Stranded asset implications of the Paris Agreement in Latin America and the Caribbean

We use the Global Change Assessment Model (GCAM) to analyze the composition and magnitude of stranded assets in the LAC power sector. GCAM is an open-source, global IAM which captures important interactions between the global economic, energy, agriculture, and land-use systems [40–43] (supplementary figure 1 is available online at stacks.iop.org/ERL/15/044026/mmedia). Dynamic-recursive models of each system are linked through markets and paired with a reduced-form atmosphere-carbon-cycle-climate model called Hector [44].

GCAM contains 32 geopolitical regions and operates in five-year time steps from 2010 (the last base year) to 2100. LAC is divided into seven model regions, four of which (Argentina, Brazil, Colombia, and Mexico) represent individual countries (see supplementary table 1 for a breakdown of countries contained in each GCAM LAC region). Key inputs which drive model results include socioeconomic assumptions (population, labor participation rates, and labor productivity growth rates for each geopolitical region) and representations of the physical world (resources, biophysical processes like net primary productivity), technologies, and policy. In each model period, the model solves for the equilibrium prices and quantities of various energy, agricultural, and greenhouse gas (GHG) markets at either the global or regional level. GCAM tracks emissions of 24 GHGs and air pollutants endogenously based on activity in the energy, agriculture, and land-use systems (supplementary note 1). The model tracks electricity generation by technology vintage (supplementary note 2), which allows the quantification of stranded assets in monetary terms.

We explore four global GHG mitigation scenarios to assess the implications of the Paris Agreement on stranded assets in LAC (table 1). These scenarios vary in terms of near-term and long-term mitigation stringency. To represent the Paris Agreement's long-term temperature goals, we constrain the cumulative CO₂ emissions budgets over the century (2011–2100) to levels that are consistent with limiting mean global surface temperature increase to 2 °C (1000 GtCO₂) or 1.5 °C (400 GtCO₂) [45]. Country-level NDCs are aggregated to the GCAM region level in a manner consistent with previous studies [46] (supplementary note 3; supplementary table 2). Our representation of the NDCs assumes economy-wide mitigation, implemented through a carbon price. Real-world measures will differ from this approach. Regardless, our results are meant to be illustrative and our idealized implementation is sufficient to illustrate the key points raised in this paper. In the long-term, the regional allocation of emissions within the emissions budget is based on global least-cost mitigation and can vary across scenarios.

In the Straight-to-2 °C and Straight-to-1.5 °C scenarios, countries are assumed to achieve their Copenhagen pledges [46] through 2020 before pursuing least-cost mitigation efforts (as implemented by a global carbon price) starting from 2021 to meet the emissions budgets. In contrast, in the NDCs-to-2 °C and the NDCs-to-1.5 °C scenarios, the same emissions budgets are assumed to be achieved while countries mitigate according to their NDCs until 2030, after which global least-cost mitigation is employed. Since it has been demonstrated that the NDCs collectively produce higher emissions than global least-cost mitigation pathways [47, 48], the Straight-to scenarios provide more flexibility to act earlier in decarbonizing the economy and minimize overall financial implications [6].

GCAM tracks electricity generation by technology vintage. Generation in a vintage's initial year of operation represents full utilization and the generation for a vintage can never exceed full utilization. If a technology cannot cover its operating costs, it is retired before the end of its lifetime. We employ a logistical method to retire power production capacity when variable costs approach the price received for power. An s-curve function defines the fraction of power plants which must retire when the variable cost of operation exceeds the market price of electricity (supplementary note 2). These retirements are tracked as a reduction of the vintage's generation capacity.

Our estimate of the value of retired production capacity is based on the original investment to bring the capacity on line and its expected physical lifetime at the time of installation, as well as the fraction of the vintage retired. The cost to bring the vintage of capital on line is the original overnight capital cost (supplementary table 3). The financial value of installed capacity is assumed to decline linearly with time from its initial overnight capital cost. That is, the economic value of the capital stock in subsequent years is the original capital cost times the fraction of the capital stock's foregone useful life (supplementary figure 2). In other words, the foregone value of a prematurely retired power plant is calculated as the total capital cost of the asset times the fraction of expected (physical) lifetime (supplementary table 4) foregone due to premature retirement. This can be expressed as:

SV = OCC * ((EL − AL)/EL), where:

SV = stranded value,

OCC = overnight capital costs,

EL = expected lifetime, and

AL = actual lifetime.

