Introduction

Hydrogen is ubiquitous at the Earth’s surface, and locally high surface fluxes of hydrogen have been known for decades, particularly at midocean ridges (Zgonnik, 2020). However, the potential for substantial subsurface hydrogen resources has been dismissed because hydrogen is highly reactive and diffusive (Apps and van der Kamp, 1993). Recent modeling has challenged this assumption, suggesting that thousands to billions of megatons of hydrogen may reside in the subsurface (Ellis and Gelman, 2024). Although much of this hydrogen likely is inaccessible because it is too deep, too far offshore, or in small accumulations that are not economically viable, even a small fraction could fill a substantial portion of the projected hydrogen demand in the coming years (Ellis and Gelman, 2024).

Until recently, subsurface hydrogen discoveries were generally unintentional and most likely went unreported because drilling mud gas logging tools do not routinely measure hydrogen, and hydrogen commonly is used as a carrier gas in laboratory gas analyses, preventing analysis of natural concentrations. In a world that historically has sought petroleum, discoveries of hydrogen-saturated formation fluids, if recognized at all, were generally considered a failure, and were ultimately abandoned. Therefore, despite the millions of exploratory and production boreholes that have been drilled throughout the United States (app. 1, fig. 1.1), the possibility of hydrogen as a new potential fluid target remains.

Recent discoveries of hydrogen across the globe include the Bourakebougou field in Mali (Prinzhofer and others, 2018; Maiga and others, 2023,56 2024) and in the Yorke Peninsula in South Australia (Boreham and others, 2021), as well as isolated but high concentrations observed in North America; other parts of Africa, Europe, Asia; and other parts of Australia (Zgonnik [2020] and references therein). Hydrogen also has been associated with subsurface helium discoveries (Milkov, 2022). Within the United States, hydrogen has recently been reported in Arizona, California, Kansas, Montana and Nebraska (U.S. Geological Survey [USGS], 2019; Zgonnik, 2020). Observations of hydrogen in concentrations of more than 0.5 mole percent (mol%) are shown in figure 1, illustrating a general scattering of hydrogen occurrences throughout the United States (USGS, 2019; Zgonnik, 2020; Brennan and others, 202113). Particularly high concentrations have been found in the midcontinent region, around the Four Corners area (Arizona, Colorado, New Mexico, and Utah), and along the coast of California. However, little has been published on establishing the controls on the locations of these occurrences and thus, predictability remains low.

To facilitate continued exploration of this new potential energy resource, research into ways to identify regions that may be favorable for hydrogen exploration is underway. In 2021, Boreham and others (2021) published a radiolytic production rate map for Australia, delivering the first regional-scale map depicting relative favorability for hydrogen throughout that country. However, that map relied heavily on surface geology extrapolated to a depth of one kilometer (km), which may be of limited utility due to lithologic heterogeneity along such distances, and the map did not fully integrate the spatial distribution of hydrogen generation sources with other potential factors, such as reservoir or seal potential. In particular, the use of geophysical potential fields data may provide an understanding of geology at depths greater than 1 km. Considering several additional sources of hydrogen generation and integrating a variety of surface and subsurface datasets, this report presents the first publicly available map of geologic hydrogen prospectivity for the conterminous United States.

Purpose and Scope

Geologic evaluations of subsurface resources are initiated on broad spatial scales that narrow with increasing knowledge. Studies begin with regional or “play” assessments that serve to prioritize areas for future, more focused efforts (Lottaroli and others, 2018). Because geologic hydrogen is a resource recently discovered, our methodology is designed for the regional—or even continental—scale, and our purpose is to identify areas of the conterminous United States for more indepth research.

In defining the geologic hydrogen system and the components necessary for potential accumulations to exist, this report distinguishes between the “play” scale and the “prospect” scale. At the play scale, we do not try to define where or at what depth specific hydrogen accumulations may exist, but instead, focus on the spatial layering of processes that are required for accumulations to occur generally. Processes such as the generation of geologic hydrogen, its broad migration direction, reservoir porosity for its storage, and seal capacity for prevention of leakage to the surface are tractable to evaluate at regional scales. Processes that define or are related to localized prospects may include subsurface traps; reservoir temperature; biotic or abiotic consumption; or the detailed migration, mixing, and phase behavior with other fluids (water, hydrocarbon, nitrogen, and so on). Although these latter processes are also relevant and important at the play scale, they are difficult to map accurately on broad spatial scales and thus we have not incorporated them into our 2024 methodology. Future work may address this difficulty as it becomes more feasible and as more is learned about the subsurface hydrogen system.

Hydrogen System Components

For purposes of this report, the geologic hydrogen system at the continental scale is defined as comprising three primary components: source, reservoir, and seal. Future work may add a fourth component addressing the preservation of hydrogen, because it is highly reactive. The source component is defined by the mechanism for the generation of geologic hydrogen and incorporates possible vertical and lateral fluid migration. The reservoir component is defined by the presence of porosity in the subsurface to store hydrogen. The seal component is defined by the presence of low-permeability rocks paired with potential reservoirs to prevent seepage and loss of hydrogen to the surface.

