Committed emissions and the risk of stranded assets from power plants in Latin America and the Caribbean

Latin America and the Caribbean (LAC) has the least carbon-intensive electricity sector of any region in the world, as hydropower remains the largest source of electricity. But are current plans consistent with the international climate change goals laid out in the Paris Agreement? In this paper, we assess committed CO2 emissions from existing and planned power plants in LAC. Those are the carbon emissions that would result from the operation of fossil-fueled power plants during their typical lifetime. Committed emissions from existing power plants are close to 6.9 Gt of CO2. Building and operating all power plants that are announced, authorized, being procured, or under construction would result in 6.7 Gt of CO2 of additional commitments (for a total of 13.6 Gt of CO2). Committed emissions are above average IPCC assessments of cumulative emissions from power generation in LAC consistent with international temperature targets. To meet average carbon budgets from IPCC, 10%–16% of existing fossil-fueled power plants would need to be closed before the end of their technical lifespan. Our results suggest that building more fossil-fueled power plants in the region could jeopardize the achievement of the Paris Agreement temperature targets.


Introduction
Latin America and the Caribbean (LAC) has the least carbon-intensive electricity sector of all regions in the world, thanks to the highest share of hydroelectricity in the world (IEA 2018a). But this is changing. Hydropower generation has scaled down its percentage in the power mix from 58% in 2009 to 50% in 2016 (IEA 2018a). Utilization rates have been reduced by droughts, and capacity additions have slowed down due to social and environmental concerns, and increasing capital cost (Pereira de Lucena et al 2011, Soito and Freitas 2011, IRENA 2016, Van Vliet et al 2016, de Queiroz et al 2019.
Natural-gas-based power generation has generally filled the gap, sustained by abundant and competitive supply, turning it into the second source in the power mix (IRENA 2016, Yépez-García et al 2018. While installation of wind, solar, geothermal, and bioenergy and waste power plants is growing rapidly, representing 57% of renewable capacity addition in 2017, it still represents only 6.5% of total capacity (ENERDATA 2019a). In the absence of changes in public policies and/or market design, natural gas and coal could play an increasingly important role in the electricity mix (Calderón et al 2016, Clarke et al 2016, Lucena et al 2016, van der Zwaan et al 2016.
All countries in the region have presented nationally determined contributions (NDC) that include emission reductions in the power sector as part of their contribution to the Paris Agreement (World Bank 2019). But current energy planning is only partially consistent with commitments, and would result in the addition of new fossil fuel power plants in the region (OLADE 2018, Cadena 2019. Worse, current NDCs are not aligned with the temperature targets of the Paris Agreement (Iyer et al 2015, UNEP 2017, Binsted et al 2019, so that even if countries did implement their NDCs, they would continue to add more fossil fuel power plants than what would be consistent with the achievement of the international temperature targets. To keep climate change impacts on development in check (Hallegatte et al 2015), global leaders have agreed to pursue efforts to limit global warming well below 2°C, and as close to 1.5°C as possible (United Nations 2015). Either target requires reaching net zero emissions of CO 2 globally (Fay et al 2015, Rogelj et al 2015, Sachs et al 2016 and in LAC (Vergara et al 2016, Paredes 2017. In particular, stabilizing climate change requires that all regions switch to carbon-free electricity before 2050 (Williams et al 2012, Audoly et al 2018, Davis et al 2018. The IPCC's special report on global warming of 1.5°C finds that by 2050, the net carbon content of the power sector should fall to close to 0 and renewable supply should represent 70% of the electricity mix (Huppmann et al 2018a).
Long-term decarbonization goals matter for energy infrastructure planning because power plants lifetime may range from 30 to 50 years (Fay et al 2015, Millar et al 2016, Sachs et al 2016, Grubb et al 2018.
