Next Article in Journal
Predicting Business Failure with the XGBoost Algorithm: The Role of Environmental Risk
Previous Article in Journal
Research on the Impact of Large-Scale Photovoltaic Development on Regional Economic Growth—A Case Study of Qinghai Province
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sustainability
Volume 17
Issue 11
10.3390/su17114946
sustainability-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Assessing the Effects of Hydraulic Fracturing on Streamflow Through Coupled Human–Hydrological Modeling

by
Zhulu Lin
1,
Tong Lin
2,
Haochi Zheng
3 and
Siew Hoon Lim
4,*
1
Department of Agricultural and Biosystems Engineering, North Dakota State University, Fargo, ND 58108, USA
2
Environmental and Conservation Sciences Program, North Dakota State University, Fargo, ND 58108, USA
3
Department of Earth System Science and Policy, University of North Dakota, Grand Forks, ND 58202, USA
4
Department of Agribusiness and Applied Economics, North Dakota State University, Fargo, ND 58108, USA
*
Author to whom correspondence should be addressed.
Sustainability 2025, 17(11), 4946; https://doi.org/10.3390/su17114946
Submission received: 23 April 2025 / Revised: 20 May 2025 / Accepted: 26 May 2025 / Published: 28 May 2025
Browse Figures
Versions Notes

Abstract

:
The Devonian–Mississippian Bakken Formation in western North Dakota (USA) is one of the largest hydraulic fracturing oil fields in the world. Streamflow analysis showed that the average seven-day low flows in the region surprisingly increased during the recent oil boom. The increase, ranging from 88% to 3648%, was largely due to the fact that the region had received 20% more precipitation than normal during that time period. To study the impact of hydraulic fracturing at Bakken on regional streamflow under normal precipitation and other scenarios, we integrated a socioeconomic agent-based model that simulates the hydraulic fracturing water uses with a hydrological model that simulates the streamflow in the Little Muddy River in the region. Our results showed that compared to the existing (baseline) scenario, the average seven-day low flows in the Little Muddy River decreased from 18% to 88%, while the annual average flows did not change much under drier to normal precipitation scenarios. Our research also finds that climate factors and water management policies were more influential than hydraulic fracturing and population growth. The emergency water management policies implemented at the peak of shale oil development had mitigated the hydraulic fracturing impact on regional streamflow at low-flow conditions and improved water resource sustainability in the region.

1. Introduction

The Devonian–Mississippian Bakken Formation in western North Dakota (USA) is one of the largest unconventional oil fields in the world. The U.S. Geological Survey (USGS) estimated a total of 1200 Mm3 (7.4 billion barrels) of recoverable oil reserves in the region [1]. The oil production in the Bakken region expanded exponentially during 2008–2014 as a result of advanced hydraulic fracturing and drilling technologies and favorable oil prices, increasing more than nine times from 21,600 m3 (136,000 barrels) per day in December 2007 to 195,000 m3 (1.3 million barrels) per day in December 2014. The decline in oil prices since 2015 has led to a slowdown in new hydraulic fracturing activities in the region, and oil production has stabilized around 160,000 m3 (~1 million barrels) per day [2]. However, with the U.S. energy policy shifting toward emphasizing domestic production and energy independence, oil development in the Bakken region is expected to increase in the near future. Hence, sustainable water management and practices are of great importance for the region.
The climate in western North Dakota is characterized as semiarid with annual precipitation averaging about 14–16 inches [3,4]. The expansion of the oil industry in the region has led to tremendous increases in freshwater demand [5,6,7,8]. Nevertheless, the impact of Bakken development on streamflow in western North Dakota was surprisingly limited during the peak of the recent oil boom from 2008 to 2014 [9]. Lin et al. [9] analyzed daily streamflow recordings from 12 USGS stream gages located along nine small-to-medium streams in the region and found that the annual average seven-day low flows during 2008–2014 were all greater than those during 2000–2007 (i.e., immediately before the oil boom) and the percentages of the increases ranged from 88 to 3648%. Lin et al. [9,10] attributed this result to two major reasons. First, during 2008–2014, the region received 22.2% more precipitation than the 30-year normal (1981–2010). Second, the state of North Dakota adopted several emergency water policies to adaptively manage regional water resources to meet the acute increase in water demand for hydraulic fracturing. These policies include authorizing the Western Area Water Supply (WAWS) Project, implementing the “In-Lieu-Of Irrigation” (ILOI) program, temporarily relaxing Lake Sakakawea surplus water restrictions, and accelerating the issuance of temporary surface water permits. Details about these policies can be found in Lin et al. [9,10].
An important question that warrants further investigation is: How would the increased water demand from the Bakken oilfield affect water-related sustainability, especially regional streamflow, if none of these emergency policies had been implemented, or if precipitation in the region eventually returns to normal or even swings back to drier conditions? Our study aims to answer these questions through coupled human–hydrological modeling.
The coupled human–hydrological modeling approach has gained popularity in recent decades to tackle the complex issues of water resource management for sustainable uses, with growing recognition among scientists and policymakers of the importance of incorporating the interactions between human actions and natural systems into the policy-making process [11]. Agent-based models (ABMs) have been extensively used to simulate human decision making when modeling socio-hydrological issues related to sustainable water resource management [12,13,14].
Due to its ability to represent complex human decisions and actions at different scales, ABM has been a popular candidate for integration with surface water and/or groundwater models in many coupled human–hydrological modeling studies [14]. Kent [15] developed a software interface by coupling ABM and SWAT (Soil and Water Assessment Tool) to study the impacts of hypothetical policy initiatives, farmers’ adoption of best management practices and economic returns on crop yield, runoff, tile flow, and nutrient loads in a river basin in East Central Illinois (USA). Khan et al. [16] developed a coupled ABM-SWAT framework to study how prioritizing water uses for food production, hydropower generation, and ecosystem health would affect streamflow in the Mekong River basin (Southeast Asia) and the Niger River basin (West Africa).
ABMs have recently been coupled with integrated surface water (e.g., SWAT) and groundwater (e.g., MODFLOW or Modular Groundwater Flow Model) models to tackle the hydrological complexities of surface water and groundwater interactions. Javansalehi and Shourian [17] linked an ABM with the SWAT-MODFLOW hydrological model to study the impacts of climate change on agriculture and water resource sustainability by investigating the interplay between farmer actions and water resources. Du et al. [18] developed a tightly coupled and fully integrated ABM-GSFLOW modeling framework (GSFLOW stands for Groundwater and Surface-water FLOW model that is based on the integration of the USGS Precipitation-Runoff Modeling System and MODFLOW) to study the spatial heterogeneity and temporal dynamics of human–hydrological interactions in the Heihe River basin in China. Additional studies of coupling ABMs and groundwater models were reviewed by Canales et al. [19].
Lin et al. [20] developed a spatially explicit ABM to simulate the dynamics of the depot-based water allocation system for hydraulic fracturing water distribution in the Bakken region (USA). A water depot is a business that sells water to oil companies for hydraulic fracturing in the Bakken region, and the water depots are owned by individuals or institutions that have secured state-issued water permits to access the water supply. Energy companies obtain hydraulic fracturing operation-related water by trucking it from water depots to their oil and gas wells in western North Dakota [10] while recognizing that the permanent water permit granting process takes months or even years to complete and that industrial uses, such as hydraulic fracturing, are given low priority—ranking second to last.
Over the past decades, many studies have focused on the environmental concerns, especially related to water, of hydraulic fracturing in the Bakken region. Scanlon et al. [8] and Horner et al. [21] estimated the water footprint of hydraulic fracturing at Bakken, including the quantity of fresh water used for hydraulic fracturing and other related activities and the volume of produced wastewater, as well as their management and disposal. Lauer et al. [22] and Cozzarelli et al. [23] studied the chemistry of wastewater produced from hydraulic fracturing and its effects and environmental risks associated with potential spills into the environment. Lin et al. [9] conducted a comprehensive analysis of the historical data of water use, precipitation, streamflow, and groundwater level in the region and concluded that hydraulic fracturing had a limited impact on the regional water resources due to the compound factors of wet weather patterns and adoption of multiple emergency water management policies.
Lin et al. [10,20], respectively, developed lumped and spatially explicit socioeconomic ABMs to simulate the dynamics of the water depot-based water allocation and distribution system in western North Dakota and how different factors, such as water demand, climate, and management policies, would affect its efficiency in water distribution for oil development activities. However, they did not study how these different environmental and socioeconomic factors would affect the regional water resources through the unique water allocation and distribution system.
Therefore, in this case study, we will couple the spatially explicit socioeconomic ABM [20] with the hydrological SWAT to study the impacts of Bakken oil development on regional streamflow under different scenarios of hydraulic fracturing, climate, population growth, and water management policies. Through the coupled human–hydrological modeling framework and carefully designed scenarios, we will be able to parse out the effects of hydraulic fracturing on streamflow from other confounding factors. Although the Bakken region in western North Dakota has its distinctive features, our modeling framework and findings will assist policymakers and water practitioners in other regions in developing adaptive management strategies and policies to mitigate future climate change to meet the unprecedented water demand associated with hydraulic fracturing [5,6,24].

