Soil surface flux measurements are a reliable means for assessing fugitive gas migration across soils and seasons

In recent years, concerns have arisen surrounding production of oil and gas and fugitive CH₄ emissions from aging wells (Abboud et al., 2020; Schiffner et al., 2021). As wells are drilled to match demand, and existing wells degrade with age, the impact of fugitive CH₄ emissions may become more prolific (Dusseault et al., 2014; Ingraffea et al., 2014; Johnson et al., 2017).

Fugitive gas migration (GM) is a form of wellbore leakage that occurs when natural gas, largely composed of CH₄, seeps past the outermost casing string of a well and infiltrates the surrounding environment. Leakage is generally due to a loss of structural integrity of protective zonal isolations along a wellbore (Watson and Bachu, 2007; Watson and Bachu, 2008; Bachu, 2017). Migrating gas from the wellbore can impact groundwater, soils, and result in emissions to the atmosphere.

Because natural soils are highly variable, it can be a challenge to detect and quantify GM. Subtle changes in the local soil environment can strongly influence measurements, even over short distances, and introduce substantial uncertainty. In the field, there are 2 common measurement techniques:

1. Measurement of in-soil gas concentration, known as the “bar hole” method (Marrin, 1988; Schroder et al., 2016). This method involves a shallow well drilled into the soil to measure the concentration of gaseous species at a given depth (often 50 cm).

2. Measurement of soil surface gas flux using flux chambers (Erno and Schmitz, 1996; Kang et al., 2014; Schout et al., 2019). Flux chambers measure the concentration change over time of a gaseous species at the soil surface over a defined surface area, allowing for the calculation of mass transfer per unit area per unit time (flux rate), which can be readily converted to a volumetric emission rate.

Currently, the in-soil concentration technique is the industry standard in Canada. This method is recommended by the Alberta Energy Regulator (AER; Figure 1) and mandated by the British Columbia Oil and Gas Commission (BCOGC) for Well Abandonment GM assessments (Alberta Energy and Utilities Board, 2003; AER, 2021a, 2021b; BCOGC, 2021). In-soil measurements can serve to identify the spatial extent of a gas plume in the subsurface. Soil probes can also be placed at different depths to build vertical gas concentration profiles and give insight on the behavior of soil processes such as oxidation and biological production (Riveros-Iregui et al., 2008). However, in-soil techniques are intrusive, and probe gas can be contaminated with atmospheric air during installation and sampling, particularly when large volumes are drawn. In-soil gas concentration can vary widely due to the immense spatiotemporal variability of soil properties like texture, porosity and temperature, as well as moisture content and microbial activity (Marrin, 1988; Pumpanen et al., 2003).

In-soil CH₄ measurements are used to determine GM incidence, but they cannot directly measure GM severity as a volumetric emission rate. This would require the calculation of soil gas diffusivity and/or permeability, which is prone to significant error. As a result, GM emission rates in Canada are generally not reported volumetrically, as would be standard for other fugitive CH₄ sources.

Soil surface flux measurements are not the regulatory standard in Canada, but they have been used in various GM studies (Erno and Schmitz, 1996; Kang et al., 2014; Schout et al., 2019; Fleming et al., 2021). Flux chamber surveys are nonintrusive and determine gas flux rates directly by measuring the rate at which concentration changes inside the chamber headspace (Riveros-Iregui et al., 2008; Pumpanen et al., 2004; Mora and Raich, 2007; Nickerson and Risk, 2009; Nickerson et al., 2014). Chambers can be moved around the surface to help delineate the spatial distribution of gas at a well site (Forde et al., 2018; Forde et al., 2019). Flux chamber measurements are affected by factors such as wind, landscape morphology, and interactions with site vegetation (Nickerson and Risk, 2013; Schroder et al., 2016). In cold climates, heavy snow cover and frozen soils can negatively impact measurement reliability (Pirk et al, 2016), though in more moderate winter settings, these systems can be used successfully. Temperature inversions and pressure differentials at the soil-atmosphere boundary as well as variable barometric pressure can also influence the magnitude of surface flux rates (Klusman and Jaacks, 1987; Forde et al., 2019).