The source component of the hydrogen system has received the most attention in previous studies and several potential mechanisms have been proposed for the presence of hydrogen in the mid-to-upper crust (Sherwood Lollar and others, 2014). For example, serpentinization, or the oxidation of iron in the presence of water, is the main source of hydrogen in midocean ridge settings (Frost and Beard, 2007; Liu and others, 2023), in subduction zones within the mantle wedge or slab itself (Guillot and others, 2001), and may occur when ultramafic rocks are in contact with water at moderately high temperature (200–315 degrees Celsius [°C]; McCollom and Bach, 2009). The radiolysis, or molecular splitting of groundwater through contact with radiogenic rocks, is another commonly cited source of hydrogen. Helium is also generated by decay of radiogenic nuclides in similar rocks and may be colocated with hydrogen; figure 1.2 in appendix 1 shows the subsurface helium distribution in the conterminous United States. Radiolysis, however, tends to rely on potassium-thorium-uranium series decay from crystalline, granitoid-type rocks and, therefore, requires long residence times (hundreds of millions to billions of years) for an effective volume of hydrogen generation. Some authors have also proposed that hydrogen is generated from overmature, organic-rich rocks (Horsfield and others, 2022; Mahlstedt and others, 2022; Boreham and others, 2023; Hanson and Hanson, 2024), although this hypothesis remains difficult to represent on broad spatial scales because of a lack of sampling of very high maturity rocks since they have limited utility for hydrocarbon exploration. Alternatively, hydrogen is likely to be consumed by lower-maturity, organic-rich rocks during thermal maturation and petroleum generation (Apps and van der Kamp, 1993; Smith and others, 2005) or adsorbed in some clay minerals that may be present in fine-grained mudrocks (Wang and others, 2023). In both cases, the depths of organic-rich rocks become critically important (optimal conditions require great depths for hydrogen generation and shallow depths for hydrogen consumption) and may be addressed more tractably at basin or subregional spatial scales. Poorly defined deep sources of hydrogen may exist from the primordial Earth formation, with slow degassing during Earth’s history, or may consist of hydrogen of indeterminate origin that has migrated through the lithosphere to mid- or upper-crustal levels along deep crustal-scale faults (Zgonnik, 2020).

As a first approach, we have simplified the source subcomponent mechanisms to include the following:

• Serpentinization-type reactions based on the identification of ultramafic rocks but without consideration of water or temperature ranges;

• Radiolysis of water based on the identification of Archean or Proterozoic granitoids, without consideration of the presence of water; and

• Deep sources of hydrogen.

The details of these modeled subcomponents are provided in the “Methods” section of this report. Additionally, buoyant migration of subsurface fluids allows for the transportation of hydrogen from source to reservoir and allows a broad footprint of reservoirs to access hydrogen that is generated downdip. Long-distance lateral migration is well established in petroleum systems (Tissot and Welte, 1984; Hunt, 199640). Here, we use a similar approach to model the migration pathways that may be expected from hydrogen source rocks, which realistically expands the spatial reach of these source rocks and incorporates lateral migration into the source subcomponents.

Reservoir and seal components are likely to be broadly comparable to those components of hydrocarbon and other subsurface natural gases. Because the diffusivity of hydrogen is similar to that of helium (Jacops and others, 2017), hydrogen may be sealed by similar rock facies and in similar trapping configurations. Because most helium and hydrocarbon gases are in sedimentary reservoirs with sedimentary seals, the presence of sedimentary strata is given preference as a first proxy for the presence of reservoir and adequate sealing facies. Lateral and vertical migration of hydrogen from source rocks (some of which may be quite deep, for example, serpentinization sources) is also modeled using methodologies developed for other subsurface gases where long distance migration is well-established (Tissot and Welte, 1984; Hunt, 1996).

Methods

The spatial distributions of hydrogen system components are integrated based on presence to identify highly prospective regions. However, because there is uncertainty in the presence of these components in the subsurface, a chance of sufficiency (COS) methodology was selected to establish fractional probabilities of each component’s presence and effectiveness, which is explained in more detail in the “Assigning Chances of Sufficiency” and “Risk Analysis” sections of this report. “Sufficiency” is defined as the viability of a hydrogen system component to contribute to a subsurface occurrence of hydrogen. Because no volumes are known yet for subsurface hydrogen occurrences, this definition is not benchmarked to any specific volume. The minimum concentration for an occurrence is 0.5 mole percent (mol%; fig. 1). The COS similarly does not imply a statistical likelihood of subsurface hydrogen being found. For example, a source COS of 0.5 does not imply that 50 percent of wells in a particular location will successfully observe hydrogen or even a hydrogen source. Rather, this COS metric is used in this early exploratory stage to establish a relative scale on which to evaluate prospectivity from one location to another, based on the stacking of hydrogen system components. Similar knowledge-based quantitative probability methods have been used in petroleum evaluations (White, 1993; Otis and Schneidermann, 1997), geothermal studies (Ito and others, 2017), seabed mineral deposit studies (Juliani and Ellefmo, 2019), and even subsurface radioactive waste disposal studies (Malvić and others, 2020).