To assess the impact of long-lived infrastructure on climate change, Davis and Socolow (2014) introduced the concept of committed carbon emissions in existing infrastructure. Those are the carbon emissions that would result from the operation of existing fossilfueled power plants and other carbon-intensive equipment during their typical lifetime and with typical utilization rates. The same concept has been applied to planned power plants, that is plants that are announced, authorized, being procured, or under construction (Shearer et al 2017, Edenhofer et al 2018, Pfeiffer et al 2018. Here, we assess committed emissions from operational and planned power plants in LAC. We use the power plan tracker (PPT) database from ENERDATA (ENERDATA 2019a), which provides information on power plants classified by fuel type, age, capacity, historical output, and operational status as of January 2019.
We compare committed emissions with a wide range of carbon budgets for the LAC power sector. We define those as the sum of gross CO 2 emissions from the LAC power sector in the scenarios of the world economy that keep global warming in the 1.5°C-2°C range reported in the IPCC Special Report on 1.5°C (2018) 5 . The IPCC considers pathways generated using a variety of modeling paradigms; different technology assumptions-in particular, exploring the impact of whether carbon dioxide removal can offset emissions from power generation and different costs and potentials for non-CO 2 emission reductions-, discount rates, interpretations of temperature targets-peak or long-term warming (Huppmann et al 2018a(Huppmann et al , 2018b. We consider all those pathways.
We find that committed emissions from the existing power sector in LAC amount to 6.9 GtCO 2 . This commitment is within the range of LAC power carbon budgets consistent with 1.5°C, which we find to be 1.9 to 13.5 GtCO 2 . However, committed emissions are greater than the median of 1.5°C-compliant carbon budgets reported in the IPCC database, and greater than the 60th percentile of 2°C carbon budgets in the same database. If all planned power plants are built, we find that committed emissions would rise to 13.6 GtCO 2 , which is more than 90% of LAC power carbon budgets reported in any scenario consistent with 1.5°C or 2°C published by the IPCC.
These findings suggest that to meet the average allowable carbon budget for 2°C (6.2 GtCO 2 ) or 1.5°C (5.8 GtCO 2 ), utilities in the region would need to close prematurely 10%-16% of the existing fossilfueled capacity, respectively, or reduce the utilization rate of existing plants to the same effect. Doing so could be politically difficult, as policies that result in the closure or reduced utilization of power plant would diminish the financial value of those assets, i.e. they would create stranded assets 7 . Closing down power plants would also result in sudden losses of jobs for the workers and communities who depend on those assets. Both impacts are politically difficult to manage because they create concentrated losses on homogenous groups that can easily organize to protest the reforms (Olson 1977, Trebilcock 2014, and because they can go against policy objectives of social inclusion (Hallegatte et al 2013, Jenkins 2014, Bertram et al 2015, Nemet et al 2017, Vogt-Schilb and Hallegatte 2017, Gambhir et al 2018, ILO 2018, Rozenberg et al 2018. This paper is part of a growing literature that quantifies committed emissions in energy infrastructure (Davis et al 2010, Pfeiffer et al 2018, 2016, Tong et al 2019. This literature has focused on global 5 In IPCC scenarios where carbon capture and storage is used to produce negative emissions in the power sector, we add net emissions from power generation and captured emissions to compute our gross carbon budgets (see Methods and data). Carbon budgets are computed over the lifetime of all the existing and planned power plants, that is 2019-2054. 6 See footnote 5. 7 The words stranded assets are used in the literature on climate change to describe various things (Caldecott 2017): assets that are lost because of the impact of climate change itself, fossil fuel resources that cannot be burnt into the atmosphere if a given climate target is to be reached, also called unburnable carbon (Solano-Rodriguez et al 2019); and man-made capital that has to be retired early because of climate policies, such as coal power plants that become unprofitable after a carbon price is implemented. In this paper we focus on power plants. With this definition, closing a power plant or reducing its utilization rate for environmental reasons creates stranded assets for its owners irrespective of whether the power plant had already reached its financial payback period. A power plant is stranded if and when it could technically be used longer and produce more revenues, but is closed due to environmental (or other) policy. emissions, or on showing that coal power plants under construction globally (Edenhofer et al 2018), or even just in India (Shearer et al 2017), would make a significative contribution to global emissions. In this paper, we focus in LAC, a region that was home to only 5% of global CO 2 emissions in 2016 (IEA 2018b). Unlike global commitments from coal, the committed emissions we find in LAC are not a game changer for the global climate change agenda. But our results show that international temperatures targets do matter to LAC energy planners: existing plans would surpass most of LAC's power carbon budgets, and adding fossil fuel power plants may increase the risk of stranded assets in LAC.
