1. Introduction
A framework for analyzing the spatiotemporal distribution of produced water (PW) demand is needed to improve PW management. Production of oil and gas and the incidental PW in the Permian Basin of New Mexico and Texas are expected to continue for the foreseeable future, as the Basin was identified as the largest unconventional oil play in the world [
1]. Despite an increase of PW reuse within the oil and gas industry for hydraulic fracturing, PW volumes are estimated to exceed the water demand for hydraulic fracturing by 2.6 times in the Permian Delaware Basin [
1]. Persistent drought, decreasing water availability, increasing demand for water, and impending stringent regulations on saltwater disposal (SWD) wells are all driving the need for exploring new PW management strategies. Current research on the beneficial reuse of treated PW suggests that treatment options are available for cleaning different qualities of PW for fit-for-purpose uses outside the oil and gas industry [
2]. However, the practice is limited by the high cost for transportation, treatment, and energy, as well as acceptability by the public; the need for developing regulations for water quality standards for the treated PW; and the rules for PW discharge, handling, transport, storage, and recycling [
3].
Currently, PW volumes and demand for treated PW are often estimated at large spatial scales. For example, the global daily PW quantity was approximated to be 33.4 million m
3 (210 million barrels) in 1999 [
4]. Veil [
5] surveyed oil and gas producing states to estimate annual PW volumes at the state scale. In the United States, the total PW volume generated from oil and gas wells in 2017 was estimated at approximately 11 million m
3 (69 million barrels) per day, with 41% of the PW generated in Texas (4.3 million m
3 or 27.1 million barrels per day) and 4% of the PW generated in New Mexico (0.38 million m
3 or 2.4 million barrels per day). In New Mexico, PW volume data are reported to the Oil Conservation Division monthly for each well with a unique identification number assigned by the American Petroleum Institute (API), and are incorporated into publicly accessible web-based applications for viewing data at the township scale [
6,
7].
Quantifying the demand for potential uses of treated PW outside of the oil and gas industry is important for addressing water shortage, assessing the feasibility of PW reuse, and understanding the impact of PW reuse on the overall water budget. Scanlon, et al. [
8] estimated the irrigation demand as exceeding PW volumes by five times and municipal demand exceeding by two times within the major oil and gas plays of the United States. Although PW reuse within the oil and gas industry for hydraulic fracturing in some regions may reduce opportunities for beneficial reuse outside the industry, Scanlon, et al. [
8] suggest PW volumes far exceed the industry demand in the Permian Basin. Approximating the volume of PW and potential demand at large spatial scales is useful for understanding the magnitude of PW challenges; however, it does not provide the granularity for making management decisions at the local scale where management occurs. An analysis of localized water demand compared to PW volumes is needed to better assess the potential for use of treated PW outside of the oil and gas industry.
Demand is commonly used as an economic concept describing a consumer’s desire to purchase, and a willingness to pay a certain price for a good or service. Research on water demand is typically performed within the economic context for residential use (e.g., [
9]) and industrial use (e.g., [
10]). Water demand can also be viewed from a physical standpoint as the volume of water required to sustain an activity or system. Ye et al. [
11] provided an example of this second view where they estimated the ecological water demand of natural vegetation within a river basin in China. The second definition is more similar to water accounting than the classic supply-demand definition. From a PW management perspective, both economic and physical demands for treated PW ultimately need to be evaluated. Another consideration of demand might be from the oil and gas industry perspective where PW demand is related to the location and requirement for onsite reuse, other potential uses, and disposal options.
The potential physical demand and the location of the demand for treated PW impact the feasibility of potential reuse outside of the oil and gas industry. While current regulations in New Mexico do not allow any PW reuse outside the oil and gas industry, there is a need to understand the spatial and temporal distribution of demand for treated PW in anticipation of new regulations, and as water scarcity continues to be a major concern in the state. Previous research identified several potential end-uses for treated PW in Lea and Eddy counties, New Mexico, and established some initial guidance on developing mapping applications for exploring potential reuse endpoints [
12], but did not quantify the respective volumes of water demand. Current efforts are underway to provide public access to PW quality and quantity data aggregated at the quarter-township scale through a web mapping application for illustrating the spatiotemporal trends of PW in southeastern New Mexico [
7]; however, this application does not assess potential demand outlets.