2. Materials and Methods

2.1. Study Area

The Little Muddy River is a tributary of the Missouri River in the United States. It rises in the prairie country of northern Williams County (North Dakota) and flows west, then south, joining the Missouri River near Williston, North Dakota, the center of the Bakken shale oil development. The Little Muddy River watershed is located in a core area of four North Dakota counties (i.e., Dunn, McKenzie, Mountrail, and Williams Counties) where more than 85% of the horizontal hydraulic fracturing wells were drilled [9] (see Figure S1 in the Supplement Materials). The watershed under study drains an area of 2415 km2. Although the average annual seven-day low flows of the Little Muddy River increased by about 88% during the peak of the oil boom between 2008 and 2014, this increase was the least compared to other surface water sources in the region [9]. It appeared that the Little Muddy River had been affected by the oil development activities in the Bakken region, therefore, it was selected to study the impact of unconventional oil development on streamflow in western North Dakota.
The Missouri River and Lake Sakakawea were the most frequently used surface water sources for hydraulic fracturing in the Bakken region. However, these two surface waters are unlikely to be affected by hydraulic fracturing water uses compared to other surface water sources because of their massive streamflow and water volume. For example, the Missouri River accounts for approximately 96% of annual stream flows in North Dakota [25]. The Little Muddy River was the second most frequently used surface water source, following the Missouri River and Lake Sakakawea system. Since 2012, a total of 22 temporary and one conditional industrial water permit have been issued for the Little Muddy River. There were 13 water depots built along the river, among which eight were within our study area (see Figure 1). From 2012 to 2015, 23 industrial water permits were issued on the Little Muddy River, and the cumulative industrial water uses reached 2.0 Mm3 (1639 ac-ft). The number of issued permits and the total amount of water used for hydraulic fracturing from the Little Muddy River both ranked second, only behind the Missouri River and the Lake Sakakawea system.

