Geothermal Energy and Greenhouse Gas Emissions: A Systematic Review

Abstract

This systematic review evaluates life cycle greenhouse gas emissions from geothermal electricity generation, with a focus on hydrothermal flash, hydrothermal binary, and enhanced geothermal systems. By screening 169 studies published since 2017, 24 met the inclusion criteria and were analyzed alongside Pre-2017 benchmarks. Results indicate median life cycle emissions of 43 g CO₂ eq/kWh for hydrothermal flash, 36 g CO₂ eq/kWh for enhanced geothermal, and 30 g CO₂ eq/kWh for hydrothermal binary. Compared with fossil fuels, these values remain substantially lower, underscoring the role of geothermal in low-carbon energy transitions. Evidence suggests that emissions are declining during the construction phase due to advances in drilling efficiency and modular plant design, while operational emissions remain stable, reflecting the limited deployment of carbon capture. End-of-life recycling varies between technologies. These findings highlight geothermal energy’s capacity to support resilient, circular, and sustainable power systems.

Keywords

Geothermal Energy - Greenhouse Gas Emissions - Life Cycle Assessment - Circular Economy - Energy Transition

Significance

This review synthesizes life cycle assessments of geothermal energy technologies to evaluate the associated greenhouse gas emissions. The study assesses mature geothermal technologies, such as hydrothermal flash, binary geothermal, and enhanced geothermal, highlighting advances in construction efficiency, the impact of carbon capture, and opportunities for recycling materials at end of life. The findings show the potential of geothermal power to support low-carbon, circular, and resilient energy transitions.

Introduction

Global warming is the driving force for energy systems to minimize life cycle emissions and embed circularity principles. While geothermal energy offers stable low-carbon electricity, a systematic understanding of its greenhouse gas (GHG) emissions and circular resource opportunities is lacking. With temperatures rising at a rate nearly four times the global average,[1] the region’s fragile ecosystem has already seen global warming’s dangerous effects in action. For one, rising temperatures have led to the melting of Arctic glaciers, which have accounted for 35% of global sea-level rise,[2] an extremely dangerous reality. Moreover, global warming has caused an increase in wildfires, posing a negative feedback loop, as GHGs are released into the atmosphere while simultaneously being caused by global warming in the first place.[2] The warming has led to desertification of significant amounts of dryland, impacting livelihoods worldwide.[3]

These are only a handful of the harmful implications of rising temperature levels being felt across the planet, and can all be tied back to the dangers of GHG emissions.

In order to reduce GHG emissions, an effort has been made to incorporate renewable energy sources, which have far fewer GHG emissions than traditional fossil fuels like coal, oil, and natural gas, into electric grids throughout the world. Of these newly incorporated renewable energy sources, geothermal energy production has proven to be promising and even advantageous over other renewable sources such as wind and solar due to its dispatchability. A process designed to generate energy harnessed from the hot and stable temperatures below the surface of the earth, geothermal energy production keeps a constant year-round baseload providing grids with consistent electricity and buildings and homes with reliable heating.

In an operation in which geothermal water is pumped into the ground, contact with rocks and groundwater of high temperatures[4] heats geothermal water, producing steam to propel a steam turbine. It is this process that generates what is known as “geothermal energy,” one of the most reliable sources of renewable energy out there.

Given its status as not only a renewable energy source but also one that is capable of providing a constant baseload, geothermal energy is considered to be an incredibly reliable source of clean energy. However, despite the consensus that geothermal power plants have lower GHG emissions than those which rely on fossil fuels, estimates for the lifetime GHG emissions of geothermal energy plants vary tremendously across the literature.

In 2017, researchers Annika Eberle, Garvin Heath, Scott Nicholson, and Alberta Carpenter of the National Renewable Energy Laboratory published a report investigating the lifetime GHG emissions (in g CO₂ eq/kWh) of three prominent geothermal technologies: enhanced geothermal systems (EGS) binary, hydrothermal (HT) flash, and HT binary, using data from the literature available at the time.[5]

Although this previous gives benchmark estimations of geothermal life cycle emissions, this review did not capture the rapid expansion of studies published after 2017, leaving recent advances in drilling efficiency, carbon capture, and end-of-life recycling unaddressed. In their report, Eberle et al. systematically compiled and analyzed data, providing a clear foundation for policy decision-making, technology development, and circular design strategies.[5]

However, in contrast with many of the other more prominent renewable sources, estimates of lifetime and life cycle GHG emissions of geothermal power plants remain inconsistent due to variations in parameters of plants such as size, location, and temperature gradient, and differences in the technology and technological advances of the geothermal plants in the first place.[5]

Geological formations can be good indicators, however, and the most common forms of geothermal electricity generation are as follows:

• HT binary: Geothermal water is pumped up from deep below the earth’s surface where it gets heated, before going through a heat exchanger, which transfers the heat to a separate working fluid. This converts the working fluid into steam, which spins a turbine, before being condensed again back into a working fluid. This process typically takes place when the below-earth temperature is between 107 and 182 °C.[6]

• HT flash: After being heated by earth's high internal temperature, geothermal water gets pumped to the surface and separated into hot steam (which spins the turbine) and cool liquid which then gets pumped back into the earth, alongside the steam after being condensed back into a liquid. This process typically takes place when the below-earth temperature is greater than 182 °C.[7]

• EGS binary: The layer below the earth’s surface is hydraulically stimulated to create an artificial reservoir for geothermal water to flow through, created where there is insufficient natural permeability but still high temperatures below the earth’s surface. Once geothermal water is injected into the ground, energy is generated through a binary cycle.[8]

This review will begin by compiling GHG emissions of geothermal plants, followed by the methodology used to collect data from published studies. After presentation of the studies used to develop the datasets analyzed in this review, limitations will be discussed, and findings and results regarding GHG emissions of geothermal plants will be presented. The discussion of the findings of this systematic review will conclude the report.

