Natural source zone depletion predicts finite longevity for petroleum products in soil and groundwater

Abstract

From environmental concerns to economics, managing inadvertent releases of petroleum products in soils and groundwater is a significant concern. Here, estimates of specific petroleum product volumes ranging from 80,000 to 800,000 L/hectare are developed using cryogenic coring techniques at 5 sites. Heat generated from natural petroleum hydrocarbon oxidation at 10 sites provides daily estimates of Natural Source Zone Depletion (NSZD) over 0.43 to 3.07 years. NSZD rates ranging from 4,000 to 20,000 L/hectare/year are relatively constant in time and apparently independent of specific product volumes, soil permeabilities, and depth. A weak correlation between NSZD rates and source zone temperatures suggests that microbially mediated oxidation is controlled by electron transport from electron donors (hydrocarbons) to atmospheric oxygen (terminal electron acceptor). Data-driven regression of daily NSZD rates yields source zone longevity estimates of one to four decades. At two sites with active product recovery, NSZD outperformed recent conventional recovery efforts by 96% and 99%. These findings demonstrate the importance of NSZD as a key remedy component at mature sites, limitations of conventional recovery, and the potential of enhancing NSZD rates. Source zone longevity is a new quantifiable metric, that can support informed decisions for managing petroleum products in soils and groundwater.

Introduction

There is good news in the world of soil and groundwater remediation. Natural Attenuation (NA) commonly mitigates risks associated with many contaminants in groundwater plumes^1,2. Furthermore, broad recognition has come to the fact that Natural Source Zone Depletion (NSZD) can drive consequential depletion of petroleum liquids in source zones^3,4. Per the National Research Council² source zones are shallow subsurface bodies containing NAPLs. Evermore, cognitively, or not, we are relying on the assimilative capacities of subsurface settings to mitigate risks associated with contaminants in subsurface plumes and source zones.

Herein we focus on forecasting remaining refined petroleum liquids, defined as “products”, in source zones given ongoing NSZD. Common petroleum products include gasoline, diesel, and jet fuel. Crude oils, having varied compositions and API gravities, are not considered herein. The primary mechanism for product NSZD is methanogenesis^4–6. Following Fig. 1, critical aspects of methanogenesis include (1) methanogens converting petroleum hydrocarbons into methane and carbon dioxide in source zones, (2) produced gases moving upward through the vadose zone via advection and diffusion^3,7,8, and (3) methane in the vadose zone being converted to carbon dioxide, water and heat, via a mixture of aerobic and anerobic oxidation of methane^5,9. Lines of evidence for product NSZD include (4) observed fluxes of gas phase carbon dioxide, methane, and oxygen in the vadose zone^3,10, (5) heat generated via in-situ oxidation of hydrocarbons¹¹, (6) the presence of robust microbial communities relying on petroleum hydrocarbons as electron donors⁵, and (7) assimilative capacities of NSZD constraining lateral movement of product bodies¹².

Fig. 1.

Conceptualization of primary processes governing NSZD in soils and groundwater systems composed of transmissive and low permeability zones^3,13,14.

NSZD rates fall in a relatively narrow range of 1,000s to 10,000s L/hectare per year, independent of the types of petroleum LNAPL or hydrogeologic settings,^4,15. Narrow ranges of NSZD rates, given diverse physical conditions, raises the intriguing question of what controls NSZD rates. While the kinetics of methanogenesis can be slow, methanogenesis is not limited by the availability of an electron acceptor such as oxygen or sulfate⁴. Critically, NSZD occurs throughout NAPL bodies¹⁴. In contrast the efficacy of active remedies, much like upstream oil and gas production, are often constrained by poor sweep efficiencies about recovery systems leading to localized depletion of petroleum liquids^15,16. All too often, active remedies create localized “holes” in source zones, analogous to Swiss cheese, potentially doing little to reduce overall risks or site-scale longevity of source zones.