2.2. ABM for Water Depots

Lin et al. [20] developed a spatially explicit ABM to simulate the dynamics of nearly 600 individual water depots for hydraulic fracturing water use distributions in western North Dakota. The detailed development of the ABM is described in [20], while the ABM assumptions, model calibration, and the distribution of water depots are briefly summarized here.
The entities of the ABM consist of four types of water-depot agents and one regulator agent. The four types of water-depot agents include (1) permanent agents, which are water depots with permanent industrial water permits, (2) temporary agents, which are water depots with temporary industrial water permits, (3) irrigation transferred agents, which are water depots with a portion of their permanent irrigation water permits temporarily transferred to industrial water use, and (4) municipal/co-ops agents, which are water depots owned by local towns or through the WAWS project (see Figure S2). The regulator agent is the North Dakota Office of the State Engineer, the state water management agency. Readers may refer to Lin et al. [20] for detailed descriptions of these agents and how their behaviors are modeled in the ABM.
The overall assumption of the ABM is that the goal of all agents is to maintain a balance between water supply and demand for hydraulic fracturing, which the state agency can use to manage the regional water resources for long-term sustainable use. The primary exogenous driver for the model is the water demand for hydraulic fracturing. The simulation area covers the 16 western counties in western North Dakota that have the most hydraulic fracturing activities. In the ABM, we used the institution theory [12] to model the regulator’s behavior while employing evolutionary programming [26] to simulate the water-depot agents’ decision-making process when applying for water permits and cognitive maps [27] to simulate their ability and willingness for water sale competition. Influence boundaries were set for all water-depot agents to restrict their competitive behavior toward their neighboring agents, but not for non-neighboring ones.
All water depot agents are in discrete units, and their attributes are represented as state variables, which include physical location (longitude and latitude), permit type, approved water use, source, water quality, capital investment, operation fees, water price, water sale, profit, passion, effort, network, and operation history. The simulation period is 2007–2014, with a yearly simulation time step.
The model is initialized with the area’s spatial data and the water depot locations in 2007. At each time step, the county-level hydraulic fracturing water demands are calculated against the industrial water supplies. The water demand in each county is the sum of the water uses for all horizontal wells located in that county, while the industrial water supply is the total amount of water sold by all water depots located in that county. If the annual hydraulic fracturing water demand is greater than the current year’s water supply through water depots, the permit application, regulation, and competition submodels will be called to generate more water-depot agents [20]. Otherwise, the model will continue to the next time step. The agent-based model was calibrated against real-world water-use recordings in 2007–2014 and was able to simulate the number of water depots as well as the spatial locations and water uses of different types of water depots with reasonable success at the 8 km × 8 km cell and county levels.

2.3. SWAT Model Development

To develop the SWAT model (SWAT2012, Rev. 635) for the Little Muddy River watershed, we used a 10 m DEM to delineate the watershed into 21 sub-basins (Figure 1) and classified the watershed into three slope classes: 0–5%, 5–10%, and >10%. We used the USGS National Land Cover Database and STASTSGO databases to classify land use and soil types, respectively. Any land use, soil type, or slope class with an area larger than 10% was included in the model, which was further divided into 154 Hydrologic Response Units (HRUs) of a unique combination of land use, soil type, and slope class. Daily precipitation, maximum and minimum temperature, solar radiation, relative humidity, and wind speed data were obtained from Climate Forecast System Reanalysis [28,29], and weather data from six local stations were used in this study (Figure 1). For each HRU, the water balance, potential evapotranspiration, and surface runoff were calculated by the variable storage coefficient method [30], Hargreaves method [31], and the Soil Conservation Service (SCS) Curve Number method, respectively.
The SWAT model simulation period was from 1 January 2004, to 31 December 2014, with 2004–2011 for model calibration and 2012–2014 for validation. The USGS stream gage station below Cow Creek near Williston (#06331000) was set as the outlet of the watershed under study (Figure 1), and the daily streamflow recorded at this station was downloaded from the USGS National Water Information System [32] to be compared with the model-simulated streamflow during the calibration and validation periods. In the model validation periods (2012–2014), we added six point sources to represent the water use of the water depots that withdrew water from the Little Muddy River from 2012 to 2014 since industrial water permits were issued in the Little Muddy River beginning in 2012. Since we can only add one point source per sub-basin in SWAT, we combined the yearly water uses of all water depots located in the same sub-basin into a single yearly time series with negative values for the point source in that sub-basin. The water depots in the Little Muddy River watershed were spread across six sub-basins (Figure 1), hence, we added six point sources with negative values in the SWAT model.
We used the SUFI2 (Sequential Uncertainty Fitting Procedure 2) algorithm in the SWAT-Cup 2012 [33] to calibrate the SWAT model’s simulated streamflow against the daily observations of stream discharge. We ran three iterations of SUFI2 with each iteration having a maximum of 1000 model runs. After each iteration, we removed the insensitive parameters from the model calibration process, and used the confidence intervals obtained in this iteration to set the new ranges of parameter values for the remaining parameters in the next iteration. After the model was calibrated, the parameter values obtained in the last iteration were applied for model validation. The adjusted model parameters using SUFI2 are listed in Table 1.
We used the Nash–Sutcliffe efficiency (NSE) coefficient and the percent of bias (PBias) to quantify the goodness of fit for model calibration and validation. NSE ranges from −∞ to 1 with a value closer to 1 indicating a better match of model simulations and observations. PBias indicates the model’s overall tendency for under-prediction or over-prediction. A negative PBias indicates an overall trend of under-prediction for streamflow from the model, while a positive PBias indicates an overall over-prediction from the model.

2.4. Model Coupling and Scenario Analysis

After the SWAT model was calibrated, it was then coupled with the agent-based model of water depots to evaluate the impact of hydraulic fracturing on the streamflow of the Little Muddy River under different scenarios. As shown in Figure 2, the hydraulic fracturing water uses under different scenarios were first simulated using the agent-based model of water depots; next, the hydraulic fracturing water uses by the existing and newly generated water depots withdrawing water from the Little Muddy River were simulated as negative point sources in the SWAT model for the Little Muddy River watershed. Thus, the potential impact of the hydraulic fracturing water use on regional surface water resources was reflected in the changes in the simulated streamflow of the Little Muddy River by comparison against the baseline scenario.
Table 2 lists thirteen scenarios in addition to the baseline scenario (i.e., Scenario 0), under which both ABM and SWAT were validated (2012–2014) with hydraulic fracturing, population, precipitation, and state regulation remaining unchanged. We designed Scenarios I-1, I-2, and I-3 to evaluate the impact of the increased hydraulic water demand, Scenarios II-1, II-2, and II-3 to evaluate the impact of regional population growth, and Scenarios III-1, III-2, and III-3 to evaluate the impact of precipitation changes. It is worth noting that the baseline scenario of precipitation was 22.2% higher than the regional 30-year normal [8]. Therefore, we should consider Scenario III-1 (precipitation ↓ 10%) as 10% wetter-than-normal years, Scenario III-2 (precipitation ↓ 20%) as normal precipitation years, and Scenario III-3 (precipitation ↓ 40%) as approximately 20% drier-than-normal years. We also designed Scenarios IV-1, IV-2, and IV-3 to evaluate the impact of three emergency water management policies implemented during the Bakken oil boom. Since the hydraulic fracturing water use and (increased) precipitation compounded each other in the baseline scenario, we hence designed a compound scenario (i.e., Scenario V) to represent a situation where there is no hydraulic fracturing water demand and the precipitation is equivalent to the 30-year normal level (or 20% less than the baseline scenario). The changes listed in Table 2 represent the only modifications made to the simulation scenarios while holding other factors at the same levels as in the baseline scenario.

