Introduction

The U.S. Fish and Wildlife Service (USFWS) oversees more than 560 National Wildlife Refuges, many with a Comprehensive Conservation Plan (CCP) outlining goals and objectives to help manage and protect resources such as wildlife and their habitat. One of the resource management objectives of the Cabeza Prieta National Wildlife Refuge (CPNWR) in southern Arizona as stated in the CCP is to sustain a population of 500 to 700 desert bighorn sheep (Ovis canadensis mexicana; USFWS, 2006), which necessitates ensuring free-standing water is available for their consumption. CPNWR is located within the Sonoran Desert where the timing and small amount of annual rainfall limit the availability of natural water supplies for desert bighorn sheep. Water availability is particularly important during the summer months when desert bighorn sheep most commonly visit managed water resources (Harris and others, 2020). To increase water availability for desert bighorn sheep and other wildlife, CPNWR has constructed new, or enhanced existing, bedrock pools that collect rainfall and runoff. The pools are referred to commonly as tanks or tinajas. Modified tanks are naturally occurring water catchments which have been modified by engineering. Constructed tanks are commonly referred to as developed tanks. The water storage basins of developed tanks, created by excavation of rock walls in a runoff catchment, provide water sources in areas that previously lacked these resources (Broyles, 1997). Natural tanks were modified in many ways, generally expanding water storage basins by excavation of adjacent rock walls in natural runoff catchments (Broyles, 1997; fig. 1). The enclosed rock structures of these modified tanks have increased local water storage capacity hundreds to thousands of times beyond their natural capacity, permitting some to become perennial resources (Broyles, 1997).

The relatively low regional precipitation (fig. 2; National Oceanic and Atmospheric Administration [NOAA], 2022) and high evapotranspiration at CPNWR precludes some tanks from filling to their maximum water storage capacity or maintaining a year-round water supply. Therefore, management of tanks as perennial water sources requires some tanks to be mechanically filled by hauling water by truck and trailer. Water hauling operations can be expensive and have effects on wildlife and designated wilderness, such as habitat disruption and increased erosion. In keeping with CPNWR’s goal of wilderness stewardship, also prioritized in the CCP, site visits for resource improvements using motorized vehicles and additional mechanized equipment are to be minimal within wilderness (USFWS, 2006). Management decisions on how often and how much to fill tanks are informed by monitoring estimates of stored water volumes. However, the geometries of the tanks are complex, making estimates of stored water volumes challenging to measure or calculate. Accurate documentation of tank geometry and simplification of calculating volume estimates could better inform management decisions to efficiently meet both wildlife and wilderness stewardship objectives.

The U.S. Geological Survey (USGS) Southwest Biological Science Center, in cooperation with the USFWS during February 2022, employed ground-based light detection and ranging (lidar) to help meet resource management objectives at the CPNWR. Ground-based lidar remote sensing is an efficient and accurate survey tool to measure complex topography and geologic structures like bedrock tanks. Ground-based lidar systems collect millions of spatially explicit topographic measurements in minutes with sub-centimeter (cm) precision. These high-resolution topographic data allow for accurate representations of natural surfaces (Caster and others, 2022) and measurements of confining volumes (Fabbri and others, 2017) without the need to simplify surface geometry into common three-dimensional shapes, such as cylinders. For water catchments, such as tanks, topographic data allow for efficient estimation of how water storage volumes change with water surface elevation (stage height), which is termed a stage-storage model. The lidar-derived data acquired in this study are provided in the associated data release (Caster and others, 2024).

Objectives

In this report we present the results for ground-based lidar survey and analysis of three tanks identified as a priority by CPNWR managers. The purpose of this work was to (1) accurately characterize the complex structure of each tank, and (2) develop a stage-storage model for estimating water storage volumes from simple observations of water surface-stage heights in each tank. Results of this study will permit the refuge to: (1) easily estimate water available for wildlife at any point of time, (2) interpret tank recharge following rainstorms, and (3) make important and potentially expensive management decisions more effective and economical regarding whether and when to transport water via vehicles to mechanically refill the tanks in designated wilderness.

Study Area

CPNWR is in the Sonoran region of the Basin and Range Province of North America, an extensive system of block-fault mountains separated by alluvial valleys. The Sonoran climate is a hot desert, with low temperatures averaging above 0 degrees Celsius (°C) (32 degrees Fahrenheit [°F]) year-round and high temperatures in summer months averaging around 41 °C (107 °F; fig. 2; NOAA, 2022). The distribution of intra-annual precipitation within the region is bimodal, with most of the rainfall occurring from frontal storm systems during the winter months and high intensity convective storms in the summer months that are associated with the North American Monsoon (fig. 2). Although intra-annual rainfall within the region peaks during the summer monsoon season, the high temperature extremes, and hours of sunlight during the summer result in high evapotranspiration potential that limits effective precipitation.