This research addresses a gap in the literature by developing a framework for setting the analysis at the same scale in which PW management occurs and then estimating the physical demand for water use categories that may be outlets for treated PW. Using Lea and Eddy counties, New Mexico, as a case study, the framework investigates the historical trends in water use, the physical demand for different water use categories, and estimates the PW supply. The spatial datasets used in this framework (i.e., land use/land cover, road networks, oil and gas wells, and SWD wells) provide a common foundation for a transferable method used to estimate the demand of potential outlets for treated PW to reduce disposal volumes and promote solutions to meet the water demand in the southwest United States. The framework in this study assumes no regulatory and cost barriers to avoid limiting any potential options because treatment technologies continue to improve, and regulations are currently being considered for treated PW reuse. The framework also allows for a description of the geographic distribution of water demand at three different spatial scales to demonstrate the effects of data aggregation on estimating water demand.
3. Results and Discussion
The volume results from this analysis are presented in different units relevant to the various audiences and to facilitate a better cross-discipline understanding (
Table 3). Agriculture in New Mexico typically measures water volume in terms of acre-feet, the oil and gas industry typically measures volume in barrels, and cubic meters is common for the scientific community.
The spatial and temporal differences between irrigated agriculture, dust suppression, power generation, and river augmentation—spatially distributed demand versus point demand—made it difficult to produce a map summarizing the total potential demand per grid cell. Therefore, two different approaches were used for estimating the potential demand. The first approach used spatially disaggregated data to estimate the potential demand for agriculture and dust suppression per grid cell. The second approach used a single point of demand for either power plant or river augmentation demand to calculate the collection area needed to meet different percentages of the total demand based on the minimum distance to the sources of PW. Agriculture had the largest demand of all the potential uses of treated PW; however, the demand varied in time and location. A limitation is that neither of these two analyses included a treatment facility or transportation infrastructure location.
Two pieces of critical information are still needed to improve the usefulness of the methods and information presented in this study. The location of a PW treatment facility is a critical need for improving the usefulness of the information presented in this study because it would serve as both a collection and distribution point. Once a treatment facility is located, the methodology presented could be improved and optimized for flows into, and out of, the treatment facility with the associated costs. The location of pipeline infrastructure used for transporting raw PW and treated PW would be a dramatic improvement for modeling the PW management, as it could effectively reduce the relative distance to the point of supply to the point of demand.
3.1. Broad Historical Water Demand
The broad historical demand provided a reference point for developing the framework. The trends of broad water use demand for Eddy and Lea counties show that the largest water use by category is irrigated agriculture (
Figure 4), which is an order of magnitude larger than the second largest water use category and more than two orders of magnitude larger than the smallest categories. Reservoir evaporation accounts for a substantial portion of the water use in Eddy County but is zero in Lea County. Public water supply and mining are the next two largest water use categories in both counties. Agricultural water demand is decreasing in both counties, a reflection of the retirement of agricultural land in Eddy County to meet the Pecos River Compact deliveries and a rapidly decreasing groundwater supply in Lea County [
30]. The mining category includes water used in potash, coal, metals, and oil and gas, as well as water used for watering vegetation during mine reclamation. There is a 179% decline in the mining water use category in Lea County from 2000 until the present, despite the increase in oil and gas activity during the same period. Some possible causes for this decline include errors in data reporting, mining water being imported from Texas, water reuse and recycling, and/or a reduction of potash mining activity. Tracking the water used within the oil and gas industry remains a challenge in New Mexico even as new regulations require the type of source water used in drilling.
3.2. Granularity
An assessment of the granularity was used to determine three spatial scales of analysis to be used within the framework for analyzing the spatial distribution of PW supply and demand. Descriptive statistics were calculated for the minimum distance from each oil or gas well to the nearest SWD well for each county, and they also considered all the wells in both counties because some wells may be in one county, while the nearest SWD well could be in the other county. The results suggest a majority of oil and gas wells have a SWD well within two miles. The mean distance between oil and gas production wells and the nearest SWD well was 1.51 miles and 1.53 miles for Eddy and Lea counties, respectively. The data do not follow a normal distribution and are skewed to the right of the histogram, suggesting the use of the median distance as a better representation of the central tendency (
Figure 5). The median distances were 1.1 miles for Eddy County, 1.34 miles for Lea County, and 1.21 miles combined (
Table 4). The minimum distance was less than 0.01 miles, likely because some production wells have an onsite SDW. The maximum distance was 16.06 miles in Eddy County and 8.02 miles in Lea. One shortcoming of this analysis is that it does not consider production wells or SWD wells on the other side of the Texas border, which could potentially shift the statistics. However, even if those wells were included, there are few data readily available describing how much PW is moved across the state borders. Based on the results of the descriptive statistics, we produced three grid scales (1.1, 1.5, and 3.0 miles) which capture the distance between oil and gas wells to SWD wells for approximately 45%, 62%, and 91% of the wells in Eddy and Lea counties, respectively.