The hydrogen prospectivity map of the conterminous United States integrates 21 layers of geological and geophysical data. Table 1 outlines the 3 primary model components, the 3 source subcomponents, the 21 individual layers, the distribution of the COS assigned to each layer, the spatial buffers applied, and the layers that incorporate lateral migration—all of which inform the geologic hydrogen model. Raster grids for each of these layers are provided in the associated USGS data release (Hearon, 2025). The COS for each layer is assigned based on interpretations of the geologic hydrogen model and broad knowledge of the geology of the conterminous United States.

Serpentinization as a Hydrogen Source

Reduction of water by oxidation of iron-containing minerals in ultramafic rocks (that is, serpentinization-type reactions) is one of the most widely cited mechanisms for the formation of natural hydrogen (Gaucher, 2020; Zgonnik, 2020; Blay-Roger and others, 20248). The modeled serpentinization subcomponent (SP) comprises five layers (SP1 to SP5) predicated on the presence of subsurface ultramafic rocks (fig. 2). Layer SP1 is the area defined by the east coast magnetic anomaly (ECMA), which is expressed as an offshore, high-amplitude, positive magnetic anomaly extending from the coast of Georgia to the coast of Nova Scotia. Previous studies suggest that the anomaly may be caused by the presence of subsurface ultramafic rocks or serpentinized mantle material associated with the boundary between rifted continental and oceanic crust (McBride and Nelson, 1988; Wu and others, 2006; Biari and others, 2017). Layer SP2 is defined by areas where ultramafic rocks are present at the surface (Krevor and others, 2009; Faure, 2010; Horton and others, 2017) and are inferred to potentially extend into the subsurface. This layer includes broad coverage of ultramafic rocks, including banded iron formations, which have been proposed as additional hydrogen source rocks beyond traditional peridotitic lithologies (Geymond and others, 2022, 2023). Layer SP3 contains areas that are underlain by the midcontinent rift, the southern Oklahoma aulacogen, and the Reelfoot rift (Elling and others, 2022). These areas represent three failed Proterozoic midcontinental rifts that are defined by prominent, high Bouguer gravity anomalies. Rift settings are known for isostatic shallowing of the crust-mantle boundary, promoting the presence of shallow ultramafic rocks and commonly featuring exhumed mantle (McKenzie, 1978; Boillot and others, 1980; Manatschal, 2004; Gillard and others, 2019).

Layer SP4 outlines subsurface regions that may contain ultramafic rocks based on geophysical potential fields data. Isostatic gravity data (Kucks, 1999) and magnetic data (Bankey and others, 2002) were considered where overlapping positive anomalies in both datasets were initially hypothesized to indicate the presence of potentially serpentinized, dense ultramafic rocks at depth. Because potential fields signals can be generated by numerous nonunique sources, this filter is being used to target dense and magnetic rocks in general, which can include serpentinites (which may be lower density than their unserpentinized precursors, but still are dense relative to granitoids and most sedimentary rocks) as well as banded iron formations. To better identify sources at midcrustal depths, the magnetic data are a 500 km high-pass filtered dataset (Bankey and others, 2002), which removes long wavelengths. Two magnetic maps were then generated: (1) a reduced-to-pole dataset and (2) a dataset with a pseudogravity transformation (Blakely, 1995; Oakey and Saltus, 2016). The latter was created to better highlight magnetic source bodies and provide a more comparable wavelength spectrum to the gravity data. Two index maps were created to display gravity and magnetic data using both datasets (fig. 3) and were used to highlight regions with positive gravity and positive magnetic signatures which may indicate greater potential for subsurface ultramafic lithologies. The regions from layer SP3 (failed rifts) were used to check for expected positive gravity-magnetic signatures but were observed to contain positive gravity, but mixed magnetic anomalies. This observation aligns with that of Bonnemains and others (2016) in which the magnetic signatures of serpentinized ophiolites can depend on the temperature and thus mineralogy of the serpentinization reaction, as well as the degree of serpentinization. Given the ambiguous magnetic signature of serpentinites, only isostatic gravity was ultimately considered in the model for layer SP4. Furthermore, although Laramide-associated basement uplifts (for example, the Wind River Range, Bighorn Mountains, Laramie Mountains, and Granite Mountains in Wyoming, and Black Hills in South Dakota and Wyoming) display strong, positive isostatic gravity anomalies, these mountains were expected to be predominantly felsic, yet remain in the model results; we note that rare ultramafic rocks have been observed in these ranges (Potts, 1972; Manzer and Heimlich, 1974; Harper, 1986), and hypothesize that more may exist at depth, accounting for the strong positive gravity anomaly. The midcontinent rift, southern Oklahoma aulacogen, and Reelfoot rift defined in SP3 were then removed from layer SP4 so that they are not represented twice in this analysis.