3. Results

3.1. SWAT Calibration

Figure 3 shows the results of model calibration and validation for the coupled ABM-SWAT models. NSE was 0.77 for model calibration and 0.74 for validation, respectively, and PBias was 7.8% for model calibration and 12.7% for validation, respectively. It should be noted that during the model calibration period (2004–2011), no water was withdrawn from the Little Muddy River for hydraulic fracturing water use. During the model validation period (2012–2014) (i.e., the baseline scenario), the water withdrawal from the Little Muddy River for hydraulic fracturing water use was simulated using the coupled ABM for water depots. Then, the sum of the water uses sold by different water depots in the same sub-basin was fed into the calibrated SWAT model as a negative point source per sub-basin. As shown by graphical comparison and goodness-of-fit statistics, the coupled ABM-SWAT models can simulate the effects of hydraulic fracturing water use on streamflow in the Little Muddy River fairly well, although the average streamflow was slightly overestimated.

3.2. Streamflow Impacts

Table 3 lists the annual average flow and the average annual seven-day low flows under the thirteen different scenarios, and the graphical comparisons of daily streamflow time series in the Little Muddy River under different scenarios were plotted in Figures S3–S7. While the average annual seven-day low flows decreased from 18% to 88% in all scenarios, the annual average flows did not change much (from −0.02% to −4.2%) except under Scenario III and Scenario V, when precipitation decreased by 10–40% compared to that in the baseline scenario. As mentioned earlier, the baseline scenario of precipitation was 22.2% higher than the regional 30-year average. Therefore, Scenario III-2 (precipitation ↓ 20%) and Scenario V (No HF + normal precipitation) were considered normal years in terms of precipitation.
Table 3 shows that under Scenario I, when the hydraulic fracturing water demand increased from 50% to 200% while keeping other factors at the current level, the average seven-day low flow decreased by 18.1% to 42.4%, but the impact on the annual average flow was negligible. Under Scenario II, when the population in the region was projected to increase by 5% to 20% while keeping other factors unchanged, the average seven-day low flows decreased by 24.8% to 58.4%, and the impact on the annual average flow was again negligible.
However, under Scenario III, when the region’s precipitation decreased by 10% to 40%, the impacts on both the annual average flow and the average seven-day flow were significant, with the annual average flow decreasing by 27.9% to 81.2% and the seven-day low flow decreasing by 69.3% to 88.7%. Under Scenario IV, when a certain water management policy was not implemented, the annual average flow would decrease by 1.1% to 4.2%, while the average seven-day low flow would decrease by 71% to 88%, with the impact similar to that of the precipitation decrease in Scenario III.
To assess the impact of hydraulic fracturing water demand alone on the Little Muddy River streamflow under a normal precipitation scenario, we compared streamflow under Scenario III-2 and Scenario V because under these two scenarios, the region received normal precipitation, with or without hydraulic fracturing activities. Table 3 shows that the annual average flow under Scenario V (no hydraulic fracturing) was only 0.3 percentage points higher than that under Scenario III-2 (with the current level of hydraulic fracturing activities). In other words, the impact of hydraulic fracturing on the annual average flow was minimal. However, its impact on the average seven-day low flow was a different story: the average annual seven-day low flow under Scenario V was 14 percentage points higher than that under Scenario III-2.

