Stranded assets and early closures in global coal mining under 1.5∘C

With the Glasgow Climate Pact 2021, the global community has committed explicitly to phasing down coal consumption. Yet the coal supply sector continues to develop new capacities, despite the risk of asset stranding. This article presents the first assessment of the implications of 1.5∘C mitigation pathways for the coal mining industry. Based on open coal mine data and a new version of the open coal sector model COALMOD-World, the prospects for individual coal mining regions and their risk of early mine closures and asset stranding are analyzed. Results show that global cumulative production capacity from operating thermal coal mines exceed the remaining consumption values for 2020 through 2050 by more than 50%. This supply-consumption discrepancy would hit Russia and the USA especially hard, causing the stranding of around 80% of operating capacities in each case. But the early closure of operating coal mines would affect all of the world’s major thermal coal producing regions, with most regions seeing more than three-fourths of their mine capacity closing early by 2030. Stranded assets from operating coal mines would total some USD2015 120 to 150 billion until 2050, with an additional USD2015 100 billion should currently proposed new coal mining projects be realized. If demand declines in accordance with 1.5∘C pathways, new coal mines or mine extensions would be redundant in all coal regions. Although the stranded asset value of mines is relatively small compared to that of the coal power plant sector, early closures would especially affect workers and local communities. Thus, efforts are urgently needed to ensure a just transition in coal mining regions and to address excess operating and proposed coal supply capacities that continue to fuel global warming.


Introduction
In the first major international agreement of its kind, the COP26 Glasgow Climate Pact explicitly seeks to phase-down (unabated) global coal-fired power generation. This goal is urgently needed in order to limit global warming to 1.5 • C, which requires a phase-out of most coal use within the next two to three decades [1,2]. Yet the coal supply sector and national energy planning in many countries still seem oblivious to the need to cut future coal production [3,4]. Annual global investments in coal supply have remained continuously high in recent years [5], and are set to remain at current levels for the coming decades [4,6].
The decoupling of coal supply capacities from future admissible levels of coal consumption could place coal mines at an increased risk of becoming stranded assets [7,8]. While it is well known that coal-fired power plants face a substantial risk of asset stranding [9][10][11], the potential asset stranding of coal mines has received limited attention so far [12]. Compared to oil and gas production, capital expenditure costs make up only a minor share of the total production costs for coal mining [13]. Nevertheless, in the past coal miners have experienced large asset value write-downs due to excess investments and coal price drops [14]. Apart from the financial risks they pose to owners and investors, early closures of coal mines affect coal workers and local the economies of coal-producing regions [7,15].
Moreover, excess long-lasting coal supply infrastructure can cause lock-in effects, hindering the transition away from coal [12,16,17]. Trout et al [18, p 7] find that the emissions from developed coal reserves alone-that is, the coal stored in running mines and those under development-would exhaust almost 80% of the remaining 1.5 • C emission budget. In order to keep the 1.5 • C target within reach, approximately 88% of the total global coal reserves need to remain unextracted until the end of the century [19, p 233]. Actively limiting coal supply could serve as an important complement to demandside climate policies [20,21]. This is the first study to offer a detailed assessment of the implications of 1.5 • C mitigation pathways for the global coal mining sector. This analysis focuses on investments, premature mine closures, and the stranding of production assets in coal regions. Identifying misaligned coal supply developments that contradict mitigation efforts can help governments to address these and reduce lock-in risks. A better understanding of future regional coal production developments can also enable national and regional governments to address inevitable structural changes in affected regions early on, thereby reducing the negative socio-economic impacts of coal phase-outs [15,22].
This study also provides a new version of the open partial equilibrium model of the global steam coal sector COALMOD-World (CMW). The model uses data from the open-source coal mine database 'Global Coal Mine Tracker' (GCMT) provided by Global Energy Monitor [23] to approximate and consider lifetimes of existing and new coal mines. This new version of CMW now also allows for the assessment of remaining coal supply capacities, investments, and asset stranding. Previous model versions either entirely overlooked or only partially considered the retirement of existing infrastructure [24][25][26].
The focus of the study is on steam coal-that is, thermal coal excluding lignite. Worldwide, steam coal comprises approximately 3/4 of current total coal consumed, and about 90% of thermal coal consumed [27]. The 1.5 • C compatible coal demand scenarios are based on the IPCC [1] mitigation pathways, representing coal consumption in scenarios of the category 'C1: limit warming to 1.5 • C (>50%) with no or limited overshoot' [28]. To illustrate recent production planning in coal sector [compare 4], a coal demand scenario, based on the International Energy Agency (IEA) [29] 'Stated Policy Scenario' (STEPS), is added.

