·MWh −1 , and 41% (9.6 TW) is found at an LCOE equal to or below 25 EUR·MWh −1 . All of the zones exhibit installable capacity in all of the LCOE ranges due to suboptimal terrain characteristics such as elevation, tilt, or shadows.
A different way to look at the PV potential is by considering PV parks' FLH as a proportion of the total potential by zone, as presented in Figure 8. This makes it apparent that 2300 FLH is the predominant condition nationally, which explains the good PV generation potential across the country. On the other hand, it also indicates that within zones, there is a large difference between the highest and lowest FLH yields. Differences of around 800 to 500 FLH or +/− 12 EUR·MWh −1 between the best and worst locations can be seen in the California peninsula or the Gulf of Mexico, respectively. The differences are a consequence of particular local conditions, such as irradiance or temperature. The latitude effect discussed previously in the PV potential is even more clearly shown. The northern zones, by contrast, such as the northwestern 8% of the placements, yield 2300 FLH, whereas, in more central zones, such as the Gulf, Isthmus, or Central, the share of the same FLH batch is over 40%. Similarly, the tropical weather effect also slightly reduces the share of this FLH batch in the Yucatan peninsula (37%) compared to other zones at similar latitudes, such as the Central or Gulf (40% and 50%, respectively) regions. Unlike the wind turbines simulation, PV technologies were simulated using a single PV model, and therefore a table with an average PV park design is not provided.  A different way to look at the PV potential is by considering PV parks' FLH as a proportion of the total potential by zone, as presented in Figure 8. This makes it apparent that 2300 FLH is the predominant condition nationally, which explains the good PV generation potential across the country. On the other hand, it also indicates that within zones, there is a large difference between the highest and lowest FLH yields. Differences of  A different way to look at the PV potential is by considering PV parks' FLH as a proportion of the total potential by zone, as presented in Figure 8. This makes it apparent that 2300 FLH is the predominant condition nationally, which explains the good PV generation potential across the country. On the other hand, it also indicates that within zones, there is a large difference between the highest and lowest FLH yields. Differences of zones, by contrast, such as the northwestern 8% of the placements, yield 2300 FLH, whereas, in more central zones, such as the Gulf, Isthmus, or Central, the share of the same FLH batch is over 40%. Similarly, the tropical weather effect also slightly reduces the share of this FLH batch in the Yucatan peninsula (37%) compared to other zones at similar latitudes, such as the Central or Gulf (40% and 50%, respectively) regions. Unlike the wind turbines simulation, PV technologies were simulated using a single PV model, and therefore a table with an average PV park design is not provided.

Discussion
This section focuses on differences with the most relevant previous VRES assessment and the implications of the chosen methodology. Additionally, a brief discussion about the validation of the computational models by their authors is offered. Finally, possible ways by which the assessment can be further improved in the future are presented.
In comparison to other VRES assessments in Mexico, only five of the regional studies mentioned in the literature review can be compared to some extent to this study due to their mixed, and sometimes unstated, methodologies. Jaramillo et al. [16] concluded that 750 kW wind turbines in Baja California Sur would operate on average for 2200 FLH, producing LCOEs from 40-60 EUR‧MWh -1 (1.1 USD to EUR conversion rate used [6]). In comparison, this assessment demonstrates an average of almost 3000 FLH (25% more), with LCOEs ranging from 46-50 EUR‧MWh -1 . The results reported herein largely coincide with Jaramillo et al.'s, despite the latter having arrived 15 years ago, but with vast differences in turbine design. Jaramillo et al. drew on comparable costs, such as a CAPEX of 1000 EUR‧kW -1 and OPEX equivalent to 1% of the CAPEX, with an additional five years of economic life. The CAPEX cost estimation function utilized here also includes costs associated with grid connection and balance of system costs that were not stated as considered in Jaramillo et al. and whose impact on the LCOE could offset the performance gains of future turbine designs. In another example, the assessment for the state of Veracruz by Hernández Escobedo [18] deviates considerably from the results obtained in this assessment. The techno-economic average energy yield per turbine in this study in the same state increased by around 19% in comparison to the results of Hernández Escobedo. However, if merely a simple state average, as reported by Hernández Escobedo, is considered, a 60% yield reduction per turbine is seen here, despite the modeling of futuristic turbine designs. Further investigation and the reproducibility of Hernández Escobedo's work is not possible as non-open source software was employed. Furthermore, Hernández Escobedo [22] also assessed another Mexican region, namely the Baja California peninsula, where a comparison was also not possible, as it is an ineligible area due to environmental

Discussion
This section focuses on differences with the most relevant previous VRES assessment and the implications of the chosen methodology. Additionally, a brief discussion about the validation of the computational models by their authors is offered. Finally, possible ways by which the assessment can be further improved in the future are presented.