4. Discussion

Figure 4 shows visual comparisons of annual average streamflow and average annual seven-day low flows under some key representative scenarios, including the baseline, 100% increase for hydraulic fracturing water demand, 10% increase for population growth, and 20% decrease for precipitation, three management scenarios, and the compound scenario. It is interesting to note that the average annual seven-day low flow of the Little Muddy River under Scenario III-2 (precipitation ↓ 20%) was 87% lower than that under the baseline scenario (see also Table 3). We recall that the average annual seven-day low flow in the Little Muddy River (USGS stream gauge #06331000) increased by 88% under the baseline scenario when the region received 22% more precipitation than a normal year [9]. In other words, the increase of the average annual seven-day flow in the Little Muddy River during 2008–2014 would not have happened if the region had received a normal amount of precipitation according to our simulation.
Comparing streamflow under Scenario IV shows that the “surplus water” policy [21] for Lake Sakakawea (Scenario IV-3) was the most important policy implemented during the Bakken oil boom, which had the highest effects on both the annual average flow and the average seven-day flow. This is not surprising because Lake Sakakawea at the Missouri River is the single most dominant water source in the region. The “surplus water” policy refers to the decision made by the U.S. Army Corps of Engineers in December 2010 to relax the restriction on the surplus water in Lake Sakakawea, allowing 123 Mm3 (32.5 billion gallons) of annual surplus water to be withdrawn from the lake for short-term uses. This policy was valid for five years with the possibility of renewal for another five years [21]. Mainly because of this policy, the Office of the State Engineer started to issue considerably more temporary water permits for industrial water use starting in 2012. A temporary water permit does not grant a water right to the permit holder, but allows them to use or sell the approved amount of water for industrial uses such as hydraulic fracturing. Since it does not grant a water right, the approval process of issuing temporary water permits was greatly expedited to meet the acute water demand from hydraulic fracturing [10]. Without this policy in place, there would be more temporary permits issued or a larger amount of water approved for water depots in other medium-sized rivers such as the Little Missouri River, which in turn would have resulted in a greater impact on low flows in these rivers.
Figure 4 shows that Scenarios IV-1 and IV-2 would generate an effect similar to that of Scenario IV-3 but to a lesser degree. In 2011, the North Dakota state legislature authorized the WAWS project to sell up to 20% of the project water to the oil industry to fund the construction of the WAWS project. By 2014, about 20% of the total water used for hydraulic fracturing was supplied through this project [9]. In the ILOI program, the farmers who hold irrigation permits can temporarily forgo part of their water permits and sell the forgone portion of water to oil companies for hydraulic fracturing. However, this transfer of water use from irrigation to industry is temporary (usually one year) and needs approval from the state. Similarly, by 2014, the ILOI program accounted for about 25% of the total water use for hydraulic fracturing at Bakken [10]. Since the WAWS project mainly serves the northwest region of the Bakken region, where the Little Muddy River watershed is located, we found that it had a larger impact on streamflow in the Little Muddy River than the ILOI program, which affected the entire western North Dakota.
Furthermore, Figure 4 also shows that climate factors and policy changes (Scenarios III-2 and IV) are generally more influential factors than hydraulic fracturing water demand and regional population growth (Scenarios I-2 and II-2) in affecting streamflow in the small-to-medium streams in the region, as represented by the Little Muddy River.
It is worth noting that the impact of increased hydraulic fracturing (Scenario I) is the smallest among all the factors considered in this study. The impact of the water demand created by the hypothetical population increase in the region (Scenario II) is even greater than the increased hydraulic fracturing activities. Lin et al. [9] showed that the increased population of temporary oilfield service workers contributed additional domestic water use, which was equivalent to ~15% of the annual industrial water use for the shale oil development in the Bakken. Therefore, when assessing the effects of hydraulic fracturing on streamflow in the region, one should not overlook the impacts of socioeconomic changes brought by the oil development [8]. By coupling a hydrological model with ABM, we are able to incorporate these socioeconomic factors into the analysis.
However, our analysis was not without limitations. The human–hydrological coupling demonstrated in this study was unidirectional. In other words, we only considered the effects of the decisions made by the water depot agents and the regulator agent on the streamflow of the Little Muddy River, but not in the other direction. The agent-based model was developed for the entire Bakken region in western North Dakota, while the SWAT model was developed only for the Little Muddy River watershed, a small portion of the region (see the inset of Figure 1). To develop a two-way, fully coupled ABM-SWAT model such as in [16,18], we would need to redesign a regulation submodel for the ABM to consider the level of water abundance in individual rivers or streams. In the current ABM, the regulation submodel only differentiates the water abundance of Lake Sakakawea/Missouri River from other rivers/streams in the region, while assuming the same level of difficulty for obtaining a water permit from all other surface waters. The assumption is also independent of the streamflow condition of the river or stream under consideration.
Another limitation is that our study did not explicitly account for model uncertainty, including the uncertainty of model parameters in ABM and SWAT, although both models were calibrated against historical data, to provide a confidence region for models’ predictions and the subsequent scenario analysis. The coupled human–hydrological modeling framework could be further strengthened through adopting formal uncertainty analysis methods such as GLUE (Generalized Likelihood Uncertainty Estimation) or Bayesian methods to comprehensively evaluate parameter uncertainties and improve the reliability of model predictions and the results of scenario analysis.
Despite these limitations, our coupled ABM-SWAT modeling study has shown that the emergency water management policies implemented at the peak of shale oil development were instrumental in mitigating hydraulic fracturing’s impact on regional streamflow, especially under low-flow conditions. It should be noted that the agent-based model was developed specifically for the unique water depot-based water distribution system in the Bakken region of western North Dakota, and that we have mainly focused on the impact of hydraulic fracturing water demand, population growth, precipitation changes, and water management policies on regional streamflow because these factors are considered by the water depot agents and the regulator agent when they make decisions. Other potential factors such as temperature and land use changes that may have additional effects on streamflow in other places are not modeled in this ABM since they are not included the agents’ decision-making process as revealed through our interviews with water depot owners and the state geohydrologists [34]. Although the Bakken region in western North Dakota has its distinctive features, the human–hydrological modeling framework and its findings can assist and inform policymakers and water practitioners in other places to develop adaptive management strategies that can proactively mitigate future climate changes and meet the water demands for hydraulic fracturing and other energy development activities.

5. Conclusions

  • A socioeconomic agent-based model that simulates hydraulic fracturing water use was integrated with a watershed-scale hydrological model to study the impact of hydraulic fracturing at Bakken on streamflow in the small-to-medium streams in the region under different scenarios such as hydraulic fracturing water demand, population growth, precipitation changes, and water management policies.
  • Except for precipitation change scenarios, all other scenarios did not have much impact on annual average flow; but all scenarios would significantly decrease the annual average seven-day low flows in the Little Muddy River (a medium-sized stream in the Bakken region), ranging from 18% to 88%.
  • The impact of increased hydraulic fracturing was the smallest among all the factors considered in this study. In general, climate factors as represented by precipitation and policy changes were more influential factors than hydraulic fracturing water demand and regional population growth in affecting streamflow in the Little Muddy River.
  • The emergency water management policies implemented at the peak of shale oil development had mitigated the hydraulic fracturing impact on regional streamflow at low-flow conditions and improved water resource sustainability in the region.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/su17114946/s1, Figure S1: The study area and horizontal oil wells in western North Dakota (USA). Figure S2: The study area and water depots in western North Dakota (USA). Figure S3: Comparison of streamflow in the Little Muddy River under the Baseline Scenario and Scenario I. Figure S4: Comparison of streamflow in the Little Muddy River under the Baseline Scenario and Scenario II. Figure S5: Comparison of streamflow in the Little Muddy River under the Baseline Scenario and Scenario III. Figure S6: Comparison of streamflow in the Little Muddy River under the Baseline Scenario and Scenario IV. Figure S7: Comparison of streamflow in the Little Muddy River under the Baseline Scenario, Scenario V, and Scenario III-2.

