Gas Emissions and Subsurface Architecture of Fault‐Controlled Geothermal Systems: A Case Study of the North Abaya Geothermal Area

East Africa hosts significant reserves of untapped geothermal energy. Exploration has focused on geologically young (<1 Ma) silicic calderas, yet there are many sites of geothermal potential where there is no clear link to an active volcano. The origin and architecture of these systems are poorly understood. Here, we combine remote sensing and field observations to investigate a fault‐controlled geothermal play located north of Lake Abaya in the Main Ethiopian Rift. Soil gas CO2 and temperature surveys were used to examine permeable pathways and showed elevated values along a ∼110 m high fault, which marks the western edge of the Abaya graben. Ground temperatures are particularly elevated where multiple intersecting faults form a wedged horst structure. This illustrates that both deep penetrating graben bounding faults and near‐surface fault intersections control the ascent of hydrothermal fluids and gases. Total CO2 emissions along the graben fault are ∼300 t d−1; a value comparable to the total CO2 emission from silicic caldera volcanoes. Fumarole gases show δ13C of −6.4‰ to −3.8‰ and air‐corrected 3He/4He values of 3.84–4.11 RA, indicating a magmatic source originating from an admixture of upper mantle and crustal helium. Although our model of the North Abaya geothermal system requires a deep intrusive heat source, we find no ground deformation evidence for volcanic unrest or recent volcanism along the graben fault. This represents a key advantage over the active silicic calderas that typically host these resources and suggests that fault‐controlled geothermal systems offer viable prospects for geothermal exploration.

volatiles but no surface volcanism. These areas are associated with tectonic faulting and include the Natron and Magadi basins at the Kenya-Tanzania border (Lee et al., , 2017Muirhead et al., 2016) and the Habilo area NW of Fantale in the Main Ethiopian Rift (MER) (Hunt et al., 2017). To date, few studies have investigated the architecture and potential of these fault-controlled geothermal systems. However, when compared to the silicic calderas (which pose considerable volcanic eruption hazards that could lead to damage and disruption of geothermal infrastructure, Fontijn et al., 2018;Clarke et al., 2020;Tierz et al., 2020), these fault-controlled geothermal areas are potentially much lower risk and therefore safer options for exploration, investment and development.
Here, we bring together new remote sensing and gas emission data from the North Abaya geothermal area in the MER, which is a fault-controlled geothermal play that shows no surface volcanic edifice. North Abaya is consistently highlighted as one of Ethiopia's key geothermal prospects (Burnside et al., 2021), and although previous studies have investigated regional volcano-tectonic activity (Corti et al., 2013;Ogden et al., 2021) and documented surface geothermal manifestations (i.e., hot springs and fumaroles, Craig et al., 1977;Chernet, 2011;Minissale et al., 2017) there is no conceptual model to understand the heat source, fluid flow, gas emissions and geothermal potential of the resource. We show that a large offset graben bounding fault plays a key role in channeling gas and hydrothermal fluids toward the surface, and while there is no evidence for volcanism along the fault, a deep magmatic heat source is still required. Gas emissions along this fault zone are comparable to sites of proven geothermal resources in Ethiopia (e.g., the silicic caldera of Aluto) and while there are fundamental differences between the North Abaya geothermal play and the silicic calderas, our conceptual model suggests that there is great potential for further exploration and development.

The Main Ethiopian Rift (MER)
The MER (Figure 1) is the northernmost segment of the EARS and accommodates E-W extension between the Nubia and Somalia Plates (Corti, 2009). The region is extending at 4-6 mm yr −1 (e.g., Saria et al., 2014) and this is accommodated by both faulting and magmatic intrusion (Corti et al., 2013;Ebinger, 2005;Keir et al., 2006). The MER is subdivided into northern (NMER), central (CMER), and southern (SMER) segments and there is a broad consensus that rift maturity decreases southwards (Agostini et al., 2011). One of the most fundamental differences is the style and location of volcano-tectonic activity. In the NMER, border faults, with large vertical offsets >100 m, define the boundaries of the rift but are largely abandoned (Casey et al., 2006;Keir et al., 2006;Wolfenden et al., 2004). Active seismicity and magmatic intrusion in the NMER is instead focused along the axis of the rift (Ebinger & Casey, 2001;Keir et al., 2006;Kendall et al., 2005) in a region commonly referred to as the Wonji Fault Belt (e.g., Boccaletti et al., 1999;Gibson, 1969;Mohr, 1967). In the SMER, the focus of this study, geological and geophysical data show very different patterns and indicate that border faults still accommodate significant extension (Corti et al., 2013;Kogan et al., 2012;Philippon et al., 2014). Recent volcanism is colocated with border faults along the rift margins (Corti et al., 2013;Rooney et al., 2011), again, contrasting with the focused axial magmatism observed in more northerly segments of the MER.
Bimodal volcanism due to rift-related magmatic intrusion is abundant throughout the MER. Mafic lava flows and scoria cone fields are abundant, as are highly evolved peralkaline rhyolitic complexes (Gibson, 1969;Hutchison et al., 2015: Hunt et al., 2019. Primitive mantle-derived melts are stored at depths of ∼15 km beneath the surface where they form dyke complexes and undergo mafic fractionation (Iddon & Edmonds, 2020). Such dykes are thought to then undergo either rapid transition to the surface where they erupt as monogenetic cinder cones (Mazzarini et al., 2013;Rooney et al., 2011) or are focused toward shallow (∼5 km deep) silicic magma bodies where they undergo more extensive crystal fractionation to form trachytic and rhyolitic melts (Iddon et al., 2019;Peccerillo et al., 2003;Rooney, Hart, et al., 2012). These erupt to form a thick pile of coalescing rhyolitic lava flows and domes, pumice cones, and pyroclastic deposits (Hunt et al., 2019;Hutchison et al., 2015Hutchison et al., , 2016.