Approximately 93 percent of the refuge (1,255 square miles [mi²]; 3,250 square kilometers [km²]) is designated wilderness containing 24 developed water sources for wildlife. Of these water sources, the refuge identified three tanks as important for desert bighorn sheep for this study: (1) Buckhorn Tank located in the Cabeza Prieta Mountains, (2) Senita Tank in the Drift Hills, and (3) Eagle Tank in the Sierra Pinta (fig. 3). Buckhorn and Eagle Tanks are modified natural tanks whereas Senita Tank is a developed tank (in the sense of Broyles, 1997). However, all three are in natural catchments with man-made shallow bore tunnels (boreholes) into rock walls. All three tanks are also located downstream of a small man-made sediment retention dam constructed during initial modification or development to help reduce sediment infilling and increase water retention in the tank (figs. 1 and 3).

Buckhorn Tank is a modified natural tank located within the National Hydrography Dataset (NHD; Seaber and others, 1987) sub-watershed (HUC-12) of Tule Well, which drains local precipitation southwest into Mexico. The upstream contributing area for Buckhorn Tank is approximately 0.12 km² (30 acres) of rocky, mountainous terrain with an average slope of 27 degrees and runoff channels up to half a kilometer in length (about 0.3-mile long, fig. 3). The tank was initially modified in 1949, which included construction of a concrete liner wall to seal off a rock fault and elevate the pour point by 2 feet (ft) (USFWS, 1950). In 1976, the tank was completely cleaned out, the sediment trap was partially cleaned out, and additional improvements were made. Improvements included sealing the tank with a Thoroseal and Acryl 60 plaster mix and construction of a small masonry dam at the outside edge of the concrete liner wall to prevent sediment from backflowing into the tank (Dodson, 1976a; USFWS, 1977); this higher pour point also increased the tank’s capacity. The tank was completely cleaned out and the sediment trap was partially cleaned out again in 2003 (Tracey Martin, Arizona Desert Bighorn Sheep Society, personal commun., September 9, 2022). Buckhorn tank has the largest capacity of the three tanks included in this study, which was originally reported as 30,250 liters (L) (7,991.2 gallons [gal]) (USFWS, 1950) and after addition of the masonry rock dam estimated to be 37,850 L (9,998.9 gal) by Broyles (1997).

Senita Tank is a developed tank in the Mohawk wash NHD HUC-12 sub-watershed, which drains northeast. Relative to the other tanks studied, Senita Tank has the smallest upstream contributing area of approximately 0.12 km² (3.8 acres). This upstream contributing area also has the shallowest average slope of 21.6 degrees and shortest runoff channels (0.15 kilometers [km], [492 ft]; fig. 3). Senita Tank was constructed in 1958 (USFWS, 1958b) and improved in 1974 (USFWS, 1975) to have a capacity of 26,495 L (6,999.2 gal) as reported by Broyles (1997). Improvements included removal of debris and application of a Thoroseal and Acryl 60 plaster mix to seal the tank (Dodson, 1976b). In 1976, the upper runoff catchment and small masonry dam at the entrance were sealed with the plaster mix (Dodson, 1976b).

Eagle Tank is a modified natural tank in the Hummingbird Canyon NHD HUC-12 sub-watershed, which drains east into the Mohawk Valley. Eagle Tank has an upstream contributing area of approximately 0.18 km² (45 acres) with an average slope of 35 degrees and runoff channels over half a kilometer in length (fig. 3). The tank was modified in 1957 to a capacity originally reported as 24,605 liters (6,500.0 gal) (USFWS, 1958a) and later estimated as 24,602 liters (6,499.2 gal) by Broyles (1997). No evidence has been found on site or in the CPNWR’s annual narratives indicating the tank and upstream sediment retention dam have been improved or cleaned out since construction.