Finally, layer SP5 represents potential serpentinization sources throughout the entire United States that are unknown and have not been explicitly accounted for in layers SP1–SP4. Thus, this layer is given an evenly applied nonzero COS across the Nation.

Radiolysis as a Hydrogen Source

Along with serpentinization, radiolysis of water by ancient, radiogenic rocks also is commonly considered a major source of geologic hydrogen (Sherwood Lollar and others, 2014; Gaucher, 2020; Zgonnik, 2020). The modeled radiolysis subcomponent (RD) is split into five layers (RD1 to RD5) that represent the presence of radiogenic subsurface rocks (fig. 4). Layer RD1 contains areas with known subsurface uranium deposits and RD2 displays areas that are favorable for the concentration of uranium in the subsurface. These layers were compiled by the USGS from published sources and are available from the U.S. Energy Information Administration (U.S. Energy Information Administration, 2020a, b91). Because uranium decay can cause the radiolysis of water, both layers were used but have different COS values.

Layer RD3 covers regions that are underlain by the Archean and Proterozoic cratonic platform (Marshak and others, 2016). The granites and granitoids of the North American craton are commonly enriched in radioactive elements such as uranium and thorium and are sufficiently old to allow for abundant decay of these long-lived radionuclides. Layer RD4 contains geologically younger regions that are not underlain by the craton and thus is an inverse layer from RD3. The area outlined in RD4 is underlain by terranes of the western and eastern North American cordillera as well as the southeastern coastal plain that were accreted throughout the Phanerozoic (Marshak and others, 2016). These rocks are highly variable in lithology and age and, therefore, also highly variable in the concentration of radioactive elements and time for decay. Thus, a lower COS is assigned to RD4 regions relative to those regions in RD3.

Finally, layer RD5 represents areas with Phanerozoic granites and granitoids (Horton and others, 2017). These rocks are common along the western cordillera of the United States, and may be enriched in uranium and thorium, but have not had the long residence time needed for abundant radioactive decay.

Deep Sources of Hydrogen

The deep-sourced subcomponent captures a potential supply of geologic hydrogen that is derived from sources in the lower crust, or possibly deeper in the Earth’s interior. The exact generation mechanism for this inferred source is not known but may be the decomposition or dehydration melting of dense, hydrous silicate phases introduced by subducted slabs to the deep mantle (Williams and Hemley, 2001; Hirschmann, 2006; Ohira and others, 2014) or the outgassing of primordial hydrogen from the core and mantle (Williams and Hemley, 2001; Loewen and others, 2019; Olson and Sharp, 2022). The deep source subcomponent (DP) is split into four layers (DP1 to DP4; fig. 5). Layer DP1 includes areas of mapped, kilometer-depth scale surface faults (McCafferty and others, 2023). Faults can connect to and provide pathways for geologic hydrogen to migrate from lower crustal sources into reservoirs for storage. Layer DP2 consists of areas of Proterozoic suture zones (Whitmeyer and Karlstrom, 2007). These areas are interpreted to represent significant internal lithospheric boundaries that may promote fluid migration from depth. Phanerozoic suture zones, such as the Coast Range, San Andreas Fault, the various Appalachian and Ouachita belts, and the Suwannee suture zone (among others) may also represent similar lithospheric boundaries. These areas were not included in initial mapping because of the lack of a unified data layer; however, these areas can be incorporated into this workflow when available. Layer DP3 captures regions of high heat flow (Blackwell and others, 2011), defined here as greater than 80±10 milliwatts per square meter (mW/m²). These high heat flow regions are commonly associated with geothermal fluid circulation and may promote hydrogen migration from depth. Volcanic fluids and geothermal springs have been observed to contain hydrogen as well as hydrogen-metabolizing organisms (Spear and others, 2005; Lindsay and others, 2019; Zgonnik, 2020). Lastly, layer DP4 represents potential deep sources of hydrogen throughout the United States that are unknown and have not been explicitly accounted for in layers DP1, DP2, and DP3. Similar to SP5, DP4 is given an evenly applied, nonzero COS across the United States.

Lateral Migration of Subsurface Fluids

Migration of subsurface fluids is governed by porosity and permeability pathways within and between rock units, including permeability induced by naturally occurring fractures and faults. Furthermore, in multiphase immiscible fluid systems (such as hydrocarbon and water or a hydrogen gaseous phase and hydrogen-saturated water), buoyancy forces also drive migration. These processes are complex and can occur in both sedimentary settings, as has been shown abundantly in petroleum systems (Tissot and Welte, 1984; Hunt, 199640), as well as in metamorphic and igneous rocks, as has been demonstrated primarily in geothermal applications (Ingebritsen and Manning, 2010; Lamur and others, 2017). The direction of fluid flow also is complex (Bjorlykke, 2010), and a detailed examination is beyond the scope of this preliminary favorability mapping study. To determine favorable areas for possible hydrogen accumulations, we focus on the flow directions associated with the long-distance lateral migration that commonly occurs in sedimentary basins. For this report, this long-distance lateral migration is defined as migration across map distances of up to 250 km. Simple methods for mapping lateral fluid flow are well established from petroleum and hydrologic studies and include drainage area delineation (Hubbert, 1953) and flow line mapping (Tóth, 1962), both of which are used for this report.