Volcanic and Tectonic Features of the North Abaya Geothermal Area
The North Abaya geothermal area is located in the Soddo area of the SMER (Corti et al., 2013; Figures 1 and 2a). Here, like many areas of the MER, surface geology is typified by NE-SW trending faults and bimodal volcanism (primarily mafic scoria cone fields and larger, 3-15 km wide, silicic centers and calderas). Mapping of regional fault structures by Chernet (2011) and Corti et al. (2013) identified a prominent network of right-stepping en echelon normal or oblique faults, with vertical offsets generally <100 m, and lengths of 1-10 km. Corti et al. (2018) showed that rift architecture in this sector of the SMER is asymmetric, with the eastern side of the rift characterized by a single large offset border fault (the Agere Selam escarpment), while the western side is characterized by the aforementioned closely spaced, small offset normal faults. Corti et al. (2018) described the western margin in the Soddo area as a rift-ward dipping monocline where a dense concentration of boundary faults results in a staggered uplift toward the rift flank over 10-20 km (Figure 2b; Corti et al., 2013). Across this rift border zone, where our study area is located, three prominent graben structures are observed and from west to east these are the Salewa Dore-Hako graben, the Abaya graben and the Chewkare graben (Figure 2c and Section 4.1). Tectonic activity in this region has occurred through the Late Pleistocene-Holocene (Corti et al., 2013) and there is ongoing seismicity associated with this border fault zone (Ogden et al., 2021).
Recent volcanism is concentrated in the Salewa Dore-Hako Graben and includes isolated mafic scoria cones and the 3.5 km wide, 250 m high Salewa Dore-Hako rhyolite complex (Figure 2a). Both the mafic and silicic vents show alignment with nearby graben faults, indicating important fault controls on magma ascent (noted previously by Corti et al., 2013, andat other Ethiopian rift volcanoes, cf. Hutchison et al., 2015;Hunt et al., 2020). Although robust age constraints on volcanism in North Abaya are currently lacking, the colocation of fumaroles at mafic and silicic vents attests to very recent, most likely Holocene, eruptive activity probably coincident with episodes of faulting (Corti et al., 2013).
Various geothermal features have been identified in the North Abaya area (summarized in Figure 2a). Of these the most vigorous thermal manifestations (hot springs and fumaroles with temperatures up to ∼95°C, Chernet, 2011) are found 1-2 km north of Lake Abaya along a major fault that bounds the Abaya graben (herein referred to as West Abaya graben fault (WAGF), Figure 2b). Steam vents have also been observed at the summit of the Salewa The MER is divided into northern, central, and southern segments (NMER, CMER, SMER) and fault structures (modified after Agostini et al., 2011) are shown as blue lines. Volcanic centers are shown by yellow triangles, lakes are shaded blue and large settlements are shown by black squares. The Abaya volcanic field is located within the white rectangle. The globe inset shows major plate boundaries in red and the region covered by the DEM as a blue square.

Remote Sensing
To map volcanic landforms and tectonic structures in North Abaya, we used a 12.5 m digital elevation model (DEM) from the Japanese ALOS satellite (ALOS PALSAR). These data were combined with previous fault data bases of Agostini et al. (2011) and Corti et al. (2013). Ground deformation is frequently observed at geothermal sites in Ethiopia (e.g., Biggs et al., 2011;Birhanu et al., 2018;Hutchison et al., 2016;Lloyd et al., 2018b) and for Abaya, we evaluated this using satellite radar interferometry (InSAR). Recently, Albino and Biggs (2021) used Sentinel-1 data provided by the European Space Agency (ESA) to generate an InSAR time series along the entire EARS for the period 2015-2020. Here we explore a subset of the data from North Abaya and for full details of the processing, the reader should refer to Albino and Biggs (2021).

Soil CO 2 Flux and Temperature Surveys
Soil CO 2 flux and temperature were measured in January-February 2019. These months are the driest period in Ethiopia and provide the most stable meteorological and environmental conditions (i.e., there was no significant precipitation nor changes in surface vegetation during our study). The objectives of our survey were first, to transect major rift faults in the SE of the study area and second, to generate detailed maps of soil degassing along the WAGF and the flank of the SDHRC (the main volcanic edifice in the study area, Figure 2a). The typical spatial resolution of our sampling was 50 m and in total 757 measurements were made. CO 2 flux was measured via the accumulation chamber technique (Chiodini et al., 1998;Parkinson, 1981). We used a West systems flux meter with an inbuilt LICOR-LI820 CO 2 concentration sensor and the flux measurement was based on the rate of CO 2 increase in the chamber over a 2-min measurement period. Repeat measurements were typically within 10%-25% and showed better precision in high flux zones. Several low-and high-flux sites were revisited throughout the duration of the campaign, and we found that repeated measurements at the same site were within the above measurement uncertainty. These values are similar to previous geothermal studies (e.g., Hutchison et al., 2015) and are  typical of the instrument reproducibility and natural variations in emission rates (Carapezza & Granieri, 2004;Chiodini et al., 1998;Giammanco et al., 2007;Viveiros et al., 2010).
At each station, we measured soil temperature using a Digi-Sense Type K thermocouple probe and an Oakton Temp 10 Thermocouple Thermometer inserted into 50 cm soil depth. To create the measurement hole, we used a sledgehammer to drive a 50 cm metal bar into the ground before inserting the thermocouple. Note that in all cases CO 2 flux was measured before the 50 cm hole was made and since penetration through the soil causes frictional heat, the probe was left in place until a stable temperature reading was obtained. In some locations the 50 cm depth could not be reached (due to stones or bedrock) and so the depth of the probe was recorded as well as the temperature (Data Set S1).
In areas of volcanic-hydrothermal gas emissions, there are usually multiple sources of CO 2 (i.e., magmatic gases variably mixed with biogenic and terrestrial background). Chiodini et al. (1998) applied a graphical statistical approach (originally by Sinclair, 1974) to demonstrate that volcanically influenced degassing areas often show bimodal CO 2 flux populations, and that these populations can be distinguished using the inflection points on logarithmic probability plots. In our study, we used probability plots and the inflection points to identify different background and volcanic-hydrothermal populations in both our temperature and CO 2 flux data sets (Section 4.3). To produce maps of CO 2 flux and temperature, we used a sequential Gaussian simulation (sGs) method (Cardellini et al., 2003). These methods allow the user to undertake hundreds of realizations of the survey grid and are particularly useful for CO 2 because they constrain the uncertainty in the total flux. sGs methods require a high sampling density and hence were attempted only on the Western flank of the Abaya Graben and the flank of the SDHRC. Three hundred sGs were performed using the sgsim simulation tool (Deutsch & Journel, 1998) in the Stanford Geostatistical Modeling Software (SGeMS) package (Remy et al., 2009). To generate maps, we calculated the arithmetic mean of each individual cell across all simulations and for CO 2 , we calculated the total flux of each simulation and used this to compute the mean and standard deviation of all simulations and assess total CO 2 emission and its uncertainty.