Materials and Methods

Field Surveys

We surveyed each tank, and the adjacent area upstream and downstream, using ground-based lidar and real-time kinematic global positioning system and global navigation satellite systems (RTK-GPS/GNSS) (fig. 4). The lidar surveys used a Riegl VZ1000 laser scanner with a mounted Nikon D810 digital camera (fig. 4) to acquire photo-colorized topographic measurements with a median density greater than (>) 500 points-per-square meter. We conducted lidar scans from multiple stationary positions to better capture the complexity of the terrain, registering each scan position with four to eight stationary reflective ground control targets. The positions of the ground control targets were surveyed with a Trimble R8 RTK-GPS/GNSS rover and base station receiver pair and were used to georeference the lidar topographic data. National Geodetic Survey (NGS; https://geodesy.noaa.gov/index.shtml) published survey control benchmarks were not available within the vicinity of the sites. A location was selected at each site to deploy the base station receiver and the spatial data were referenced to the National Spatial Reference System (NSRS) by post-processing of at least 3 hours of static GNSS observables through the NGS’s Online Positioning User Service (OPUS, https://www.ngs.noaa.gov/OPUS/). The results classify the data as USGS Level III GNSS positioning (Rydlund and Densmore, 2012). Vertical coordinate information of the RTK-GPS/GNSS survey and georeferenced lidar data is referenced to the North American Vertical Datum of 1988 (NAVD 88) and used the geoid model Geoid18. Horizontal coordinate information is referenced to the 2011 adjustment of the North American Datum of 1983 (NAD 83 [2011]) and used the cartesian coordinate system NAD 83 (2011), Universal Transverse Mercator (UTM) zone 12N (Rydlund and Densmore, 2012).

Prior to each survey, CPNWR staff pumped out and temporarily stored water from each tank and then removed sediment that had infilled since modification, development, or subsequent improvement. Next, to measure stage height, CPNWR staff installed fixed rods marked from 0 ft to 12 ft (Buckhorn and Eagle Tanks) or 14 ft (Senita Tank) at 1/10-ft intervals from the bottom of each tank (fig. 4A). These preparations allowed the survey to characterize the full capacity of each tank with lidar and provided a height reference for estimating stage-storage capacity relationships.

Data Analysis

The lidar echo return measurements, termed a point cloud, were prepared for analysis by first colorizing and georeferencing the data in RiSCAN PRO software (http://www.riegl.com/products/software-packages/riscan-pro/; accessed August 1, 2024). Average scan-to-scan registration error between targets was 0.5 cm. Lidar-to-GPS registration errors averaged 0.7 cm. The average GPS solution error was 1.4 cm, and the average horizontal and vertical GPS position error is 0.6 cm and 1.2 cm, respectively. We thus estimate the total positional uncertainty as 2.1 cm based on analysis of the sum of squared errors in quadrature, that is:

$\sqrt{{0.5}^2+{0.7}^2+{1.4}^2+{0.6}^2+{1.2}^2}=2.1 \text{cm}$

(1)

We segmented the georeferenced point cloud data to classify data points as either bare earth, vegetation, or our survey equipment, including the tank stage height rods, using an iterative height filter and manual selection. We then used the bare-earth dataset (which included the rock surfaces inside the tank) to analyze the geometric dimensions and storage capacity of each dry tank as well as the associated sediment retention dam directly upstream.

Tank Stage-Storage Capacity

Using the bare-earth point cloud of each tank, we evaluated the maximum storage capacity as well as stage-storage capacity for 1 ft, 2 ft, 4 ft, 6 ft, 8 ft, and 10 ft stage heights (fig. 5B). The maximum tank storage capacity was defined by the area below the pour point elevation at the downstream side of the tank basin. The pour point, or the outlet for water draining out of the tank, was defined by a concrete wall surrounding the main basins for both Senita and Eagle Tanks. The Buckhorn Tank maximum storage capacity included the area above the concrete liner enclosed by the small masonry dam added in 1976. Stage-storage capacity was defined by the elevations of the six stage heights (that is 1, 2, 4, 6, 8, and 10 ft) from the installed stage height rods within the tanks.

Stage-storage capacity can be calculated directly from the bare-earth classified points or from derived GIS products, such as an interpolated surface raster (grid cells) and triangulated-irregular networks (TINs). Each method has the potential to introduce interpolation errors. To account for this, we carried out point-based, raster-based, and TIN-based volume calculations for each stage height and report the mean and 95 percent confidence interval of the mean of the three methods for each tank.

The point and raster-based methods are the most common approaches to estimate reservoir volume, in which basin elevations are subtracted from water fill elevations at each measurement point and the differences are summed to estimate a volume for the entire reservoir. These methods have the benefit of providing a spatially explicit representation of stage heights that allow for validation and visualization. We carried out the point-based storage capacity estimate in CloudCompare (version 2.10.2; CloudCompare, 2019) by first separating the tank point cloud below the maximum fill elevation into two datasets, one representing the highest elevation, hereafter the ceiling, and one representing the lowest elevation submerged by water, hereafter the ground (fig. 5D). At higher stage heights, a portion of the tank’s overlying rock was lower than the stage elevation. In these cases, we used the maximum elevation of the rock within a 5-cm moving window as the ceiling. In other cases, the ceiling was defined by the maximum or stage height elevation. The ground dataset was defined by the minimum elevation within a 5-cm moving window. The ground was subtracted from the ceiling to get a positive volume reported in cubic meters (m³).