Author Contributions

Conceptualization, Z.L., H.Z. and S.H.L.; methodology, Z.L. and T.L.; software, T.L. and Z.L.; validation, Z.L., T.L., H.Z. and S.H.L.; formal analysis, T.L., Z.L., H.Z. and S.H.L.; investigation, T.L.; resources, Z.L., H.Z. and S.H.L.; data curation, T.L.; writing—original draft preparation, Z.L.; writing—review and editing, H.Z. and S.H.L.; visualization, T.L. and Z.L.; supervision, Z.L.; project administration, Z.L.; funding acquisition, Z.L. and S.H.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded in part by the National Science Foundation (grant number ICER-1413954), the North Dakota Department of Water Resources, and the North Dakota Water Resources Research Institute. Siew Hoon Lim’s work was supported by the USDA National Institute of Food and Agriculture’s Multistate Project NE 2249.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Conflicts of Interest

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analysis, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

References

  1. Gaswirth, S.B.; Marra, K.R.; Cook, T.A.; Charpentier, R.R.; Gautier, D.L.; Higley, D.K.; Klett, T.R.; Lewan, M.D.; Lillis, P.G.; Schenk, C.J.; et al. Assessment of Undiscovered Oil Resources in the Bakken and Three Forks Formations, Williston Province, Montana, North Dakota, and South Dakota; USGS Fact Sheet 2013–3013; US Geological Survey: Reston, VA, USA, 2013; 4p. Available online: https://pubs.usgs.gov/fs/2013/3013/ (accessed on 1 April 2024).
  2. North Dakota Industrial Commission Monthly Production Report Index. Available online: https://www.dmr.nd.gov/oilgas/mprindex.asp (accessed on 25 May 2025).
  3. Shaver, R. Availability and Quality of Surface and Groundwater Resources in West-Central and Southwest North Dakota; North Dakota Water Resource Investigation: Bismarck, ND, USA, 2012. Available online: https://www.swc.nd.gov/info_edu/reports_and_publications/pdfs/wr_investigations/wr53_report.pdf (accessed on 25 May 2025).
  4. Carter, J.M.; Macek-Rowland, K.M.; Thamke, J.N.; Delzer, G.C. Estimating National Water Use Associated with Unconventional Oil and Gas Development. U.S.; Geological Survey Fact Sheet 2016–3032; US Geological Survey: Reston, VA, USA, 2016; p. 6. Available online: https://pubs.er.usgs.gov/publication/fs20163032 (accessed on 1 April 2024).
  5. Scanlon, B.R.; Ikonnikova, S.; Yang, Q.; Reedy, R.C. Will water issues constrain oil and gas production in the United States? Environ. Sci. Technol. 2020, 54, 3510–3519. [Google Scholar] [CrossRef] [PubMed]
  6. Scanlon, B.R.; Reedy, R.C.; Wolaver, B.D. Assessing cumulative water impacts from shale oil and gas production: Permian Basin case study. Sci. Total Environ. 2022, 811, 152306. [Google Scholar] [CrossRef] [PubMed]
  7. Scanlon, B.R.; Reedy, R.C.; Male, F.; Hove, M. Managing the Increasing Water Footprint of Hydraulic Fracturing in the Bakken Play, United States. Environ. Sci. Technol. 2016, 50, 10273–10281. [Google Scholar] [CrossRef] [PubMed]
  8. Cwiak, C.L.; Avon, N.; Kellen, C.; Mott, P.C.; Niday, O.M.; Schulz, K.M.; Sink, J.G.; Webb, T.B.J. The New Normal: The Direct and Indirect Impacts of Oil Drilling and Production on the Emergency Management Function in North Dakota; North Dakota State University: Fargo, ND, USA, 2015; Available online: https://www.ndsu.edu/fileadmin/emgt/FINALThe_New_Normal_January_2015_a.pdf (accessed on 25 May 2025).
  9. Lin, Z.; Lin, T.; Lim, S.H.; Hove, M.H.; Schuh, W.M. Impacts of Bakken Shale Oil Development on Regional Water Uses and Supply. J. Am. Water Resour. Assoc. (JAWRA) 2018, 54, 225–239. [Google Scholar] [CrossRef]
  10. Lin, Z.; Lim, S.H.; Lin, T.; Borders, M. Using agent-based modeling for water resources management in the Bakken region. J. Water Resour. Plann. Manag. 2020, 146, 05019020. [Google Scholar] [CrossRef]
  11. Wada, Y.; Bierkens, M.F.P.; de Roo, A.; Dirmeyer, P.A.; Famiglietti, J.S.; Hanasaki, N.; Konar, M.; Liu, J.; Müller Schmied, H.; Oki, T.; et al. Human–water interface in hydrological modelling: Current status and future directions. Hydrol. Earth Syst. Sci. 2017, 21, 4169–4193. [Google Scholar] [CrossRef]
  12. An, L. Modeling human decisions in coupled human and natural systems: Review of agent-based models. Ecol. Model. 2012, 229, 25–36. [Google Scholar] [CrossRef]
  13. Berglund, E.Z. Using agent-based modeling for water resources planning and management. J. Water Resour. Plann. Manag. 2015, 141, 04015025. [Google Scholar] [CrossRef]
  14. Alam, M.F.; McClain, M.; Sikka, A.; Pande, S. Understanding human–water feedbacks of interventions in agricultural systems with agent based models: A review. Environ. Res. Lett. 2022, 17, 103003. [Google Scholar] [CrossRef]
  15. Kent, G. Modeling and Analysis of Management for an Agro-Ecosystem Using an Agent-Based Model Interface for the Soil and Water Assessment Tool (SWAT). Master’s Thesis, University of Illinois at Urbana-Champaign, Champaign, IL, USA, 2014. [Google Scholar]
  16. Khan, H.F.; Yang, Y.C.E.; Xie, H.; Ringler, C. A coupled modeling framework for sustainable watershed management in transboundary river basins. Hydrol. Earth Syst. Sci. 2017, 21, 6275–6288. [Google Scholar] [CrossRef]
  17. Javansalehi, M.; Shourian, M. Assessing the impacts of climate change on agriculture and water systems via coupled human-hydrological modeling. Agric. Water Manag. 2024, 300, 108919. [Google Scholar] [CrossRef]