Our methodology extends the one developed by Johnson et al [23] by applying asset depreciation.

Energy and industry CO₂ emissions continue to rise until 2030 in the NDCs-to-2 °C and NDCs-to-1.5 °C scenarios, both globally and in LAC (figure 1). Globally, emissions in the NDCs scenarios are 17% and 77% higher than the Straight-to-2 °C and Straight-to-1.5 °C scenarios in 2030; in LAC, this emissions gap is 18% and 62%, respectively (see supplementary note 4 for a discussion of LAC Emissions Pathways). These pathways are consistent with previous findings that there is a substantial gap globally between the NDC pledges and least-cost emissions pathways [47–49]. The higher near-term emissions in the NDCs scenarios entail steeper reductions beyond 2030, since the NDCs and the Straight-to scenarios are constructed to achieve the same emissions budget over the century. The near-term trend and emissions gap are similar in the power sector, which fully decarbonizes by 2050 in each mitigation scenario (see supplementary figure 4 for information on the LAC power sector in the Reference scenario).

Zoom In Zoom Out Reset image size

It is notable that LAC reaches net-negative energy and industry CO₂ emissions by 2050 in both of the 1.5 °C scenarios, while global emissions in those scenarios remain positive through mid-century. This result is influenced by the use of a uniform global carbon price to achieve the cumulative emissions budget. Under such a regime, emission-reduction efforts are directed toward lowest cost, irrespective of the source of emissions. Since LAC's energy system is presently less carbon-intensive than the average for the rest of the world, it is able to reach net-zero emissions more quickly than regions which have more carbon-intensive infrastructure already locked in place. Negative emissions come from bioenergy with carbon capture and sequestration (BECCS); because LAC has a higher share of bioenergy in primary energy consumption than the rest of the world of the world historically, it tends to deploy BECCS more quickly at a given carbon price in our modeling framework. However, the potential for large-scale BECCS deployment is highly uncertain for a variety of reasons [50]; the sensitivity analysis in section 3.4 includes scenarios in which CCS (and therefore BECCS) is unavailable. (For further discussion on LAC emissions pathways and the role of BECCS, see supplementary notes 4 and 5.)

The four mitigation scenarios are characterized by a major transformation of the energy system by 2050 (supplementary figure 3), including increased energy efficiency and conservation, a transition from emitting fossil-fuel technologies (such as oil and gas power and petroleum based transportation) to low- and non-carbon emitting technologies (including renewable electricity, carbon capture and sequestration, and liquid biofuels), and a shift in the type of investments throughout the energy system.

Here, we focus on stranded assets and investments in the power sector, an important sector in the context of climate change mitigation [4], as a conservative measure of the scale and value of stranded assets in LAC. Across the mitigation scenarios explored in this study, between 60 GW (Straight-to-2 °C) and 128 GW (NDCs-to-1.5 °C) of fossil-fuel power plants are prematurely retired before the end of their physical lifetimes in the LAC power sector from 2021 to 2050 (figure 2; supplementary table 5). These amounts are equivalent to 15%–33% of the total installed capacity in 2015 in LAC (approximately 393 GW) [51]. Since the NDCs-to-1.5 °C scenario requires the fastest reductions in CO₂ emissions, the magnitude of stranded assets in that scenario is also greatest, resulting in nearly 50% more stranding than the Straight-to-1.5 °C scenario and more than double the stranded capacity observed in the Straight-to-2 °C scenario. Most of this stranding occurs between 2031 and 2035, when LAC energy and industry emissions are reduced by nearly 75%. Over 80 GW of capacity is prematurely retired over this five-year period.