To provide a single migration surface for modeling lateral hydrogen migration in sedimentary settings throughout the United States, an elevation map of basement rocks was created. This surface is defined as the base of the stratigraphic section in regions with sedimentary surface geology and is otherwise coincident with topography in regions with metamorphic or igneous surface geology. The map was created by subtracting the total thickness of sedimentary rocks (Frezon and others, 1983) from the GEBCO 2023 digital elevation model (GEBCO Bathymetric Compilation Group, 2024). Because lateral migration was only modeled in sedimentary settings for this study, regions that have metamorphic or igneous rocks cropping out were ultimately removed from the map. The resultant basement elevation surface is shown in figure 6.

Map-based migration analysis began with the delineation of drainage areas and was followed by the creation of flow lines in all areas with sedimentary cover (fig. 6). The migration pathways were created with SLB PetroMod version 2020.1 (SLB, 2020), using the PetroCharge Express tool, and were modeled on the basement elevation surface described in the preceding paragraph; a high-resolution version of these flow lines is shown in fig. 1.3 in appendix 1. During lateral migration, hydrogen is assumed to be a separate phase that is buoyant relative to groundwater and is consumed biotically and abiotically (Gregory and others, 2019); this consumption is accounted for in the model by linearly decreasing the COS as lateral migration distance increases from 0 to 250 km from the source.

For each source layer that explicitly represents a subsurface source (identified in table 1 as “Lateral migration”), the following workflow was developed:

1. 1. Source presence defined by the layer is clipped to only include source in areas with sedimentary cover.

2. 2. A spatial buffer was applied, as given in table 1, to provide coarser spatial resolution to point features.

3. 3. A Matlab script (app. 2) was developed that first created a list of grid cells with source presence, and then:

   1. A. Identified flow lines within a 4-km radius of a source cell,

   2. B. Calculated migration distance along the vertices of the flow line, and

   3. C. Populated a resultant flow line point set with migration distance.

4. 4. A resulting pointset with migration distances for the layer was then interpolated using the ArcGIS Empirical Bayesian Kriging tool (Krivoruchko and Gribov, 2019) to provide a raster map of migration distance.

5. 5. The migration distance grid was then converted to a COS grid by linearly interpolating between 0 and 250 km distances corresponding to a COS of 1 and 0, respectively.

6. 6. Any migration distance beyond 250 km was given a COS of zero.

Reservoir and Seal

Although a source of hydrogen is critically important for determining where hydrogen may be flowing in the subsurface, a porous reservoir that is sealed by low permeability caprock is likely to be required for economic accumulations to be realized. This analysis treats reservoir and seal on a continental scale and, therefore, the main discriminator for predicting the potential for lithologies with sufficient reservoir quality is simply the presence or absence of sedimentary strata. Although fractured crystalline rock may be permeable as is demonstrated in geothermal applications (Lamur and others, 2017), or as has been observed in relatively rare petroleum accumulations (Petford and McCaffrey, 2003), the same geologic processes that led to those fractures likely preclude the existence of a competent seal.

The reservoir component (RS) focuses on the presence of predictable porosity and permeability and is composed of three layers (RS1 to RS3; fig. 7). Layer RS1 indicates the presence of sedimentary strata based on surface geology; layer RS2 is the inverse (Horton and others, 2017), indicating regions with metamorphic or igneous rocks at the surface and therefore likely an absence of sedimentary strata in the subsurface as well. The COS assigned to RS1 is higher than that assigned to RS2 (table 1), representing the higher likelihood of sufficient reservoir when sedimentary rocks are present. Finally, many regions of the United States have sedimentary rocks at the surface, but the overall sedimentary cover may be thin (less than a few kilometers), leading to a marginally lower likelihood of a viable hydrogen reservoir. Layer RS3, which highlights regions within well-defined sedimentary basins (Frezon and Finn, 1988), is included to capture the benefit of having a thick sedimentary column and, therefore, many potentially viable reservoirs available.