The average number of oil and gas wells in proximity to a SWD is also important to consider because SWD wells could be considered good collection points since they already have pipeline and/or road infrastructure networks built to receive PW. The average number of oil and gas wells within three radial buffers of SWD wells was calculated (
Table 5). The average number of wells within 1.1 miles of SWDs was approximately 2, with a maximum of 8. There was an average of approximately 2.7 wells within 1.5 miles with a maximum of 9. At 3 miles, the average is approximately 6.7 wells within the buffer and a maximum of 21. The 3-mile buffer captures 91% of the oil and gas wells in Eddy and Lea counties. The average distance between SWD wells was 1.35 miles with a maximum distance of 8.7 miles.
In the context of PW management, assessing the granularity provides a systematic method for determining a scale useful for data aggregation to address challenges in transportation needs. Transportation of PW through trucking and/or pipelines is often discussed in scientific literature and at industry meetings as one of the main costs of PW management. A 2019 report by the Ground Water Protection Council provides some insight, where an estimated 20-mile roundtrip was used as an example for trucking distance from source water to the well location [
31]. Our analysis improves on previous analyses that aggregated PW volumes at the township scale of 6-mile grid [
6,
32], which is insufficient for making management decisions.
3.3. Produced Water Supply
The estimated annual supply of treated PW was 22,855,016 m3 (18,536 acre-feet) in Eddy County and 22,605,859 m3 (18,334 acre-feet) in Lea County. The estimated annual total available volume of treated PW in the two counties was 45,460,875 m3 (36,870 acre-feet) which is much lower than the total reported volumes because of the assumptions of 58% reused within the oil and gas industry and a 50% water recovery through thermal desalination processes. These volumes could increase or decrease depending on the water quality, treatment recovery, and the volume of water reused within the oil and gas industry for hydraulic fracturing and enhanced oil recovery.
The estimated supply of available PW is variable throughout the study region. Maps were produced to illustrate the spatial variability of supply in Eddy and Lea counties at three grid scales (
Figure 6). The symbology was first classified using the Jenks natural breaks method on the 1.1-mile grid data in order to equally distribute map errors, and then the largest class was increased to capture the effect of increased volumes at the 3-mile grid resolution. The maps illustrate the effect of aggregation on resolving the granular data—as the grid cell size increases, the ability to resolve the location of supply decreases. The higher spatial resolution shows more of the checkerboard pattern from created by the differences in land ownership and management status. The symbology was deliberately held constant for the different grid cell sizes to further illustrate this point. The largest volume class in the symbology has a range larger than the other seven classes combined to capture the increased aggregated volume at the 3-mile grid resolution creating map errors in the 1.1-mile grid map for the largest volume class. There are grid cells in the 3-mile where the lower classes are still represented, and reclassifying the symbology would increase the error in the lower volume classes. The result of the different levels in aggregation provides evidence that assessing the granularity is important for producing information at an appropriate scale for PW management decisions; thus, the remaining analyses are presented at the 1.1-mile grid cell.
An important point that should not be understated is the cost of treating the estimated volume of water. The PW treatment costs depend on the PW quality and water quality requirements of end users. PW in the Permian Basin can contain approximately 1400 compounds, including BTEX (benzene, toluene, ethylbenzene, and xylene), various suspended particles, dissolved solids, soluble and insoluble organic compounds, naturally occurring radionuclides, and production chemicals (e.g., acids, corrosion inhibitors, surfactants) [
33]. The average total dissolved solids (TDS) concentration of PW in the Permian Basin is approximately 90–100 g/L [
34], which requires thermal distillation treatment. It is estimated that the costs to recover a net 45,460,875 m
3 treated PW could be between USD 266 and USD 496 million, with unit treatment costs ranging between USD 2.92/m
3 and USD 5.45/m
3 for PW with a salt level of 90 g/L from Osipi et al. [
35] and Xu et al. [
36]. Additional information on the costs and removal efficiencies of technologies for treating various qualities of PW can be found in Ma et al. [
37] and Geza et al. [
38]. Although providing a complete cost analysis is outside the scope of this study, some additional context can be provided by highlighting the SWD well disposal costs for the same volume of PW would likely be between USD 71 million and USD 715 million using the disposal cost for SWD wells ranging from USD 0.25/barrel (
$1.57/m
3) and USD 2.50/barrel (
$15.72/m
3) published in an industry magazine [
39].