Zoom In Zoom Out Reset image size

Natural gas and oil power plants without carbon capture and sequestration storage (CCS) represent the largest fraction of prematurely retired capacity in our scenarios in LAC. In our scenarios, natural gas without CCS accounts for about 45% of stranded capacity in the NDCs-to-2 °C scenario and about 54% of stranded capacity in the NDCs-to-1.5 °C scenario, calling to question natural gas's role as a 'bridge fuel' [52]. To meet the growing demand for electricity, between 751 and 967 GW of new capacity are installed during the 30 year period from 2021 to 2050 (figure 2). These capacity additions are roughly 1.9–2.5 times the total electricity generation capacity in LAC in 2015 [51, 53]. As anticipated, scenarios with the 1.5 °C cumulative emissions budget require more capacity additions than the 2 °C scenarios. This is because achieving the more stringent budget in the 1.5 °C scenarios requires the electricity sector to both (1) decarbonize faster by replacing carbon-intensive plants with new low-carbon capacity and (2) produce more electricity overall, so that end-use sectors can reduce emissions by switching their energy use to electricity.

Similarly, the NDCs scenarios require more new installations overall than the Straight-To scenarios, although the timing of these installations is delayed. Greater near-term mitigation spreads the new installation requirements more evenly across time; low-carbon power installations in the near-term both lower near-term emissions and reduce the rate at which emissions must be curtailed post-2030 (figure 1), limiting the need for greater investments post-2030 to 'catch up' with the cumulative emissions budget [54].

An assessment of the costs associated with these premature retirements and new installations helps illuminate the economic implications of climate change mitigation. Scale is the primary driver of investment costs, with each scenario averaging about $2.5–2.7 billion per GW of new generation capacity (figure 3). Investment costs for the Straight-to-2 °C scenario are the lowest ($1.9 trillion across LAC between 2021 and 2050; see supplementary table 6), while those for the NDCs-to-1.5 °C scenario are the highest (nearly $2.6 trillion). These results are consistent with the insights from figures 1 and 2—while the NDCs are insufficient to limit warming to 2 °C, they imply even more challenges for limiting warming to 1.5 °C [7]. Overall, investment requirements in the NDCs-to-1.5 °C scenario represent about 0.8% of LAC's projected (exogenously specified) GDP from 2021 to 2050, and reach as high as 2.1% of GDP for the 2031–2035 period.

Zoom In Zoom Out Reset image size

Similarly, the costs of stranded capacity (hereafter referred to as 'stranding costs'; see methods) are highest in the NDCs-to-1.5 °C scenario, with cumulative costs of $90 billion between 2021 and 2050 (figure 4). These costs are two-thirds higher than the Straight-to-1.5 °C scenario and more than double the costs in the Straight-to-2 °C scenario. In contrast, the difference in stranding costs between the Straight-to-2 °C and NDCs-to-2 °C scenarios is about 37% ($13 billion USD) over the course of 30 years. For the 1.5 °C scenarios in particular, the timing of asset stranding is driven by their vastly different emissions pathways (figure 1). While the Straight-to-1.5 °C scenario requires more mitigation in the near-term, the NDCs-to-1.5 °C scenario requires a rapid decline in emissions post-2030. This in turn results in much more dramatic stranding of assets in the NDCs-to-1.5 °C scenario compared to the Straight-to-1.5 °C scenario. In other words, the value of strengthening near-term ambition is even greater for a 1.5 °C temperature target [10].

Zoom In Zoom Out Reset image size

An interesting result from our analysis is that the cost of stranding coal technologies is the greatest across scenarios, even though these technologies make up a relatively small percentage of the total capacity stranded (figures 4, 2). This is because (1) coal power plants are more capital intensive than gas and oil plants, and (2) coal power plants are assumed to have longer lifetimes (60 years) than gas and oil plants (45 years) and hence the economic value of coal power plants depreciates slower than gas and oil plants. In addition, oil with CCS plants contribute significantly to stranding costs for the 1.5 °C scenarios between 2041 and 2050 because not only are these plants capital intensive, but they are also relatively new. Hence, although a small amount of capacity is retired, the plants have comparably large economic value when their operations cease.

To assess the extent to which our estimates of power sector investment and stranding costs are influenced by key modeling parameters, we conduct a sensitivity analysis on assumptions about (1) political willingness to avoid stranded assets, (2) technology availability, and (3) the role of land-use change (LUC) in mitigation (table 2). In total, all combinations of these assumptions result in 36 sensitivity cases (18 cases per temperature target).