We used the ceiling and ground point-cloud datasets to create 5-cm resolution digital elevation models (DEMs) that were used in the raster-based stage-storage capacity estimate. A 5-cm resolution was selected for DEM analysis as it provided sufficient detail to capture surface complexity while reducing noise introduced by differences within the horizontal footprint of the ceiling and ground. The DEMs were imported to Esri ArcGIS version 10.7 and differenced as they were with the point cloud datasets (Caster and others, 2022). The resulting raster was summarized using zonal statistics within the two-dimensional planar footprint of the tank’s ground DEM to estimate storage capacity.

We calculated the TIN-based volume estimate in Maptek PointStudio (Maptek PointStudio version 8.1; https://www.maptek.com/; accessed August 1, 2024) as a solid model of a reservoir. A solid model is created by defining a central location within the void space of the point cloud, which is used as a node to build facets at the distal extents of the point cloud. Solid model construction requires non-overlapping features from the perspective of the central node to accurately define radial distance relationships that are used for volume estimates. We subset the point cloud by stage height to eliminate overlapping areas and then combined the products to ensure the most accurate capacity estimates (see for example, fig. 5E).

We developed a stage-storage model for each tank as a best fit relationship between the mean of the volume estimates from the three methods and the stage heights. In addition to linear regression models, we evaluated power, exponent, log, and second-order polynomial model fitting functions. For models with an R-square value >0.9, we selected the best model as the most parsimonious equation with the lowest root mean square error (RMSE). We additionally developed a stage and water storage surface area model for each tank and those are presented in appendix 1.

Sediment Retention Dam Storage

In addition to the tank fill volumes, we investigated potential sediment trapped behind the upstream dam and calculated the remaining storage capacity of the impounded reservoir. During field surveys, we did not conduct subsurface testing to estimate sediment depths or porosities behind the sediment retention dams. To approximate sediment volume, we interpolated a channel bed using the slope and geometry of the surrounding banks and channel above and below the dam to estimate a hypothetical base elevation. Using the point-based method (described in subsection “Tank Stage-Storage Capacity”), we then subtracted the interpolated base from the point cloud measurements at the channel surface for the area we identified as unconsolidated sediment (fig. 6). This estimate of sediment volume approximates the combined volume of space filled by sediment and by pores (voids) in the sediment deposit that can be potentially filled with air or water.

Next, we used the elevation at the top of the dam to estimate the remaining void space above the current sediment surface but below the dam that could, for example, be filled in the future by sediment or water. To do this, we first subtracted the surface elevations for the area upstream of the dam from the elevation at the top of the dam. We next summarized the positive values within the channel bed to estimate the remaining storage capacity of the sediment retention dams upstream of each of the tanks.

Results

The lidar point cloud for each survey site covered the tank and a portion of the up and downstream area leading to and from the tanks (for example, fig. 7) that are provided in a separate data release (Caster and others, 2024). This included about 38 percent of the upstream contributing watershed area for Buckhorn Tank, around 90 percent of Senita Tank’s contributing watershed, and less than 5 percent of Eagle Tank’s contributing watershed. At all three sites, the lidar survey captured almost all the channel draining from the tank to an arroyo in the downstream valley, as well as the staging area for water trucks and the temporary water storage tanks. Visual comparisons of this survey data with the most recent aerial photographs acquired in 2019 for the U.S. Department of Agriculture (USDA) National Agricultural Imagery Program (NAIP) (Earth Resources Observation and Science [EROS] Center, 2018; https://earthexplorer.usgs.gov) demonstrated that each survey aligns well with other national datasets and could be useful for further watershed analysis.

Tank Capacity

The mean-stage storage capacities for each tank are provided as tabular data in the following subsections “Buckhorn Tank,” “Senita Tank,” and “Eagle Tank” (tables 1–3). In the tables, we present the mean capacities calculated from the point, raster, and TIN-based measurement methods in m³, liters (L), and gallons (gal; tables 1–3). Error estimates in the tables are unitless, provided as percentages of the mean stage-storage capacities, so that they can be more easily applied to any given unit of capacity. For ease of application in the field by the refuge managers, we present the results of the stage-storage capacity, the upper and lower bounds of the 95 percent confidence intervals, and the stage-storage model equations in gallons.