Section 2 presents the methods and data. Section 3 provides results. Section 4 discusses those results and concludes.

Methods and data
We define committed emissions as the emissions that will occur over the remaining lifespan of a fossil-fuelburning electric generator 8 . We focus on generators, defined as devices that generate electrical power for use in an external circuit. A plant consists in one or more generators.

Carbon emissions per generator from ENERDATA and IEA
We compute committed emissions in two basic steps. In the first step, we assess current emissions by generator. We decompose CO 2 emissions F (tCO 2 yr −1 ), as the product of capacity C (GW), utilization rate E C where E is electricity output (GWh yr −1 ), and carbon intensity of electricity generated F E (tCO 2 GWh −1 ). We assume utilization rates and carbon intensities to be constant over time. To make the most of the data available, each quantity is computed per country i,fuel f , and status s: We take existing and planned capacities C i f s , , from the Power Plant Tracker (see appendix 4, appencies are available in the online supplementary materialis available online at stacks.iop.org/ERL/14/124096/ mmedia ) (ENERDATA 2019b). The PPT reports unit status, date of commissioning, fuel type, net capacity, electricity output and localization in January 2019. The database reports 14 816 generators in LAC, 34% of which (5048) are fossil-fuel-based (oil; coal, peat and oil shale; and natural gas). We focus on fossil fuel plants, as the others do not commit CO 2 emissions.
The PPT classifies generators in operational, announced, authorized, bidding process and under construction, stopped, canceled, mothballed, and synchronized statuses. We qualify as planned the generators under the announced 9 , authorized 10 , bidding process and under construction statuses. Those do not currently emit carbon dioxide but will do so starting at their commission date. Operational and synchronized units are included in the existing status. Those are already emitting carbon dioxide. Table 1 summarizes the amount of generator per category.
We take electricity output E i,f per country and energy type from ENERDATA (2019a) and ENER-DATA (2019b). These two sources are slightly inconsistent. The total (bottom-up) sum of power generation listed in Power Plant Tracker (ENERDATA 2019b) does not match national statistics of power generation per country and fuel (ENERDATA 2019a). In total, fossil-fuel-based generation reported in PPT for 2016 (450 TWh) represents 67% of total electricity production from national statistics (665 TWh) 11 . We solve this issue at the country and fuel level. In most cases, the sum from PTT is lower than the reported national statistic. One reason is that PPT does not report any electricity output for some generators. Another is that for some flex-fuel plants, PPT reports only generation from the main fuel. We fill missing generation data using averages per country and fuel, then scale up production from all plants to match production from national statistics.
(We implicitly assume no bias for not reporting any type of power plants.) In very rare cases, production from PPT is slightly larger than production reported in ENERDATA (2019a). For those cases, we scale down linearly the electricity output in the PPT database to match the statistics. We take CO 2 emissions by country and fuel F from ENERDATA (2019a). Since the last year fully reported for CO 2 emissions is 2016, we compute the carbon intensity of electricity per country and fuel based on electricity output for 2016 reported in ENERDATA (ENERDATA 2019a, 2019b). We latter test the sensitivity of our results to the data sources chosen.