Methodology

Ensuring comparability with previous work,[5] the review adopted the same schematics for the screening framework, updated to include studies published after 2017. The goal was to establish a new, updated dataset of geothermal life cycle GHG emissions consistent with prior methodology. Following Eberle et al.,[5] the review process included two separate screens, outlined in [Tables 1 and 2]. First titles and abstracts for relevance. The second stage was a full text evaluation against predefined inclusion and exclusion criteria. A total of 169 unique studies were screened in this process, each with a unique DOI. If a study did not pass the first screen (failed to meet at least one of the criteria), it was not reviewed in the second screen. Please note that many of the studies failed to meet multiple criteria, meaning that there is significant overlap between the number of studies that failed criteria. Please also note that the studies that failed to meet the first criterion of the first screen (“full text is available and written in English”) were not tested against any of the other criteria given their lack of availability.

Table 1 Criteria used to perform first screen adjusted for the research, based on Eberle et al.[5] accompanied by the number of studies that failed each criterion.

Criteria for first screen	Failed screen
Full text is available and is written in English	4
Was published after 2017	0
Was published as one of the following: An archival journal article or trade journal article greater than three published pages in length A conference proceeding greater than five double-spaced pages in length A book or book chapter, thesis, dissertation, or report	9
Covers geothermal energy, reporting electricity as an end product	63
Reports quantitative results from an LCA or reviews results from multiple LCAs	85

Table 2 Criteria used to perform the second screen, following Eberle et al.[5] accompanied by the number of studies that failed each criterion.

	Criteria	Failed screen
Quality	Uses a currently accepted LCA method (e.g., follows guideline 14040 from the International Organization for Standardization [ISO 2006])	10
	Employs a relevant impact assessment method (e.g., selected impact categories, category indicators, and characterization methods)	9
	Evaluates at least two life-cycle stages	6
Transparency and completeness	Reports their method transparently with regard to key parameters, assumptions, and methods (e.g., defines a system boundary)	3
	Provides a numerical description of the system characteristics (e.g., plant size or well depth)	3
	Reports the environmental-impact estimates quantitatively	1
	If appropriate, reports the name of LCA software or database used for the analysis	9
	Provides citations for data sources	0
	Reports a unique estimate of the result (i.e., the result is not cited from prior work)	6
	Provides enough information to scale results by plant generation	6
Relevance	Evaluates a modern or near-future system (e.g., supercritical carbon dioxide [CO2] as a working fluid). Geo-pressured geothermal systems are excluded.	0

Given the large number of studies screened, this process was supported by use of a large language model to increase consistency. Making use of advanced reasoning to screen all papers against this rigid criterion ensured consistency in the screening of each piece of literature, and allowed this process to be conducted in a timely fashion.

Studies that passed both of the screening stages, and reported life cycle GHG emissions of geothermal electricity plants, covering at least one of the following technologies: EGS binary, HT flash, and HT binary, were included. For each eligible study, data were extracted for life cycle GHG emissions (g CO₂ eq/kWh) and where available, data on the following life cycle phases: construction phase, operational phase, and end-of-life phase were collected.

Extracted values were organized into three datasets:

• Post-2017 dataset (this review),

• Pre-2017 dataset,[5]

• Combined dataset aggregating both sources.

After data were collected in each of these categories, the following statistics were determined for both the total GHG emissions of each technology, and the life cycle phases within each, for both the “Post-2017” dataset and the “Combined” dataset: minimum value, 25th percentile, median, 75th percentile, and maximum value. Using these statistics, box plots for each technology’s total GHG emissions and life cycle phase emissions were created.

Results and Discussion

The systematic literature review screened 169 studies published since 2017, identifying diverse geothermal technologies including EGS binary, HT flash, HT binary, dry steam, CO₂-plume geothermal (CPG), hybrid geothermal plants, and geothermal heat pumps. For consistency with previous life cycle analyses, this review focuses on the three most widely studied technologies: EGS, HT flash, and HT binary.

Following the rigorous two-stage screening criteria, 32 studies met all the inclusion criteria, and 24 of these studies reported life cycle GHG emissions suitable for the quantitative analysis. The 24 studies cover EGS binary, HT flash, and HT binary, and all represent geothermal scenarios.