Methods and data
To assess the global coal supply sector in a 1.5 • C world, a scenario analysis using a new version (v2) of the open COALMOD-World model is performed. CMW is a comprehensive model of the global steam coal market, covering about 90% of global steam coal production and consumption (see supplementary material S.1). It represents coal producers and exporters as profit-maximizing players with model-endogenous decisions on new capacity investments. Previous model versions have been widely used for academic studies, assessing international coal market dynamics and various climate policy implications for the supply, trade, and consumption of coal [compare 24,25,[30][31][32][33]. CMW and related model data is openly available for use and adaptation [26]. Few other models of the global coal market exist. These include the EIA [34] National Energy Modeling System with its Coal Market Module (CMM), which can be used to calculate coal flows and prices. The CMM can be used to calculate annual coal production, trade and consumption volumes, as well as coal prices. Auger et al [7] develop a model based on Paulus and Trüby [35] and Trüby and Paulus [36], which covers investments and retirements of coal mines, however, relies on proprietary data to be run. The new version of CMW applied here is based on openly available code and data. It considers the limited lifetimes of existing and new coal mines, which makes it possible to assess investments, remaining capacities, and stranded assets in the coal mining sector. The data on coal mines' lifetimes comes from the GCMT [23], the first comprehensive open-access database to provide worldwide data on coal mines.

Lifetime of coal mines
New (proposed) coal mines have an average lifetime of 29 years, according to the GCMT (where data is available, and excluding China) 4 . The reported lifetimes are generally in the range of six to 50 years per mine. In contrast, Baruya [37,p 75] assumes an amortization time of 20 years for coal mine projects. However, according to coal sector experts, major mines rarely operate for fewer than 20 years due to the high end-of-life costs for such mines (personal communication, March to May 2021). Based on this information, average lifetime of new mines is set here as 25 years. This is a rather conservative estimate for the technical lifetime of a mine; however, it is meant to represent the entire techno-economic lifetime of coal mine assets.
The GCMT also provides limited information on remaining lifetimes of operating coal mines. The GCMT covers approximately 800 (540 outside of China) operating steam coal mines (excluding lignite and pure met coal mines), with a total annual production of about 4.3 Gt (2.7 Gt outside of China). However, a value for 'Reported Life of Mine (Remaining)' is only given for 99 mines outside of China. By contrast, a 'Reserve to Production Ratio (R/P)' value is available for 388 mines. However, while R/P values provide information on available physical resources, the technical lifetime of a mine also dependsamong other things-on the available infrastructure and physical accessibility of reserves. To improve the representation of regionally disaggregated remaining lifetime estimates, the relationship of remaining mine lifetime to R/P for operational thermal coal mines is estimated with a logarithmic regression function and applied to GCMT entries, providing only R/P values (for details and resulting estimates of regional remaining lifetimes, see supplementary material S1.2).

New features of the COALMOD-World version 2.0
The latest version of CMW features a new retirement mechanism for coal production capacities, which yields an improved representation of available mining capacity and necessary investments. In CMW, a coal production asset comprises all equipment and capital stock at coal mines that enable the extraction of a certain amount of coal per year. The investment in coal production capacity represents the initial investment needed to make available the new coal mining capacity. The previous model version allowed mine capacities to be idle for indefinite times and be restarted without additional costs. However, as experts noted, keeping inactive mines in operational conditions is very costly and is rarely done for more than 1-1.5 years. Thus, the new retirement mechanism is based on coal mine ages and lifetimes, not prior production (for details, see supplementary material S1.1).