Herein, we advance a novel approach to continuously forecasting remaining refined petroleum products in soils and groundwater. Our focus is on NSZD as the primary process governing losses. We begin by estimating the amount of product present at a specified time (t₀) using cryogenic coring techniques¹⁷. Cryogenic coring holds the promise of reliable estimates of remaining product via (1) limiting drainage of NAPLs from core during recovery, (2) preventing samples from dropping out of sample systems during recovery, and (3) avoiding collections of non-representative expanded flowing sands. NAPL losses are quantified continuously through time by measuring the amount of heat generated via oxidation of petroleum hydrocarbons using multiple level in-situ temperature sensors linked to cloud-based data storage-analytics-visualization platforms¹¹. Estimates of remaining product are updated daily by subtracting cumulative NSZD losses from the amount of product present at t₀. Regressions of remaining product versus time are used to forecast times when essentially all remaining products will be depleted. Our vision is that forecasted times to near complete depletion of product source zones will help inform sound decisions for managing petroleum products in soils and groundwater.

Methods

Metrics for product volumes and NSZD rates

The amount of product present in the subsurface are described in terms of specific product volumes. A specific product volume is the volume of product (L³) per horizontal area (L²) of a source zone. Units used for specific product volumes include liters/hectare and meters. NSZD rates are defined in terms of changes in specific product volumes per time. Units used for NSZD rates are liters/hectare/year and meters/year. Length based specific volumes and NSZD rates follow the common metric of tracking product thicknesses at sites (albeit in monitoring wells). Specific product volumes and NSZD rates provide a basis for comparing the performance of NSZD to the performance of active remedies.

Sites

We consider 10 sites with historical releases of products in soils and groundwater. Site attributes are presented in Table 1. Five of the sites have specific product volume estimates from cryogenic coring and daily NSZD rates based on heat generated from NSZD. Based on a 10% standard deviation from triplicate cryogenic core collected at Site C, specific product volumes are reported to one significant figure. Five of the sites have continuous NSZD rates, without specific product volumes. Two of the sites have production data from active remedies.

Table 1.

Study site labels and attributes.

Site	Location	Geology/Qualitative Soil Permeability	Depth to Source Zones from Grade (m)	Product(s)	Concurrent Active Recovery (L/hectare /year) and (m/year)	Specific Product Volume (L/hectare) and m) at t0
A	USA	Mississippi River overbank silts and sands / Low	~ 1	Gasoline and diesel	4	800,000 0.08
B	USA	Coastal eolian sand dunes / High	~ 10	Intermediate Distillate Diluent	1,000	400,000 0.04
C	USA	Glacial valley train sand and gravel / High	~ 8	Gasoline and diesel	NA	400,000 0.04
D	USA	Braided fluvial sands and gravels / High	~ 1	Gasoline and diesel	NA	80,000 0.008
E	USA	Missouri River overbank silts and sands / Low	~ 5	Gasoline and diesel	NA	400,000 0.04
F – Paved	USA	Poorly sorted Intermountain basin fill / Moderate	~ 5	Gasoline and diesel	NA	NA
F –Unpaved	USA	Poorly sorted Intermountain basin fill / Moderate	~ 5	Gasoline and diesel	NA	NA
G	USA	Poorly sorted fine-grained coastal basin fill / Moderate	~ 4	Gasoline and diesel	NA	NA
H	USA	Poorly sorted Intermountain basin fill / Moderate	~ 3	Gasoline and diesel	NA	NA
I	USA	Gulf coast coastal plain sands and silts / Moderate	~ 2	Natural gas condensate	NA	NA

Cryogenic coring for specific product volumes

Photographs of cryogenically collected core are presented in Fig. 2(a-c). Cryogenic coring was conducted at all sites by Drilling Engineers, Fort Collins, Colorado. Drilling was conducted using a CME-55 drill rig and hollow stem augers (11 cm ID, 23 cm OD). Cores were collected in clear 5.7 cm ID PVC liners in continuous drives of 67 cm using a modified CME continuous sample barrel. Sample barrel modifications included an insulated internal dual wall cooling barrel and insulated lines for fully contained delivery and exhaust of liquid nitrogen. Core collection covered the entire interval where product was present. Rotating hollow stem augers with fixed sample barrels were advanced leading to “overcut” cores being pushed into soil liners.