Overview of Estimates of Life Cycle GHG Emissions from Three Geothermal Technologies

[Table 3] displays each of 1the studies used in the creation of the “Post-2017” dataset and the values they report. Please note that some of the studies used in this dataset provide estimates of GHG emissions from multiple different geothermal plants or scenarios (as represented by the paper’s respective estimate count), often with different parameters (e.g., location, lifetime, size, etc.). Given these variations, some of the parameter values are listed as ranges, encompassing all of those used by scenarios within each paper. Additionally, some studies provide certain parameters for some sites but not others, and in this table, values may only be referring to the cases whose parameters are given.

Table 3 Studies used in the “Post-2017” Dataset and their reported values.

Author(s)	Year	Estimate count	Location	Impact assessment method	Lifetime (years)	Size (MW)	Depth (km)	Temp. (C)
EGS Binary
Li et al.[20]	2024	4	France, Europe	ReCiPe 2016	15–30	0.02–4.1	2.5–6.518	90–168
Paulillo et al.[9]	2020	10	United Kingdom	ILCD (International Life Cycle Data System)	30	>1	4.5	175
Paulillo et al.[28]	2021	1	–	Environmental Footprint 2.0 (EF2.0)	20–40	0.4–11	2.5–6	–
Paulillo et al.[29]	2022	1	United Kingdom	Environmental Footprint 2.0 (EF2.0)	30	5.7	4–4.25	–
Pratiwi et al.[30]	2018	4	France	ILCD 2011, 2007 GWP 100 of the IPCC	25	2.59–3.74	2.708–3.513	150–170
Strojny et al.[31]	2024	3	Poland, Norway	ReCiPe 2016	33	0.8–13	4.2–4.45	–
Wang et al.[32]	2020	1	China	CML 2002 and IPCC 2007	30	3.3	3.705	236
Zuffi et al.[23]	2022	1	United Kingdom	Product Environmental Footprint 2.0	–	1	–	175
HT Flash
Basosi et al.[33]	2020	3	Italy	ReCiPe 2016 and ILCD 2011 Midpoint+	30	20	3–4.5	196
Díaz-Ramírez et al.[34]	2023	1	Iceland	ReCiPe 2016	30	303	2.346	–
González-García et al.[35]	2022	1	Mexico	ReCiPe 2016	30	85	–	180
Karlsdottir et al.[19]	2020	2	Iceland	CML 2001 + CED	30	303.3	–	–
Karlsdottir et al.[36]	2020	1	Iceland	Cumulative Energy Demand (CED)	30	303	–	–
Kjeld et al.[37]	2022	1	Iceland	Kjeld et al. 2022, Table 14	40	90	>2	250
Mainar-Toledo et al.[38]	2023	1	Iceland	ReCiPe 2016 Midpoint	30	120	–	200
Paulillo et al.[39]	2019	2	Iceland	ILCD (International Life Cycle Data System)	30	270–303.3	2.15	–
Pratama et al.[40]	2023	1	Indonesia	GWP100 CO2 eq	30	–	–	210
Rossi et al.[41]	2023	5	Italy	Environmental Footprint 3.0 (EF3.0)	30	20	–	–
Tosti et al.[21]	2020	1	Italy	ILCD 2011 Midpoint+	40	61	3	300–350
Wang et al.[31]	2020	2	China	CML 2002 and IPCC 2007	30	0.3–24	0.3–0.8	91–160
Zuffi et al.[23]	2022	2	Iceland, Italy	Product Environmental Footprint 2.0	–	60–303.3	–	180–200
HT Binary
Dawo et al.[42]	2021	4	Germany	GWP100 CO2 eq	30	1.2–1.94	5.24	120
Lohse[43]	2018	3	Germany	GWP100 CO2 eq	15	3	3–4.4	120–150
Menberg et al.[44]	2021	3	Germany	IMPACT 2002+	30	5.5	8.664	138
Menberg et al.[45]	2023	4	Germany	IMPACT 2002+	30	5.5	8.664	138
Wang et al.[32]	2020	1	China	CML 2002 and IPCC 2007	30	0.4	3.216	110
Zuffi et al.[23]	2022	1	Italy	Product Environmental Footprint 2.0	–	10	–	180
Zuffi et al.[46]	2024	1	Italy	ReCiPe 2016	30	10	–	180

As is seen in [Table 3], the dataset spans global case studies, with a lifespan ranging from 15 to 40 years, plant size of 0.02–303 MW installed capacity and well depth of 0.3–8 km. The resource temperature ranges from 90 to 350°C, underscoring the technological and geological diversity that drives the emission variability.

Limitations of this Work

Given the nature of this report and having a dataset composed entirely of values from separate, independently conducted studies, there are several limitations and areas of ambiguity. For one, making use of a large language model to screen literature made way for the inherent errors that come with using artificial intelligence in any capacity. For each article that passed the screening from the artificial intelligence, however, the article was then read and reviewed to ensure that it did, in fact, meet all of the criteria for use in the dataset. Furthermore, it was ensured that no publications were duplicated in the dataset.

This “Post-2017” dataset is also composed of values from plants with variation in size, depth, temperature, location, and lifetime studied. In this regard, there is an opportunity for harmonization through the normalization of parameters, but that is outside the scope of this report. Also similarly to Eberle et al.,[5] many of the results within each category come from a single source, often even providing data on different scenarios for the same plant. For example, 10 out of the 25 estimates for EGS binary come from Paulillo et al., which provide 10 different scenarios for the United Downs Deep Geothermal Power Project.[9] Using multiple pieces of data from one study could introduce a bias, in that the same methods may be used to generate data for different scenarios, or they may have similar parameters.