Coal demand scenarios
To increase the robustness of results and to represent the uncertainty of 1.5 • C compatible coal demand scenarios, a lower bound, a central, and an upper bound 1.5 • C coal demand scenario are included. These three 1.5 • C coal demand scenarios represent 0.25, 0.5 (median), and 0.75 percentile values for annual coal consumption in the power sector of IPCC [1] scenarios that limit global warming to 1.5 • C (>50%) with no or limited overshoot. In the baseline 1.5 • C scenarios presented in this study, coal consumption of power plants with Carbon Capture, Transport and Storage (CCS) is excluded, as the availability of CCS and its application to coal-fired power plants remains uncertain [compare [38][39][40][41]. In supplementary material S3, results for scenarios including coal demand of power plants with CCS are included for comprehensiveness, and to assess possible effects of CCS for stranded assets.
To illustrate current coal production plans in the coal sector [4] and allow for the comparison of the 1.5 • C scenarios with such a scenario, the WEO19-STEPS scenario is added. The latter represents a state of continuous high global coal demand through to the year 2050. Coal demand growth rates for this scenario are based on the IEA [29] STEPS scenario. Further details on the scenarios are provided in supplementary material S2 and all scenario input data files are available in Hauenstein [42].

Excess capacities, early closures, and stranded assets in coal mining
Excess coal production capacities and stranded assets in the case of 1.5 • C pathways are calculated for the period 2020-2050 in this study. These calculations are based on CMW data for the lifetimes and investment costs (overnight capital cost, OCC) of capacities, and on CMW results for available mining capacities (PCap af ) and realized production (Prod af ) in year a in production node f. Early closures are calculated for the year 2030, representing general trends towards the end of the current decade. The share of early closures in year a is defined here as the share of idle capacities in year a that also do not resume production in later model years.
Excess capacities are calculated as the difference between available capacity and realized production over the considered time period (2020-2050) for the production node f (equation (1)) In this study the monetary value of stranded assets is calculated based on the non-recovered share of initial OCC (equation (2)), following an approach similar to that of Edwards et al [9]. A full recovery of the OCC is assumed when the mine produces at full capacity for 25 years (lifetime of new mines). For coal production capacities that started production before 2020, a total lifetime of 25 years and production at full capacity for the time prior to 2020 is assumed. Thus, only the share of OCC that is assigned to the remaining lifetime from 2020 on (RL) can potentially become stranded. Node specific OCC figures are taken from the CMW input data [26,42] Stranded Calculating stranded assets based on recovered OCC of initial investments over an assets lifetime is a widely used approach [9,12]. However, financial asset devaluation can be significantly higher than just the value of unrecovered OCC, as the asset's value also depends on expected future cash flows [13,14]. Thus, the stranded asset values calculated in this study are relatively low-end estimates.

New coal mines redundant globally under 1.5 • C
Global steam coal consumption drops by 88% (central 1.5 • C scenario; values for the upper and lower bound 1.5 • C scenarios given in brackets, here 75%-93%) in the 1.5 • C scenario between 2020 and 2030 ( figure 1, top panel). With CCS available, this rapid decline is slowed down only slightly, with consumption dropping by 85% (70%-91%; see supplementary material S3 for results on scenarios with CCS). This is in stark contrast to the WEO19-STEPS scenario, with coal consumption even slightly increasing until 2030, and only decreasing marginally thereafter. For all major producing regions, a 1.5 • C demand path means drastically reduced output compared to a scenario with continuously high demand (lower three panels of figure 1). However, the speed of the decline varies among producing countries. Of the remaining global coal production in 2030, 86% (85%-92%) come from only three countries: China, Indonesia, and India. China and India produce exclusively for their domestic markets, which are the last two remaining major consumers in 2030, while Indonesia continues to produce for both domestic and export use. The amount of internationally traded steam coal is reduced by three-fourths (57%-82%) between 2020 and 2030, eroding export opportunities for most traditional exporting countries (see figure S5 in supplementary material S3). Only Indonesia remains as major exporter in 2030 due to its low production costs and its central position in the Asian market, where consumption is increasingly concentrated.
Investments in coal mine additions become redundant in the 1.5 • C scenarios (figure 2). In all coal-producing regions the operating coal mining capacities suffice to meet the remaining demand until 2050. Additional sensitivity runs show that  figure 2), with some regional shifts in capacity, mostly towards Indonesia and India.
The currently proposed new steam coal mining capacities amount to about 1400 Mtpa, with large shares in traditional coal exporting countries (Australia, Russia, and South Africa). However, if coal consumption is reduced to 1.5 • C-compatible levels, all of these new projects face a substantial risk of asset stranding, as there is no demand for additional coal supplies. The asset value at risk of stranding-the overnight capital cost (OCC) of all proposed new mine projects-sums up to about USD 2015 100 billion (figure 2). If capacity investments continue to follow the WEO19-STEPS path, cumulative OCC spending on new coal mines amounts to USD 2015 380 billion by the year 2050. However, not only new coal mines are at risk of stranding, as the following section shows.