The annual total PW volumes in this study are underestimated by approximately 9.8% compared to data reported by the New Mexico Oil Conservation Division (NMOCD) for the similar time period. NMOCD reported 238,853,304 m
3 (193,638 acre-feet) in Eddy and Lea counties in 2021 compared to the 216,563,845 m
3 (175,571 acre-feet) estimated in this study. PW volume is constantly changing throughout the production cycle and a more elegant process such as a dynamic hybrid agent-based and system dynamics model as discussed in Langarudi et al. [
40] could increase the accuracy of the estimates for improving management decisions. Another improvement to this analysis would be the inclusion of decline curves for each well based on the geologic formation the well is producing from.
3.4. Irrigation Water Demand
The irrigated agricultural water demand for Eddy and Lea counties was estimated using growing degree days, crop coefficients, and local reference evapotranspiration values. The growing season primarily occurs between March and November—the highest intensity of irrigation water demand occurring between June and August (
Figure 7). The highest demand for a grid cell was 649 acre-feet in June and the mean demand per grid cell was 40.2 acre-feet. Approximately 23% of the grid cells had <1 acre-feet of demand, suggesting misclassification in the USDA imagery. Alfalfa was the dominant crop for Eddy County and cotton was the dominant crop in Lea County; alfalfa is showing a declining trend. The spatial distribution of the irrigation demand is focused along the Pecos River in Eddy County and where numerous center pivots pump groundwater from the Ogallala aquifer in eastern Lea County. The highest demand occurs in June through August at the peak of the growing season.
The estimated total annual irrigated agriculture water demand is 170,944 acre-feet and 343,915 acre-feet for Eddy and Lea counties. The estimate for Eddy County is similar to the annual estimates of 164,861 acre-feet of consumptive use provided in Magnuson et al. [
19]; however, the estimate for Lea County is more than double the 117,366 acre-feet provided in Magnuson et al. [
19]. It is unclear as to why the Lea County estimate diverges so much from the official State report since the estimates are both made using the USDA Crop Data Layer as a basis. Two main differences are the purpose of the reports and the year of data used. Magnuson et al. [
19] reported water use only in terms of withdrawals from surface and groundwater, whereas the current study is looking at total crop water demand. Another difference could be the current study was run for 2021, whereas Magnuson et al. [
19] based their report on 2015 data and note that the eastern part of New Mexico received precipitation that was significantly higher than average, which would cause the withdrawal values to be lower.
Nonetheless, the trends in agriculture production suggest that Lea County would have a much larger irrigation water demand. In 2018, the acres of land in farms in Eddy County (1,087,902 acres) were 78% less than the land in farms in Lea County (1,938,321 acres), and the total agriculture income was USD 109 million for Eddy and USD 226 million for Lea [
41]. Although the larger livestock population in Lea accounts for some of the differences in agricultural income, the larger agriculture acreage reported by NMDA suggests Magnuson et al. [
19] could be underestimating the consumptive use in Lea County. In our analysis of irrigated acreage based on the 2020 Cropland Data Layer, we estimate that Eddy County had 37,538 acres and Lea County 73,991 acres. The relative intensity of demand presented in this study is a useful starting point for informing PW management decisions, and the future goal should be working towards better quantifying absolute volumes of irrigation water demand. Additionally, future efforts in quantifying irrigation water demand should include spatiotemporal precipitation data such as the gridded PRISM climate data.
3.5. Dust Suppression
The purpose of using water for dust suppression is to meet an optimum moisture and compaction goal [
42]; however, dust suppression on rural roads is a challenge in arid environments because unpaved roads produce dust given the persistent dry conditions. Treated PW could provide a viable alternative for dust suppression because the sources, assuming PW treatment facilities would be located within the oil and gas fields, would be closer to the haul roads, the salt in brackish water helps bind the clay to roadbeds—improving the longevity of the treatment—and the demand for dust control is year-round.
An estimate of 9884 miles in Eddy and 8776 miles in Lea County have a potential treated PW demand for dust suppression. Using the assumptions of treating every mile at least once annually and at a rate of 3388 gallons/mile (7.969 m
3/km), the total dust suppression demand was estimated to be 33,488,465 gallons/year (126,768 m
3/year) for Eddy County and 29,444,153 gallons/year (111,458 m
3/year). The dust suppression frequency is difficult to predict and the assumption of one treatment per year for every mile is likely conservative for some locations while likely overestimated for grid cells with a lower concentration of road segments. The demand is spatially distributed throughout both counties and the mean demand is approximately 10,072 gallons/year for Eddy and 8405 gallons/year for Lea County per 1.1-mile grid cell (
Figure 8). The map legend breaks are at 8000-gallon equal intervals, representing an approximate two fully loaded 4000-gallon tanker trucks. The grid cells with the highest demand occur along the edge of the Delaware Basin where there is also a high concentration of oil and gas wells, and therefore more roads. The estimated dust suppression water demand could be improved with more specific information on the width of the roads, the average daily traffic, and road quality.
3.6. Power Plant Demand
Energy and water are closely linked through the concept of energy–water nexus, which includes oil and gas production as well as electricity generation. Thermoelectric power generation requires large volumes of water for steam generation and cooling. The total volume of power plant water demand was assumed to be 5,516,122 m
3 (4472 acre-feet), and the minimum distance to a grid cell centroid was used to cumulatively calculate the area required to meet 100, 50, and 25% of the demand based on the supply estimates (
Figure 9).
The collection area was 821 mi2, 455 mi2, and 252 mi2 for meeting 100, 50, and 25% of the demand. The mean annual supply per grid cell within the calculated collection area was 8113.5 m3 (2,143,360 gallons), 7304.8 m3 (1,929,724 gallons), and 6642.7 m3 (1,754,816 gallons) for meeting 100, 50, and 25% of the demand. A limitation to the method used for determining the collection area for meeting the demand for the power plant water or the Pecos River augmentation demand is that it is not optimized for clustering of PW sources, potential anisotropy in the data, the quality of the PW, or pipeline delivery networks that could change where the water would be sourced from and the total collection area.
3.7. Pecos River Augmentation
The Brantley Reservoir was used as the hypothetical point of delivery for treated PW into the Pecos River. The collection area to meet 100, 50, and 25% of the demand is 1488 mi
2, 1048 mi
2, and 667 mi
2, respectively (
Figure 10).
The mean annual volume for the 100, 50, and 25% collection area is 8.1, 5.8, and 4.5 acre-feet/1.1-mile grid cell. Many of the grid cells have less than 2 acre-feet of PW annually, which alludes to the difficulty in establishing a collection area while balancing the logistical challenges of transporting water from the oil and gas wells to a PW treatment plant and then to a delivery point on the Pecos.
Similar to the power plant demand analysis, this analysis was not optimized for clusters of supply, nor did it include a pipeline network. Any planning for a system to meet these demands should explore these considerations. Another consideration for the Pecos River augmentation would be setting multiple points of delivery to meet the demand.
Any release to the Pecos River would need to be in coordination with the Interstate Stream Commission (ISC), Carlsbad Irrigation District (CID), the Pecos Valley Artesian Conservancy District (PVACD), and the Fort Sumner Irrigation District (FSID). Supplementation of the Pecos River Basin could benefit the National Environmental Policy Act (NEPA) and the Endangered Species Act (ESA). In years where there is an abundant supply in the reservoirs, treating PW for river augmentation would not be an optimal beneficial use.
The potential use of treated PW for augmenting stream flows in New Mexico might become economically feasible as treatment and transportation costs approach the current costs associated with meeting compact requirements. In 2003, New Mexico implemented an agricultural land retirement program that cost the state more than $100 million to purchase water rights in order to increase the deliveries by an average of 10,000 acre-feet per year. Another costly practice currently employed is to pump groundwater in order to meet Pecos River Compact deliveries in years of delivery shortages, which has the negative impact of lowering the water table and affecting the freshwater supply for agriculture. Thus, using treated PW to help meet the Pecos River Compact requirements could relieve some of the pressure on the state and irrigation districts that are responsible for ensuring the deliveries to Texas.