Fig. 2.

Soil core collected using cryogenic coring techniques: (a) frozen core in core barrel, (b) frozen core in PVC sample liner, (c) expanded image of frozen water in the core, and (d) typical as-built details for temperature sensors.

Once the samples were in the liners, liquid nitrogen was circulated through the dual wall cooling barrel for 4–6 min. Liquid nitrogen was supplied via portable 100-liter Dewars. After in situ freezing (1) the lined sample barrel containing the frozen core was brought to the surface, (2) the liner containing the frozen core was pulled from the sample barrel, (3) the recovered core was capped and labeled, (4) sample recovery was recorded (length recovered/length of the drive), and (5) core was promptly placed in onsite coolers filled with dry ice (-78 °C). Coolers with frozen cores and dry ice were shipped overnight to a Laboratory at Colorado State University and promptly transferred to a -80 °C freezer.

Frozen cores were subsampled and analyzed at Colorado State University. Frozen core (-80 °C) were cut into 2.6 cm-thick subsamples (“hockey pucks”) at intervals of 7.5 to 15 cm. The cores were cut using a chop saw (DeWalt) equipped with a circular 36 cm diamond-tipped masonry blade. Core subsamples were (1) visually logged by a professional geologist (2) photographed under visible and UV light (3) used to resolve physical properties including porosity and fluid saturations, and (4) analyzed to quantify total concentrations of methane, benzene, gasoline range organics (GRO), and diesel range organic (DRO). Further details regarding collection and analysis of cryogenic cores are described in journal articles and reports^17–20.

Factors motivating the use of cryogenic coring techniques include in situ retention of pore fluids, preservation of volatile compounds, preservation of RNA^20,21, and reliable estimation of LNAPL mass-in-place. Preservation of the core attributes of concern is facilitated by (1) preventing losses of unconsolidated soil during sampling recovery and (2) temporally freezing the formation below the sampled interval, helping to stabilized flowing sands¹⁷.

Continuous NSZD rates from temperature data

Strings of 8 to 23 multiple level temperature and ORP sensors were installed in the cryogenic coring holes from near grade through the interval impacted by product. Intervals between sensors range from 0.1524 to 1.24 m. Multiple level temperature data provides a basis for resolving heat generated through oxidation of product constituents and correspondingly, NSZD rates. Much like composting organic waste or a burning wood in a woodstove, NSZD produces heat at rates proportional to the quantity of material that is oxidized¹⁷.

Sensor hardware was provided by S3NSE Technologies Inc. (Fort Collins, Colorado). Each string has digital temperature sensors (± 0.1 °C; DS18B20, Adafruit Industries). All sensors are on a single four-strand data wire. Sensors are mounted on solid PVC rod or 5.08 cm ID PVC pipe. Annular space between the rods or PVC pipe and the boring wall are filled with (1) a uniform fine-grained quartz sand or (2) natural formation collapse. A 1-meter hydrated bentonite seal is placed at the top.

Data from each sensor is uploaded hourly via cellular connections to Ubidots™; a cloud-based data storage, analytics, and visualization platform. Data collection and wireless communication is controlled by a Particle Electron™ microprocessor (E260KIT, Particle Industries, Inc.). Transformations of temperature data to NSZD rates are based on subsurface temperatures being a function of surface heating and cooling and heat associated with NSZD. Given subsurface temperature at two points, a system of two-equation two-unknown is iteratively solved and tested to resolve heat associated with surface heating and cooling and NSZD. Complete documentation of methods used to collect temperature data and resolve daily NSZD rates are described in prior peer-reviewed study¹¹. Compared with NSZD rates derived from thermal methods, gas-flux approaches yield estimates within a similar overall range^4,11,22. Typical as built details, pictures, and temperature data from arrays sensors are presented in Fig. 2(d).