There are numerous studies that use in-soil and surficial techniques to monitor gases moving to atmosphere (Riveros-Iregui et al., 2008, Pumpanen et al., 2003; Chadwick et al., 2009; Mathieson et al., 2010; Jenkins et al., 2012; Romanak et al., 2012; Schroder et al., 2016). However, the current literature lacks a direct comparison of the relative effectiveness and sensitivity of each technique and their application to GM affected oil and gas well sites, including the potential for false positive or negative detections.

In this study, we compared in-soil concentration and surface flux measurements as methods for detecting fugitive GM and its severity. We hypothesize that surface flux rates will yield more useful and accurate information on GM because conservation of mass dictates that CH₄ leakage into the bottom of a shallow soil should generally equal CH₄ out to the atmosphere. We further hypothesize that in-soil CH₄ concentration will vary more as a function of soil texture and environmental conditions.

We used a one-dimensional (1D) diffusive gas transport model to examine how different soil textures (clay, loam, sand), leak rate of CH₄ into the soil profile, and soil environmental conditions (water content, temperature) affect in-soil CH₄ concentration and soil surface CH₄ flux rates in a near-surface soil system. A model environment allowed us to explore a wider range of scenarios than we could see in the field. We focused our study on surface soils receiving inputs of GM because the surface soil is where both flux chamber and in-soil measurement methods are applied, and because in the first meter of soil, we can largely ignore the effects of lateral and low-frequency pressure advection.

Lateral migration of CH₄ has been demonstrated in GM scenarios where low gas permeability confining layers in the subsurface induce horizontal flow (Forde et al., 2018). However, unsaturated, near-surface soils above approximately 1 m are still thought to be dominated by vertical, 1D diffusive transport where gases exploit preferential pathways formed by macro pores, root decay, and heavy rain percolation (Huggett, 1975; Kühne et al., 2012; Creelman et al., 2013).

Other studies including Lyman et al. (2020) and Forde et al. (2019) have suggested that synoptic, low frequency variations in atmospheric pressure may have significant impacts on surface CH₄ flux rates, particularly where GM originates from the deep subsurface. Our study assumed constant CH₄ leak rates into a shallow synthetic system and was intended to capture the variability produced by measurement methodology alone. In the very near-surface soil system, higher frequency pressure variations are more likely to interrupt the diffusive regime, such as when wind speed is high (Bowling and Massman, 2011). Furthermore, longer timescale, low frequency pressure variations would have a limited effect on near-surface soils because of their limited pore gas storage volume and limited capacity for storage flux. Near-surface perturbations in soils and other porous media can be modeled using diffusive–advective approaches, although 1D diffusive models have also been shown to represent such systems with a high degree of realism (Graham and Risk, 2018).

Our 1D diffusive soil gas transport model was a variant of a model used in previous studies, including Nickerson and Risk (2009). The model domain comprised a 1 m deep soil profile subdivided into 10 uniform layers (0.1 m depth) with an overlying atmospheric layer. We simulated CH₄ leaking into the bottom of the soil profile, plus natural production of CH₄ within the profile, and the diffusive transport of both CH₄ sources through the soil and out to the atmosphere. The primary model outputs were:

1. In-soil concentration of CH₄ within each soil layer (μmol mol^–1 or ppm).

2. Flux of CH₄ between soil layers, and from the uppermost soil layer to the atmospheric layer (μmol m^–2 s^–1).

We used the model to understand what a service provider should expect to observe under field conditions when using in-soil or surface flux measurements to make assessments of GM incidence and severity. The CH₄ leak rate into the soil profile, soil texture, and soil environmental conditions (water content, temperature) were varied to test the effects on soil CH₄ transport behaviors and distribution within the profile, and to the atmosphere.

There were 8 parameters defined in the model:

1. Atmospheric CH₄ concentration (C_a).

2. Biological CH₄ production (P) within the soil profile.

3. e-folding depth (η); biological CH₄ production is concentrated in the uppermost portion of the soil and η describes the depth at which P falls to 1/e of its value at the soil surface (Bowling et al., 2015).

4. CH₄ Leak Rate (L_R); volumetric flow rate of CH₄ gas entering the bottom of the soil profile.

5. Porosity (ϕ); each of the 3 soil textures was defined by its air-filled porosity.

6. Soil CH₄ Diffusivity Rate (D_R); diffusion coefficient for each soil texture, derived from CH₄ diffusion in media with porosity ϕ versus diffusion in free air and water (Millington and Quirk, 1961; McCarthy and Johnson, 1995).

7. Depth (z) of the soil profile.

8. Duration of the simulation (O).

For a complete description of model parameters and equations, refer to Supplemental Information (SI), for parameters and units, see Figure 2.

Modeling was conducted under 3 different scenarios:

1. Homogenous steady state (all parameters constant with depth). Homogenous conditions allowed us to observe 1D diffusive transport in a generalized shallow soil system.

2. Nonhomogenous steady state (P, η, ϕ, and D_R variable with depth). Nonhomogenous scenarios simulated natural soil layering, where biological CH₄ production (and e-folding depth), porosity, and CH₄ diffusivity rate vary within the soil profile.

3. Transient conditions (P, η, ϕ, and D_R variable in time). In transient scenarios, parameters were varied to reflect soil environmental conditions (water content, temperature) of 5 Alberta ecoregions, the province with the highest density of oil and gas production (Downing and Pettapiece, 2006; Turetsky et al., 2014). Variation in biological CH₄ production simulated annual and diel temperature changes in each ecoregion as the two are strongly coupled (Davidson and Janssens, 2006). Variation in porosity simulated changes to soil water content driven by precipitation patterns in each ecoregion. As water content reduced or increased air-filled pore space, the CH₄ diffusivity rate was affected.

GM affected soils with active CH₄ leak rates (L_R > 0) were compared to soils without GM (L_R = 0), to evaluate the effectiveness of both measurement methodologies in a simulated natural baseline scenario, where we attempted to differentiate between CH₄ signals from GM seepage and natural production.

We also evaluated the impacts in unique conditions such as developed muskeg soils. For further details on scenarios, boundary conditions, and governing equations, see Figure 2 and SI.

While biological CH₄ production (methanogenesis) was modeled, methanotrophy and certain other chemical reactions that occur within the soil profile were not considered. This has the potential to bias our model outputs toward higher in-soil CH₄ concentrations and surface CH₄ flux rates than would be the case with oxidation to CO₂ (Smith et al., 2018), but only to a modest degree given the shallow soil profile being tested. We also did not consider ebullition, a form of nondiffusive transport that can dominate in saturated muskeg soils.

The theoretical model shows that parameters C_a, η, P, L_R, and z had a positive relationship with in-soil CH₄ concentration. Soil D_R was inversely related to in-soil concentration, and as D_R increased, in-soil CH₄ decreased because soils could more readily transport CH₄ molecules to atmosphere. The only parameter that did not influence in-soil concentration at steady state was ϕ, as it simply represents temporary gas holding capacity of the soil.

Surface CH₄ flux rates showed fewer dependencies on model parameters, and only P and L_R affected flux rates at steady state. Soil ϕ and D_R were unimportant because, by definition, flux-in must equal flux-out, this being the principal aspect of mass conservation. Changes to in-soil concentration are simply the mechanism by which soil adjusts its internal concentration gradient to satisfy conservation of mass. First order relationships between flux and input parameters as identified in Table 1 reflect well-known relationships that exist in diffusive soil systems as governed by Fick’s second law and compared to other soil gas transport models and field studies (Pumpanen et al., 2003; Nickerson and Risk, 2009; Creelman et al., 2013; Smith et al., 2018).

Figure 3a displays in-soil CH₄ concentration results from steady state homogenous scenarios, where all parameters were constant through depth. Three soil textures were tested: clay, loam, and sand, with D_R = 10^–7, 10^–6, 10^–5 (m² s^–1) respectively. The L_R into the bottom of the soil profile was constant at 1.0 m³ CH₄ day^–1 and applied to each texture until the system reached equilibrium, at which point in-soil concentrations were measured through depth. In the more diffusive sand, in-soil CH₄ concentration barely exceeded 1,000 ppm, whereas in less diffusive clay, in-soil concentration was 2 orders of magnitude higher, consistent with the difference in D_R, at 100,000 ppm (Figure 3a).

Consider in-soil GM surveys conducted at 2 well sites with differing soil textures and equal amounts of CH₄ seepage into the upper soil column. Measurements taken at the clay site might reveal highly elevated in-soil CH₄ concentration which could be interpreted as a severe GM problem, causing a disproportionate number of resources to be allocated for remediation of the site. In contrast, measurements at the sand site would barely exceed detection threshold and may cause emissions to the atmosphere to be overlooked.