1.5 • C requires the premature retirement of existing coal mines
The remaining cumulative production capacity from existing steam coal mines amounts to roughly 87 Gt between 2020 and 2050. However, the remaining coal production between 2020 and 2050 in the 1.5 • C scenarios amounts to 32 Gt (26-39 Gt) only (figure 3). More than half of the remaining global cumulative capacity would need to remain unused. This would affect all major coal-producing countries, yet with some variation, for example, due to remaining domestic demand and access to the remaining demand centers in Asia.
Russia and the USA would incur the largest shares of excess capacities, with 78% (73%-83%) and 83% (79%-85%), respectively. The USA already have large excess capacities today, due to continued investments in new coal mining capacities and at the same time collapsing domestic demand in recent years [14]. Domestic U.S. coal demand continues to decline rapidly under 1.5 • C scenarios, and exporting U.S. coal to the remaining demand centers in Asia is not competitive due to limited export infrastructure along the U.S. West Coast [30,33]. Russia has adopted the political objective to become one of the largest coal exporting countries and has significantly expanded its coal exports and related infrastructure in the last years [43,44]. However, Russian coal mines are mostly far from the coast, which implies high transportation costs to reach export markets and makes Russian coal exports noncompetitive in case of low global coal prices.
Indonesia would face the lowest overall capacity stranding of all major producing countries (47%, 36%-55%) due to its low production costs and proximity to the remaining demand centers. However, with Indonesian exports mostly going to China, capacity stranding in Indonesia could increase as well, in case China succeeds to increase its level of selfsufficiency [compare 45]. China and India-the two countries with the highest remaining demand, and  currently the largest producers of coal worldwide, both of which serve only their domestic marketswould be affected by their domestic demand declining faster than their coal mining capacities are depleting.
Large shares of the existing coal mines in all major coal producing regions would have to close before reaching their retirement age ( figure 4(a)). Only in Indonesia would fewer than 60% of mines have to close early by 2030. About one third of Indonesia's current coal mining capacity will have reached its retirement age or exhausted reserves by 2030 and therefore will retire anyways. Other regions with an aging coal mining fleet and decreasing capacities due to regular retirements of mines include Kazakhstan and some Chinese and Indian regions. In all other coal mining regions, three-fourths or more of the remaining mining capacities would have to close prematurely by 2030. Exports of all major exporting countries, except Indonesia, would drop by more than 80% between 2020 and 2030. Remaining exports would almost exclusively go to the Asian market, where Indonesia is situated best to outcompete other coal exporting countries like Colombia and the USA, which are far from this market, but also Australia, Russia, and South Africa, which can supply coal only at higher costs than the low-cost Indonesian mines.
Thus, currently operating capacities are also at risk of becoming stranded assets. Early closures of steam coal mines between 2020 and 2050 could lead to stranded assets of some USD 2015 140 billion (USD 2015 120-150 billion) ( figure 4(b)). Considering absolute asset volumes, asset stranding of existing mines would be highly concentrated in China and the USA. In China it is the sheer size of the coal sector that leads to these large volumes. The coal mine fleet in the USA is still relatively large, and at the same time would lose its market shares comparatively quickly. Furthermore, OCC for coal mines are rather high in some Chinese and U.S. coal mining regions with high shares of underground mines (e.g. in Appalachia) [compare 26], thus, also the asset value per unit of production capacity is higher. In India and Indonesia OCC of coal mines are generally lower, but also less production capacities would be stranded. Of the currently proposed new coal mines a relatively large share is situated in the coal exporting countries Australia, Russia, and South Africa (see figure 2). If realized, these capacities could face a particularly high risk of stranding due to the collapsing export market.