Different sensitivity cases result in different investment and stranding outcomes (figure 5, supplementary figure 6) and CO₂ emissions pathways (supplementary figure 7). For example, increased political willingness to avoid stranded assets reduces power sector stranded asset costs by 27 billion USD over the twenty-year period from 2031 to 2050 (NDCs-to-2 °C), holding all other parameters at their central assumption values (figure 5). However, increased stranding avoidance in the power sector shifts emissions mitigation to other sectors such as refining (supplementary figure 8) and requires higher carbon prices (17% in 2050) to achieve emissions mitigation goals (supplementary table 8). This heightened refining sector mitigation in turn requires a significant increase in biofuels, intensifying the competition for cropland and raising food prices (by 9% in 2050; see supplementary table 9). Although conducting a detailed evaluation of the implications of food price increases for consumers is beyond the scope of this study, our results suggest that avoiding stranding in the energy sector could have important implications for other sustainable development priorities.

Zoom In Zoom Out Reset image size

Additionally, our analysis suggests that the NDCs-to-1.5 °C scenarios are not feasible in cases with limited technology availability, in particular, without CCS technologies (supplementary figure 6). This is because, by the time the current NDCs are implemented in 2030, cumulative global CO₂ emissions are already 300 GtCO₂ higher than the century-wide 1.5 °C budget (supplementary figure 7). Without CCS and hence CO₂ removal from bioenergy with carbon capture and sequestration (BECCS; supplementary note 5), the level of net-negative emissions needed in the second half of the century to bring cumulative global CO₂ emissions back below the 1.5 °C budget by 2100 is simply not feasible in the GCAM modeling paradigm. While this finding about the importance of CCS for 1.5 °C scenarios is consistent with recent studies [55, 56], it is also important to note that all of the feasible NDCs-to-1.5 °C scenarios in this study imply a temporary overshoot of the 1.5 °C temperature target, since the emissions budget is always exceeded before being met by the end-of-century.

While limited technology availability renders the NDCs-to-1.5 °C scenarios infeasible within our modeling paradigm, the NDCs-to-2 °C scenarios were found to be feasible, albeit at higher costs. Reducing technology availability from Full Tech to No CCS increases power sector stranded asset costs (2031–2050) by $16 billion (figure 5). Despite these higher stranding costs, power sector CO₂ emissions remain higher in the No CCS sensitivity cases compared to their Full Tech counterparts. Further, with CCS unavailable, power sector mitigation cannot be easily displaced to the refining sector. Instead, additional mitigation effort in LAC is shifted from the energy system to the land system, which results in food price increases of 44% (in 2050) due to increased competition for land for afforestation [57, 58]. Mitigation that cannot be shifted to the land sector is accomplished by energy efficiency and conservation, with primary energy consumption, final energy consumption, and passenger vehicle miles traveled reduced 23%, 11%, and 7% respectively in the No CCS case compared with the Full Tech case in 2050 (supplementary figure 9, supplementary table 10).

Furthermore, while changes to the role of land-use in mitigation tend to have small average impacts on power sector stranding costs in LAC (figure 5), the range of potential impacts is fairly large (supplementary figure 10). Shifting to a high role for LUC in mitigation (High LUC) can either increase or decrease stranding costs, depending on assumptions about technology availability. Increasing the role of land-use in mitigation results in a shift towards afforestation and away from bioenergy production (supplementary figure 11). With full technology availability, this decreased availability of bioenergy feedstocks limits the role of BECCS as a mitigation strategy (supplementary figure 11) in both the refining and power sectors. While increased afforestation means that less emissions reductions are required of the energy system overall, a greater share of energy system emission reductions must come from the power sector in the High LUC mitigation cases, because refining (and end-use sectors which consume a large amount of liquid fuels) have fewer mitigation options without BECCS. More carbon-intensive power plants are forced to prematurely retire, increasing stranding costs, and more low-carbon generators are installed to compensate for these premature retirements and to electrify end-use sectors with fewer low-carbon liquid fuels available. Conversely, with limited technology availability, bioenergy is a less important mitigation strategy to begin with; increasing the role of LUC in mitigation helps shift emissions reductions away from the energy system and results in lower stranding costs.