The seal component defines the presence of predictable, low-permeability rocks paired with reservoirs that can keep hydrogen in the subsurface and prevent seepage to the surface. This component is composed of four seal (SL) layers (SL1 to SL4; fig. 8). Layer SL1 defines the presence of evaporites in the subsurface (Ege, 1985) as salt layers provide notably good sealing facies (Beauheim and Roberts, 2002). Layers SL2, SL3, and SL4 are the same as the reservoir layers RS1, RS2, and RS3 because the presence or absence of sedimentary strata is taken as the primary driver for the possibility of a viable sealing facies in the subsurface. However, the COS values are lower when these layers were used in the seal component, reflecting an increased general sense of risk in seal over reservoir. Furthermore, thicker sedimentary strata associated with basins are interpreted to offer the benefit of (1) sufficient burial to compact rocks to low porosity and (2) a higher likelihood of one or more potential sealing formations throughout the stratigraphic column.

Assigning Chances of Sufficiency

The COS values for each component and subcomponent were assigned using an expert elicitation, or expert-driven, approach using a peer review panel (O’Hagan and others, 2006). During this panel, the authors presented all data layers and gained feedback on the methods and COS confidence ranking between components and subcomponents. The panel audience was two geoscience subsurface data specialists, a geoscientist with extensive experience conducting petroleum assessments in the public sector, a geoscientist with extensive experience conducting petroleum play and prospect assessments in the private sector, and two geophysicists with extensive backgrounds in the public and private sectors using potential fields data.

The final COS values for each hydrogen system component are illustrated with an uncertainty matrix (fig. 9). Uncertainty matrices are widely used to represent the effect and uncertainty in a model parameter (Fugelli and Olsen, 2005; Speziale and Geneletti, 2014). As provided in table 1, subcomponent layers are assigned a median COS and an uncertainty range represented by high and low COS endmembers. To determine these values, the data layer is considered in terms of its sufficiency for the hydrogen system, as well as confidence that includes uncertainty in the input data layer itself and uncertainty in its sufficiency median value. Recommended COS values and ranges are shown in an uncertainty matrix format in figure 9A. The COS values in this matrix were obtained during an iterative process of balancing confidence in data layers with their importance for the hydrogen system. Model results are less sensitive to the absolute COS value and instead depend more on relative COS differences assigned to various model layers. Increasing knowledge of the hydrogen system may result in modifications to these preliminary COS values. Uncertainty ranges assigned in figure 9A generally indicate increasing (but symmetrical) ranges with decreasing confidence. The exception is the lowest left domain (low confidence, low sufficiency); this uncertainty range is only 0.1 to prevent negative numbers while maintaining symmetrical distributions. To illustrate the methodology, source subcomponent data layers are plotted in figure 9B. The reservoir and seal component layers followed the same methodology.

Risk Analysis

All three primary model components—source, reservoir, and seal—are required for a successful subsurface accumulation of geologic hydrogen. Mathematically, we quantified the likelihood of sufficiency of each of these primary components with a COS as described in the beginning of the “Methods” section of this report. The COS values for the primary model components are shown in table 1. The total hydrogen prospectivity, P, is then described by:

P = COS

_SC

× COS

_RS

× COS

_SL

,

(1)

where

P

is the value for the total hydrogen prospectivity;

COS

is the chance of sufficiency;

SC

is the COS for the source components;

RS

is the COS for the reservoir component; and

SL

is the COS for the seal component.

The COS is a probability, ranging in value from 0 (impossible) to 1 (certain). Thus, if any primary component has a COS of zero (impossible), then the total prospectivity in that location will also be zero.

Within each primary hydrogen system component, multiple layers have been considered that may contribute to the sufficiency of that component. We treat these using the following relations for each primary hydrogen system component:

COS

_SC

= 1 − [(1−COS

_SP

) × (1−COS

_RD

) × (1−COS

_DP

)],

(2)

COS

_RS

= 1 − [(1−COS

_RS1

) × (1−COS

_RS2

) × (1−COS

_RS3

)], and

(3)

COS

_SL

= 1 − [(1−COS

_SL1

) × (1−COS

_SL2

) × (1−COS

_SL3

) × (1−COS

_SL4

)],

(4)

where

SP

is the value for the source subcomponent of serpentinization;

RD

is the value for the source subcomponent of radiolysis; and

DP

is the value for the source subcomponent of deep sources.

The COS values for the source subcomponent are provided in table 1. The COS maps for equations 2, 3, and 4 are shown in figure 10.

Finally, since each source subcomponent itself incorporates several layers, these are defined by:

COS

_SP

= 1 − [(1−COS

_SP1

) × (1−COS

_SP2

) × (1−COS

_SP3

) × (1−COS

_SP4

) × (1−COS

_SP5

)],

(5)

COS

_RD

= 1 − [(1−COS

_RD1

) × (1−COS

_RD2

) × (1−COS

_RD3

) × (1−COS

_RD4

) × (1−COS

_RD5

)], and

(6)

COS

_DP

= 1 − [(1−COS

_DP1

) × (1−COS

_DP2

) × (1−COS

_DP3

) × (1−COS

_DP4

)],

(7)

where the subscripts in equations 5, 6, and 7 are described for each layer in the “Serpentinization as a Hydrogen Source,” “Radiolysis as a Hydrogen Source,” and “Deep Sources of Hydrogen” subsections of this report. The COS maps are shown for equation 5 (fig. 2E), equation 6 (fig. 4F), and equation 7 (fig. 5D).

Equations 2–7 all incorporate multiple layers by multiplying chances of insufficiency, which is defined as 1–COS, rather than COS directly. The COS for each hydrogen system component or subcomponent is then recovered by subtracting the cumulative chance of insufficiency. This method allows multiple layers to contribute by increasing the COS where they spatially overlap, and unlike the primary components, not all layers must exist simultaneously for an accumulation to occur. Using this methodology, multiple overlapping layers results in only an incremental improvement in the overall component or subcomponent COS. For example, for the seal component in a region where sedimentary rocks are present, the additional presence of salt improves the overall COS_RS from 0.8 to 0.82 (for the midcase COS values in table 1). Furthermore, if that region is within a well-defined sedimentary basin, the COS_RS improves to 0.874.

All layers were mapped on an initial 1-km square grid; the final risk analysis was performed on a slightly coarser 5-km square grid. A Matlab script to solve equations 1–7 is provided in appendix 2. Equations 1–7 can also be easily solved using raster calculators in geographic information systems (GIS) software.

Results

The potential for a viable hydrogen source is shown as a COS map in figure 10A. This map illustrates the important role that the North American craton plays in providing a general source of old (Archean and Proterozoic) granitoids in the central part of the United States that are potential sources of radiolytic hydrogen. Overall, the highest COS regions for hydrogen sources occur where serpentinization sources (mapped surface or subsurface ultramafic rocks) coincide with the North American craton, thus providing multiple possible hydrogen generation mechanisms. Examples of these areas include southern Oklahoma and the Texas Panhandle (the southern Oklahoma aulacogen), and northeastern Kansas, southeastern Nebraska, west-central Iowa, eastern Minnesota, and Michigan (the midcontinent rift). A few high COS source regions that do not overly the North American craton include the coast of central and northern California, which contain ophiolite sequences, and the Salton Sea area of southeastern California and southwestern Arizona. The source COS map is particularly important for exploration efforts that focus on capturing actively seeping or migrating hydrogen, or for stimulated hydrogen generation (Osselin and others, 2022), since reservoirs and seals are unnecessary in those scenarios (Prinzhofer and Cacas-Stentz, 2023).

The reservoir and seal COS maps are shown in figures 10B and 10C, respectively. Both primary components, which are mapped on the continental scale, mainly depend on the occurrence of sedimentary strata. Although fractured crystalline (igneous and metamorphic) rocks may be locally porous, reservoirs are assigned a higher COS when sedimentary rocks are present in the subsurface because sedimentary rocks typically exhibit higher porosity and more predictable permeability. In general, although reservoir quality is critically important for the development of fields and placement of wells, we do not view reservoir quality as particularly risky on the continental or play scale, which are focused on regions of high prospectivity but not the specific placement of wells. The seal component is viewed as generally riskier than the reservoir component but follows similar logic and the map is created using similar underlying data. Hydrogen is more diffusive than methane and heavier hydrocarbons, and competent seals are critical for retaining subsurface accumulations (Maiga and others, 2024).

A stochastic uncertainty analysis was performed, varying the COS of the primary components and their underlying subcomponents, which generally results in a bulk shift to higher or lower COS for the composite map with only small variations from region to region (fig. 1.11). The final median (P50) hydrogen prospectivity map of the conterminous United States is shown in figure 11. Reflecting the significant uncertainty remaining in subsurface hydrogen potential, there are no areas in the United States with a COS of either 1 (100 percent certain sufficiency) or 0 (100 percent certainty of insufficiency). Instead, values range from 0.15 to 0.85, with a mean of 0.54 (refer to fig. 11). A COS of 0.5 generally indicates equal chances of sufficiency and insufficiency and, like a coin flip, indicates little or no preferential weight in either outcome. The COS values are generally higher than 0.5 in the region of the North American craton that contains sedimentary subsurface rocks. Overall, the highest COS values for hydrogen prospectivity occur in Michigan; parts of Kansas; north Texas; southern Oklahoma; on the western side of the Williston Basin of North Dakota and eastern Montana; and near the Four Corners region of Arizona, Utah, Colorado, and New Mexico (fig. 11). Additionally, the region intersecting western Kentucky and southern Illinois overlies the New Madrid seismic zone, the ancient Reelfoot rift, and a mineralized fluorspar district associated with serpentinized ultramafic intrusions of Permian age (Smith-Schmitz and Appold, 2022) and has hydrogen prospectivity values reaching 0.85.