In comparison to other VRES assessments in Mexico, only five of the regional studies mentioned in the literature review can be compared to some extent to this study due to their mixed, and sometimes unstated, methodologies. Jaramillo et al. [16] concluded that 750 kW wind turbines in Baja California Sur would operate on average for 2200 FLH, producing LCOEs from 40-60 EUR·MWh −1 (1.1 USD to EUR conversion rate used [6]). In comparison, this assessment demonstrates an average of almost 3000 FLH (25% more), with LCOEs ranging from 46-50 EUR·MWh −1 . The results reported herein largely coincide with Jaramillo et al.'s, despite the latter having arrived 15 years ago, but with vast differences in turbine design. Jaramillo et al. drew on comparable costs, such as a CAPEX of 1000 EUR·kW −1 and OPEX equivalent to 1% of the CAPEX, with an additional five years of economic life. The CAPEX cost estimation function utilized here also includes costs associated with grid connection and balance of system costs that were not stated as considered in Jaramillo et al. and whose impact on the LCOE could offset the performance gains of future turbine designs. In another example, the assessment for the state of Veracruz by Hernández Escobedo [18] deviates considerably from the results obtained in this assessment. The techno-economic average energy yield per turbine in this study in the same state increased by around 19% in comparison to the results of Hernández Escobedo. However, if merely a simple state average, as reported by Hernández Escobedo, is considered, a 60% yield reduction per turbine is seen here, despite the modeling of futuristic turbine designs. Further investigation and the reproducibility of Hernández Escobedo's work is not possible as non-open source software was employed. Furthermore, Hernández Escobedo [22] also assessed another Mexican region, namely the Baja California peninsula, where a comparison was also not possible, as it is an ineligible area due to environmental protection constraints according to the LEA used in this study. Rodriguez Hernandez's [23] assessment of the area of Mexico City concluded that there is virtually no techno-economic wind potential, which is consistent with the findings of this study. The potentials in the AZEL assessment [25] are most comparable to those in this study. Table 4 summarizes the comparison between them. In general, the AZEL estimates the total installable capacity for both open-field PV and onshore wind. This is a direct consequence of the more extensive LEA employed here that significantly reduces the maximum installable VRES capacities compared to AZEL, which only considers 11 constraints for onshore and 13 for open-field PV. Therefore, the maximum installable capacities in the AZEL assessment appear to be larger. Nevertheless, when looking at FLH or generation per installment, in this study, the performance always increased. The reason for this is that AZEL employed a power density factor to calculate national potential, whereas, in this study, the power generation was simulated for each VRES placement. Additionally, the improved VRES performance and optimal wind turbine designs were not considered in AZEL, which explains the poorer individual performance and reduces the potential gap between both assessments. Consequently, the generation and capacity potentials differ. Another topic is the validation of the models used. In general, the validation of the models was carried out using international datasets, comparisons with other, and, when available, measured data. Here, only the final deviation percentage and conclusions of the validation are included. For more information, please refer to the original sources.
The authors of the GLAES model declare that eight studies in Europe containing LEA were replicated, and the average land differences were compared. The average land deviation found was 3.51%. As mentioned in Section 2.1, most of the input datasets were previously employed and were kept unchanged. Exceptions are official datasets published by Mexican authorities that are presumable more accurate than international sources. For these reasons, a similar land deviation can be expected and is considered acceptable for a national-wise assessment such as for Mexico by the authors of this assessment. For the simulation procedure, the work of Ryberg [32] offers time series visual comparisons against measured data, regional potential, and turbine design comparisons for the validation of onshore wind potential. The author concludes that "the model does an acceptable job at recreating the measured generation". The same model was used also by assessing the offshore wind potential in Europe by Caglayan et al. [35], where a similar conclusion was reached. In regards to the PV simulation scheme, Ryberg [32] compared the simulation results to those found by Pfenninger et al. [74] and the Global Solar Atlas [54]. The results of the comparison show an approximate 5% temporal deviation against Pfenninger and 6% spatial deviation with Global Solar Atlas. Given the complexity of the validation procedure of both wind and PV, the reader is again encouraged to go to the original source for more information.
The VRES assessment results can vary significantly due to a number of assumptions made by their authors in the LEA, VRES simulation procedure, or techno-economical parameters incorporated. Nevertheless, new constraints could gain relevance or current ones lose it as VRES technologies are progressively deployed and increasingly interact with other activities. Consequently, the area available for VRES installations and ultimately the VRES generation potential will change as environmental, physical, sociopolitical, and economic conditions do.
A way to improve the assessment is by accessing higher-quality data. Even though the most up-to-date databases were used herein, data measurements can always be improved in the future and return results that are more precise. Other unevaluated issues arising in this assessment are disruptive technologies that can totally change the power system by revealing new energy potentials. Unfortunately, no model can predict the occurrence of these, much less their consequences. Only foreseeable technological improvements on current designs through 2050 have been captured by the design of synthetic power curves for wind turbines and PV panels with improved performance, as estimated by Ryberg [32], are considered. Increasing the simulations' resolution (temporally and spatially) is beneficial for increasing the accuracy of the results, although its exact extent is not yet known. Hourly evaluation and 100 m spatial resolution is currently a widely accepted resolution for a country assessment and can be performed in a reasonable amount of time due to its lower computational complexity. Local conditions, such as noise tolerance, shadows, or roughness-lengths, are also left unevaluated as a trade-off.

Conclusions
In light of the results of this study, several conclusions can be drawn. Firstly, openfield PV is by far the most abundant and cheaper VRES in the country. The generation potential of PV energy is up to 10 times larger than the wind energy potential and is present practically all across the country. Most of the LCOEs from PV energy are below 30 EUR·MWh −1 , which is less than half of most of the wind potential. Moreover, based on the generally good PV potential across the country, its deployment is likely to be highly decentralized. The development of the power system should anticipate adequate legal, economic, and technological frameworks in order to make the best use of these. A revision of the applicability of the LEA constraints and buffer zones is also recommended to capture changes in social preferences over time towards wind and PV technologies.
Secondly, the onshore wind potential is regional. Around 81% of the best wind energy potential is concentrated in three mountainous zones. This potential is enough to satisfy around 4 to 5 times the current energy demand. Given this concentrated distribution, there must be a strong emphasis on the development of future energy infrastructure, such as transmission lines, energy storage, and even renewable fuel production, in these wind energy-rich regions in order to incorporate more wind energy into the national energy system. Off-grid connections are possible too; however, the wind potential is likely to be underutilized given the low demand that is usually found in these remote areas. In contrast, bringing interconnected renewable energy projects in there would help incorporate more renewable energy into the national electricity mix. Such projects can also help overcome the other two energy challenges, such as better electricity access in remote areas and meeting Mexico's environmental commitments.
Thirdly, in the center of the country, where the political boundaries create smaller and more densely-populated areas, there is the least techno-economic VRES potential. This limited renewable potential comes exclusively from open-field PV. Mexico City, for example, exhibits no techno-economic wind potential but around 260 MW of PV at 28 EUR·MWh −1 , which is limited by land availability. At the same time, the central regions have high power consumption rates, which could increase the proclivity to deploy the existing PV potential as soon as possible. For this reason, the implementation of a higher-resolution regional model is recommended, as well as a more region-specific analysis favoring optimal PV potential utilization. Finally, apart from the possible methodological improvements presented in the discussion section, incorporating inputs from local stakeholders in future assessments will be highly beneficial. Key players can help to better understand additional local challenges and opportunities in terms of land use, technology costs and access, legal and environmental considerations, etc., that are hard to represent in a nationwide analysis. Future studies can incorporate local data inputs when appropriate and to validate the assessments' results to make more robust conclusions, thus, reducing the resistance to implementing new sustainable strategies.

Conflicts of Interest:
The authors have no conflict of interest to declare.

Appendix A. Earthquakes and Tsunamis as LEA Constraints
Several studies have investigated how wind turbines would behave in response to earthquakes. In 2003, Lavassas et al. [78] carried out an analysis and developed a design for a 1 MW, steel, 44 m-high wind turbine tower that corresponded to the Greek Anti-Seismic Code EAK 2000. It was concluded that seismic activity would only play a significant role in seismically hazardous areas with a combination of medium and soft placement soil. Similarly, Ritschel et al. [79] performed two types of computational simulations: one, typically used for buildings and called "modal," as described by Clough et al. [80]; and one code specifically for wind turbines called "Flex5" [81]. The simulation was carried out to generate accelerograms in accordance with the European standard Eurocode 8 [82] for a Nordex N80 (onshore, 80 m rotor diameter, 60 m hub height, 2.5 MW output) that was installed in a seismic region in Ryuy-Cho, Japan, a site with a peak ground acceleration of 0.3 g. According to Ritschel et al., although both models returned slightly different results, with the Flex5 one featuring more bending, simulated seismic shakes were negated by the designed turbine models. Moreover, according to a 2016 literature review on wind turbines and seismic hazards by Kastanos et al. [83], there exists a consensus in the engineering community that the loads caused by earthquakes of importance for wind turbines are horizontal and wind-driven and that there is a "self-insolated effect" during the most destructive portion of an earthquake. Kastanos et al. supported their conclusions by drawing on observations of earthquakes in 1986 near an onshore wind farm in North Palm Springs, California, and in 2011 at the Kamisu offshore wind farm phase 1 in Tohoku, Japan.
Contrary to reports on hurricanes damaging wind turbines, the evidence of the effects of earthquakes can be exemplified clearly by the cases in Japan and California with offshore and onshore wind farms, respectively. In both instances, no turbine damage was reported. During the 2011 Great East Japan earthquake and tsunami (measuring 9.1 on the Richter Moment Magnitude Scale [84]), the Kamisu offshore wind farm contained seven Hitach HTW 2.0-80 wind turbines of 2 MW and with 60 m hub heights. Although the power was interrupted by grid checking after the tsunami, the wind farm continued providing power for three days following the tsunami, with no damage reported [85]. Matsunobu et al. [85] investigated the case and found that an "extreme wave + wind" is much more severe in terms of structural stress than a "Tsunami + power production" scenario, based on simulations of the Kamisu offshore phase 1 and the actual 2011 tsunami, which confirmed that the mono-pile wind turbine structures can withstand seismic vibrations. These results and observations were implemented in the design and construction of phase 2 of the same project. In this context, Bhattacharya et al. [86] noted that the use of offshore wind farms can even increase the seismic resilience of nuclear power plants, which are often close to the sea. They could also supply the power for cooling systems and would not fail for the same reasons as other emergency systems and, therefore, could help avoid a cascade of catastrophic consequences.
In the case of onshore wind, a study by Kübler et al. [87] from Swiss Re reports that in the Painted Hills onshore wind farm in California, none of the 65 wind turbines was destroyed by the earthquake in 2011. However, 48 of them had to undergo minor repairs. The wind farm is in the Riverside country, which is regarded as having a 75% likelihood of an M7.0 • earthquake occurring within the next 30 years. For more local cases, Oaxaca, Mexico is a site at which wind farms have operated since 1994 [88]. On the 7th and 19th of September 2017, two notable earthquakes occurred at this site; the first measured M8.2 • and the second M7.1 • [89]. In both instances, none of the 1200 wind turbines were damaged [89]. Again, the power supply was also cut due to afflictions on the power lines, urban areas, and security grid issues [89].
Nevertheless, given the limited experience of only the last few decades and that not enough on-site experiments have been conducted, it cannot be concluded that wind turbines are earthquake-proof. This hazard analysis should be taken into consideration with the increased risk of a common failure of wind turbines [83]. In the case of solar technologies, during the 2011 earthquake and tsunami in Japan, existing solar panel operations were shut down due to grid problems rather than the destruction of the panels [90]. Table A1. Land constraints and buffer zones for the placement of VRES technologies in Mexico.

Group
No.