Bulk Gas Chemistry and Carbon Isotopes
Dry gas samples from gas-rich springs and fumaroles were collected in 9 ml preevacuated tubes. These samples were then analyzed for bulk chemistry and C isotopes (δ 13 C) at the Department of Earth and Planetary Sciences at the University of New Mexico (UNM). Gas chromatography (GC) and quadrupole mass spectrometry (QMS) were used to measure CH 4 -CO 2 -H 2 -CO and Ar-He-N 2 -O 2 concentrations, respectively. The analytical setup used at UNM is identical to that of Lee et al. (2017) and is described in detail therein. Experimental errors for the GC and the QMS are <2 and <1% respectively. Due to the collection of samples in glass vials with rubber septa, He and H 2 are likely to rapidly diffuse out of the vials and results for these gas species are not representative of the composition of the gas discharge.
Samples with the highest concentrations of CO 2 were selected for δ 13 C analysis using a Thermo Scientific Delta Ray Infrared Spectrometer. The samples were diluted using the capillary dilution system provided by the manufacturer and introduced into the inlet of the instrument through a needle and capillary. In order to compensate for pressure decrease during analyses, a pure N 2 gas was connected to the vial. Calibration was performed prior to and following the analyses with a commercially available calibration gas and all CO 2 -δ 13 C measurements are shown in delta notation as per mil values (δ‰) relative to Vienna Pee Dee belemnite (VPDB). Our measurements are characterized by a δ 13 C standard error of ±0.1‰.

Helium Isotopes
For helium isotope measurements gases were collected in Cu-tubes from moderate to high temperature (60°C-95°C) bubbling springs and fumaroles. At each locality, samples were collected using (3/8-inch) Cu-tubes connected at one end with Tygon tubing fitted with an inverted funnel, which was inserted into the source of gas manifestation (e.g., Kennedy et al., 1985;Weiss, 1968). The other end of copper tube was fitted with a second section of Tygon tubing, a second copper tube and a third section of Tygon tubing, which was submerged in water to ensure a one-way flow of gas through the sampling apparatus, thus minimizing air contamination. Approximately 1-2 hr were taken to flush the entire sampling apparatus before sealing both ends of the copper tubes using specially designed stainless-steel clamps that create a cold-weld in the Cu-tubing, thus sealing sample gas inside the tube for transport to the laboratory.
Noble gas isotope analysis was conducted using a dual mass spectrometer setup, interfaced to a dedicated extraction and purification system at the University of Oxford (Barry et al., 2016). In brief, gases collected in Cu-tubes were transferred to the extraction and purification line where reactive gases were removed by exposing gases to a titanium sponge held at 950°C. The titanium sponge was cooled for 15 min to room temperature before gases were expanded to a dual hot (SAES GP-50) and cold (SAES NP-10) getter system, held at 250°C and room temperature, respectively. A small aliquot of gases was segregated for preliminary analysis on a quadrupole mass spectrometer. He and Ne isotopes were then concentrated using a series of cryogenic traps; heavy noble gases (Ar-Kr-Xe) were frozen down at 15 K on a stainless-steel finger and the He and Ne were frozen down at 19 K on a cold finger filled with charcoal. The temperature on the charcoal finger was then raised to 34 K to release only He, which was inlet into a Helix SFT mass spectrometer. Following He analysis, the temperature on the charcoal cryogenic trap was raised to 90 K to release Ne, which was inlet into an ARGUS VI mass spectrometer. Uncertainties on helium isotopes and He/Ne values are less than 3%.

Tectonics and Recent Volcanism
Fault structures in the North Abaya study area are shown in Figure 2. We mapped three major graben structures (the Salewa Dore-Hako graben, the Abaya graben and Chewkare graben) as well as regional faults that define an overall NNE-SSW trend. Graben bounding faults show the greatest displacement ( Figure 2c) with the WAGF marked by the largest vertical offset of ∼110 m. Overall, the North Abaya faults display a right-stepping en echelon pattern (Corti et al., 2013), which leads to various intersecting fault zones. This is particularly well developed at the southern end of the WAGF where a ∼100 m high wedge-shaped horst block is observed ( Figure 2b). We refer to this area as the Abaya horst and a field photograph from the west of this structure looking south shows that this is a site of fumarolic activity where hydrothermal upwelling has led to surface alteration and the formation of bright red clays. On the ground, these fumaroles display a WNW-ESE alignment, which suggests that these vents are linked to WNW-ESE structures orthogonal to the WAGF. Importantly, we found field evidence for the existence of such faults along the shore of Lake Abaya, where a WNW-ESE fault with a throw of 2-3 m was observed ( Figure 3b). We suggest that although major NNE-SSW rift-aligned tectonic faults accommodate the bulk of extension, their en echelon fabric creates numerous intersecting fault sets that may develop into highly fractured "gridded" fault zones (as seen in the Abaya horst).
Our mapping of volcanic vents shows that these are mainly located in the Salewa Dore-Hako graben (in agreement with previous studies, Corti et al., 2013). Mafic scoria cones define a ∼20 km long NNE-SSW trend, while silicic vents are focused at the ∼5 km long ∼N-S oriented SDHRC which comprises overlapping obsidian coulees. The overall NNE-SSW alignment of the mafic vents suggests a feeder dyke of similar orientation (Corti et al., 2013), while the more chemically evolved silicic volcanism is indicative of an upper crustal (∼5 km deep) magma reservoir as observed elsewhere in the rift (e.g., Aluto, Hutchison et al., 2016;Gleeson et al., 2017, Iddon et al., 2019. Fumaroles are observed at the SDHRC, but are much weaker than those observed along the WAGF (Figure 3a).

Ground Deformation
The results of a 2015-2020 Sentinel-1 InSAR survey for the North Abaya region are shown in Figure 4. The map shows the mean line of sight velocity (in cm yr −1 ) relative to a reference area that is located >30 km east of the study area and well distanced from any volcanic or tectonic features. Deformation rates in the North Abaya geothermal area are on the order of 0 to −0.5 cm yr −1 . Albino et al. (2022) investigated limits of detection in this Sentinel-1 time series and demonstrated that linear deformation rates must be greater than 0.5 cm yr −1 to be detected over this 5-year period. At North Abaya, deformation rates ( Figure 4) clearly fall . Ground deformation north of Lake Abaya during the period 2015-2020. The results are shown as mean line of sight velocity in cm/yr, relative to a representative reference area (labeled REF) that is well distanced from any volcanic and/or tectonic features (and assumed not to be deforming). Note that SDH indicates the Salewa Dore-Hako rhyolitic complex. Volcanic ground deformation usually results in uplift-subsidence of at least ±2 cm/yr. In this case, no such deformation signals are detected. below the limits of detection and are negligible when compared to uplift/subsidence signals that typify other East African volcanoes that host geothermal resources (typically >2 cm yr −1 , Albino & Biggs, 2021). Thus, our data show that there was no significant deformation in the North Abaya study area during the 2015-2020 period.