Discussion
Trout et al [18] have shown that, in order to limit global warming to 1.5 • C, a substantial share of developed coal reserves would have to remain unused. This study further fleshes out that analysis, assessing the implications of a 1.5 • C policy environment for the steam coal mining sector and its exposure to the risk of asset stranding. Global coal consumption in line with the 1.5 • C target would lead to a significant reduction of coal output and cause excess capacities across all major coal producing regions. The finding that any new coal mining capacities after 2020 would be redundant is in line with IEA [51]. Also for scenarios with continuously high coal consumption, results for investments in new coal mines in this study align with the results of other studies [compare 7,29].
Not only new coal projects but also currently operating mines would be significantly impacted by a required rapid reduction in coal demand. About 63% (56%-70%) of their cumulative capacity until 2050 would have to remain unused, and more than three-fourths of all operating capacities in almost all coal mining regions would need to retire early by 2030. This is about three times the amount of capacities affected by premature retirements Auger et al [7] find in case of gradually declining coal consumption (−60% by 2040 compared to today's consumption). Global coal consumption is already largely phased-out by 2035 in 1.5 • C scenarios due to coal's high emission intensity and the relatively good subsitutability of coal, particularly in the power sector [1]. If CCS would become generally available for coal-fired power plants, it could only prevent early coal mine closures and asset stranding in the range of 1%-8% (compared to results for scenarios without CCS). This is due to the relatively late and limited deployment of CCS on existing coal-fired power plants in most 1.5 • C scenarios [28]. This could change, if CCS is rolled out very fast [52,53], however, past experiences with CCS cast doubt on its near-term large-scale deployment on coal-fired power plants due to high costs and technical difficulties [40,41,54].
The stranded asset value (unrecovered OCC) from currently operating coal mines alone could represent about USD 2015 140 billion. If all of the currently proposed new coal mines were to be built, this could add another USD 2015 100 billion of stranded asset value. So far, few assessments of stranded assets in the upstream coal sector exist [12,55], and used definitions for stranded assets vary. For 2 • C scenarios, CTI [56] and IRENA [57, p 26] estimate coal supply capital expenditure costs (capex) and asset values at risk of stranding to amount to some USD 220 and 250 billion, respectively, including the value of maintenance capex [56] and coal reserves [57]. Assessing differences in financial values, the cumulative value of stranded coal supply assets until 2035 in 2 • C scenarios could amount to USD 2016 300 billion to one trillion, depending on the applied discount rate [58]. This reflects that the financial value of coal supply assets (based on expected future cash flows) can be significantly higher than the capex of new coal mine capacities [13]. Thus, results of this study should be considered a lower end estimate for the value of stranded coal supply assets under 1.5 • C. In the downstream coal power sector, asset values at risk of stranding are significantly larger than for upstream coal supply, with around USD 900-1400 billion (unrecovered OCC only) in 1.5 • C scenarios [9]. Long lifetimes (around 50 years) and relatively high capex of coal-fired power plants (1300-3000 USD kW −1 ) [59] compared to coal mine assets cause these higher asset values at risk.
Although the value of possible stranded assets in the upstream and downstream coal sector differ, they are strongly linked by their dependence on future global coal-fired power generation pathways, with coal-fired power plants being the largest consumers of coal. However, regional differences arise. Net exporters of coal, like Australia or Indonesia, have a high risk of asset stranding in the upstream coal supply sector, while net importers, like the EU or Japan, could be affected mostly by asset stranding in the coal power sector [compare 9]. For China and India, with their large coal mining capacities and power plant fleets, up-and downstream asset stranding risk is high, and more than 50% of all coal assets at risk of stranding are located in these two countries. In both countries, coal supply is largely in the hand of state-owned or controlled companies, while in countries, like Australia and the USA, private companies dominate [56,60].
For both, up-and downstream coal sector, the volume of assets at risk of stranding can be significantly reduced if immediate action is taken and investments in additional capacities are prevented [compare 7,9]. Introducing moratoria on new coal mines could be one option [31,61]. Particularly, for coal exporting countries with a high amount of proposed new capacities, like Australia or Russia, this could be an effective measure to limit stranded assets. Additionally, restricting access to finance and insurance could increase project costs and eventually limit the realization of new coal projects [62]. Beyond reducing the risk of asset stranding, limiting coal supply could complement demand side climate policies, aiming to reduce coal consumption, by increasing the relative cost of coal and by contributing to prevent carbon leakage [20,63,64].
However, the coal mine asset value at risk of stranding is only one side of the medal. Coal mine closures can lead to significant economic downturns in affected regions due to de-industrialization, unemployment and outmigration [15,65,66]. Globally, there are currently around 4.7 million direct jobs in coal mining, concentrated in China and India, and possibly even more informal and dependent indirect jobs [67]. and coal rents contribute about 0.2%, or USD 160 billion (year 2019) to global annual GDP [68].
This highlights the importance of adequate measures to compensate for economic and socioeconomic effects of coal mine closures. Renewable energy and other energy transition related industries could offer alternative employment and economic development opportunities [69,70]. National and international funds to provide structural change assistance to affected regions, as well as stakeholder integration in policy processes can furthermore contribute to increase stakeholder consent to phasing out coal production [71,72].
Limitations to this study include the meager availability of coal sector data. Although Global Energy Monitor [23] provides the first comprehensive open database for coal mines with the GCMT, the data on coal mine lifetimes is still quite limited-even more so on OCC for coal mines. While the results obtained for required capacity additions across the 1.5 • C scenarios are robust to further reductions of remaining lifetimes, the amount obtained for stranded assets is sensitive to assumptions on OCC values. Furthermore, the results of this study are based on intertemporal optimization with perfect foresight. In contrast, real-world coal markets are characterized by cyclical effects, with periods of excess investment in mining capacities when coal prices rise, and early write-offs of uneconomic capacities when coal prices dip [14]. Thus, the volume of stranded coal supply assets could turn out even higher.