For all three tanks, the TIN-based solid volume method was consistently the smallest estimate of storage capacity (tables 1–3). The point and raster-based surface differencing methods produced more similar capacity estimates, though the point-based method estimates tended to be larger (tables 1–3). These differences were relatively small and resulted in an average standard error (SE) of measurements for Buckhorn, Senita, and Eagle Tanks of 1.8 percent, 2.3 percent, and 1.7 percent, respectively. The upper and lower bounds of estimated capacity as represented by the 95 percent confidence intervals for Buckhorn, Senita, and Eagle averaged plus or minus (±) 3.6 percent, 4.5 percent, and 3.4 percent of stage-storage capacity, respectively.

Buckhorn Tank

Buckhorn Tank, a modified natural tank, had the largest capacity of the three tanks included in this report, with a maximum storage capacity of 9,108.730±252.312 gal (table 1, fig. 8). The original portion of the tank, the volume of the rock borehole below the top of the concrete liner, represented a stage height of approximately 9 ft with a storage capacity of 7,192.540±201.391 gal (table 1, fig. 8). Using these measurements and the additional stage capacities in table 1, stage storage was best modeled using either a power or polynomial equation, both of which had coefficient of determination (R²) values >0.99. However, the second-order polynomial equation, BHS_gal (Buckhorn Tank Stage-Storage Model in water-storage units of gallons; fig. 8), had the lowest RMSE at 59.9 gal with an average residual <1 percent of measured capacity. Applying the BHS_gal stage-storage model, we approximate the stage heights at which Buckhorn Tank is at half and one-quarter capacity (that is, 50 percent and 25 percent full, respectively) are 6.4 ft and 3.8 ft, respectively. The model predicts that stage-storage drops below 100 gal at a stage height of 0.5 ft.

Table 1.    

Results of stage-storage volume estimation for Buckhorn Tank, a modified natural tank in Cabeza Prieta National Wildlife Refuge, Arizona.

[Mean stage-storage capacity represents the average storage capacity of the point, raster, and triangulated irregular network (TIN)-based measurement methods (described in “Tank Stage-Storage Capacity” section) at each stage height determined by a fixed rod within the tank. Note that the 9.0-foot (ft) stage height corresponds with the original capacity of the tank when filled to the top of the concrete liner, and the maximum capacity includes the shallow entrance into the tank, enclosed by a small masonry dam, above the liner. To facilitate use of each metric, the standard error (SE) and 95 percent (%) confidence interval (CI) are provided as a percentage of the mean stage-storage capacity. For ease of reference, mean stage-storage capacity is provided in the original measurement units: ft, foot; m³, cubic meter; L, liter; gal, U.S. gallon. For the graph of the stage-storage model see figure 8.]

Table 1.    Results of stage-storage volume estimation for Buckhorn Tank, a modified natural tank in Cabeza Prieta National Wildlife Refuge, Arizona.

Stage height (ft)	Point method (m3)	Raster method (m3)	TIN Method (m3)	Mean stage-storage capacity					SE (%)	95% CI (±)
				(m3)		(L)		(gal)
1.0	1.35800	1.35840	1.25000	1.32213	1,322.13		349.270		2.23	4.37
2.0	3.61900	3.61531	3.36000	3.53144	3,531.44		932.908		1.98	3.88
4.0	9.43400	9.34812	8.54000	9.10737	9,107.37		2,405.912		2.55	5.00
6.0	16.29800	16.12961	15.14000	15.85587	15,855.87		4,188.677		1.86	3.65
8.0	24.07400	23.91019	22.62000	23.53473	23,534.73		6,217.217		1.60	3.13
9.0	28.00600	27.31420	26.36000	27.22673	27,226.73		7,192.540		1.43	2.80
10.0	33.29000	33.10571	31.35000	32.58190	32,581.90		8,607.226		1.55	3.04
10.4a	35.46200	34.57891	33.40000	34.48030	34,480.30		9,108.730		1.41	2.77

^a

A stage height of 10.4 ft is the maximum stage-storage capacity.

Senita Tank

Senita Tank is a developed tank that is 485 gal smaller than Buckhorn Tank, with a maximum storage capacity of 8,623.308±226.793 gal (table 2, fig. 9) at an estimated stage height of 13 ft. Using the stage capacities in table 2, we modeled stage storage using a second-order polynomial equation, SNS_gal (Senita Tank Stage-Storage Model in water storage units of gallons; fig. 9), that had an R² value of 0.9991. SNS_gal had an RMSE of 84.4 gal with an average residual of 12.8 percent of measured capacity (fig. 9). Residual errors for SNS_gal were greatest at 1 ft, where the model underpredicts storage capacity by 61 percent (30.9 gal). At 2 ft stage heights and higher, the model better predicts stage storage with an average residual of 4.7 percent. Applying the SNS_gal stage-storage model, we approximate the stage heights at which Senita Tank is at half and one-quarter capacity are 8.8 ft and 5.9 ft, respectively. The model predicts that stage-storage drops below 100 gal at a stage height of 1.2 ft.