Additionally, many of the LCA studies used in creation of this “Post-2017” dataset presented data in the form of imprecise graphs in which values for life cycle phase emissions, or even total emissions in some cases, were not explicitly given. This meant that some values had to be interpreted from these figures for use in the dataset, opening the door for some potential inaccuracies.

Another limitation of this work is that many of the studies used for the “Post-2017” dataset provide data on the same plant, with slight variations in parameters, leading to different values for GHG emissions. This has the potential to give certain plants more weight than others in the data, simply by virtue of different scenarios being explored within one study.

Finally, some of the data points used are those of general models described within their respective paper, often with values derived from literature, in a similar manner as this report. In this way, there was some overlap between reports cited in the dataset of other papers used, and those used in this report’s “Post-2017” dataset. However, given that these papers reviewed LCA studies and provided unique quantitative results of their own, these studies did pass the screening process and were therefore fit to be used in the development of this report’s dataset.

Summary of Life Cycle GHG Emissions

[Fig. 1a–c] displays the summary plots of total GHG emissions of each respective geothermal technology, plotted as the “Pre-2017” dataset from Eberle et al.,[5] the “Post-2017” dataset with all of the data collected for this report, and the “Combined” dataset that aggregates data from both datasets. For all of the box plots in [Figs. 1–5], the top and bottom whiskers represent the maximum and minimum GHG emission values collected, respectively, the top and bottom of the box represent the 75th and 25th percentiles of GHG emissions, respectively, and the line in the box represents the median value for GHG emissions. All GHG emissions are represented in g CO₂ eq/kWh.

Fig. 1 Distributions of reported GHG emissions of geothermal technologies (a) EGS binary, (b) HT flash, and (c) HT binary, disaggregated by dataset.

[Fig. 1a] displays the distributions of EGS binary total lifetime GHG emissions, which have increased by 21% from 32 g CO₂ eq/kWh (Pre-2017) to 39 g CO₂ eq/kWh. It should be noted that the interquartile range of the newly collected dataset extends lower than that of previous studies. The “Combined” dataset reports a median of 36 g CO₂ eq/kWh.

[Fig. 1b] displays the distributions of HT flash total lifetime GHG emissions, which highlights a reduction of 51% from 47 g CO₂ eq/kWh (Pre-2017) down to 23 g CO₂ eq/kWh. In contradiction to the lowering of the median, the maximum has increased dramatically by 249% from 245 g CO₂ eq/kWh (Pre-2017) up to 855 g CO₂ eq/kWh. The “Combined” dataset reports a median of 43 g CO₂ eq/kWh.

[Fig. 1c] displays the distributions of HT binary total lifetime GHG emissions indicates a 27% increase from 11 g CO₂ eq/kWh (Pre-2017) up to 36 g CO₂ eq/kWh. It should also be noted that there is a drastic decrease in the interquartile range in the “Post-2017” data, from the “Pre-2017” data, with the middle 50% encompassing only the higher values in that of the “Pre-2017” dataset. The “Combined” dataset reports a median of 30 g CO₂ eq/kWh.

EGS Binary GHG Emissions Disaggregated by Life Cycle Phase

[Fig. 2] elaborates on the corresponding subfigures of [Fig. 1], this time displaying the distributions of the GHG emissions of EGS binary plants during their three life cycle phases: construction, operation, and end of life. Please note that the data for each of these individual life cycle phases was explicitly noted as corresponding to this specific phase in its original literature. The same is true for the life cycle phase emissions plots for the remaining geothermal technologies.

Fig. 2 Distributions of reported GHG emissions of EGS binary life cycle phases (a) construction, (b) operation, and (c) end of life, disaggregated by dataset.

The distributions of EGS binary construction phase GHG emissions is displayed in [Fig. 2a] showing a 47% decrease from 40 g CO₂ eq/kWh (Pre-2017) down to 21 g CO₂ eq/kWh. This represents a substantial decrease in reported GHG emissions values of EGS binary plants during their construction phase since 2017, with this median value being below the minimum reported before 2017. It should also be noted that the interquartile range has remained of similar length, but the range of values covered has also dropped significantly. However, the maximum has increased beyond the maximum of the “Pre-2017” dataset. The “Combined” dataset reports a median of 28 g CO₂ eq/kWh during the construction phase of an EGS binary geothermal plant.

[Fig. 2b] displays the distributions of EGS binary operation phase GHG emissions, which has remained stable, with an increase from 2 g CO₂ eq/kWh (Pre-2017) up to a median of 3 g CO₂ eq/kWh. The distributions of EGS binary operation phase GHG emissions remains pretty consistent between these two datasets, although the maximum value reported in the “Post-2017” dataset greatly exceeds that of the “Pre-2017” dataset, reaching 24 g CO₂ eq/kWh. The “Combined” dataset reports a median of 3 g CO₂ eq/kWh during the operation phase of an EGS binary geothermal plant.