Discussion

Recent exploration in the United States has focused on the region of the midcontinent rift in Kansas, Nebraska, and Iowa. A four-way, structurally elevated closure exists at the basement level called the Nemaha High where high concentrations of hydrogen were observed in several wells in the 1980s through the early 2000s. Previous studies concluded that the hydrogen in these wells was sourced from kimberlites west of the feature (Coveney and others, 1987) or from iron oxidation within the underlying fractured Archean and Proterozoic basement (Guélard and others, 2017). Our analysis also supports the alternative hypotheses that hydrogen originates from the serpentinization of deep-seated, ultramafic rocks in the core of the midcontinent rift or from the migration of deep-sourced hydrogen along the rift border faults. In either of these scenarios, migration pathways generated using a basement structure map (described in the “Lateral Migration of Subsurface Fluids” subsection of this report) indicate that updip lateral migration focuses fluid flow directly from the midcontinent rift towards tested well locations (fig. 12A). Similarly, high concentrations of hydrogen (96 mole percent hydrogen [mol% H₂]) have been observed in the Hofmann #3 well in Iowa (Moore and Sigler, 1987). This well location is along the west flank of the midcontinent rift and in an area where lateral migration flow lines converge before flowing updip to the northwest (fig. 12A). The regions along the midcontinent rift, especially those regions where migration is focusing, are likely favorable for hydrogen and have some of the highest composite COS in our prospectivity analysis. However, reliance on crystalline rocks for the reservoir and seal components reduces prospectivity for accumulations in the farthest northern part of the midcontinent rift in Minnesota (fig. 11).

Although hydrogen occurrences have been discovered in wells targeting other natural gases or fluids (for example, water, helium, and hydrocarbons), other detections have been made with soil gas surveys. Recently, high concentrations of hydrogen in soil gas have been observed in shallow, ellipsoidal surface depressions known colloquially as “fairy circles” or Carolina bays (Zgonnik and others, 2015). These features have been theorized to be formed by processes as varied as impact craters; thermokarst lakes; or wind, wave, or groundwater processes; however, their depth and aspect ratio preclude compatibility with impact craters and thermokarst lake processes (Lundine and Trembanis, 2021). Although these features are not always associated with hydrogen degassing, a correlation between these features and elevated surface hydrogen observations has been widely supported, for example in Australia (Frery and others, 2021), Brazil (Moretti and others, 2021), Mali (Prinzhofer and others, 2018), Namibia (Moretti and others, 2022), Russia (Larin and others, 2015), and elsewhere in the United States (Zgonnik and others, 2015). Fluxes have been noted to vary rapidly throughout daily cycles (Moretti and others, 2021). The mechanism underlying the association has not been thoroughly established, although shallow transport models have sought to explain their periodicity (Myagkiy and others, 2020). In the United States, Carolina bays occur in some parts of Nebraska (known locally as the Nebraska “rainwater basins” or in Kansas as playa-lunette systems) and extensively along the eastern seaboard (Lundine and Trembanis, 2021).

The eastern seaboard of the United States is a passive margin created during the rifting of Pangea. Globally, many rifted margins are often associated with serpentinized mantle, and previous researchers have hypothesized that exhumed or otherwise serpentinized mantle rocks may exist offshore along the eastern seaboard of the United States at depth (Biari and others, 2017). We used the ECMA to delineate the rifted edge of continental crust and propose it as a potential spatial proxy for a source of subsurface hydrogen due to serpentinization. Lateral migration flow pathways from this potential source are shown in figure 12B and indicate flow updip from offshore to onshore. Many flow lines, and the drainage areas that encapsulate them (fig. 6), overlie the region of Carolina bays in the United States, from northern Florida to southern New Jersey (Lundine and Trembanis, 2021). North of the Baltimore Canyon Trough, a major basement depression off the coast of New Jersey (Grow and others, 1988), flow is redirected outboard, rather than inboard; this transition coincides with the disappearance of Carolina bays in the northeastern States. We hypothesize that Carolina bays, with sporadic hydrogen degassing, may represent the far inboard surface seepage of hydrogen that has migrated long-distance from deep, offshore serpentinized mantle-derived source rocks. If this is true, subsurface reservoirs may contain hydrogen that was trapped along this route near the coastline and offshore. Accordingly, relative increases in COS are shown in our prospectivity map, particularly along the coast of North Carolina, eastern Maryland, Delaware, and southern New Jersey, where the ECMA is particularly close to the shoreline (figs. 11 and 12). This hypothesis can be further tested by examining subsurface stratigraphic relations from offshore to onshore to characterize potential carrier beds with sufficient porosity and permeability to facilitate this migration.

Summary

Although much remains unknown and untested in the geologic hydrogen system, this study seeks to integrate basic concepts for hydrogen generation, migration, and storage to provide a useful and publicly available methodology for continental-scale mapping of geologic hydrogen prospectivity. The method has been applied in the conterminous United States and provides critical information to guide further detailed studies. Our methodology leverages decades of exploration tools and concept development from other subsurface resources to quantify spatial trends in prospectivity. As more is learned about the hydrogen system, this flexible and open methodology can be easily adapted to consider alternative or additional ideas and can be applied at a variety of spatial scales in other regions of the Earth and potentially other planets.