Table 2.    

Results of stage-storage volume estimation for Senita Tank, a developed tank in Cabeza Prieta National Wildlife Refuge, Arizona.

[Mean stage-storage capacity represents the average storage capacity of the point, raster, and triangulated irregular network (TIN)-based measurement methods (described in section “Tank Stage-Storage Capacity”) at each stage height determined by a fixed rod within the tank. To facilitate use of each metric, the standard error (SE) and 95 percent (%) confidence interval (CI) are provided as a percentage of the mean stage-storage capacity. For ease of reference, mean storage capacity is provided in the original measurement units: ft, foot; m³, cubic meter; L, liter; gal, U.S. gallon. For the graph of the stage-storage model see figure 9.]

Table 2.    Results of stage-storage volume estimation for Senita Tank, a developed tank in Cabeza Prieta National Wildlife Refuge, Arizona.

Stage height (ft)	Point method (m3)	Raster method (m3)	TIN method (m3)	Mean stage-storage capacity			SE (%)	95% CI (±)
				(m3)	(L)	(gal)
1.0	0.33200	0.30179	0.28000	0.30460	304.60	80.467	4.04	7.92
2.0	1.11500	1.04192	0.97000	1.04231	1,042.31	275.349	3.28	6.43
4.0	4.06800	3.91528	3.62000	3.86776	3,867.76	1,021.754	2.78	5.44
6.0	8.85000	8.64088	8.08000	8.52363	8,523.63	2,251.704	2.20	4.32
8.0	14.73700	14.61608	13.91000	14.42103	14,421.03	3,809.632	1.46	2.86
10.0	20.24000	20.30913	19.57000	20.03971	20,039.71	5,293.930	0.96	1.88
13.0a	33.16600	33.19235	31.57000	32.64278	32,642.78	8,623.308	1.34	2.63

^a

A stage height of 13.0 ft is the maximum stage-storage capacity.

Eagle Tank

Eagle Tank is a modified natural tank and the smallest tank examined in this report, with a maximum storage capacity of 6,039.603±120.188 gal (table 3, fig. 10) at an estimated stage height of 10.6 ft. Using the stage capacities in table 3, we modeled stage-storage using a 2nd order polynomial equation, EGS_gal (Eagle Tank Stage Storage-Model in water storage units of gallons; fig. 10), that had an R² value of 0.9987. EGS_gal had an RMSE of 79.7 gal with an average residual of 3.1 percent of measured capacity (fig. 10). Applying the EGS_gal stage storage-model, we approximate the stage heights at which Eagle Tank is at half and one-quarter capacity are 7.6 ft and 5.3 ft, respectively. The model predicts that stage storage drops below 100 gal at a stage height of 1.3 ft.

Table 3.    

Results of stage-storage volume estimation for Eagle Tank, a modified natural tank in Cabeza Prieta National Wildlife Refuge, Arizona.

[Mean stage-storage capacity represents the average storage capacity of the point, raster, and triangulated irregular network (TIN)-based measurement methods (described in section “Tank Stage-Storage Capacity”) at each stage height determined by a fixed rod within the tank. To facilitate use of each metric, the standard error (SE) and 95 percent (%) confidence interval (CI) are provided as a percentage of the mean stage-storage capacity. For ease of reference, mean storage capacity is provided in the original measurement units: ft, foot; m³, cubic meter, L, liter; gal; U.S. gallon. For the graph of the stage-storage model see figure 10.]

Table 3.    Results of stage-storage volume estimation for Eagle Tank, a modified natural tank in Cabeza Prieta National Wildlife Refuge, Arizona.

Stage height (ft)	Point method (m3)	Raster method (m3)	TIN method (m3)	Mean stage-storage capacity			SE (%)	95% CI (±)
				(m3)	(L)	(gal)
1.0	0.21000	0.19698	0.18000	0.19566	195.66	51.688	3.62	7.10
2.0	0.79000	0.75671	0.71000	0.75224	752.24	198.721	2.52	4.94
4.0	3.38400	3.30867	3.15000	3.28089	3,280.89	866.719	1.72	3.36
6.0	7.53100	7.49012	7.21000	7.41037	7,410.37	1,957.612	1.11	2.18
8.0	12.86600	12.84596	12.35000	12.68732	12,687.32	3,351.635	1.09	2.13
10.0	19.66200	19.49516	18.85000	19.33572	19,335.72	5,107.956	1.05	2.05
10.6a	23.22400	23.06318	22.30000	22.86239	22,862.39	6,039.603	1.02	1.99

^a

A stage height of 10.6 ft is the maximum stage-storage capacity.

Discussion

The U.S. Fish and Wildlife Service estimated the maximum water storage capacity of Buckhorn and Eagle Tanks after they were modified in 1949 and 1957, respectively, but there is not a similar record that we are aware of for Senita Tank after it was developed in 1958 (table 5). Maximum water storage capacity was later estimated for all three tanks by Broyles (1997) using a common geometric simplification termed conical volume. Broyles’ (1997) estimate of maximum water storage capacity for Eagle Tank only differed by 0.8 gal from the estimate by USFWS (1958a; table 5). However, Broyles’ (1997) estimate for maximum water storage capacity of Buckhorn Tank differed by 2,007.7 gal from the estimate by USFWS (1950; table 5). This discrepancy is likely due to the initial project lacking the small masonry dam at the edge of the concrete liner, which increased the capacity of the tank. Our estimate of capacity at 9 ft, likely the original maximum storage capacity of the tank, is 10 percent lower than the 1950 estimate.

We achieved Objective 1 by using ground-based lidar to refine these estimates of water storage capacities. We measured Buckhorn and Eagle Tanks to have a 9 percent and 7 percent smaller capacity, respectively, and Senita Tank to have a 23 percent greater capacity than previously reported by Broyles (1997; table 5). The percent differences in capacity estimates for Buckhorn and Eagle compared to Broyles (1997) meets our expectations for comparing volume calculation methods that incorporate different levels of geometric complexity, as indicated by percent standard error and 95-percent confidence interval among three different measurement methods used in this study. However, the 23 percent difference between the ground-based lidar volume estimate and the conical volume for Senita Tank previously reported by Broyles (1997) is greater than our expectation for methodological differences. We think that the difference is because a small-walled catchment that was used to convey runoff was included in our ground-based lidar calculations but may not have been considered by Broyles (1997; fig. 11). Using a stage height of 11 ft, which mostly excludes the walled catchment in the lidar data, models a stage-storage capacity of 6,421.041 gal (24,306.28 L), which is 8 percent different from conical volume estimates (fig. 11). It is also possible that a portion of the catchment was filled with sediment during the survey by Broyles (1997). These results show that the accuracy of the previous estimates of the maximum storage capacity for 49 perennial water sources by Broyles (1997) needs to be considered carefully prior to contemporary use. Factors such as tank geometric complexity may produce significant differences between lidar-based and conical volume estimation methodologies and documented or undocumented changes to tank capacity caused by sedimentation or management modifications may further increase these discrepancies.

To achieve Objective 2, we built a stage-storage model for each tank using the ground-based lidar survey and the recently installed stage height rods. Stage-storage capacity below the maximum capacity had not been formally estimated with any other measurements prior to this study. Stage storage was predicted using a second-order polynomial equation for each tank that provided estimations within the 95-percent confidence interval bounds for all except the smallest stage heights at Senita Tank. For that tank, the stage-storage model equation in SNS_gal underestimated storage capacity at a 1 ft height and slightly overestimated capacity up to a 4 ft stage height. Despite these differences, all three models provided an accurate estimate of stage-storage capacity from measured stage heights and will be useful tools for monitoring changes in water volumes for tank water storage and management needs.

We measured storage capacity after sediment was removed from each tank. Thus, the relationship between stage storage and stage height will change with future sedimentation. During pre-survey preparation, CPNWR staff removed sediment from each tank, with the greatest volume removed from Buckhorn Tank. In 1976, an estimated 5 yd³ (3.8 m³) of sediment was removed from Buckhorn (Dodson, 1976a). For both Buckhorn and Eagle Tanks, unconsolidated sediment had almost filled the channel to capacity behind the sediment retention dams located directly upstream of the tanks. Because little water can pool behind the dams and allow sediment to settle out, runoff water at these two tanks likely also contains sediment. A rock overhang above Eagle Tank causes runoff water to pool in an external basin separated from the excavated borehole by a short concrete wall containing several drainpipes, effectively serving as a secondary sediment retention basin and minimizing the amount of sediment deposited in the borehole. Using the stage-storage model, sediment filling the 1 ft stage height would reduce the maximum storage capacity by 4 percent, 1 percent, and 1 percent for Buckhorn, Senita, and Eagle Tanks, respectively. Sediment filling the 4 ft stage height would reduce the maximum fill capacity by 26 percent, 13 percent, and 14 percent for Buckhorn, Senita, and Eagle, respectively. Sediment over-topping the retention dam and filling Senita Tank is currently of low concern owing to the relatively large capacity remaining behind the dam, and the secondary sediment retention basin at Eagle Tank seems to minimize sedimentation. As supported by the quantity of sediment removed from Buckhorn Tank by CPNWR staff prior to surveying, however, sediment deposited by runoff water in this tank has the potential to greatly limit water storage capacity. For example, sedimentation since 2003 had limited the volume of Buckhorn Tank to about 50 percent of maximum capacity by July 2019. Refuge managers could use the estimates of sedimentation rates in this study (“Sediment Storage” section) to determine when and with what frequency to conduct periodic maintenance to clean the accumulated sediment out of the tanks to maximize water storage capacity.

The frequency with which floods contribute to sedimentation and tank recharge are not known. Broyles (1997) identified all three tanks as perennial water sources in the late 1990s, but CPNWR staff currently supplement this water supply by periodically hauling water to the tanks by truck and trailer. We do not yet have observations on changes in tank stage-storage volumes owing to factors like evaporation (appendix 1), leaks, or consumption by wildlife, but we expect that the natural fill volume and recharge rate for each tank is currently declining owing to decreases in precipitation and increases in air temperature in recent decades. Since 1970, annual rainfall has declined, and air temperatures have increased (fig. 12) using Sen’s Slope estimate of an assessment of precipitation and air temperature reported for the nearest National Oceanographic and Atmospheric Administration Cooperative Observer Program (NOAA–COOP) weather station Tacna 3 NE USC00028396 (NOAA, 2022), located approximately 30 miles north of Buckhorn Tank. The Sen’s Slope estimate is a non-parametric statistical test used to identify the magnitude and direction of temporal trends that is complemented by the Mann-Kendall test, used to assess the trend’s significance (Sen, 1968). Although winter rains have apparently not significantly changed over the last five decades since 1970, summer rains during the North American Monsoon have decreased at the 99 percent confidence interval (fig. 12). Additionally, temperatures within the region have significantly increased at the 99.9-percent confidence interval both annually and during the monsoon season (fig. 12). These climatic changes in rainfall patterns and temperatures may reduce natural water availability during seasonal periods when wildlife is more likely to visit tanks (Harris and others, 2020), thus increasing the importance of maintaining these resources.

Conclusion

In cooperation with the U.S. Fish and Wildlife Service (USFWS), the U.S. Geological Survey (USGS) Southwest Biological Science Center used ground-based lidar to measure the water storage capacity for three tanks in the Cabeza Prieta National Wildlife Refuge (CPNWR), Arizona, which are important water sources for desert bighorn sheep (Ovis canadensis mexicana). In addition to the maximum storage capacity, we also estimated storage capacity for at least six stage heights below the maximum capacity to develop a stage-storage model for each tank. The lidar surveys of each tank and estimated contours for each stage height are provided in an associated data release (Caster and others, 2024). Of the three tanks in this report, Buckhorn Tank had the largest maximum water storage capacity at 9,108.730 gallons (gal) (34,480.30 liters [L]), followed by Senita Tank at 8,623.308 gal (32,642.78 L), and then Eagle Tank at 6,039.603 gal (22,862.39 L). The resulting stage-storage models provide storage capacity estimates that are typically within the 95 percent confidence bounds of measured capacities and are accurate enough for refuge managers to monitor water volumes as needed. The ability to consistently monitor loss and recharge within these tanks is important for assessing water availability for desert bighorn sheep, though the accuracy of these estimates can change over time if sediment is deposited within the tanks and reduces water storage capacity. At Buckhorn Tank, a large volume of sediment was removed by CPNWR staff prior to the survey, demonstrating that sedimentation within the tank has occurred in the past and will likely occur in the future.

The confined and complex cave-like geometries of the tanks made ground-based lidar an invaluable tool for evaluating the three tanks within this report. Each tank had at least a portion of its storage confined within shallow boreholes within bedrock that were easily mapped by ground survey. Thus, we note the utility of ground-based lidar relative to, for example, airborne lidar which would not have adequately surveyed the interior of the borehole tanks but could be a different and additionally important resource to provide topographic and geomorphic context of the surrounding watersheds. These three tanks represented an immediate priority for water resource monitoring within the available time and budget of this project; however, these methods are applicable to other natural, modified, and developed water resources within the region. Development of storage capacity estimates, and stage-storage models from 3-dimensional ground-based lidar data, can be useful tools for understanding water storage, recharge, and loss in these habitats managed for desert wildlife.