CO 2 Degassing and Soil Temperatures
Maps of CO 2 flux and soil temperatures are shown in Figures 5 and 6, respectively. CO 2 flux ranged from 0.2 to 6,020 g m −2 d −1 and showed elevated values along the WAGF, with the highest values observed around the northern wedge of the Abaya horst. A few elevated CO 2 flux values were observed at the SDHRC and along graben bounding faults, but within the center of the grabens, fluxes were low ( Figure 5). Soil temperatures ( Figure 6) were between 18.3°C and 98.5°C. They also show high values associated with the WAGF, but unlike CO 2 , elevated temperatures were only observed at the northern wedge of the Abaya horst rather than along the length of the fault. Temperatures were low (<45°C) on the SDHRC except for some weak fumaroles located on the top of the complex.
To evaluate the existence of different CO 2 and temperature sources, we examined probability plots (Figure 7). It is expected that when data consist of a single log-normal population, this will plot as a straight line, and when there are multiple log-normal populations, these will plot as curves of overlapping populations defined by inflection points (Chiodini et al., 1998). Our CO 2 flux and soil temperature data show clear inflection points at values of 28.2 g m −2 d −1 and 40°C, respectively. We interpret these two populations as (a) a magmatic-hydrothermal source (associated with high temperatures and high CO 2 flux) and (b) a background source (associated with low temperatures and low CO 2 flux, and most likely derived from biogenic and/or deep magmatic/mantle sources). In Figures 8 and 9 we show transects along and across the major faults and volcanic areas of the study area, and we use the upper value of the background population to help identify areas of significant magmatic-hydrothermal input.
Before looking at transects in detail, it is important to point out two notable features of the background population. First, while background CO 2 flux at North Abaya (0.2-28.2 g m −2 d −1 ) is within the range of typical nonvolcanically influenced soil (10-30 g m −2 d −1 , Mielnick & Dugas, 2000;Rey et al., 2002;Cardellini et al., 2003) it is higher and more variable than other sites in the MER (i.e., 0.5-6.0 g m −2 d −1 at 10-20 km distance from Aluto volcano, Hutchison et al., 2015). Second, within the background CO 2 flux population, we noted minor inflections at 12.6 and 2.0 g m −2 d −1 (Figure 7a). The first feature is explained by the fact that North Abaya is much more vegetated than the area surrounding Aluto (the former is adjacent to a large lake and within a major river catchment) and therefore the soil is richer in organic material and hence biogenic CO 2 flux is almost certainly higher. The second feature, regarding multiple minor inflections in the log-probability plot, suggests that there may be several background CO 2 flux populations. Interestingly, some of the more elevated background values do appear to be associated with rift aligned faults (e.g., the red labeled fault in Figure 9, C-C' shows a CO 2 flux 26 g m −2 d −1 ). An explanation is that these faults have enhanced permeability and a greater deep magmatic/mantle CO 2 flux that is unrelated to any near surface volcanic-hydrothermal system. We did not collect C isotope samples for these different background sites; therefore, we cannot be certain whether there is a stronger biogenic fingerprint at Abaya, and whether elevated background values associated with faults display a magmatic signature. For completeness, we include these inflection points in the background population in Figures 8 and 9, and we emphasize that while the high temperature, high CO 2 flux magmatic-hydrothermal source is clearly defined, there is undoubtedly a range of biogenic and deep mantle/ magmatic CO 2 sources captured in the background population.
Given this broad categorization of magmatic-hydrothermal and background populations, we can take a more detailed look at spatial variations in soil CO 2 flux and temperature using transects (shown in map view in Figures 5 and 6, and as plots in Figures 8 and 9). Transect A-A'-A" shows CO 2 flux and temperature from south to north along the WAGF (Figure 8). CO 2 flux is variable along the fault but shows highest values at the Abaya horst and a maximum value of 6,020 g m −2 d −1 in the area ∼500 m north of this structure. Maximum CO 2 flux values then show a general northward decrease to more typical background values. It is important to note that between ∼1,500 and ∼2,300 m along the profile we were unable to access the hanging wall of the fault because of surface water, and so the lack of high CO 2 in this area likely reflects a gap in sampling rather than a genuine decrease in CO 2 flux in this section of the fault. Soil temperature also shows highest values at the Abaya horst (up to 98.5°C at the fumaroles in Figure 3a). Notably, between 2,300 and 3,500 m there is elevated CO 2 but only a few temperatures >40°C (Figure 8). This demonstrates that CO 2 flux and soil temperature are not always correlated and therefore CO 2 and hydrothermal steam do not always travel together.
Transect B-B' includes several fault structures in the Salewa Dore-Hako graben and then rises eastward on the flank of the SDHRC (Figure 9). No significant temperature anomalies were detected along this profile. CO 2 flux showed no evidence for elevated values across the intragraben faults but did show several elevated values up to 176 g m −2 d −1 on the volcanic complex. These maximum CO 2 values are an order of magnitude lower than those obtained on the WAGF (Figure 8). The elevated CO 2 flux on the volcanic complex appears to be localized ( Figure 5) and does not show an obvious NNE-SSW (fault controlled) trend, arguing against a tectonic control on CO 2 degassing. Instead, there appears to be a closer relationship between CO 2 flux and elevation, with peak values in CO 2 approximately centered on a topographic high.
Transect C-C' covers the eastern escarpment of the Salewa Dore-Hako graben as well as a regional fault to the west of this ( Figure 5). Although the vertical offset for these faults are broadly comparable (∼20 and ∼15 m for the graben fault and regional fault, respectively) there is a marked contrast in their CO 2 emission with the regional fault showing subtle variation in background values and the graben fault showing a ∼400 m wide zone of elevated values (up to 58 g m −2 d −1 ). This finding suggests that the graben bounding faults provide more permeable conduits for deeper magmatic-hydrothermal gases.
Our high-density survey grid along the WAGF and the flank of the SDHRC allowed us to construct maps of CO 2 flux and temperature using sGs methods ( Figure 10). Omnidirectional variograms of normal scores were computed for each area and the modeled variogram and parameters are shown beside each map. The results mirror the trends in the underlying data (i.e., Figures 5 and 6) and clearly indicate highest CO 2 flux and temperatures along the WAGF, particularly at the wedge of the Abaya graben. For CO 2 we calculated the total flux (shown as mean ± standard deviation) and the key finding is that while the two survey sites cover a similar area (3.8 and 3.9 km 2 ), the total flux along the WAGF is 294 ± 71 t d −1 which is 10× greater than that on the flank rhyolitic complex (29 ± 3 t d −1 ).