Forecasting

Specific product volumes at the time of core collection are calculated as

1

where M is the number of sub-core sections for a boring location, is the LNAPL saturation (L³ LNAPL /L³ pore space), is porosity (L³ pore space/L³ porous media), is the representative height (L) of the subsample and is the time when the core was collected, set to zero for all sites.

Estimates of the specific product volumes through time are calculated as

2

where is the remaining specific product volume (L) at time t (T), is the number of times steps from to , is the location specific product volume (L) at , and is the observed NSZD rate (L/T) for the period . Following upstream oil and gas decline curve analyses techniques²³, regression of vs. time provides a basis for forecasting remaining product as a function of time including the time to near complete product depletion. Key assumptions associated with Eq. 2 include: 1) no significant net inflow or outflow of product, 2), no heat source other than surface heating-cooling and NSZD, and 3) no significant ongoing measures to deplete product. Where active recovery occurs uniformly through areas of concern and does not impact NSZD rates assumption 3 can be addressed by superimposing losses associated with NSZD with active remedies²⁴. Mathematically, cumulative site wide product depletion (L³) are defined as

3

where is the area in which NSZD is occurring at time step , is the number of times steps in which active recovery is measured, is the average site wide product depletion rate for time step . Lastly the fraction of ongoing product depletion attributable to NSZD can be estimated as

4

Results

Figure 3 presents decline plots (remaining product vs. time) and estimates of product longevity for the five sites with initial specific product volumes. The first column in Fig. 3 uses identical data ranges on the-x axis (time) and y-axis (specific product volumes). Identical data ranges provide a basis for inter-site comparisons of initial specific product volumes, NSZD rates (slope = ) and times to product depletion (x intercept). Following Table 1, initial specific LNAPL volumes range over one order of magnitude from to liters/hectare ( to meter).

Fig. 3.

Sites with estimates of time zero specific LNAPL volumes. First column, decline plots with matching axis. Second column, decline plots with axis set to data ranges. Third column, estimates of time to near complete product depletion based on linear regressions of decline data from increasing (daily) periods.

The second column in Fig. 3 and, all plots in Fig. 4 without specific LNAPL volumes, present decline plots using the periods of record for each site. Plotting data across periods of record facilitates evaluation of temporal changes in NSZD rates through time. NSZD rates based on linear regression of the data in Figs. 3 and 4 range from to liters/hectare/year ( to meter/year). Values are reported to one significant figure based on the accuracy of thermal methods for estimating NSZD rates¹¹. Times to near complete product depletion range from 13.7 to 41.5 years. Rounding to 2 significant figures, R² values for the linear regressions range from 0.91 to 1.00 with a mean and standard deviation of 0.98 and 0.0057, respectively.

Fig. 4.

Specific product volume decline plots for sites without specific product volumes.

The third column in Fig. 3 presents evolving estimates of product longevity based on linear regressions using 1 day increases in the regressed data sets. Initial forecasts are based on regressions using 2 days of data (0.0055 years). Subsequent forecasts add one day of records to the regressed record with a maximum period of 3.07 years. Estimates of product longevity at times less than 0.5 year are large. Large early time longevity estimates are attributed to low estimates of NSZD rates using short records of subsurface temperature. Methods used to estimate NSZD rates from temperature data require approximately 3 months of daily temperature data to “condition” the computational algorithm to antecedent temperature conditions¹¹.

Figure 5 presents a summary of results including NSZD rates, specific product volumes at the time of cryogenic coring (), and estimated times to near complete product depletion. R² values for linear regressions of product decline data fall in the range of 0.912 to 0.997 with a mean value of 0.981. Figure 6 explores relationships between NSZD rates and key site attributes including specific product volumes, qualitative soil permeabilities, depth to source zones, and mean source zone temperatures. The observation that NSZD rates at Site F are similar between paved and unpaved areas is unexpected. This finding highlights the need for further investigation into the effects of surface barriers on NSZD rates.

Fig. 5.

Summary of results including NSZD rates, specific product volumes at the time of cryogenic core collection, and estimated time to near complete product depletion (i.e., units are presented in the legend).

Fig. 6.

NSZD rates as a function of select site attributes including: (a) qualitative permeabilities, (b) specific product volume, (c) depth to source zone, and (d) source zone temperature. Qualitative permeabilities follow Low = silts, Moderate = sand, and High = coarse sand-gravel.

Discussion

What controls NSZD rates

Following Figs. 3 and 4, NSZD rates (are largely constant over periods of record, all the while that specific product volumes are declining. Inter-site comparisons in Fig. 5 advance a narrow range of NSZD rates, to liters/hectare/year ( to meter/year). Near constant NSZD rates while specific LNAPL volumes are decreasing and the narrow range of NSZD rates for sites with differing specific LNAPL volumes leads to a hypothesis that NSZD is a zero-order process with respect to specific LNAPL volumes. Other physical attributes that might govern NSZD rates are considered in Fig. 6. Data in Fig. 6 including p-values show no clear correlations between NSZD rates and soil permeability, specific product volumes, or depth to source zones.

Returning to Fig. 1, the first step in product NSZD is a biologically mediated disproportionation reaction in which petroleum hydrocarbons are concurrently oxidized to carbon dioxide and reduced to methane via a fermentation reaction⁶. Using decane as a model hydrocarbon, Gibbs free energy and enthalpy values for the methanogenesis step are − 266 and + 120 kJ/mol, respectively²⁵. Following Fig. 1, heat from anerobic and aerobic oxidation of methane in the vadose zone is a likely source of heat for endothermic methanogenesis reactions in the saturated zone. Corresponding to methanogenesis being an endothermic process, a modest correlation is seen between monthly NSZD rates and mean temperatures in source zones (Fig. 6 panel d).

Results, and the conceptual model in Fig. 1, provide a basis for a novel hypothesis that NSZD rates are controlled by transport of electrons through a cascade of anaerobic and aerobic reactions, moving electrons from electron donors in source zones (hydrocarbons) to the ultimate electron acceptor, oxygen. Following Le Chatelier’s principle, reaction rates are governed by ratios of reactants and products²⁶. Depletion of reactants in source zones via shuttling electrons to oxygen sustains NSZD rates. Transport of electron donors and acceptors may be the key factor driving assimilation of contaminants in soils and groundwater^3,27. Notably, in the case of petroleum in deep rocks, hydrocarbons can exist for 10s to 100s of millions of years. Arguably, the longevity of petroleum hydrocarbons in deep rocks is due to the lack of viable electron acceptors and limited dissipation of reactants leading to near equilibrium conditions between reactants and products.

Longevity of source zones

Resolving NSZD rates through time is an emerging topic of interest. Three sets of soil cores were collected over decades leading to estimates of long-term NSZD rates at an active refinery²⁸. NSZD rates were predicated on changes in specific product volumes between three sampling events. The accuracy of the forecasts is constrained by varied core collection methods, the possibility of new releases during the period of study and only having three sets of cores corresponding to two measured NSZD rates.

Herein we go further to forecasting future NSZD rates based on regression of hundreds of daily NSZD rates. Furthermore, given estimates of specific LNAPL volumes, forecasts for the longevity of remaining products are advanced. Much like weather forecasts and petroleum reservoir decline curve analyses²⁸, continuously collected new data are used to routinely update forecasts. As the time to the forecast date approaches zero, the accuracy of the forecast is anticipated to converge to the true answer. In the current vernacular, we are using data driven methods versus static models predicated on mathematical approximations of complex processes with fixed inputs. Per Figs. 3 and 5, our forecasted longevities of source zones, to one significant figure, fall in the range of one to four decades.

A simple analog for product source zone longevity is water standing on asphalt on a warm day. Given constant physical conditions (including no sources) evaporative loss rates will be largely constant across the water body and the shallowest parts of the pool will disappear first. In the case of the pool, the primary benefit of active removal of water from the pool would be reduced longevity.

How forecasts can support decisions

Decisions for managing product source zones have been driven by mandates to remove free product to the maximum extent practicable²⁹. Common methods for removing product are outlined by the US Environmental Protection Agency²⁴. Unfortunately, it appears that we have missed the fact that active product recovery measures at mature sites can be of limited consequence relative to losses associated with NSZD. Applying Eqs. 3 and 4, Table 1 data, and Fig. 5 data, ongoing product depletion attributable to NSZD at Sites A and B are 96 and 99%, respectively. Correspondingly, optimistically assuming continuous active recovery with uniform depletion across source zones, active recovery would reduce source zone longevity by 4 and 1%, respectively.

Given the mandate to remove free product to the maximum extent practicable, Enhanced NSZD (ENSZD) appears to be a high value proposition. Building on Fig. 1, NSZD can be enhanced by delivering stoichiometrically consequential masses of electron acceptors and/or depleting reaction products. Plausible ENSZD options include soil venting, addition of sulfate, and phytoremediation. Perhaps, providing clues to ENSZD opportunities, sites with the first and third highest NSZD rates (A and G) have high densities of trees. The site with the second highest NSZD rate (Site B) is an aeolian sand dune deposit with a high degree of natural venting. ENSZD has the potential to proportionally reduce the longevity of product source zones. Arguably, efforts to reduce overall longevity of product source zones, via ENSZD or other, should focus on areas with the largest specific LNAPL volumes and correspondingly, longevity. Areas with small specific LNAPL volumes may be well-suited to NSZD only remedies.

Source zone longevity is a new metric that can support sound decisions for managing petroleum products in soils and groundwater. A modern corollary to removing free product to the maximum extent practicable is to reduce source zone longevities, to the extent practicable. Once source zones are depleted it becomes plausible to begin to contemplate what it will take to address remaining aqueous and sorbed phase hydrocarbons of concern in transmissive and low permeability zones.

Uncertainties

Our source zone longevity forecasts are based on measured specific product volumes and NSZD rates with accuracies of one significant figure. Based on limited data, hypotheses regarding zero order processes, and a conceptualized cascade of electron acceptors, we anticipate that NSZD rates will remain near constant over the lives of source zones. This hypothesis remains to be proven. Like long-term weather forecasts, our forecasts are based on future conditions that may change. To address potential uncertainties in our forecasts we continuously update our forecast using new data. In time our forecasts are likely to converge to the truth.

Further work is needed to verify our methods and the validity of our forecasts. First, long-term NSZD rates need to be independently verified by collecting and analyzing soil cores through time (e.g. 5-year intervals) using rigorous core collection and analytical techniques. Temporal estimates of specific product volumes need to be based on rigorous and consistent methods such as cryogenic coring. Forecasts presented in Fig. 3, including the associated data regressions, provide a basis for anticipating remaining specific LNAPL volumes at defined locations and future times. Secondly, the visions of product source zones being fully depleted by NSZD needs to be verified by collecting soil cores from a subset of the innumerable closed sites where product has historically disappeared from monitoring wells. Thirdly, long-term records of NSZD rates need to be collected using methods that accurately capture dynamic NSZD rates. Given high frequency measurements and low long-term costs, temperature-based methods for measuring NSZD rates are likely to be the most promising approach for acquiring long-term records of NSZD rates. Notably, once product is fully depleted our attention may need to move on to the longevity of aqueous and sorbed phase hydrocarbons in soils and groundwater.

Conclusions

In a novel approach, we advance methods to forecast the longevity of petroleum products in soils and groundwater given on going NSZD. Our data shows that NSZD by itself holds the promise of depleting product source zones in periods ranging from one to four decades. Furthermore, our work advances (1) source zone longevity as a metric for making sound decisions regarding management of petroleum products in soils and groundwater and (2) ENSZD as a promising approach to reducing source zone longevities to the extent practicable.

Author contributions

Kayvan Karimi Askarani: Conceptualization, field work, data collection and management, data analysis, writing – original draft, review & editing, and data visualizationTom Sale: Conceptualization, field work lead, writing – original draft, review & editing.

Funding

Described work was funded by the University Consortium for Field Focused Groundwater Research using unrestricted gift funds and the ESTCP program.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to the fact that the data were collected by Colorado State University, are owned by them, and are confidential, but are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.