Remaining lifetime of generators
The second step to compute committed emissions is to project the remaining lifetime of each generator. The PPT provides a date of commissioning for most generators. We fill data gaps with the averages at country, technology and unit status level. In addition, there are 23 fossil-fuel-based generators (for a total of 6.2 GW or 3.7% of 2019 capacity) that classify as planned, but for which the reported date of commissioning is in the past. For those, we give priority to the status reported and set the commissioning date to 2019. (Note that the commissioning date does not impact our estimates of committed emissions as long as it is not in the past.) We assume the lifetime of power generators to be 37, 35 and 32 years for coal, natural gas and oil technologies, respectively, following Davis and Socolow (2014). (We later perform a sensitivity analysis on these assumptions.) The PPT reports 251 operating fossil-fuel based generators older than that (for a total of 19.7 GW or 11.6% of 2019 capacity). For those, we assumed their lifespan is extended by 5 years more, following Pfeiffer et al (2018) 12 . Table 2 summarizes the assumed lifespan and average carbon intensity of electricity in LAC by technology. Appendix 2 reports carbon intensity by country and technology.

Correcting for missing countries
The PPT covers only 18 Latin American countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Jamaica, Mexico, Panama, Paraguay, Peru, Trinidad and Tobago, Uruguay, and Venezuela. According to ENERDATA (2019a) these countries are responsible for 94% of carbon emissions from electricity generation in LAC. We create a 'rest of LAC 13 ' aggregate to which we assign the missing emissions per fuel type, with average age taken from the other countries reported in PPT.

Carbon budgets from IPCC
To assess carbon budgets available for power generation in LAC, we rely on the IAMC 1.5°C public database hosted by IIASA (Huppmann et al 2018a). This database contains an ensemble of quantitative, model-based climate change mitigation pathways consistent with 1.5°C and 2°C warming supporting the IPCC's special report on 1.5°C (IPCC 2018, Huppmann et al 2018b). The IPCC uses seven categories of scenarios, grouped by their likelihood to satisfy different temperature targets. Table 3 provides a classification of the pathways reported.
Many climate-stabilization trajectories reported by the IPCC feature negative emissions in the power sector in the second half of the century. One key technology to produce electricity with negative net GHG emissions is bio-energy with carbon capture and storage (BECCS); it relies on the burning of biomass in power plants in connection with the long-term storage of resulting CO 2 (Smith et al 2016, Williamson 2016. When BECCS is available, the least-cost strategy to achieve global carbon neutrality is to eventually generate negative-emission electricity thereby offsetting previous overshoot emissions or emissions from other sectors of the economy that are more difficult to decarbonize (Audoly et al 2018).
We thus need to use two variables from the IPCC database to compute carbon budgets: CO 2 emissions of electricity supply, which reports emissions net from any carbon dioxide removal, and the separate carbon sequestration in the electricity supply. We compute gross CO 2 emissions from the power sector as the sum of net CO 2 emissions from electricity supply 14 and carbon sequestration in the electricity supply sector 15 (see also appendix 6). To compute total budgets, we simply sum these two variables, between 2019 and 2064-which is the year when the last planned unit would operate under normal conditions according to  . If instead of filling up the missing date of commissioning based on averages at country, technology and unit status level, we would have used a distribution of entry dates, this peak would be smoothed over time. Also, most plants under construction or authorized in the ENERDATA database report a commissioning date of 2019, which may reflect a bias in the way the data is reported. The peak in 2019 is further influenced by our decisions to 'correct' to 2019 the commissioned date of units that appear as 'planned' but with a commissioning date in the past in the PPT. None of those peaks affect our estimates of total committed emissions.
The PPT reports 456 planned fossil-based generators, summing to 102 GW or 61% of current fossilfueled capacity in the region. Most planned fossil fuel power plants are natural gas plants (87 GW), followed by coal, peat and oil shale (13.5 GW) and oil (2.1 GW). Brazil leads the fossil-based pipeline, with 38 GW of natural gas, 4.8 GW of coal, and 0.9 GW of oil. Mexico and Chile have in their planned pipelines 22 GW and   We plot data from 1990, however, the database includes units which started to operate before that. 6.7 GW of natural gas capacity, respectively. Committed emissions from the pipeline are dominated by natural gas (63%), followed by coal (26%).
In terms of committed emissions, we find that the continued operation of existing capacity over its remaining lifetime at current utilization rates would result in 6.9 GtCO 2 of emissions through the coming decades. Most committed emissions from operational generators in LAC come from natural gas (52%). This contrasts with the global situation, where coal generators are the main contributors of committed emissions (Pfeiffer et al 2018). Figure 2 shows projected emissions through time by fuel and status (appendix 3 shows projections by country). Projected emissions increase at an average annual rate of 13% between 2018 and 2030 as planned power plants are built and start to operate. Meanwhile, projected emissions from the operational plants decrease at an average annual rate of 2.9% as existing plants reach the end of their lifetime and are decommissioned. Additions to the capital stock are higher than retirements over this period. Committed emissions from operational generators decrease to zero by 2054, as the last planned generator will start to operate in 2030. In total, building all planned power plants would add 6.7 GtCO 2 of committed emissions. Figure 3 provides details of committed emissions from both existing and planned power plants by country. Mexico, Argentina, and Brazil lead committed emissions from operational generators, at 1.8, 1 and 0.9 GtCO 2 , respectively. If planned plants are built, Brazil would become the top contributor to committed  emissions in the region, with 2.7 GtCO 2 , almost tripling committed emissions from its operational generators. Mexico would add 1.2 GtCO 2 ; Chile would add 0.9 GtCO 2 and become the third largest committer in the region. Brazil, Colombia, and Dominican Republic are the countries where building the planned plants add most emissions relative to committed emissions from operational plants (at 3.1, 2.1 and 1.8 times the operational committed emissions, respectively, while the average among other countries is 0.6). If both Brazil and Colombia's committed emissions from planned power plants where only 60% of committed emissions from existing power plants, committed emissions from planned power plants in the region would sum to 4.3 GtCO 2 (and total committed emissions would sum to 11.2 GtCO 2 ). Figure 4 shows the same information by year of commissioning and fuel (see appendix 5). Each bar in figure 4(A) corresponds to committed emissions from power plants added at a specific year in the past. Committed emissions added by the generators in operation in the 90s come primarily from coal. In LAC, natural gas started to gain importance in the late 90s and it turned into the main contributor of committed emissions from 2001 onward. Figure 4(B) plots the same information in a cumulative fashion. It shows that while committed emissions have roughly grown linearly over the last two decades, building all the power plants that appear as planned in the PPT would roughly double committed emissions in only four years. (Again, our assessment does not feature a prediction of how much of the units planned in the PPT will be actually built.) Committed emissions from plants under construction and bidding status sum 4.3 GtCO 2 , while authorized and announced would add 1.4 GtCO 2 and 0.9 GtCO 2 , respectively. More than half (62%) of committed emissions from planned power plants come from natural gas generators, which would add 4.2 GtCO 2 . The largest chunks would be added by Brazil  Our finding of 6.7 GtCO 2 is slightly higher than, but close to the 6.0 GtCO 2 reported by Pfeiffer et al (2018) for planned fossil fuel generators in LAC. We interpret this closeness as a sign of the robustness of the approach. This small difference could come from more projects being planned between the moment Pfeiffer at al collected their data and January 2019, and when we collected ours in January 2019. Unfortunately, the databases used in both studies are paywalled, and our data does not contain a date when projects where announced, so we cannot verify that. The difference is small enough that part of it could also come from different gaps in the data. Pfeiffer et al merge five databases for generators allocating in the planned pipeline the generators under construction or planned statuses in early 2017. They use emission factors from individual fuels and historic heat rated from the IEA. Conversely, we use the PPT database comprising the planned pipeline to announced, authorized, bidding process and under construction statuses in early 2019. We calculate emission factors from the country dashboard from ENERDATA (2019a). Figure 5 plots committed emissions against current emissions. For instance, the green dots on the right indicates that Mexico today emits 120 MtCO 2 yr −1 from the power sector. But existing power plants will emit about 1.8 GtCO 2 over their lifetime and adding planned power plants would bring this number to 3 GtCO 2 . The Brazilian case is the most contrasting. Today, Brazil emits 42 MtCO 2 yr −1 . However, committed emissions from existing plants will be 0.9 GtCO 2 over their lifespan. This number will scale up to 3.6 GtCO 2 if the planned power plants are fully implemented. In other words, committed emissions from existing and planned generators in Brazil represents 87 years of CO 2 emissions. Map 1 shows that Brazil is the most extreme case according to that metric. On average in the region, committed emissions from existing and planned power plants sum to 34 years of current emissions.
As committed emissions would grow if planned power plants are built, so would the average carbon content of electricity. Table 4 shows the carbon intensity of electricity generation of the top four countries CO 2 emitters in LAC in 2012 (OECD 2015) 2018 (ENERDATA 2019b) and 2030. We calculated the electricity output from the full set of operational and planned technologies (both renewable and fossil fuel) based on PPT capacities and the ratio between current electricity and capacity (ENERDATA 2019b). If the planned plants are fully implemented in Brazil, the carbon intensity of the electricity would be 134gCO 2 kWh −1 , which is 61% higher than the current intensity. Figure 6 shows the range of carbon budgets for the LAC power sector computed from the pathways gathered in Huppmann et al (2018a), using the same grouping as in table 2. The central line in the boxplot shows the median budget in that group, the rectangle shows the interquartile range, and the whiskers extend to the full range. In the scenarios compatible with 1.5°C, gross carbon budgets range from 1.1 GtCO 2 to 13.5 GtCO 2 , with an average of 5.8 GtCO 2 . In the scenarios compatible with 2°C, gross carbon budgets range between 1.7 GtCO 2 to 16 GtCO 2 , with a mean of 6.2 GtCO 2 .

Compatibility of the capital stock with remaining carbon budgets
Committed emissions from existing generators (6.9 GtCO 2 ) are thus within the range of LAC power carbon budgets consistent with 1.5°C-2°C. However, they are above 60% of 1.5°C-compliant carbon budgets reported in the IPCC database, and above 50% of 2°C carbon budgets. If all planned power plants are built, the committed emissions would surpass 85% of the carbon budget scenarios consistent with 2°C and all the scenarios consistent with 1.5 C.
These results suggests that, if the temperature targets of the Paris Agreement are to be achieved, roughly 16 52%-55% of existing and planned fossilfueled power plants in Latin America will need to be underutilized, retired early, or retrofitted with expensive CCS or efficiency upgrades.

Sensitivity of findings
We conduct a sensitivity analysis to assess the extent to which our conclusions depend on our lifespan and emission factor assumptions.
The lifetimes we used are calibrated from typical historical averages. In the private sector, payback times can be shorter than technical lifetimes. For instance, contractual terms in LAC auctions vary from 15 to 30 years, with most of countries adopting a contract term of 20 years (Mejdalani et al 2019). If power plants are used only during the typical time required for financial profitability, committed emissions would be lower. To quantify that and provide a lower bound to our estimates of committed emissions, figure 7 compares the results of committed emissions using a lifespan range between 15 and 50 years in order to perform but only technical lifespan but also shorter payback times and contractual terms in auctions.
With lifetimes of 15 years, for instance, committed emissions from both existing and planned plants would be much smaller (5.3 GtCO 2 , 40% of our best guess estimate). In fact, they would be below our estimate of committed emissions from just existing power plants used during the typical lifetimes (6.3 GtCO 2 ), and average carbon budgets from IPCC. However, committed emissions from existing generators (2.8 GtCO 2 ) would still be above 20% of 1.5°C-2°C-compliant carbon budgets, and adding planned power plants would surpass 50% of the carbon budgets consistent with 1.5°C or 2°C. We also test different data sources. We run a simulation using emission factors calculated with the CO 2 emissions from Electricity and heat production from (IEA 2018b) and Electricity output from electricity power plants from the energy balances (IEA 2018a) instead of   16 We simply report the ratio of committed emissions to the average carbon budgets, minus 100%.
ENERDATA. Using data from the IEA (and back to long our central estimates of lifetimes), committed emissions from both operational and planned pipeline jump to 13.6 GtCO 2 to 17.52 GtCO 2 (+29%), reflecting perhaps the inclusion of 'heat generation' in the scope of carbon emissions. Using the IEA as a data source would thus increase our estimate of the amount of asset stranding required to meet the average carbon budget from the IPCC. In light of the results of our sensitivity analysis, we find recent results by Tong et al (2019) for the LAC power sector to be high. They find 14.3 GtCO 2 committed just from existing power plants. One reason is that they use 40 years lifespan for all power plants, while we use 32-37 years. When using 40 years lifespan and data from the IEA to calibrate carbon emissions (as they do), we find that existing power plants in LAC commit a total of 10.5 GtCO 2 (figure 7). In addition, they assume a flat utilization rate of 53%, while we compute implicit utilization rates at the country and fuel level based on historic data (equation (1)). Using a flat 53% rate across countries and fuels, we find 11.4 GtCO 2 . We interpret the remaining difference as evidence that the database power plants that Tong et al use contains more power plants than Enerdata's (both papers use paywalled databases, making a bottom up comparison difficult). More importantly, our results concur with their conclusion that the existing stock of power plants is too large when compared to carbon budgets consistent with the Paris temperature targets. Our paper is the first to reach this conclusion at the region and sector level, comparing committed emissions with carbon budgets for a particular region and sector.

Discussion and conclusion
Our comparison of committed emissions from existing and planned power plant and total emissions in IPCC scenarios provide a crude quantification of the possible disruption to plant owners, workers, and communities that may happen during a transition to clean electricity consistent with the Paris Agreement targets. They do not quantify a fraction of power investments that would turn out to be net losses for their owners from a financial perspective (Vermeulen et al 2018)-this would require much finer data on the cost of building power plants (including terms of financing and tax structures), the cost of operating those power plants (including data on wages and fuel costs accounting for any energy subsidy or tax), and revenues from using them (including electricity wholesale prices), which we do not have access to. Lower utilization rates do not necessarily mean lower economic returns. Even at lower utilization rates, the price of power generated by fossil fuel power plants, and the value of the power reserve they may be able to provide are important parts of the equation.
Notwithstanding those limitations, our results illustrate how international climate change commitments matter to energy infrastructure planners even in developing countries with low baseline emissions. Today the power sector in LAC only emits 357 MtCO 2 per year, but implementing the totality of fossil-fueled power expansion projects reflected in ENERDATA's Power Plant Tracker would commit 6.7GtCO 2 , or 46 years of emissions. We find that 10%-16% of existing fossil-fueled power plants in the region would need to be stranded to meet average carbon budgets from IPCC. More than half of those commitments come from new planned power plants. If the planned power plants are fully implemented, the need of stranded assets to meet average carbon budgets from IPCC would range between 52% and 55%.
Ultimately, assessing the compatibility of any fossil fuel power plant addition with the temperature goals of the Paris Agreement is necessarily more complex than the simple assessments presented in this paper. The key for governments to do so might be to develop domestic long-term power generation development strategies that start from the goal of achieving net zero carbon power generation by 2050, and work backward to establish sectoral roadmaps towards that goal

Acknowledgments
The views expressed in this paper are the sole responsibility of the authors. They do not necessarily reflect the views of the Inter-American Development Bank or the countries they represent. This research received funding from IDB/French Climate Fund project RG-T3193.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.