Figure 3b shows how nonhomogenous, layered soils can give rise to complex patterns of concentration through depth, which could complicate interpretation. Nonhomogenous soils had different CH₄ concentrations at all depths (under the same L_R) compared to uniform soils that were homogenous through depth (Figure 3a), and concentrations were seen to shift depending on the relative D_R of each layer and layering order.

Steady state model scenarios demonstrate that CH₄ L_R into the soil profile was not the primary determinant of in-soil CH₄ concentration. We observed that in-soil CH₄ concentrations increased with depth as the measurement approached the CH₄ source. Overall, we found that in-soil CH₄ concentration was mostly regulated by soil texture and D_R.

Transient conditions are more reflective of the real world, where conditions change continuously (e.g., drying events, rainfall, temperature-driven biological reactions). We evaluated 2 scenarios to determine the effect of changing soil environmental conditions, specifically soil water content and temperature, on measurement of CH₄ using in-soil concentration and surface flux methods:

1. The effect of single ecoregion (Fescue Grassland [FG]) soil environmental conditions applied to 3 soil textures (clay, loam, and sand).

2. The effect of multiple ecoregion (Mixed Grasslands, Moist Mixed Grasslands, FGs, Boreal Transition, and Peace Lowlands) soil environmental conditions applied to one soil texture (loam).

Figure 4 depicts the model results for in-soil CH₄ concentration and surface CH₄ flux rates from different soil textures (clay, loam, and sand) under soil environmental conditions of Alberta’s FG ecoregion. Two constant L_R, 0.1 and 1.0 m³ CH₄ day^–1, were applied. Following the regulatory guidance for real-world GM surveys, in-soil concentration was measured at a depth of 50 cm (Figure 1).

We observed that changes to in-soil CH₄ concentration were driven by soil environmental conditions, as D_R varied in response to soil water content from precipitation, and biological CH₄ production varied with temperature (Figure 4a). In contrast, flux measurements from all soil textures varied to a much lesser degree as conditions changed.

Figure 5 demonstrates the effect of variation in soil water content and temperature reflecting the seasonal patterns of 5 Alberta ecoregions on in-soil CH₄ concentration and surface CH₄ flux rates in 1 soil texture, loam. Two constant L_R were applied, 0.1 and 1.0 m³ CH₄ day^–1. For this transient scenario, in-soil measurements from all ecoregions varied up to 79% (Figure 5a and b). Surface flux rates, though seasonal, were much less variable at around 20% (Figure 5c and d).

Transient model results demonstrate that in-soil CH₄ concentrations were more affected by changing soil environmental conditions than CH₄ flux rates. When in-soil concentrations are measured during different times in the year (Figures 4a, 5a and b), the larger fluctuation can complicate interpretation of the results. In Figure 4a, for example, clay soil under the low L_R yielded higher in-soil CH₄ values at peak than sand soil under the high L_R at its lowest point. Similarly in Figure 5a and b, peak in-soil values from the Peace Lowlands ecoregion under the low L_R exceed low period in-soil CH₄ concentrations under the high L_R for the other ecoregions. Surface flux rates were much easier to interpret, and the high L_R always resulted in higher flux rates from all soil textures.

In the real world, temperature and precipitation patterns can vary substantially year to year, particularly in drier climates with erratic precipitation events. Adjusting for these variables would require a huge amount of information on the potential range of in-soil CH₄ that could be expected in each area and could easily result in the over or underestimation of GM severity based on local conditions at the time of the survey. It’s conceivable that misinterpretation of in-soil measurements could again lead to improper allocation of resources for remediation of GM problem sites and allow higher emitting sites to go undetected. Instead, the results suggest that flux measurements can more easily capture GM severity and require less information on seasonally variable, site-specific conditions.

We also investigated a baseline detection approach. Baseline studies are used to evaluate natural CH₄ abundance at a given site to later predict the presence or absence of GM at that location. Figure 6 displays transient model results for in-soil CH₄ concentrations (Figure 6a) and soil surface CH₄ flux rates (Figure 6b) for 3 soil textures, each with 2 possible GM conditions: (1) soils with GM, where L_R = 0.1 m³ CH₄ day^–1; (2) soils without GM, where L_R = 0.

We found that in-soil CH₄ concentrations varied widely and were controlled by texture and soil environmental conditions. Like the other transient scenarios (Figures 4 and 5), in-soil results showed that the severity of GM, or in this case, the presence of GM can be difficult to interpret. At their peak, in-soil CH₄ from clay soils without GM exceeded the lower concentration range of loam and sand textures with GM. Surface flux measurements were able to differentiate between the GM affected soil and the natural baseline across all textures and at all times of the year. Soil texture can vary over short distances in the field, and conditions change continuously, meaning that a natural baseline approach relying on in-soil measurements could lead to false negative detections when more diffusive soil textures are present.

Under current regulation, in-soil CH₄ measurements are relied upon to assess GM incidence and its severity, but the model showed that we cannot assume to understand severity from these measurements alone. We found that in-soil CH₄ concentration was strongly dependent on soil texture and environmental conditions. Surface flux rates did vary across textures and conditions, but the measurements were more useful for determining the quantity of CH₄ leaking into the soil.

If in-soil measurements were useful for evaluating leak severity, we should expect that higher in-soil concentrations be associated with elevated surface flux rates, thus indicating greater emissions to the atmosphere. Instead, we hypothesize that in-soil concentration is more reflective of soil texture and local environmental conditions.

To explore the relationship between in-soil gas concentration and surface flux rates, we analyzed soil carbon dioxide (CO₂) field measurements collected for an earlier study at the Weyburn–Midale oilfield in southeastern Saskatchewan. CO₂ and CH₄ are distinct gases, but both are subject to comparable trace gas transport mechanisms at the soil-atmosphere boundary (Conrad, 1996; Jassal et al., 2004; Wu et al., 2010). Like CH₄, CO₂ is present in surficial soils at widely varying concentrations that tend to fluctuate based on textural and seasonal conditions of the local soil environment (Bogner et al., 1997; Jassal et al., 2004; Jassal et al., 2005; Stolp et al., 2006).

In-soil CO₂ concentration and soil surface CO₂ flux were measured over 11 months (August 2011 to June 2012) as part of a screening program for potential surface effects of Enhanced Oil Recovery CO₂ injection. Three sites were sampled, each with unique land-use characteristics. Site 1 was located on farmland adjacent to the injection field, whereas Sites 2 and 3 were situated within the injected region on more industrial grounds. Soils at each site were unique in texture, vegetation coverage, compaction, drainage, and potential water content, all of which could influence CO₂ flux rates and in-soil CO₂ concentrations (Nickerson and Risk, 2013; Schroder et al., 2016). In-soil CO₂ concentration was measured monthly from bar hole wells positioned from 0 to 75 cm depth along the soil profile. Surface CO₂ flux was measured using Eosense^© automated flux chambers over a near-continuous interval (approximately 30 min). For more detailed site descriptions and methodology related to measurement hardware, see Risk et al. (2013).

Figure 7 displays 11 months of in-soil CO₂ concentration and CO₂ flux measurements at the Weyburn field region. Both measures varied seasonally and from site to site. However, flux rates varied much less between sites, and were more predictable across seasons, being persistently lower during the cold, dry winter months and higher during the warm season. In-soil CO₂ varied substantially between sites and was erratic month to month, with slight evidence of an increasing trend in winter.

Linear regression revealed only a weak to moderate relationship between in-soil concentration and surface flux values Sites 2 and 3. However, in-soil CO₂ was much lower at these locations than Site 1, where in-soil values were at times 300%–400% higher, and where we saw no associated increase in surface flux rates. Based on the underlying transport theory and our model results, it seems likely that soils at Sites 2 and 3 were simply more gas diffusive and less able to accumulate CO₂. Site 1 fits the model-described behavior of a less diffusive soil, enabling very high in-soil concentration during periods of the year in which soil water content was elevated.

These findings demonstrate the need to be wary of relying on in-soil measurements to predict the presence or severity of GM at such sites because low in-soil readings could result from a highly diffusive soil, and soil gas emissions may be very large.

Oil and gas developments in Alberta are often located in areas of water-laden muskeg, a native marshy soil. When drill sites are developed in these areas, the muskeg must be excavated and replaced with a compact, stable fill or partially excavated and infilled in a manner that ensures an acceptable degree of settlement to create a stable foundation. Infill is usually comprised of a low gas permeability clay. Depending on certain site characteristics or well pad traffic requirements, an additional foundational layer of gravel would be placed above the clay. Due to water saturation, the muskeg itself forms a low gas diffusivity layer that will likely accumulate high in-soil concentrations of CH₄. Depending on the gas diffusivity of the overlying well pad foundation layer, or the degree of water saturation in the muskeg layer, the low gas permeability material could act either as a barrier or conduit for gas flow to surface.

If the clay layer has low gas diffusivity relative to the underlying muskeg (muskeg is less saturated than normal), this foundational layer could act as a barrier to gas flow, encouraging lateral transport (Figure 8a). Alternatively, if the diffusion through the clay layer is higher relative to the underlying muskeg, this clay layer could act as a conduit and vertical diffusion will dominate the profile, transporting plumes of CH₄ that have accumulated within the muskeg to the soil surface (Figure 8b).

Based on the typical soil gas diffusivities for muskeg (e.g., 10^–9) compared to low diffusivity clay (e.g., 10^–7), the model results suggest that we should expect lower in-soil concentrations in this instance but that soil surface flux rates would be relatively unaffected. These same considerations extend to nonmuskeg well pads, which are often highly compacted (lower gas diffusivity). On the other hand, they are often backfilled with coarser substrate than adjacent areas (higher gas diffusivity). The gas diffusivity for normal well pads will likely differ from that of native soils, but the difference from native conditions will probably be less extreme than in muskeg conditions. Field studies have shown that the presence of heterogeneous and low diffusivity/permeability soils can result in surface CH₄ fluxes occurring farther away from the wellhead due to lateral gas transport (Forde et al. 2019). Increasing spatial coverage over the entire well pad when taking surface flux measurements will lead to a better representation of the extent of lateral GM.

Due to the spatiotemporal variability of soils and measurement uncertainties, the detection and quantification of fugitive GM from compromised oil and gas wells remains difficult. We sought out to model the methodological uncertainties associated with 2 GM measurement techniques.

Model results showed that in-soil CH₄ concentration measurements were strongly dependent on soil texture and environmental conditions and did not necessarily reflect the quantity of CH₄ leaking into the soil profile from below. We demonstrated how relying on in-soil values as a metric of GM incidence and severity could produce false negative readings and cause problem sites to be overlooked. Soil surface CH₄ flux measurements were much more stable across differing soil textures and changing environmental conditions, which could improve interpretation of results from GM surveys. These findings clearly demonstrate that we need to choose measurement methodology carefully, so as not to add additional uncertainty to an already variable process.

Based on the model results and supplementary field measurements, we propose that regulatory guidance could be improved in jurisdictions like Alberta, Canada, where in-soil measurement for GM is specified, but where GM issues could be better addressed and quantified with surface flux measurements using flux chambers. Increasing the quantity of reliable flux measurements at the well pad will allow for the quantification of volumetric emissions from GM, the development of inventories, and help stakeholders reduce CH₄ emissions.

All data from which results were derived for this study, including model results and field measurements, have been included as SI with this manuscript.

The supplemental files for this article can be found as follows:

Numerical Model Outputs and Figure Data. Zip

Supplemental Information. Docx

We extend our thanks to the anonymous reviewers whose time and expertise helped to greatly improve the quality of our manuscript.

Funding for this research was provided by the Natural Sciences and Engineering Research Council of Canada Collaborative Research and Development Grant to St. Francis Xavier University and Eosense.

The authors declare that they have no competing interests.

Contributed to conception and design: FH, MA, RL, DM, DR, NN.

Contributed to acquisition of data: FH, MA, DR, NN.

Contributed to analysis and interpretation of data: MA, FH, RL, DR, NN.

Drafted and/or revised the article: MA, FH, RL, DM, DR.

Approved the submitted version for publication: MA, DM, RL, FH, DR, NN.

How to cite this article: Argento, M, Henderson, F, Lewis, R, Mallyon, DA, Risk, D, Nickerson, N. 2022. Soil surface flux measurements are a reliable means for assessing fugitive gas migration across soils and seasons. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2021.00010

Domain Editor-in-Chief: Steven Allison, University of California, Irvine, CA, USA

Associate Editor: Peter Homyak, University of California, Riverside, CA, USA

Knowledge Domain: Ecology and Earth Systems