[Fig. 2c] displays the distributions of EGS binary end-of-life phase GHG emissions, the “Pre-2017” dataset reporting a median of 0.21 g CO₂ eq/kWh, and the “Post-2017” dataset reporting a median of 0.10 g CO₂ eq/kWh. Despite having median values relatively close to one another, the distributions are also very different between the two datasets. For one, the interquartile range of the “Post-2017” dataset is far greater than that of the “Pre-2017” dataset, and the overall range also spans far wider. Additionally, the “Post-2017” dataset has GHG emissions for the end-of-life phase of EGS binary plants spanning into the negatives (implying erasure of emissions during certain processes in the decommissioning phase), which is not seen at all in the “Pre-2017” dataset. Additionally, the overall spread of the “Post-2017” dataset is greater than that of the “Pre-2017” dataset. However, given that the magnitude of EGS binary end-of-life phase emissions is so small to begin with, this difference can likely be attributed to the slight variations in estimates of this small value between the “Pre-2017” and “Post-2017” studies rather than indicating a significant difference between the end-of-life emissions of EGS binary plants before and after 2017. The “Combined” dataset reports a median of 0.16 g CO₂ eq/kWh during the end-of-life phase of an EGS binary geothermal plant.

HT Flash GHG Emissions Disaggregated by Life Cycle Phase

[Fig. 3] elaborates on the corresponding subfigures of [Fig. 1], displaying the distributions of the GHG emissions of HT flash plants throughout their life cycle phases.

Fig. 3 Distributions of reported GHG emissions of HT flash life cycle phases (a) construction, (b) operation, and (c) end of life, disaggregated by dataset.

[Fig. 3a] displays the distributions of HT flash construction phase GHG emissions, showing a 60% reduction of the median from 5 g CO₂ eq/kWh (Pre-2017) down to 2 g CO₂ eq/kWh. This represents a substantial decrease in GHG emissions reported for HT flash plants during their construction phase since 2017, this value even falling below the minimum value reported past values. Moreover, these two datasets have drastically different distributions, with both the interquartile range and overall range of the “Post-2017” dataset being far greater than that of the “Pre-2017” dataset, reaching values as high as 36 g CO₂ eq/kWh, versus the “Pre-2017” maximum of only 5 g CO₂ eq/kWh. The “Combined” dataset reports a median of 4 g CO₂ eq/kWh for GHG emissions of HT flash plants during their construction.

[Fig. 3b] displays the distribution of HT flash operation phase GHG emissions indicating a 79% emission reduction from 73 g CO₂ eq/kWh (Pre-2017) down to 15 g CO₂ eq/kWh, a drastic difference between the datasets. These distributions are distinctly different from one another in that the interquartile range of the “Pre-2017” dataset is far greater than that of the “Post-2017” dataset. However, the overall range of the “Post-2017” dataset far exceeds that of the “Pre-2017” dataset, reaching a maximum of 848 g CO₂ eq/kWh. The “Combined” dataset reports a median of 22 g CO₂ eq/kWh emitted by HT flash plants during their operational phase.

[Fig. 3c] displays the distributions of end-of-life emissions of HT flash plants, the “Pre-2017” dataset not reporting any statistics due to insufficient data, and the “Post-2017” reporting a median of 0.20 g CO₂ eq/kWh. Given the lack of “Pre-2017” data, the “Combined” dataset also presents a value of 0.20 g CO₂ eq/kWh emitted by HT flash plants during their end-of-life phase.

HT Binary GHG Emissions Disaggregated by Life Cycle Phase

[Fig. 4] elaborates on the corresponding subfigures of [Fig. 1], this time displaying the distributions of the GHG emissions of HT binary plants throughout their life cycle phases.

Fig. 4 Distributions of reported GHG emissions of HT binary life cycle phases (a) construction, (b) operation, and (c) end of life, disaggregated by dataset.

[Fig. 4a] displays the distributions of the construction phase GHG emissions of HT binary plants, which have declined by 13% from 15 g CO₂ eq/kWh (Pre-2017) down to 13 g CO₂ eq/kWh. Despite these values being close to one another, the interquartile range of the “Post-2017” dataset contains larger values than the “Pre-2017” dataset and also has a larger range, reaching a maximum of 58 g CO₂ eq/kWh. The “Combined” dataset reports a median of 14 g CO₂ eq/kWh emitted by HT binary plants during their construction phase.

[Fig. 4b] displays distributions of the operational-phase GHG emissions of HT binary geothermal plants, which has been stable at 1 g CO₂ eq/kWh. Despite these values being close to one another, the interquartile range of the “Pre-2017” dataset is approximately double that of the “Post-2017” dataset, and overall maximum value exceeds that of the “Post-2017” dataset as well, leading to a larger range as well. The “Combined” dataset reports a median of 1 g CO₂ eq/kWh emitted by HT binary plants during their operational phase.

[Fig. 4c] displays the distributions of end-of-life emissions of HT binary plants, the “Pre-2017” dataset reporting a median value of 0.06 g CO₂ eq/kWh, and the “Post-2017” dataset reporting a median of 0.10 g CO₂ eq/kWh. In comparison to the distribution of the “Post-2017” dataset, that of the “Pre-2017” dataset is minuscule, consisting only of data values very close to its median. The “Post-2017” dataset, on the other hand, has values as high as 8 g CO₂ eq/kWh, greatly exceeding the median of each. The “Combined” dataset reports a median of 0.08 g CO₂ eq/kWh for GHG emissions by HT binary geothermal plants during their end-of-life phase.

Discussion

Regarding the distributions of estimated lifetime GHG emissions from each geothermal technology in the “Combined” dataset, HT flash reports the highest median GHG emissions, at 43 g CO₂ eq/kWh, followed by EGS binary with a median of 36 g CO₂ eq/kWh, then HT binary with a median of 30 g CO₂ eq/kWh. Despite slight variations in values, this ranking remains consistent with that of Eberle et al.[5] Similarly, this can be explained by the fact that the geothermal water in a binary-cycle plant remains in a closed-loop, meaning that no noncondensable gases can escape into the atmosphere. However, this is not the case in a flash plant, where no condensable gases are emitted after passing through the turbine.[5] This is supported by the fact that during its operational phase, HT flash has higher GHG emissions than both EGS and HT binary. The amount of emitted GHG is strongly dependent on geochemical properties of the geothermal brain, explaining the significant spread in [Fig. 4b].[10] Between EGS binary and HT binary plants, the greater GHG emissions attributed to EGS can be explained by the fact that a large portion of emissions associated with EGS comes from the fracturing of the reservoir, unlike HT, which makes use of naturally occurring reservoirs below the earth’s surface. This is supported by the vast emissions that come from the construction phase of an EGS binary plant, in contrast with far lower numbers associated with HT binary construction (despite still having higher construction emissions than HT flash).

Among all three geothermal technologies seen in [Figs. 2–4], it can be noted that median construction phase GHG emissions decrease over time. This trend suggests improved construction and drilling methods, which have, in general, lowered emissions of geothermal plants during this early phase of their life.[11] The construction phase strongly benefits from the learning curve of the past decades as well as from supply chain optimization. In particular, evolving modular design can help to lower emission, and improve feasibility over traditional bespoken power plants.[12] In terms of drilling, several technologies have been recently deployed or are currently under development such as high-frequency electromagnetic drilling, electric pulse drilling, and directional drilling.[13] [14] [15] [16] [17] On top of this, operational emissions of each geothermal technology have remained fairly consistent between the “Pre-2017” and “Post-2017” datasets. This trend can be explained by the fact that carbon capture technologies are currently one of the few impactful ways of lowering operational emissions of geothermal plants.[18] [19] However, these technologies are still largely in development, meaning that there are likely not enough geothermal plants with these advanced technologies to see a significant decrease in operational GHG emissions, overall.

One notable difference between the distribution of emissions using data published since 2017 and in previous studies, is the addition of negative emissions values in the end-of-life phase of EGS binary geothermal plants. These values are associated with Li et al.,[20] which includes the positive environmental impacts of recycling significant quantities of metals like steel, aluminum, and copper from production and injection pumps during the end-of-life phase of EGS binary geothermal plants, in order to mitigate some of the lifetime GHG emissions associated with the plant.[20]

Additionally, it can be seen in [Fig. 3] that in the construction and operational life cycle phases of HT flash plants, the “Post-2017” dataset possesses one value for GHG emissions, far above any presented in the “Pre-2017” dataset. This data point represents the Bagnore 3 and 4 geothermal plant in Italy, a large plant consisting of one single flash plant and an Organic Rankine Cycle (ORC), culminating in an installed capacity of 61 MW of electricity.[21] The irregularly high emissions of the Bagnore plant are the result of an extraordinary concentration of noncondensable gases, which consist of 80% of CO₂ by weight.[22] During Bagnore’s operational phase, on top of the atmospheric emissions of noncondensable gas that comes with the flash system, the plant also loses organic fluid in its ORC, taking the shape of pentane atmospheric emissions.[23] The two data points representing the Bagnore plant come from Tosti et al. and Zuffi et al.[21] [23] They present total lifetime GHG emissions values of 682 g CO₂ eq/kWh and 855 g CO₂ eq/kWh, respectively.

[Fig. 5] displays the summary box plots for GHG emissions of EGS binary, HT flash, and HT binary as reported by the “Combined” dataset developed in this report, compared with those of various other electricity-generation technologies, as reported by the U.S. Department of Energy.[24]

Fig. 5 GHG emission “Combined” statistics of geothermal technologies depicted in yellow, compared with other electricity generation technologies depicted in blue, sourced from.[24] *Please note that biopower has a reported minimum GHG emissions value of -1000 g CO₂ eq/kWh, which reaches below the bottom of the figure.

GHG emissions estimates of 36, 43, and 30 g CO₂ eq/kWh for EGS binary, HT flash, and HT binary, respectively, all lie far below the median estimates for GHG emissions of natural gas, oil, and coal, at 486, 840, and 1001 g CO₂ eq/kWh, respectively.[23] GHG median estimates for these geothermal technologies also fall below those of other renewable energy sources such as biopower and photovoltaics, at 52 and 43 g CO₂ eq/kWh, respectively.[24] However, these estimations for geothermal fall above median GHG emissions estimations for other forms of renewable energy, including concentrating solar power, hydropower, land-based wind, and offshore wind which have values of 28, 21, 12, and 19 g CO₂ eq/kWh, respectively.[24]

Conclusions and Outlooks

In this study, a systematic review of GHG emissions related to geothermal power was conducted, in which 169 geothermal environmental impact studies were evaluated for use in a dataset comprised entirely of studies published since 2017, and for use in tandem with the dataset established in Eberle et al. Of these studies, 24 passed the rigorous screening methodology established in Eberle et al., while also providing estimates for GHG emissions of geothermal electricity plants of varying technologies, including at least one of: EGS binary, HT flash, or HT binary.

Through this review and analysis, it was found that between EGS binary, HT flash, and HT binary geothermal technologies, HT flash has the highest associated GHG emissions with a median of 43 g CO₂ eq/kWh, followed by EGS binary with a median of 36 g CO₂ eq/kWh, and lastly HT binary with a median of 30 g CO₂ eq/kWh. HT flash having the highest level of emissions can be attributed to the fact that the technology does not generate power in a closed loop, thereby emitting noncondensable gases into the atmosphere. EGS binary having higher emissions than HT binary can be attributed to the higher construction emissions that come with EGS over HT, due to the hydraulic stimulation required to produce an artificial reservoir. Moreover, median construction phase GHG emissions appeared to have decreased for all technologies between the “Pre-2017” and “Post-2017” dataset, indicating improvements in construction and drilling methods to lower emissions. However, operational-phase emissions remained consistent for all three technologies between the two datasets, indicating the difficulties in reducing emissions during this stage of a geothermal plant’s life.

Outlook

Similarly to Eberle et al., this work has several limitations and there are many opportunities for further research on the subject. For one, geothermal technologies are not limited to those analyzed in the paper, and there are several emerging technologies, which will soon be accompanied by data on GHG emissions as research progresses, and as more of these plants are constructed and operated. Emerging geothermal technologies include but are not limited to: closed-loop advanced geothermal systems, CPG, magmatic geothermal, and hybrid systems. As these new geothermal technologies (which can generate power from sources beyond current geothermal technologies’ capabilities) develop further, the geothermal energy landscape will change drastically and reliable energy generation will be possible in many more locations than before. Moreover, with the development of these technologies, analysis of similar nature as this can be conducted to establish further estimates on GHG emissions of these respective technologies contributing to geothermal electricity production as a whole.

Additionally, similarly to Eberle et al., geothermal energy as a heat source was not discussed in this report, but further systematic review could be conducted to establish estimates on the GHG emissions associated with this technology as well.

Beyond the different geothermal technologies, further analysis could also be conducted using environmental impact categories other than GHG emissions, such as human toxicity and the emissions of hydrogen sulfide often associated with geothermal energy production. Another extension could be the study of water loss in “dry” systems such as closed loop or EGS.[25] [26] Furthermore, the mineral extraction of the geothermal brain and its added-value to circular economy could be analyzed.[27]

Finally, the lack of consistency between parameters like location, impact assessment method, lifetime estimate, size, depth, and temperature also lead to inconsistencies among estimates from each of these individual studies. This presents an opportunity for further analysis of this data using harmonization methods, including but not limited to finding variations in the proportion of overall emissions coming from each life cycle phase, or normalizing certain parameters to one another in analysis.

Contributorsʼ Statement

S.H. provided investigation, ideation, conceptualization, writing – original draft, and writing – reviewing and editing. M.W. provided resources, supervision, and writing – reviewing and editing.

Conflict of Interest

The authors declare that they have no conflict of interest.

• References

• 1 Almonte MP. Vulnerability in the Arctic in the Context of Climate Change and Uncertainty. The Arctic Institute - Center for Circumpolar Security Studies; https://www.thearcticinstitute.org/vulnerability-arctic-context-climate-change-uncertainty/ (accessed July 2025)
• 2 WWF. Climate Change. WWF Arctic; https://www.arcticwwf.org/threats/climate-change/ (accessed July 2025)
• 3 D’Odorico P, Bhattachan A, Davis KF, Ravi S, Runyan CW. Adv Water Resour 2013; 51: 326-344
• 4 U.S. EIA. Geothermal Explained. Eia.gov; https://www.eia.gov/energyexplained/geothermal/ (accessed July 2025)
• 5 Eberle A, Heath G, Nicholson S, Carpenter A. Systematic Review of Life Cycle Greenhouse Gas Emissions from Geothermal Electricity. National Renewable Energy Laboratory; 2017. https://docs.nrel.gov/docs/fy17osti/68474.pdf
• 6 NREL. Geothermal Electricity Production Basics | NREL. Nrel.gov. https://www.nrel.gov/research/re-geo-elec-production (accessed July 2025)
• 7 U.S. DOE. Electricity Generation. Energy.gov; https://www.energy.gov/eere/geothermal/electricity-generation (accessed July 2025)
• 8 U.S. DOE. What Is an Enhanced Geothermal System (EGS)?. U.S. Department of Energy; 2016. https://www.energy.gov/sites/prod/files/2016/05/f31/EGS%20Fact%20Sheet%20May%202016.pdf
• 9 Paulillo A, Cotton L, Law R, Striolo A, Lettieri P. J Clean Prod 2020; 249: 119410
• 10 Stober I, Bucher K. Geothermal Energy. Cham: Springer International Publishing; 2021.
• 11 Isania F, Elahifar B. Geoenergy Sci Eng 2024; 246: 213638-213638
• 12 Chege P, Bardarson G, Richter A. GRC Trans 2017; 41
• 13 Brantson ET, Adu-Awuku J, Asase W. et al. J Geophys Eng 2025; 22 (05) 1245-1265
• 14 Tian J, Dheyaa A. Results Eng 2025; 25: 104416
• 15 Zhu X, Liu S, Liu W, Zhou X, Tang W. Plasma Phys. 2024 64. 10
• 16 Zhang Q, Wang G, Pan X. et al. Energies 2023; 16 (03) 1174
• 17 Lu C, Zhang T. J Appl Geophys 2020; 174: 103946-103946
• 18 Randolph JB, Saar MO, Bielicki J. Energy Procedia 2013; 37: 6625-6635
• 19 Karlsdottir MR, Heinonen J, Palsson H, Palsson OP. Geothermics 2020; 84: 101727
• 20 Li J, Ricardo R, Gallego-Schmid A, Stamford L. Sci Total Environ 2024; 907: 167843-167843
• 21 Tosti L, Ferrara N, Basosi R, Parisi ML. Energies 2020; 13 (11) 2839
• 22 Atkinson P, Celati R, Corsi R, Kucuk F, Ramey HJ. Geothermics 1978; 7 (2/4) 185-208
• 23 Zuffi C, Manfrida G, Asdrubali F, Talluri L. Geothermics 2022; 104: 102434
• 24 Wind Energy Technologies Office. How Wind Can Help Us Breathe Easier. Energy.gov. https://www.energy.gov/eere/wind/articles/how-wind-can-help-us-breathe-easier (accessed in July 2025)
• 25 Duchane DD. Geothermal Energy Production from Hot Dry Rock: Operational Testing at Tee Fenton Hill, New Mexico HDR Test Facility. Los Alamos National Laboratory; 1994. Vol. LA-UR-93-3695
• 26 Brown DW. Thirty-Fourth Workshop on Geothermal Reservoir Engineering. Stanford, California: Stanford University; 2009. Vol. SGP-TR-187
• 27 Osvald M, Kilpatrick AD, Rochelle CA, Szanyi J, Medgyes T, Kóbor B. Geofluids 2018; 2018: e6509420
• 28 Paulillo A, Kim A, Mutel C, Striolo A, Bauer C, Lettieri P. Cleaner Environ Syst 2021; 3: 100054
• 29 Paulillo A, Kim A, Mutel C, Striolo A, Bauer C, Lettieri P. Cleaner Environ Syst 2022; 6: 100086
• 30 Pratiwi A, Ravier G, Genter A. Geothermics 2018; 75: 26-39
• 31 Strojny M, Gładysz P, Andresen T, Pająk L, Starczewska M, Sowiżdżał A. Energies 2024; 17 (09) 2077
• 32 Wang Y, Du Y, Wang J, Zhao J, Deng S, Yin H. Resour Conserv Recycl 2020; 155: 104670
• 33 Basosi R, Bonciani R, Frosali D, Manfrida G, Parisi ML, Sansone F. Sustainability 2020; 12 (07) 2786
• 34 Díaz-Ramírez M, Jokull S, Zuffi C, Mainar-Toledo MD, Manfrida G. Energies 2023; 16 (09) 3616-3616
• 35 González-García H, Francke H, Göllner-Völker L. et al. Geothermics 2022; 104: 102471
• 36 Karlsdottir MR, Heinonen J, Palsson H, Palsson OP. Energies 2020; 13 (12) 3187
• 37 Kjeld A, Bjarnadottir HJ, Olafsdottir R. Geothermics 2022; 105: 102530
• 38 Mainar-Toledo MD, Díaz-Ramírez M, Egilsson SJ, Zuffi C, Manfrida G, Leiva H. Sustainability 2023; 15 (18) 13943-13943
• 39 Paulillo A, Striolo A, Lettieri P. Environ Int 2019; 133: 105226
• 40 Pratama MR, Muthia R, Purwanto WW. Carbon Neutrality 2023; 2 (01)
• 41 Rossi F, Zuffi C, Parisi ML, Fiaschi D, Manfrida G. J Clean Prod 2023; 405: 136931
• 42 Dawo F, Fleischmann J, Kaufmann F. et al. Appl Energy 2021; 288: 116661
• 43 Lohse C. Renew Energy 2018; 128: 509-519
• 44 Menberg K, Heberle F, Bott C, Brüggemann D, Bayer P. Renew Energy 2021; 167: 20-31
• 45 Menberg K, Heberle F, Uhrmann H. et al. Renewable Energy 2023; 218: 119251
• 46 Zuffi C, Ungar P, Fiaschi D, Manfrida G, Batini F. Sustainability 2024; 16 (11) 4622

Publication History

Received: 03 November 2025

Accepted after revision: 26 December 2025

Article published online:
22 January 2026