  18. Du, E.; Tian, Y.; Cai, X.; Zheng, Y.; Li, X.; Zheng, C. Exploring spatial heterogeneity and temporal dynamics of human-hydrological interactions in large river basins with intensive agriculture: A tightly coupled, fully integrated modeling approach. J. Hydrol. 2020, 591, 125313. [Google Scholar] [CrossRef]
  19. Canales, M.; Castilla-Rho, J.; Rojas, R.; Vicuña, S.; Ball, J. Agent-based models of groundwater systems: A review of an emerging approach to simulate the interactions between groundwater and society. Environ. Model. Softw. 2024, 175, 105980. [Google Scholar] [CrossRef]
  20. Lin, T.; Lin, T.; Lim, S.H.; Jia, X.; Chu, X. A spatial agent-based model for hydraulic fracturing water distribution. Front. Environ. Sci. 2022, 10, 1025559. [Google Scholar] [CrossRef]
  21. Horner, R.M.; Harto, C.B.; Jackson, R.B.; Lowry, E.R.; Brandt, A.R.; Yeskoo, T.W.; Murphy, D.J.; Clark, C.E. Water Use and Management in the Bakken Shale Oil Play in North Dakota. Environ. Sci. Technol. 2016, 50, 3275–3282. [Google Scholar] [CrossRef]
  22. Lauer, N.E.; Harkness, J.S.; Vengosh, A. Brine spills associated with unconventional oil development in North Dakota. Environ. Sci. Technol. 2016, 50, 5389–5397. [Google Scholar] [CrossRef]
  23. Cozzarelli, I.M.; Skalak, K.J.; Kent, D.B.; Engle, M.A.; Benthem, A.; Mumford, A.C.; Haase, K.; Farag, A.; Harper, D.; Nagel, S.C.; et al. Environmental signatures and effects of an oil and gas wastewater spill in the Williston Basin, North Dakota. Sci. Total Environ. 2017, 579, 1781–1793. [Google Scholar] [CrossRef] [PubMed]
  24. Zhong, C.; Zolfaghari, A.; Hou, D.; Goss, G.G.; Lanoil, B.D.; Gehman, J.; Tsang, D.C.; He, Y.; Alessi, D.S. Comparison of the hydraulic fracturing water cycle in China and North America: A critical review. Environ. Sci. Technol. 2021, 55, 7167–7185. [Google Scholar] [CrossRef]
  25. North Dakota State Water Commission. A Reference Guide to North Dakota Waters. 2014. Available online: http://www.swc.nd.gov/pdfs/water_reference_guide.pdf (accessed on 25 May 2025).
  26. Manson, S.M. Agent-based modeling and genetic programming for modeling land change in the Southern Yucatán Peninsular Region of Mexico. Agric. Ecosyst. Environ. 2005, 111, 47–62. [Google Scholar] [CrossRef]
  27. Kolkman, M.J.; Kok, M.; Van Der Veen, A. Mental model mapping as a new tool to analyse the use of information in decision-making in integrated water management. Phys. Chem. Earth Parts A/B/C 2005, 30, 317–332. [Google Scholar] [CrossRef]
  28. Fuka, D.R.; MacAllister, C.A.; Degaetano, A.T.; Easton, Z.M. Using the Climate Forecast System Reanalysis dataset to improve weather input data for watershed models. Hydrol. Proc. 2014, 28, 5613–5623. [Google Scholar] [CrossRef]
  29. Dile, Y.T.; Srinivasan, R. Evaluation of CFSR climate data for hydrologic prediction in data-scarce watersheds: An application in the Blue Nile River Basin. J. Am. Water Resour. Assoc. (JAWRA) 2014, 50, 1226–1241. [Google Scholar] [CrossRef]
  30. Williams, J.R. Flood routing with variable travel time or variable storage coefficients. Trans ASAE 1969, 12, 100–103. [Google Scholar] [CrossRef]
  31. Hargreaves, G.L.; Hargreaves, G.H.; Riley, J.P. Agricultural Benefits for Senegal River Basin. J Irrig Drain Eng. 1985, 111, 113. [Google Scholar] [CrossRef]
  32. USGS National Water Information System. Available online: http://waterdata.usgs.gov/nwis (accessed on 1 April 2024).
  33. Abbaspour, K.C. SWAT-CUP 2012: SWAT Calibration and Uncertainty Programs—A User Manual; EAWAG: Swiss Federal Institute of Aquatic Science and Technology: Dübendorf, Switzerland, 2013. [Google Scholar]
  34. Lin, T. An Agent-Based Model for Water Allocation and Management of Hydraulic Fracturing. Ph.D. Dissertation, North Dakota State University, Fargo, ND, USA, 2021. [Google Scholar]
Sustainability 17 04946 g001
Figure 1. The study area of the Little Muddy River watershed in western North Dakota (USA).
Figure 1. The study area of the Little Muddy River watershed in western North Dakota (USA).
Sustainability 17 04946 g001
Sustainability 17 04946 g002
Figure 2. Coupling of agent-based model and SWAT hydrological model. SWAT—Soil and Water Assessment Tool.
Figure 2. Coupling of agent-based model and SWAT hydrological model. SWAT—Soil and Water Assessment Tool.
Sustainability 17 04946 g002
Sustainability 17 04946 g003
Figure 3. Graphical comparison of the model-simulated (solid lines) and the observed (dots) daily streamflow in the Little Muddy River for (a) model calibration (2004–2011) and (b) model validation (2012–2014). NSE—Nash–Sutcliffe Efficiency; PBias—percent of bias.
Figure 3. Graphical comparison of the model-simulated (solid lines) and the observed (dots) daily streamflow in the Little Muddy River for (a) model calibration (2004–2011) and (b) model validation (2012–2014). NSE—Nash–Sutcliffe Efficiency; PBias—percent of bias.
Sustainability 17 04946 g003
Sustainability 17 04946 g004
Figure 4. Visual comparisons of streamflow in the Little Muddy River under representative scenarios. I-2: HF ↑ 100%, II-2: Population ↑ 10%, III-3: Precipitation ↓ 20%, IV-1: no WAWS project, IV-2: No ILOI program, IV-3: No surplus water, and V: No HF + normal precipitation. HF—hydraulic fracturing, WAWS—Western Area Water Supply, ILOI—In-Lieu-Of Irrigation.