Gas Chemistry
Compositions of gas samples are shown in Table 1. Our samples mainly contain N 2 , O 2 and CO 2 and represent air that has been flushed by gases of magmatic-hydrothermal origin. Those samples that are rich in CO 2 (up to 30%-36%) represent the most pristine magmatic gas samples. Minor gas species include He, H 2 , Ar, CH 4 , and CO, which originate from a combination of atmospheric and magmatic sources, as well as reducing reactions in the hydrothermal reservoir (i.e., CH 4 and CO, Agusto et al., 2013;Tassi et al., 2013).
New carbon isotope (δ 13 C) data for CO 2 from Abaya are compared to previous measurements from volcanic-hydrothermal systems across the East African Rift in Figure 11a. Our Abaya data show values from −6.4‰ to −3.8‰ which are almost identical to previous δ 13 C-CO 2 made at the same localities by Minissale Inflection points in the probability plot are used to identify different background and volcanic-hydrothermal populations (see Chiodini et al., 1998). For temperature, there is an obvious inflection at 40°C, which is the upper limit of the background values. For CO 2 the main inflection point is observed at 28.2 g m −2 d −1 and separates background from volcanic-hydrothermal populations. Minor inflections are observed in the background population, and we speculate that these could represent differences between faulted and nonfaulted areas in regions where there is no underlying hydrothermal system (see Section 4.3 for discussion). et al. (2017). Plotting the isotope data versus the reciprocal of CO 2 concentration in the samples reveal a crude triangular array that is usually interpreted as mixing between air, biogenic and magmatic CO 2 (Figure 11a, where end-member δ 13 C values come from Gerlach & Taylor, 1990;Javoy & Pineau, 1991;Macpherson & Mattey, 1994;Sano & Marty, 1995;Darling et al., 1995;Cheng, 1996;Chiodini et al., 2008 andTedesco et al., 2010). The Abaya gas samples were all extracted from fumarole vents and are clearly focused on the magmatic endmember. This contrasts with previous surveys of the Magadi-Natron rift basin Muirhead et al., 2020) and the Aluto geothermal system (Hutchison et al., 2016), which sampled soil gas at both fumarolic and nonfumarolic sites and therefore showed a wider spread of both δ 13 C and CO 2 concentration. Focusing on the magmatic endmember ( Figure 11b) reveals that in the most pristine magmatic gas samples, there is overlap between Abaya and Aluto (currently Ethiopia's only developed geothermal site).
Helium isotopes are shown in Figure 11c. The X-value gives the 4 He/ 20 Ne ratio of the sample relative to that measured in air and therefore gives an indication of how much air has been entrained into the sample. X-values close to 1 are air-dominated, and for our samples X-values are 6.3 and 12.8 implying limited air incorporation. Using the X-values we correct our samples for the presence of atmospheric He isotope values (after Hilton, 1996) and this yields He isotope values (R C /R A ) of 4.4 (Table 1). Again, these values are similar to previous measurements of North Abaya by Minissale et al. (2017), who found R C /R A values of 4.5-7.5 at similar locations along the WAGF. Our helium isotopic compositions are lower than MORB-like values (typically 8 ± 1 Graham, 2002) and  Figure 5). Red horizontal lines denote the maximum value for background soil temperatures and CO 2 degassing, while the blue and black dashed lines represent minor inflections in background CO 2 populations referred to in Section 4.3. CO 2 flux and temperature are greatest around the wedge of the Abaya horst (i.e., toward the southern extent of the Western Abaya graben fault). Note that between ∼1,500 and ∼2,300 m along the profile, we were unable to access the hanging wall of the fault because of surface water. The lack of high CO 2 in this area likely reflects a gap in sampling rather than a genuine decrease in CO 2 flux in this section of the fault. The data that are shown in this area are from the footwall. Figure 9. Transects of elevation, soil temperature and CO 2 flux across faults surrounding the Salewa Dore-Hako rhyolitic complex (B-B') and the eastern boundary of the Salewa Dore-Hako graben (C-C'). The vertical, black-dashed line represents the graben boundary fault, while the red lines indicate mapped regional faults. Red horizontal lines denote the maximum value for background soil temperatures and CO 2 degassing, while the blue and black dashed lines represent minor inflections in background CO 2 populations referred to in Section 4.3.

Controls on Volcanism, Hydrothermal Fluids and Degassing
The North Abaya region is dominated by NNE-SSW trending en echelon normal faults (Corti et al., 2013), which have an overall horst-graben structure (Figure 2). The Abaya and Chewkare grabens comprise the rift floor, while the Salewa Dore-Hako graben accommodates a transition from the rift shoulder toward the rift floor. The Figure 10. CO 2 flux and temperature maps derived using the sequential Gaussian simulation (sGs) approach for the Salewa Dore-Hako rhyolitic complex (a, b) and the West Abaya graben fault (c, d). Black points represent a discrete flux measurement. Omnidirectional experimental variograms (γ) are shown for each data set. The modeled variogram is shown by the blue line with the key parameters listed. most significant deformation is accommodated on the graben bounding faults and in particular the WAGF which accommodates ∼110 m of vertical offset and represents the major structure in the study area. The region either side of a fault plane is referred to as the damage zone and is usually represented by highly fractured and permeable lithologies (Bense et al., 2013). Importantly, fault damage zone width increases with fault displacement (Choi et al., 2016;Faulkner et al., 2011;Knott et al., 1996;Sperrevik et al., 2002). Given that the WAGF shows the greatest displacement, it is also expected to represent the most permeable zone, and this can explain why gas and fluid upflow is concentrated here (as evidenced by fumarolic activity, hot springs and the elevated ground temperatures and CO 2 flux, Figures 3a, 5, and 6). Although the displacement and geothermal activity on the WAGF exceed all other faults, we note that other graben bounding faults do show elevated CO 2 flux (see transect C-C' in Figure 9). While CO 2 flux is much lower, it does suggest that the graben bounding faults are key pathways for CO 2 release and that they are more permeable and/or deeper penetrating than the regional intragraben faults.
Volcanism in the study area is mainly restricted to the Salewa Dore-Hako graben (Figure 2a). Mafic scoria cones span a ∼20 km long NNE-SSW oriented trend, while silicic vents are restricted to the Salewa Dore-Hako rhyolitic edifice in the center of this segment. The elongate trend suggests feeder dyke(s) beneath the basin, in agreement with vent elongation (Corti et al., 2013). From our CO 2 flux observations, we identified that graben bounding faults act as high permeability zones. However, volcanic vents are not aligned to these faults and are instead found scattered across the center of the graben. Thus, graben bounding faults do not appear to represent preferential structures for magma ascent and our data suggest that dykes are emplaced in the center of the Salewa Dore-Hako graben and that regional intragraben faults direct magma toward the surface (cf. Corti et al., 2013).  Note. Samples were mainly collected from the West Abaya graben fault (WAGF). Although one sample is from the Salewa Dore-Hako rhyolitic complex (SDHRC). Gas concentrations are reported in mol%. Bulk gas chemistry and C isotopes were measured on samples collected in evacuated glass vials. He isotopes were measured on samples collected in Cu tubes. R/R A is the measured 3 He/ 4 He divided by the 3 He/ 4 He in air. X-value is the 4 He/ 20 Ne ratio of the sample relative to that measured in air. R C /R A is the air corrected 3 He/ 4 He for the samples using the X-values (cf. Hilton, 1996).

Figure 11.
As noted above, the WAGF shows intense surface alteration (Figure 3a) and the most elevated ground temperatures and CO 2 flux in the study area (Figures 5 and 6). High resolution topography (Figure 2b) shows that this margin of the graben is typified by multiple faults which interact and intersect and that the highest ground temperatures are focused in a wedge-shaped zone at the tip of the Abaya horst. Intersecting faults are known to enhance permeability and fluid circulation (Curewitz & Karson, 1997;Person et al., 2012) and we suggest that these interactions at the Abaya horst, as well as the large damage zone of the WAGF, explain why this is an area of intense hydrothermal upflow.
A further feature of the Abaya horst is the occurrence of ∼W-E oriented faults (Figure 3a). Although these features have minor offsets (2-3 m) compared to the NNE-SSW aligned faults (10-100 m), it is likely that they also enhance the permeability and direct fluid flow because fumaroles at the tip of the Abaya horst show a WNW-ESE alignment (orthogonal to the trend of the WAGF). Cross rift (∼W-E) oriented structures have been documented at many volcanic systems throughout the EARS (Acocella et al., 2003;Benvenuti et al., 2023;Hunt et al., 2019;Lloyd et al., 2018a;E. Robertson, Biggs, Edmonds, et al., 2016), and while we cannot ascertain the extent of these structures at Abaya, they do appear to play an important role directing fluids and gas toward the near surface. In summary, our study reveals that the WAGF controls the main upflow of hydrothermal fluids and gas from depth, and that intersecting faults enhance the permeability of this zone concentrating fluids around the wedge tip of the Abaya horst.
At the Abaya horst, we see high values of both CO 2 flux (2,000 g m −2 d −1 ) and ground temperature (>90°C), but it should be noted that this is not the case along the entire length of the WAGF. For example, around A' in our transect in Figure 8 CO 2 is significantly elevated while temperatures are only just above background values of 40°C. This indicates that CO 2 and steam transport are decoupled, and is explained by the fact that steam, which usually travels together with CO 2 , has condensed to water during transport. There are several springs and surface water bodies in the vicinity of A', adjacent to the WAGF. Our hypothesis is that steam ascending along this section of the fault intersects groundwater and condenses (i.e., when its temperature drops below 100°C, Fridriksson et al., 2006). CO 2 condenses at much lower temperatures (−78.5°C) and is therefore unaffected by groundwater interactions.
CO 2 flux is also elevated above background values at a few localities on the Salewa Dore-Hako rhyolitic center and the highest values were associated with steaming ground (with temperatures of 50°C). Our study mainly covered the SW flank of the rhyolitic center and although we also conducted transects over several rift-aligned tectonic faults in the vicinity of the edifice, none show elevated CO 2 flux or ground temperatures (Figure 9, B-B'). Previous work on Aluto volcano in the CMER (Hutchison et al., 2015) demonstrated the importance of localized permeability variations when trying to interpret diffuse CO 2 degassing. These localized variations may be associated with (a) changes in the surface lithology (Pantaleo & Walter, 2014), for example, where there are differences in permeability between dense obsidian lavas and unconsolidated pumice deposits, and (b) topographic controls on the stress field, where differences in surface loading focus permeable pathways toward topographic highs (Schöpa et al., 2011). At Aluto, high CO 2 flux values (100-1,000 g m −2 d −1 ) were often observed at topographic highs associated with piles of young volcanic deposits and indicated that the topography-induced stress field played a role focusing fluid pathways toward morphological crests (Hutchison et al., 2015). Our transects of the Salewa Dore-Hako rhyolitic center show similar correspondence between topography and CO 2 flux, and this leads us to infer a topographic control on the near-surface stress field. Our interpretation is that, like Aluto, there may be deep penetrating volcanic and/or tectonic structures which act as the main conduit of gas from depth, but once these gases enter the pile of unconsolidated volcanic material, the permeability pathways are mainly controlled by the topographic loading, and this ultimately determines the surface expression of gas emission.

Volcanic Gas Emissions: Origin and Magnitude
The bulk gases collected from fumaroles (in glass vials) are mainly dominated by N 2 and O 2 and represent air mixed with variable amounts of CO 2 (up to 36 mol.%). When we consider various sources for the CO 2 Figure 11. (a) Carbon isotopes (δ 13 C) of soil gas and fumarole samples from Ethiopian, Kenyan and Tanzania volcanic-hydrothermal systems. δ 13 C of CO 2 is plotted against the reciprocal of CO 2 concentration in the sample. The data define a triangular array defined by three endmembers: air, biogenic CO 2 (with characteristic light δ 13 C of −20‰ and −25‰) and magmatic CO 2 (with δ 13 C between −3‰ and −8‰). (b) Inset of (a) showing δ 13 C for the highest concentration samples (note the logarithmic rather than reciprocal scale). (c) He isotopes versus X-value. 3 He/ 4 He is corrected for air and given in R C /R A notation. The X-value is calculated as ( 4 He/ 20 Ne) measured /( 4 He/ 20 Ne) air and provides an assessment of how much air has been entrained into the sample. X-values close to 1 are air-dominated, and those with much higher X-values indicate that very little air has been incorporated into the sample. Endmember 3 He/ 4 He for depleted mid-ocean ridge basalts ( ( Figure 11a), it becomes evident that their high CO 2 concentrations as well as δ 13 C values of −6.4‰ to −3.8‰, require a magmatic origin for the CO 2 . The minor difference in CO 2 concentrations between our samples and those of Minissale et al. (2017) (Figure 11b) likely reflects minor differences in gas sampling setup (i.e., sealing of the sampling line and time spent flushing the line prior to sample collection). Crucially, the CO 2 -δ 13 C values are almost identical to those of Minissale et al. (2017), who also concluded a mantle-derived magmatic origin.
He isotope samples provide additional insights into the origin of gases. New air-corrected He isotope ( 3 He/ 4 He) measurements range from 3.8 to 4.1 R A . These values are lower than previously published values from Minissale et al. (2017), who observed a range of 4.4-7.5 R A for all Abaya samples (Figure 11c). New data suggest that there may be a larger crustal contribution to gases collected in 2019. When taken together, He isotope values from the region overlap with the canonical range of mid-ocean ridge basalts (MORB) (8 ± 1, Graham, 2002) and SCLM (6.1 ± 2.1, cf. Bräuer et al., 2016;Day et al., 2015;Gautheron & Moreira, 2002;Gilfillan & Ballentine, 2018). Although the highest 3 He/ 4 He values from Abaya do imply a MORB source, we cannot exclude the possibility of SCLM contributions (given the overlapping values, Figure 11c). Moreover, the fact that we see a range of values scattered to low 3 He/ 4 He is suggestive of radiogenic 4 He contributions from crustal sources (which have values of 0.02, Ozima & Podosek, 2002; Figure 11c). Given that thermal fluids sampled along the WAGF by Minissale et al. (2017) show clear evidence of water-rock interactions, it is very likely that groundwaters from crustal sources contribute radiogenic 4 He. In short, the simplest explanation of the Abaya He isotope results is that they represent an upper mantle source with a crustal 4 He overprint. This explains the scattered values and indeed similar scenarios have been proposed in other areas of immature rifting further south in the EARS (e.g., the Magadi and Natron basins in Kenya and Tanzania, Lee et al., 2017). A final observation from Abaya 3 He/ 4 He measurements is that they are very different from the plume influenced measurements from Dallol in Afar (after Darrah et al., 2013;Figure 11c). This supports geochemical evidence for a declining plume contribution SW along the MER (Rooney, Hanan, et al., 2012) and geophysical evidence for low-velocity anomalies focused beneath Afar and which have been linked to the African superplume (Bastow et al., 2010;Mulibo & Nyblade, 2013;Ritsema et al., 1999Ritsema et al., , 2011. Helium isotope data clearly support the interpretation of the MER as transition zone between upper mantle and lithospheric sources in the south and plume-influenced mantle in the north toward Afar. As noted in Section 5.2, the WAGF and the intersecting fault network represent the deepest penetrating and most permeable area of North Abaya. CO 2 flux calculations support this and show that deep C emissions are focused along the WAGF which emits ∼300 t d −1 (10× the emissions from the flank of the rhyolitic complex, ∼30 t d −1 , Figure 10). Although there are only a few CO 2 flux measurements from elsewhere in the EARS, it is evident that Abaya is an important C emitter; the flux is ∼100× that of Longnot volcano in Kenya and ∼3× that of the Oldoinyo Lengai summit area (Table 2). At Aluto volcano in the Central MER, Hutchison et al. (2015) measured CO 2 along a tectonic fault that dissects the volcanic edifice (Artu Jawe fault zone, Table 2) and found a CO 2 emission of ∼60 t d −1 over an area of 0.8 km 2 . Scaling up these values, they calculated a total CO 2 flux from Aluto volcano of 250-500 t d −1 . This flux is of similar magnitude to that of Abaya, although we emphasize that while CO 2 emissions at Aluto are associated with numerous volcano-tectonic faults spread over an area of ∼150 km 2 , the CO 2 released from Abaya is focused along a single tectonic fault. The CO 2 flux density also shows similar values between the WAGF, the Artu Jawe fault zone on Aluto and other geothermal areas (e.g., Rotorua in New Zealand and Reykjanes geothermal area in Iceland, Table 2).
From a global perspective, the CO 2 emission from Abaya compares with complexes such as Vesuvius (Italy), El Chichon (Mexico) and Teide (Canary Islands), which all have active magmatic systems. More generally, estimates of CO 2 emissions from volcanoes with detectable SO 2 plumes (after Carn et al., 2017 andFischer et al., 2019) suggest typical CO 2 fluxes between 100 and 300 kt yr −1 . Abaya's annual emission is ∼110 kt yr −1 , which again underscores that even though Abaya does not represent a conventional volcanic edifice, the CO 2 emissions are comparable to the most active volcanic systems on Earth (Table 2). Taken together with the isotopic constraints ( Figure 11), this implies that an active magmatic system must underlie Abaya. CO 2 emissions from a volcanic system are mainly a function of the mass and C content of the degassing magmatic source, and the permeability of the fracture network (Burton et al., 2013). While we cannot disentangle the role of source versus permeability in controlling CO 2 emissions, the equivalence of Abaya and Aluto (in terms of total flux and flux density, Table 1) is notable because it implies similarities in terms of magmatic heat sources and subsurface permeabilities. Given that Aluto represents a proven geothermal resource, this finding should encourage further exploration at Abaya.  Hutchison et al. (2015) Oldoinyo Lengai, Tanzania a 100 3.14 32 Koepenick et al. (1996) Magadi-Natron Basin, Kenya and Tanzania

Architecture and Geothermal Potential
In Figure 12, we present a conceptual model of the North Abaya geothermal area based on our new measurements and the previous work of Chernet (2011), Corti et al. (2013) and Minissale et al. (2017). The heat for the geothermal field is sourced from magmatic intrusions and this is supported by the occurrence of recent, likely Holocene, volcanism in the area as well as our δ 13 C and 3 He/ 4 He data andCO 2 flux constraints, which indicate a magmatic system originating from an upper mantle source (Figures 2 and 11). Hydrothermal fluids sampled from thermal springs at Abaya predominantly show a Na-HCO 3 composition and indicate significant water-rock interaction. Importantly, the oxygen (δ 18 O) and hydrogen (δD) isotopes of the Abaya hydrothermal fluids reveal a narrow range of values which parallel global and Addis Ababa meteoric water trends (Minissale et al., 2017). In contrast, surface waters from Lake Abaya and the Bilate and Humasa rivers show strong evaporation trends with positive enrichments in δ 18 O and δD. These data demonstrate that the deep fluids circulating beneath Abaya are meteoric in origin and that there is no requirement for significant interactions with the highly evaporated lakes or surface water. The narrow δ 18 O-δD range of the Abaya hydrothermal fluids implies a similar elevation (Minissale et al., 2017), and we suggest that the rift margin to the west of the geothermal field provides the most likely source area. It is worth noting that this finding is comparable to Aluto volcano, where δ 18 O measurements also reveal that despite proximity to major lakes, waters from the deep geothermal wells are meteoric in origin (>90%) and derived from rainfall on the rift margin (Darling et al., 1996;Rango et al., 2010).
The WAGF is the main tectonic structure in the North Abaya area. Its large ∼110 m vertical offset is far greater than any of the other faults and indicates a wide fault damage zone (Section 5.1) and enhanced permeability (evidenced by the highest values of soil temperature and CO 2 flux, Figures 5 and 6). Field and remote sensing observations also indicate that this is a zone of fault intersection between the main graben bounding fault and other NNE-SSW regional faults (Figure 2b), which has led to a complex en echelon fabric as well as small-scale E-W faults (Figure 3b). In short, the WAGF constitutes the main upflow zone from the deep geothermal reservoir, and the numerous fault intersections amplify permeability (cf. Curewitz & Karson, 1997;Person et al., 2012) and concentrate fluid upflow at the tip of the Abaya horst ( Figure 6).
One of the key uncertainties with our model is the architecture of the deep magmatic heat source ( Figure 12). Although recent volcanism and hence magmatic intrusion appears to be focused in the Salewa Dore-Hako Graben, the main geothermal activity is offset from this and focused on the WAGF (where there is no evidence of recent volcanic activity). We suggest that the heat source is unlikely to be associated with the SDHRC because of its realtively small volume and location ∼6 km north. We instead propose that the heat is sourced from a deeper mafic magma body ( Figure 12) due to the fact that volcanism south of the Salewa Dore-Hako (nearest to the WAGF) is basaltic and indeed the formation of rhyolitic complexes necessitates large volumes of mafic intrusions . A similar setting to North Abaya may be the Butajira volcanic field in the CMER, which shows linear clusters of scoria cones within a marginal graben structure (Corti et al., 2018;Hunt et al., 2020). Here, a magnetotelluric study by Hübert et al. (2018) showed that the lines of scoria cones at the surface are horizontally offset from a deep conductive body at ∼5-10 km depth, which can be interpreted as mafic melt. We speculate that there may be an analogous situation at North Abaya where the main magmatic intrusion (and heat source for the geothermal system) is centered on the Abaya graben but volcanic activity is horizontally offset and concentrated along dyke intrusions in the Salewa Dore-Hako graben ( Figure 12).
To date, most studies of East African geothermal plays have focused on silicic caldera complexes (Gianelli & Teklemariam, 1993;Hochstein et al., 2017;Hutchison et al., 2015;Omenda, 1998). Magnetotellurics has proved particularly powerful at imaging geothermal resources, and at Aluto and Tulu Moye (Ethiopia) these data often suggest the presence of a shallow high conductivity clay cap layer overlying a low conductivity pressurized hydrothermal system (Hübert et al., 2018;Samrock et al., 2015Samrock et al., , 2018. Magnetotellurics has been conducted at North Abaya by Reykjavik Geothermal and support the occurrence of a high conductivity clay cap rock at a depth of 0.5-2 km (Eysteinsson, pers. comm.). This is shown schematically in Figure 12 and we speculate that the WAGF provides the only major zone of permeability through this cap layer (hence why fumaroles, hot springs and degassing anomalies are largely restricted to this deep penetrating structure).
A final observation of the North Abaya geothermal area is that unlike the silicic volcanic complexes of Tulu Moye and Aluto, there have been no episodes of ground deformation detected over a period spanning 1993 to 2020 (Biggs et al., 2011;Albino & Biggs, 2021;Albino et al., 2022; Figure 4). At Tulu Moye, deformation has been linked to large-scale pressurization of the magmatic system, while at Aluto, seasonal uplift-subsidence patterns related to rainfall and pore pressure changes of the hydrothermal system are superimposed on longer-term magmatic-hydrothermal pressurization (Braddock et al., 2017;Hutchison et al., 2016). The fact that North Abaya displays neither of these trends demonstrates that (a) there have been no detectable magmatic intrusions, and (b) the geothermal reservoir does not respond to seasonal variations in rainfall. These data suggest that the mafic magmatic heat source beneath Abaya is less active and restless than the silicic mush systems of Tulu Moye and Aluto, it also suggests that the geothermal reservoir is potentially much larger than these other systems, or that timescales of groundwater flow between meteoric source and the heat source are much slower (such that seasonal variations in pore pressure are not observed). Ultimately, the lack of surface volcanism and ground deformation along the WAGF stands in stark contrast to the restless silicic calderas elsewhere in the MER (Albino & Biggs, 2021;Biggs et al., 2011), and while further monitoring of the prospect is essential, this does make it an attractive system for further exploration and development.

Conclusions and Future Opportunities
We have combined remote sensing, soil CO 2 and ground temperature surveys and gas chemistry analysis to build up a detailed understanding of the North Abaya geothermal system in the SMER. The main outcomes of our study are: 1. the North Abaya region displays a horst-graben morphology, and graben bounding faults represent the deepest penetrating, most permeable structures, 2. soil CO 2 flux and temperatures reveal that the most permeable structure is the WAGF, and that deep hydrothermal upflow is concentrated along this structure and into a zone of fault intersections, 3. even though there is no surface volcanism along the WAGF, gas emissions are ∼300 t d −1 and comparable to average values from the world's subaerial volcanoes (e.g., Vesuvius, Teide, El Chichon, Table 2), 4. gas emissions require an active magmatic system, and new C-and He-isotopes suggest an upper mantle source (SCLM and/or MORB) that is overprinted by crustal 4 He additions.
Our work presents the first detailed conceptual model of a fault-controlled magmatic geothermal resource in East Africa ( Figure 12). Further refinement and testing of our conceptual model will require additional geophysical and geochemical surveys of the WAGF. These would include focused seismic and magnetotelluric surveys to better understand the distribution of melt, geothermal fluids, and gases in the subsurface, and the orientation of fracture sets beneath the fault zone (e.g., Nowacki et al., 2018;Samrock et al., 2018;Wilks et al., 2020). Further geochemical investigations should consider measuring radon and thoron, which can provide insights into depths and speeds of geothermal upflow (Jolie et al., 2019), and additional CO 2 -δ 13 C would allow us to determine whether the different background populations inferred from log-probability plots (Section 4.3, Figure 7) show different isotopic fingerprints therefore requiring different sources.
North Abaya is a very promising site for geothermal development and in particular the northern tip of the Abaya horst, where ground temperature is highest, and permeability is enhanced by multiple intersecting faults. Compared to the active silicic caldera volcanoes that are generally seen as East Africa's best geothermal prospects, the WAGF shows no evidence of ground deformation or recent volcanism. While there are still risks of volcanic and tectonic hazards in the North Abaya area, and monitoring is to be encouraged, we suggest that fault-controlled geothermal plays, linked to deep mafic magma body sources, could represent safer and more sustainable prospects than silicic caldera volcanoes.