Conclusion
While previous studies have shown that the majority of global coal reserves need to stay in the ground if global warming is to be limited to 1.5 • C, this is the first study to assess the effects on the operative global steam coal mining sector. A 1.5 • C compatible coal demand would lead to a global slump of coal production. Any coal mine capacity additions from 2020 on would be redundant and at risk of becoming stranded assets. Furthermore, almost two-thirds of the current global cumulative production capacity needs to remain unused and the majority of coal mines need to be retired early. The value of stranded assets in the coal mining sector from operating mines alone would amount to some USD 2015 140 billion by 2050. If all of the currently proposed new coal mines and coal mine additions were to be built, this would almost double the amount of stranded assets. While this value is comparatively small (about one-sixth of stranded coal power plant assets' value), local economic dependencies on coal mining mean that declining coal production and early closures of coal mines can have severe local effects.
Stopping currently proposed coal mine capacity expansions, for example, by introducing moratoria on new capacities, could reduce the risk of coal mine asset stranding significantly. However, also excess operating coal mine capacities need to be addressed urgently in order to limit global warming to 1.5 • C. Combined coal phase-out and structural change policy packages can cushion adverse socioeconomic effects and increase legitimacy for early coal mine closures.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// doi.org/10.5281/zenodo.7086153.
The Global Coal Mine Tracker (June 2021) data is provided by Global Energy Monitor [23]. All model code can be found in Hauenstein [73]. the 39th International Energy Workshop also deserve a hearty thanks for discussions of early versions of this work. Special thanks go to Franziska Holz, Konstantin Löffler, Pao-Yu Oei, and Paola Yanguas Parra for continuous and intense discussions of this work, and their critical and always helpful feedback, as well as to Thomas Mitterecker for his excellent research assistance. I am grateful to two anonymous reviewers for valuable comments on the manuscript draft. I gratefully acknowledge funding from the Heinrich-Böll-Foundation. This work was supported by the German Ministry for Education and Research (BMBF) under grant number 01LN1704A ('CoalExit' Project). Language editing was kindly provided by the Europa-Universität Flensburg. I acknowledge financial support by Land Schleswig-Holstein within the funding programme Open Access-Publikationsfonds. All remaining errors are mine.

Declaration of interests
The author declares that he has no conflict of interest.