Figure 4. Visual comparisons of streamflow in the Little Muddy River under representative scenarios. I-2: HF ↑ 100%, II-2: Population ↑ 10%, III-3: Precipitation ↓ 20%, IV-1: no WAWS project, IV-2: No ILOI program, IV-3: No surplus water, and V: No HF + normal precipitation. HF—hydraulic fracturing, WAWS—Western Area Water Supply, ILOI—In-Lieu-Of Irrigation.
Sustainability 17 04946 g004
Table 1. Parameters and their calibrated values of the Soil and Water Assessment Tool model for the Little Muddy River watershed.
Table 1. Parameters and their calibrated values of the Soil and Water Assessment Tool model for the Little Muddy River watershed.
No.SymbolFileDefinitionAdjusting MethodParameter Values
CalibratedMinMax
1CN2.mgtSCS runoff curve numberRelative0.01−0.030.09
2ALPHA_BF.gwBaseflow recession constant (1/day)Relative−0.02−0.060.10
3GW_DELAY.gwGroundwater delay (day)Replace544.25441.69833.46
4GWQMN.gwThreshold depth of shallow aquifer for return flow to occur (mm H2O)Replace1279.1718.11287.9
5ALPHA_BNK.rteBaseflow alpha factor for bank storageReplace0.310.200.53
6CH_K1.subEffective hydraulic conductivity in tributary channel alluvium (mm/h)Replace2.64−71.0793.28
7CH_K2.rteEffective hydraulic conductivity in main channel alluvium (mm/h)Replace1.32−5.7610.45
8EPCO.hruPlant uptake compensation factorReplace0.36−0.110.39
9ESCO.hruSoil evaporation compensation factorReplace0.470.070.95
10GW_REVAP.gwGroundwater revamp coefficientReplace0.280.190.35
11RCHRG_DP.gwDeep aquifer percolation fractionReplace0.070.010.10
12SOL_AWC (all).solAvailable water capacity of all layers (mm H2O/mm Soil)Relative0.140.010.21
13SOL_K.solSaturated hydraulic conductivityRelative−0.45−0.440.45
14SOL_BD.solMoist bulk densityRelative0.400.140.45
15SURLAG.bsnSurface runoff lag timeReplace7.214.2015.47
16SFTMP.bsnSnowfall temperature (°C)Replace0.13−0.441.49
17SMTMP.bsnSnowmelt temperature (°C)Replace2.520.154.41
18TIMP.bsnSnowpack temperature lag factorReplace0.14−0.020.46
Table 2. Definitions of the scenarios for the coupled ABM-SWAT modeling.
Table 2. Definitions of the scenarios for the coupled ABM-SWAT modeling.
No.NameBrief DescriptionDefinition or Changes Made to the Coupled Models
0BaselineNo changesNo changes were made to the existing (2012–2014) levels of hydraulic fracturing, population, precipitation, and regulation
1Scenario I-1HF ↑ 50%Hydraulic fracturing (HF) water demand increased by 50%
2Scenario I-2HF ↑ 100%Hydraulic fracturing water demand increased by 100%
3Scenario I-3HF ↑ 200%Hydraulic fracturing water demand increased by 200%
4Scenario II-1Population ↑ 5%The regional population increased by 5%
5Scenario II-2Population ↑ 10%The regional population increased by 10%
6Scenario II-3Population ↑ 20%The regional population increased by 20%
7Scenario III-1Precipitation ↓ 10%Precipitation decreased by 10%
8Scenario III-2Precipitation ↓ 20%Precipitation decreased by 20%
9Scenario III-3Precipitation ↓ 40%Precipitation decreased by 40%
10Scenario IV-1No WAWS projectNorth Dakota did not authorize the Western Area Water Supply (WAWS) project
11Scenario IV-2No ILOI programThe Office of the State Engineer did not adopt the “In-Lieu-Of Irrigation” (ILOI) program
12Scenario IV-3No surplus waterThere was no relaxation of the restriction on surplus water from Lake Sakakawea
13Scenario VNo HF + normal precipitationNo hydraulic fracturing and precipitation decreased by 20%
Table 3. The annual average flow and average seven-day low flow under different scenarios.
Table 3. The annual average flow and average seven-day low flow under different scenarios.
Scenario 1Brief DescriptionAnnual Average FlowAverage 7-Day Low Flow
Value (m3/s)Change 2 (%)Value (m3/s)Change (%)
BaselineNo changes1.7772—0.0674—
Scenario I-1HF ↑ 50%1.7769−0.020.0552−18.1
Scenario I-2HF ↑ 100%1.7766−0.030.0456−32.4
Scenario I-3HF ↑ 200%1.7746−0.140.0388−42.4
Scenario II-1Population ↑ 5%1.7766−0.030.0507−24.8
Scenario II-2Population ↑ 10%1.7735−0.210.0371−45.0
Scenario II-3Population ↑ 20%1.7684−0.490.0280−58.4
Scenario III-1Precipitation ↓ 10%1.2808−27.90.0207−69.3
Scenario III-2Precipitation ↓ 20%0.6850−61.50.0088−87.0
Scenario III-3Precipitation ↓ 40%0.3333−81.20.0076−88.7
Scenario IV-1No WAWS project1.7579−1.080.0173−74.4
Scenario IV-2No ILOI program1.7573−1.120.0193−71.4
Scenario IV-3No surplus water1.7030−4.170.0082−87.8
Scenario VNo HF + normal precipitation0.6901−61.20.0181−73.1
1 These scenarios and their changes relative to the baseline are defined in Table 2. 2 Percentage of change compared to the baseline scenario.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Lin, Z.; Lin, T.; Zheng, H.; Lim, S.H. Assessing the Effects of Hydraulic Fracturing on Streamflow Through Coupled Human–Hydrological Modeling. Sustainability 2025, 17, 4946. https://doi.org/10.3390/su17114946

AMA Style

Lin Z, Lin T, Zheng H, Lim SH. Assessing the Effects of Hydraulic Fracturing on Streamflow Through Coupled Human–Hydrological Modeling. Sustainability. 2025; 17(11):4946. https://doi.org/10.3390/su17114946

Chicago/Turabian Style

Lin, Zhulu, Tong Lin, Haochi Zheng, and Siew Hoon Lim. 2025. "Assessing the Effects of Hydraulic Fracturing on Streamflow Through Coupled Human–Hydrological Modeling" Sustainability 17, no. 11: 4946. https://doi.org/10.3390/su17114946

APA Style

Lin, Z., Lin, T., Zheng, H., & Lim, S. H. (2025). Assessing the Effects of Hydraulic Fracturing on Streamflow Through Coupled Human–Hydrological Modeling. Sustainability, 17(11), 4946. https://doi.org/10.3390/su17114946

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop