Purpose and Scope

Since 2014, the National Park Service (NPS) has been aerially applying GBHs on a yearly basis to eliminate the spread of buffelgrass in the Rincon Mountain District. However, comprehensive information regarding the fate and transport of GBHs in a semiarid environment was not available at the time of the environmental assessment for their aerial application. The water resources in the Rincon Mountain District are scarce and therefore critical to the survival of native aquatic species, such as the sensitive lowland leopard frog. The potential for contamination of aquatic resources resulting from surface runoff and (or) aerial drift is a primary concern and in need of further investigation.

A 4-year U.S. Geological Survey (USGS) study was implemented to increase understanding of the occurrence, distribution, fate, and transport of glyphosate in surface water derived from past and current aerial herbicide applications. Sediment and water were sampled for glyphosate and its primary metabolite, aminomethylphosphonic acid (AMPA), during various hydrologic flow conditions in three watersheds at different stages of herbicide treatment. The potential for transport of herbicide from within the park boundary was assessed. Concentrations in water and sediment samples were related to ecologically sensitive organisms within the context of the park setting. The information gained through the study provides the Rincon Mountain District with the information needed to evaluate the specifics of the herbicide application strategy by identifying any unanticipated side effects of the application program and mitigating any negative effects of the program on the broader Rincon Mountain District aquatic ecosystem. The broader ecosystem implications are also important for other agencies when assessing their own buffelgrass remediation strategies.

Description of Study Area

The Rincon Mountain District of Saguaro National Park encompasses a diverse and rugged landscape that contains a portion of the Sonoran Desert as well as the oak and pine-forested peaks of the Rincon Mountains (fig. 2). The Rincon Mountains make up most of the terrain within the eastern park district and are one of several mountain ranges of southern Arizona. Collectively the mountain ranges are known as the “Madrean Archipelago” or “Madrean Sky Islands” and together these mountain ranges offer a mechanism for genetic exchange and support a complex and unique biodiversity. A major reason for this biodiversity is a mixing of different life zones on the relatively isolated mountain ranges, distinguished by the large elevation and climatological change within a short horizontal distance. The semiarid desert at the bottom is connected to mountain-top forest ecosystem, which promotes unique adaptive characteristics and life-zone mixing of the ecological communities (Bezy, 2016). Elevations range from 2,670 feet (ft) along the western edge of the park district to 8,666 ft at the summit of Mica Mountain. There are extensive stands of saguaro cactus, prickly pear and cholla (Opuntia spp.), ocotillo (Fouquieria splendens), various agave (Agave spp.), yucca species (Yucca spp.), and woody shrubs covering the lower elevations and bajadas. Higher elevation hillslopes are vegetated with grassland, scrubland, manzanita (Arctostaphylos spp.), oak (Quercus spp.), and juniper woodlands (Juniperus spp.). The upper elevation mountainous terrain is covered with mixed conifer forest, mostly ponderosa pine (Pinus ponderosa) and Douglas-fir (Pseudotsuga menziesii) (Powell and others, 2006).

Stream channels within the study area are mostly bedrock controlled and the geometry, sinuosity, and substrate depend on the local geologic structure and lithology of the terrain over which the stream flows. In the higher elevations, stream channels are generally narrow and high gradient, with cobble and large and small boulders that shape the channel morphology in the mountainous landscape of the study area. The lower elevation channel sections are more alluvial, with channel banks and bed substrates that are more mobile than in the higher elevations and channel morphology that is primarily controlled by stream discharge (Parker, 2006). The alluvial reaches are characterized by sand- and gravel-bed channels with discontinuous alluvial banks, and frequent expressions of bedrock sections on the channel bottom, which expose the thin alluvial depositional layer. Tinajas and pools can form in narrow bedrock sections where a fragment breaks from the bottom of the channel along joint planes or a trough has been eroded into the narrow channel bottom by continuous fluvial incision or abrasion (Parker, 2006). The carved bedrock tinajas are generally round or oval shaped, frequently occur in multiples that continuously step down a sharp channel gradient, and have a wide range of lengths and depths, ranging from a few feet to more than 20 ft across, and from less than a foot to more than 10 ft deep (Parker, 2006). The recharge and flowpaths maintaining and replenishing these tinajas have been a topic of ongoing geochemical studies in the Rincon Mountains (Eastoe and others, 2004; Eastoe, 2012).

Soils generally are thin (7 to 20 inches to bedrock) except on some ridge tops and small upland basins. The general soil classification of the ridges and bajadas for the study area is the Lampshire-Romero-rock outcrop complex (Cochran and Richardson, 2003). This soil classification unit is 45 percent Lampshire (gravelly loam), 20 percent Romero (gravelly sandy loam), 15 percent Rock outcrop, and 20 percent other soil types covering small areas (Cochran and Richardson, 2003). Lampshire-Romero is neutral to alkaline and organic matter averages 1 to 5 percent. It is well drained, with medium rates of runoff, a moderately permeable complex, and a low soil-erosion tolerance factor (Cochran and Richardson, 2003). In the study area, the highest density of buffelgrass patches grow in this soil unit and therefore it receives the most herbicide application (fig. 3).

Parker (2006) described channel alluvium and sediment transport as mostly substrate consisting of small cobble-size gravel (2–128 millimeters [mm] diameter) and sand (0.0625–2 mm diameter). Channel bottoms and tinaja sites with deposits generally contain poorly sorted sandy gravels, with a silt-clay content of less than 5 percent. The low content of fine-grained sediments (particle size less than [<] 0.0625 mm) in the Rincon Mountains indicates that hillslope sediments in the study area are nearly cohesionless, consisting of sands and gravels, and that fine silts may get trapped prior to reaching to the stream channel, thus leading to suspended sediment flow that is supply limited in fine particles, resulting in very little deposition downstream (Parker, 2006).

Most of the Rincon Mountain District is located within the larger Rillito-Pantano Wash watershed (920 square miles). Drainage of the southern- and western-facing portion of the Rincon Mountains in the park flows towards Rincon Creek (45 square miles), which in turn flows west toward Pantano Wash (fig. 1). Rincon Creek is the most consistently flowing drainage in the park. Except for a portion of the pools and tinajas, there are no perennial stream reaches in the park; streamflow is often intermittent at higher elevations and ephemeral at lower elevations. Several tinajas are considered perennial and these are crucial habitat for aquatic and terrestrial species’ survival during dry periods. Streams flow mainly in response to runoff from the intense convective rainfall that falls during the North American monsoon (July through September) and from winter precipitation and spring snowmelt. In some years, tropical moisture from decaying tropical cyclones can also cause heavy rainfall in the mountains from late September to early November. Overall, the most consistent sources of surface water are the deep tinajas shaded by canyon walls and riparian vegetation that minimize evaporation. Tinajas may receive additional inflow from springs or seeps that discharge groundwater from the fractured rock that encloses the pools (Parker, 2006). The hydrogeology of the Rincon Mountain District is complex and a complete understanding of the source waters to springs and pools has not been developed, but geochemical studies and surface-water monitoring are helping to provide this information (Eastoe and others, 2004). This greater regional understanding, however, is complicated and exacerbated by changing climate conditions, such as long-term trends toward less precipitation and warmer temperatures in an already arid climate (Zylstra and others, 2015).

Loma Verde canyon, Box Canyon, and Madrona Canyon watersheds are the focus and primary treatment areas of this study, and they are contained mostly within the Rincon Mountain District boundary. The watershed above the Loma Verde Trailhead is an unnamed canyon but referred to by the National Park Service as “Loma Verde canyon” and hereafter referred to “Loma Verde canyon” in this report. The watersheds are roughly 1.36, 6.53, and 5.70 square miles, respectively (fig. 2, table 1). The land use of these three watersheds is primarily classified as lower shrubland including desert, and the higher elevation portions of the watershed are classified as forest. A small percentage (less than 2 percent) has some degree of urbanization (low, medium, and high density), but urbanized area is lower in the watershed outside of the park boundary. Mean basin elevation for Loma Verde canyon, Box Canyon, and Madrona Canyon is 3,990, 4,290, and 5,640 ft, respectively (table 1). The terrain in the watersheds is steep with average slopes exceeding 30 percent. Much of the area draining from Madrona and Box Canyon is above 5,000 ft (64 and 33 percent, respectively). These high elevation areas enhance precipitation orographically and increase winter and spring runoff, contributing to subsurface flow of water to pools and tinajas. The mean annual precipitation for the Rincon Mountain District area is around 20 inches per year (U.S. Geological Survey, 2018; table 1), but the park measured average annual precipitation as approximately 16.8 inches from February 2014 to February 2018. Local precipitation for the sampling period was obtained from rain gages at the Madrona Ranger Station (KAZWSB-2, elevation 3,425 ft) and the Desert Learning Center (elevation 2,750 ft) (Walking Shadow Ecology, 2021). Values from the two locations were averaged. The 2014–15 winter was particularly wet with about 16.3 inches of precipitation recorded (average of Desert Learning Center and Madrona Ranger Station rain gages). Patterns during the study were representative of recent historical averages and patterns. The steep slopes and high elevation areas of the study area commonly result in rapid runoff and high energy floods capable of moving debris and substrate. Estimated magnitude and frequency of floods are similar between the Box Canyon and Madrona Canyon watersheds. The Loma Verde canyon watershed flood estimates are lower because of watershed size. The 50-percent exceedance probability peak flow (also known as the 2-year flood) ranges between 114 and 278 cubic feet per second and the 1-percent annual exceedance probability peak flow (100-year flood) ranges between 1,300 and 3,530 cubic feet per second (Paretti and others, 2014; U.S. Geological Survey, 2018).

Related Glyphosate-Based Herbicide-Use Studies for Resource Management Purposes

In the last 20 years, resource managers have been using more aggressive approaches that involve the use of herbicidal applications to manage invasive non-native plant species. Most often herbicides are used in conjunction with physical methods, such as burning, cutting, or uprooting. Herbicides are used worldwide in natural resource management settings to prepare forest sites for planting or to control competing vegetation after the tree crop has been planted, but research related to ecological implications has been limited (Rolando and others, 2017). The herbicide application in the setting of the Rincon Mountain District, where it is being used to control buffelgrass in a mostly upland desert and semiarid climate, is somewhat unique relative to other uses in the United States and other countries where it is used to control tamarisk (Tamarix spp.), giant reed (Arundo donax), and various weeds in forests and along river corridors (Feng and others, 1990; Bakke, 2001; Dixon and others, 2003; Briggs, 2009; Gresham, 2010; U.S. Department of Agriculture, 2012; Thistle and others, 2014; Rolando and others, 2017; González and others, 2017).

In U.S. Forest Service programs, GBHs are used primarily for conifer release situations, site preparation, and noxious weed control (Durkin, 2011). In silviculture and forest management, herbicides are mostly used to reduce competitive stresses exerted by undesirable weeds and other broad-leafed vegetation on small conifer plants. The GBHs are also used to control weeds in nursery settings without injuring hardwood seedlings.

Assessments have been conducted to better understand the effects of drift during ground-based glyphosate applications, especially in proximity to water (Feng and others, 1990). In one case study, Bakke (2001) reported that with one exception, there were no glyphosate detections in water and sediment sampling from riparian zones conducted after ground-based applications targeting noxious weeds in reforestation projects. A review by Rolando and others (2017) led to the conclusion that GBHs, as typically employed in planted forest management, did not pose a significant risk to humans or the terrestrial and aquatic environments.

Tamarisk and giant cane are extremely invasive in riparian communities and, in many areas of the southwestern United States, have completely replaced native vegetation during the past several decades. Governmental and non-profit agencies have tested several weed control strategies including chemical eradication with the use of the herbicides imazapyr, triclopyr, and glyphosate. Although glyphosate has proven effective at eradicating unwanted plant species in many situations and is seemingly successful as part of a restoration strategy, there have been some unanticipated issues documented such as some noxious weeds developing glyphosate resistance as well as the introduction of herbicides to river waters (Briggs, 2009; González and others, 2017).

Herbicide applications for a resource management action often are specific to the regional setting, but because the environmental conditions in which they are implemented may differ, the fate and potential effects are not known. For example, an intermittent stream treated with an herbicide in the Southwest would result in a different degradation process and environmental fate compared to a wet and colder boreal forest in the Pacific Northwest. These environmental conditions determine the concentrations that are bioavailable and the exposure potential. These differences could affect species of different regions in different ways and ultimately change the risk to a similar species living in different habitats. A common approach to addressing animal species risk is to defer to regulatory agency standards such as the U.S. Environmental Protection Agency’s acute and chronic pesticide criteria, toxic effects levels, or reregistration eligibility decision (U.S. Environmental Protection Agency, 1993, 2004). However, these blanket effect levels may not represent all the conditions described above and likely only address an upper threshold for single species, which may not be relevant to the environmental setting, exposure, or multiple species types present in specific types of habitat.

Sample Design, Site Description, and Approach

At the initiation of the current USGS study, the GBH treatment of watersheds was ongoing. The Loma Verde canyon, Box Canyon, and Madrona Canyon watersheds were at different stages of treatment regimens. The USGS study sampling focused on the most heavily treated watershed, Box Canyon (fig. 5A), and the Madrona Canyon watershed, which had not yet been treated prior to the 2014 start of the study. The Box Canyon watershed was used to monitor the occurrence and extent of glyphosate and AMPA of residual GBHs from previous treatments as well as newer 2015 to 2018 treatments (13 locations; table 4). The Madrona Canyon watershed was used as a controlled setting where samples were collected before and after the very first treatment (fig. 5B). Samples were also collected during rainfall runoff events to monitor the transport of glyphosate and AMPA from treated areas (five GBH collection locations; table 4). The Loma Verde Canyon watershed had no aerial treatment after September 2015 (figs. 5C, 6). This watershed was only sampled once in April 2017, and was used to evaluate the legacy effects of treatment (4 locations; table 4). The primary interest was to know if glyphosate and AMPA could still be detected more than a year after treatment. Primary flow paths or larger tributary catchments (greater than 0.05 square mile) were identified as areas of increased transport from spray area to downstream pools (fig. 5A–C). The larger watershed boundaries are identified in figure 5 to show flow direction and potential transport pathways from spray area to downstream areas. Pools were frequently sampled below these areas susceptible of increased GBH transport.

Water, aquatic sediment, and terrestrial sediment were collected seasonally during the study period (fig. 6) and the sampling was determined by timing of GBH treatments, seasonal climatic conditions, and flow events. The station naming convention was established by the Rincon Mountain District during pool and tinaja inventories. A number was assigned to the primary pool or tinaja and the next distinct pool increased sequentially with distance up the watershed. Some pools are closely connected with flow in a series of pools and water parameters (specific conductance, dissolved oxygen, and pH) were indistinguishable. In these instances, the primary pool number was changed to an alphanumeric name. For example, the primary pool is “Pool 01” and the related downstream Pool is named “Pool 1A.” During this study, low-flow conditions sometimes caused a shift in the previously known location of the primary pool and that resulted in a new station that included a down or upstream reference to the primary pool in the name. For example, the new pool would be named “downstream of Pool 01.” Water and aquatic sediment were mostly collected from pools during times of base flow, which could vary greatly depending on seasonal conditions. During the winter and spring, rainfall and snowmelt created base-flow conditions that would lead to continuous flow between the pools and cause homogenous water chemistry throughout the main stream channel. During dry periods of the study, very few pools were full or flowing and conditions could be different depending on recent deposition of debris and sediment, which could prevent repeat sampling at some locations.

In the Box Canyon watershed, Box Canyon pool 10 (USGS station 320949110414601; hereafter referred to as “Pool 10”) was sampled once (for isotopes only); Pools 06 and 07 (USGS stations 320945110415901 and 320946110415501, respectively; hereafter referred to as “Pool 06” and “Pool 07”) were sampled once for water and aquatic sediment; and Pools 4A, 1A, 01, and downstream of pool 01 (USGS stations 320935110421301, 320929110422401, 320926110422901, and 320923110423201, respectively; hereafter referred to the pool common names) were sampled multiple times for water and aquatic sediment (fig. 7A–G). Pool 00 (USGS station 320915110423101; hereafter referred to as “Pool 00”) was sampled frequently for water and aquatic sediment owing to the consistent presence of standing water in a small tributary to the main Box Canyon stream channel (fig. 7H). The Madrona Canyon watershed had two of the larger perennial tinajas of this study. Upper Madrona Canyon pool 09 (USGS station 320950110355701; hereafter referred to as “Pool 09”) was far from any GBH treatment and used as a reference sample, whereas Lower Madrona Canyon pool 02 (USGS station 320907110360801; hereafter referred to as “Pool 02”) was below the treatment area (fig. 8A–B). An old shallow concrete impoundment filled with sediment between Pools 02 and 09 created a downstream discharge point that was sampled on two visits and named Lower Madrona below dam (USGS station 320910110360901) (fig. 8C). Runoff was collected from different areas in the lower Box Canyon watershed and rainfall runoff flood-water samples were collected using automatic samplers in Box and Madrona Canyons (fig. 9A–D). In the Loma Verde canyon watershed, water and aquatic sediment were collected upstream from Pool 03 or Loma Verde above pool 03 (321114110420801; hereafter referred to as “above pool 03”), whereas isotope samples were collected in Loma Verde pool 03 and Loma Verde pool 01 (USGS stations 321116110421001 and 32114511042160107, respectively; hereafter referred to as “Pool 03 and Pool 01” (fig. 10A–C).

Terrestrial soil samples were collected from one location in each of the Loma Verde and Madrona Canyon watersheds and from two locations in the Box Canyon watershed (fig. 11A–C; table 4). The sampling was timed to measure conditions prior to herbicide treatment and after treatment, which provided information about the rate of dissipation on the treated soils. The timing of water and aquatic sampling relative to treatment was evaluated in conjunction with rainfall events to quantify glyphosate and AMPA that may runoff with treated sediments.

Sample and Data Collection

Data collection methods, cleaning procedures, and quality assurance followed protocols described in the USGS “National Field Manual Collection of Water-Quality Data” (U.S. Geological Survey, 2006), Wilde (2008), and Shelton and Capel (1994). Water-quality parameters, including temperature, pH, specific conductance, and dissolved oxygen, were measured as conditions permitted for each water-sampling site using an In Situ Inc. smarTROLL multiparameter meter. The multiparameter meter was calibrated daily before field measurements were collected using standards supplied by the USGS National Water Quality Laboratory (U.S. Geological Survey, 2006). Field parameters were only measured during base-flow collections.

Flood surface-water sample collection utilized Teledyne ISCO^TM Avalanche portable refrigerated automatic samplers and a Teledyne ISCO^TM 6700. A refrigerated automatic sampler was operated as a permanent deployment in the Madrona Canyon watershed and temporary deployments (refrigerated or iced) were made in the Box Canyon watershed. The automatic sampler in the Madrona Canyon watershed was configured with a carousel holding twelve 1-liter polyethylene bottles and Box Canyon had a single 2.5-gallon glass container used for a composite sample. The automatic sampler intake consisted of 3/8-inch inner diameter vinyl tubing inside conduit and affixed to a T-post. The height of the intake tubing for the permanent deployments was adjusted to 6 to 10 inches above the dry channel to minimize the pumping of bed material and to sample a representative well-mixed water column during flooding events. Samplers were initiated using a pressure transducer or liquid-level actuator. After samples were collected, the refrigerated automatic sampler kept samples chilled at 2 to 4 degrees Celsius (°C) and upon retrieval the samples were kept at this temperature until shipped and analyzed. Generally, sample bottles and tubing used for glyphosate collection would be made of polytetrafluoroethylene or glass, but this was not feasible with the automatic sample configuration. To ensure quality control, blank samples were collected with the automatic sampler to test for possible contamination. Flood samples were visibly void of fine sediments, so no suspended sediment samples were collected from surface-water sites.

Water stage height was recorded using an In Situ vented Level TROLL 700 pressure transducer or a non-vented Solinst® 3001 Levelogger pressure transducer. The data were logged continuously at 15-minute intervals and transmitted to the USGS National Water Information System (U.S. Geological Survey, 2020) or downloaded manually. The data collected via Levelogger were corrected for atmospheric pressure using a Solinst® Barologger.

Discrete water and fine sediment samples were collected in pools for analysis of glyphosate and AMPA. Terrestrial soil samples were collected along canyon ridges where herbicides were directly applied. Water samples collected for glyphosate and AMPA analysis were collected using a 20-milliliter (mL) high density polypropylene syringe. The syringe was rinsed with native water and then a Whatman 25-mm, 0.7-µm pore-size baked glass-fiber Luer-Lock filter was attached to the syringe and the plunger was slowly depressed to force several milliliters of sample water out of the syringe to rinse the filter. Once rinsed, approximately 5 mL of sample was filtered into a 20-mL baked amber glass vial. Aquatic fine-grained substrates—stream substrates (<2 mm)—consisting primarily of a mixture of clay, muck, and organics, were often difficult to gather owing to their limited presence. When collected, they consisted of 4 to 5 subsamples of fine-grained substrates collected from bottom deposits or along the margins of pools and composited. Sands and fine gravels were avoided. Fine sediments were scooped from upper depositional surfaces with a clean stainless-steel spoon and composited in a baked 125- or 250-mL amber glass jar. Terrestrial soils were scraped from the upper 2 inches of the land surface at 4 to 5 locations and composited in a 125- or 250-mL wide-mouth amber glass baked jar. All samples were shipped to the Kansas Water Science Center’s Organic Geochemistry Laboratory and analyzed for glyphosate, AMPA, and glufosinate. Glufosinate is a broad-spectrum herbicide that is used to control certain weeds in agricultural settings. Because the Rincon Mountain District does not use this compound in their mixture, no detections were observed during the study, but laboratory quality control data are presented as a negative control. Discrete isotope samples were collected from pools using a 1-liter polyethylene bottle for tritium and a 60-mL glass bottle for stable isotopes of hydrogen and oxygen (δ²H and δ¹⁸O) and shipped to the University of Miami and the USGS Reston Stable Isotope Laboratory, respectively.

Quality control samples consisted of three water blanks analyzed for glyphosate and AMPA. Automatic samplers were a major concern of possible contamination owing to the inability to use fluoropolymer or glass to collect samples. The infrequent detections in base-flow and flood environmental samples imply that contamination bias was likely not an issue. A background (no herbicide) soil sample was collected as a reference and one soil replicate was collected to determine the variability of the composite style collection approach.

Analytical Laboratory Methods

All analyses were performed by USGS laboratories following published operating procedures. Analytical methods follow Meyer and others (2009) for glyphosate and AMPA analysis, and isotopic methods follow Révész and Coplen (2008a, b).

Statistical

Censored data analysis was used to analyze the glyphosate and AMPA results. Computing summary statistics on analytical results reported below (non-detected) or above the laboratory detection limit requires a censored data analysis. The Nondetects and Data Analysis for Environmental Data (Helsel, 2012) package in R (Lee, 2015) was used to calculate the median, mean, and standard deviation. The censored regression on order statistics or “cenros” function from Nondetects and Data Analysis for Environmental Data was used to compute the summary statistics. Traditional non-parametric approaches used to graph and compute general statistics were completed using JMP^© software (SAS Institute, Inc., 2007). The dissipation kinetics of glyphosate in water were fitted using JMP version 12 software to a simple first-order three parameter exponential decay model, characterized by the following equation:

Quality Control

Field and laboratory blank water samples were all below the detection limit of 0.02 µg/L for glyphosate, AMPA, and glufosinate (table 5). Glufosinate is a separate herbicide associated with agricultural settings and is not expected to be found in the Rincon Mountain District. It was used in this study as a negative control. Sediment sample laboratory blanks consisted of a potassium hydroxide solution and all laboratory blank samples were below the method detection limit of 1.0 µg/kg.

The percent difference of the seven laboratory split replicates was zero for AMPA and glufosinate. One of the seven split replicates analyzed for glyphosate had a difference of 11 percent. A laboratory split replicate was done for one sediment and one soil sample. The glufosinate (not detected) had a 0 percent difference for both replicates, whereas the soil glyphosate and AMPA had percent difference of 6 and 12, respectively. The field concurrent replicate had a relative standard deviation that was 465 µg/kg plus or minus 14 percent for AMPA and 375 µg/kg plus or minus 25 percent for glyphosate. The high variability likely was a result of the aerial application process or uneven spraying and methods used to collect soil composite samples.

The two laboratory spikes for the water samples averaged 106, 165, and 114 percent for AMPA, glufosinate, and glyphosate, respectively. The two aquatic sediment laboratory spikes averaged 138, 220, and 87 percent for AMPA, glufosinate, and glyphosate, respectively. Standard reference standards analyzed for the sediment samples for 10, 50, and 100 µg/kg of glyphosate averaged a percent recovery of 101, 96, and 99 percent, respectively.

Blanks indicated that field and laboratory procedures and analyses were not introducing bias in the form of contamination. In addition, reproducibility in replicate samples showed little variation although most of the water samples collected during the study were below or near the detection limit. Soil samples were more variable than sediment samples, which may be related to different soil types, uneven herbicide distribution, and heterogeneous mixture of particles and organic materials. Laboratory method recovery of glufosinate in water was above 100 percent and biased high for glufosinate although no detections were observed in the study. Internal sediment samples spiked and analyzed for AMPA and glufosinate were biased high but had lower recovery for glyphosate at 83 and 90 percent.

Surface-Water Pools and Runoff

Water-quality parameters were collected during base-flow conditions with the exception of three sets of parameters collected in flood recession conditions. Mean dissolved oxygen ranged between 5.47 and 10.4 mg/L and the median pH was slightly basic at 7.2 pH units. The largest differences were observed in specific conductance. The Madrona Canyon watershed had the lowest mean concentrations at 70.9 µS/cm, whereas the Loma Verde canyon and Box Canyon watersheds mean concentrations were 395 and 180 µS/cm, respectively (table 6). Evaporation had an increasing effect on pool specific conductance, especially in the dry summer months. Seasonal differences were also related to runoff contributions diluting pool and tinaja water in the spring and fall.

The frequency of detections of glyphosate and AMPA in water samples was low for the 48 total water samples collected. Less than 15 percent of the samples had a detection of glyphosate, AMPA, or both (table 7; fig. 12A). The mean water concentrations of glyphosate and AMPA were below the laboratory detection limit of 0.02 µg/L, with the highest glyphosate concentration collected during a flood event in Box Canyon (0.20 µg/L; fig. 12B). No measurable glyphosate or AMPA was present in the one sample collected in the Loma Verde canyon watershed. Water detections were more common in the base-flow samples than in flood samples, but the flood sampling was limited in the Box Canyon watershed. Pool 00 had the greatest number of water detections and the pool is located below a steep section in a small side tributary (0.21 square mile) to the main channel of Box Canyon, on the eastern part of the watershed. Six water samples were collected in Pool 00 and all the samples had detections of AMPA and two had detections of glyphosate. The highest AMPA concentration of 0.63 µg/L was measured in Pool 00. All other base-flow water samples collected in Pools 07, 06, 4A, 1A, and 01 did not have any detections of glyphosate or AMPA.

Three flood events were sampled in the Box Canyon watershed, although two of the composites collected by the automatic samplers did not sample the rise and peak of the flood profile. The runoff following treatment can mobilize higher concentrations of the contaminant with initiation of the flood event, and the leading edge of a peak flow can be correlated with the peak concentration of contaminant (McCarthy, 2006). The February 2018 composite flood sample was the first flow after the 2017 herbicide application and the composite sample had detections of glyphosate and AMPA. The rise, peak, and portion of the recession were sampled as a composite and concentrations of 0.2 µg/L of glyphosate and 0.04 µg/L of AMPA were measured.

In the Madrona Canyon watershed, neither compound was detected in the water of Pool 02 or Pool 09 during base-flow conditions (fig. 12C). Between August 1, 2016, and March 1, 2018, eight flood events were recorded in the Madrona Canyon watershed and samples from the automatic sampler were collected at multiple points during the flood hydrograph for seven of those events (15- or 30-minute intervals collected during the rise, peak, or recession limb of the flood hydrograph). Samples collected on the rising and (or) peak portion of the flood were most often sent in for analysis to maximize the probability of detecting compounds at the highest concentration of the flood. A few flood events resulted in multiple peaks during consecutive days and multiple samples were analyzed for glyphosate and AMPA during the same flood event. Only 2 of the 20 samples analyzed during flood events had detections of glyphosate. The two samples were collected during the same flood event on July 29, 2017, with one collected on the rise and the other on the peak of the flood, and the concentrations of glyphosate were 0.05 and 0.03 µg/L, respectively. No glyphosate was detected during the recession and AMPA was not detected in any of the three samples. The previous treatment application was sprayed mid-August 2016.

Stable isotope ratios measured on 25 water samples ranged from 0.19 to −16.5 per mil and −18.3 to −122 per mil for δ¹⁸O and δ²H, respectively. Samples at the low end of the range plot near the global meteoric water line (Craig, 1961) and samples near the high end of the range plot near the local meteoric water line (developed by Eastoe and Dettman, 2016; National Park Service, 2017b; fig. 13) were determined using samples collected from rainfall in Tucson, Arizona, and the Santa Catalina Mountains. Samples that plot to the right of the meteoric water lines indicate that the water may have undergone evaporation. Additional samples were collected by Saguaro National Park in cooperation with Chris Eastoe during June 2010 and 2014 (Eastoe, 2012; National Park Service, 2017b) and represent a local meteoric water line for the study area using the following equation:

Most samples collected plotted to right of the global meteoric water line showing a slightly evaporated signal. Madrona Canyon samples follow the trajectory of the global meteoric water line whereas the Box Canyon samples more closely follow the local meteoric water line, although most samples plot to the right of the local meteoric water line with partially evaporated signal and possibly originating from higher-elevation precipitation. Some of these samples represent spring (April) samples collected in Box Canyon. The fall (November) samples from Madrona Canyon and the spring (April) samples from Loma Verde canyon show a less evaporated signal and potentially represent waters from higher elevations.

Seven sites were sampled more than one time and show variability of stable isotope values between samples (table 7). The variation in stable isotopic values may be related to different seasonal precipitation events supplying different isotopic ratio type water to the sampled pools over time. The variation in stable isotopic values over time may indicate that the water supplying the pools, prior to our sample collection, has a short residence time in the system.

Tritium is a useful tracer for determining if there is a component of water recharged during the period of nuclear bomb testing in the 1950s and 1960s, when tritium in the atmosphere peaked and then decreased during the following decades. Tritium values have stopped decreasing in recent precipitation (after 1992), and average recent values of tritium in precipitation in Tucson are 16.7 pCi/L (Eastoe and others, 2012). Tritium values from nine water samples ranged from 9.9 to 18.9 pCi/L. Values between 11.3 and 20.9 pCi/L are considered to represent water with residence time on the order of a few decades (Eastoe and others, 2012). Most of the water samples may represent more recently recharged waters on the order of years.

Aquatic Sediment

The frequency of glyphosate in aquatic sediment detections was similar to water samples, but AMPA was about 10 percent higher than water samples, detected in 25.0 percent of the 28 samples (fig. 14A). Mean aquatic sediment concentrations were 1.13 and 4.42 µg/kg for glyphosate and AMPA, respectively. No measurable glyphosate or AMPA was found in the one sample collected in the Loma Verde canyon watershed. Aquatic sediments sampled in Pool 00 in the Box Canyon watershed had the highest concentrations of the current study at 22 and 76 µg/kg of glyphosate and AMPA, respectively (fig. 14B). Pools 1A and 4A in the Box Canyon watershed also had detections of AMPA in aquatic sediment samples but they were closer to the overall mean concentration. Other than Pool 00, glyphosate was only detected in one sample at Pool 1A. In the Madrona Canyon watershed, eight aquatic sediment samples were collected and only one sample in Pool 02 had detections of glyphosate (4.10 µg/kg) and AMPA (7.70 µg/kg; fig. 14C).

Terrestrial Soil

Nineteen terrestrial soil samples were collected from areas treated with GBHs, which coincides with the greater the frequency of detection. Of the 19 soil samples collected, 89.5 percent of the soil samples collected contained glyphosate and 100 percent of the soil samples contained AMPA (fig. 15A). Mean concentrations of glyphosate and AMPA were 678 µg/kg and 1,240 µg/kg, respectively. The single terrestrial soil sample collected in the Loma Verde canyon watershed had concentrations of glyphosate (11 µg/kg) and AMPA (150 µg/kg) 592 days after the last treatment application. Terrestrial soil concentrations for both compounds were consistently higher in the Box Canyon watershed (fig. 15B). Three samples in the Box Canyon watershed exceeded 5,000 µg/kg for AMPA. The Box Canyon watershed received the greatest volume of GBHs applied and had the greatest extent of coverage. The Madrona Canyon watershed concentrations after treatment were lower than the Box Canyon watershed and never measured greater than 1,000 and 1,500 µg/kg for glyphosate and AMPA, respectively (fig. 15C).

Terrestrial soils were sampled a few times per year to measure the changes in glyphosate and AMPA after GBHs were applied. Concentrations of glyphosate and AMPA mostly decreased between applications, but there were differences between sites in the dissipation rates of compounds. Glyphosate dissipated more rapidly than AMPA, but because AMPA is the primary metabolite of glyphosate, degradation was simultaneously augmenting and compounding the total AMPA concentrations. Box Canyon north ridge (USGS station 320945110421001; hereafter referred to as the “Box Canyon north ridge”) showed more than a 99-percent decrease in glyphosate in about 10 months (2,800 to 8.8 µg/kg) and more than a 90-percent reduction in AMPA concentrations (greater than 5,000 to 410 µg/kg). A similar pattern occurred at Box Canyon south ridge (USGS station 320920110422401; hereafter referred to as the “south ridge of Box Canyon”) after about 11 months (greater than 5,000 to 26 µg/kg for glyphosate and 5,200 to 890 µg/kg for AMPA). After the first ever treatment application to the Lower Madrona Canyon east ridge (USGS station 320922110361301; hereafter referred to as the “east ridge of Madrona Canyon”), initial concentrations were lower than those measured in the Box Canyon watershed, although less volume was applied to the area. Samples collected 4 to 5 months later showed increases in both compounds, but after 9 months more than 99 percent of both compounds had dissipated.

Factors in Occurrence, Transport, and Fate of Glyphosate and Aminomethylphosphonic Acid

The climate, physiography, and soil conditions are important factors in the processes affecting the dissipation and movement of GBHs from treatment areas along the upper terraces and mountain ridges to the waters lower in the watershed. The persistence of glyphosate in an ecosystem is suspected to be primarily affected by microbial biodegradation processes, which are dependent on moisture, temperature, and organic carbon. Giesy and others (2000), Borggaard and Gimsing (2008), and Székács and Darvas (2012) described that when most of the previously listed environmental conditions and resources are limited, like those in a desert ecosystem, glyphosate dissipation will be reduced and ultimately GBHs will be more persistent. This persistence in the treatment areas was observed during the study and the longevity of GBHs was exemplified in the Loma Verde canyon watershed. Glyphosate was still present in the one sample collected 592 days after the last treatment of GBHs. The low ratio of 0.07 (11 to 150 µg/kg of glyphosate to AMPA) indicates that glyphosate was in a late stage of dissipation, but still at levels considered elevated when compared to other studies (Giesy and others, 2000; Battaglin and others, 2014).

Studies have documented that glyphosate dissipation follows a simple first-order decay model in soil (Tang and others, 2019). In the current study, glyphosate and AMPA dissipation in soil was modeled using a simple first-order decay function for samples collected at the Box Canyon south ridge location (USGS station 320920110422401). This area was only treated once during a 2-year period and samples were collected through that period to understand glyphosate dissipation. Dissipation rates and runoff processes affect the downstream Pool 00 where glyphosate and AMPA were frequently detected. The benefit of this type of modeling is to determine the potential half-life of glyphosate in the Rincon Mountain District environment and provide information about periods of higher glyphosate concentrations that could potentially be in runoff after treatment applications. The modeled first half-life at the Box Canyon south ridge site was around 75 and 125 days for glyphosate and AMPA, respectively. The equation of the glyphosate decay follows the three-parameter equation:

The decay curve is steeper for the first half-life and second order reactions level off more gradually and decay rates slow as time increases. It was estimated to take 330 and 595 days for glyphosate and AMPA concentrations, respectively, to approach a 95-percent reduction from the initially sprayed concentration. The estimated dissipation rates indicate a 2- to 3-month period in which the risk of high glyphosate concentrations transporting to downstream waters could occur. Such transport could be minimized if aerial spraying were timed with annual periods of low rainfall. This decay rate is somewhat consistent with what was observed in the Loma Verde canyon watershed 592 days after treatment although the amounts applied and the measured concentrations were greater in the Box Canyon watershed. The north ridge Box Canyon and the Lower Madrona Canyon east ridge terrestrial soil locations (USGS stations 320945110421001 and 320922110361301, respectively) also showed a similar pattern of decreasing glyphosate and AMPA between applications. The in situ decay was only modeled at the one location in the Box Canyon watershed using a small sample size. It should be noted, the second sample, taken approximately 250 days after treatment, had a large residual and actual concentrations were about twice as high as the modeled estimate indicating a lesser rate of decay in the first 200 days. Most notably, the GBH compounds in the treated terrestrial environments are persistent and last for at least a year or more. In addition, the repeat treatments are likely renewing GBH compound mass, increasing antecedent concentrations, and possibly compounding the AMPA metabolite concentrations as glyphosate degrades into AMPA. More sampling would be needed to understand soil and environmental variability to verify decay rate curves, but the results are consistent with the dissipation rates observed during other studies (Laitinen and others, 2006; Rodriguez-Gil and others, 2016; Yadav and others, 2017).

The prevailing consensus in the literature is that risk of groundwater pollution at depth is low and the most influential factors in glyphosate transport are rainfall intensity and subsequent erosion of particulates and glyphosate retained in upper 2 to 100 centimeters of the topsoil (Bowmer, 1982; Carlisle and Trevors, 1988; Feng and others, 1990; Laitinen and others, 2006; Borggaard and Gimsing, 2008; Battaglin and others, 2014; Yang and others, 2015; Napoli and others, 2015). The highest density of buffelgrass patches for all the watersheds studied was most often located on the soil type classified as the Lampshire-Romero-Rock outcrop complex (fig. 3). Dissipation rates of GBHs are affected by soil conditions and the properties of those soils. This soil type has few silts and clays, is of high permeability with low water capacity (Cochran and Richardson, 2003), has a shallow depth to bedrock, and is fragile with a high erosion factor that can promote mobilization of substrates sprayed with GBHs. In the Box Canyon watershed, these soils were treated for multiple years and the frequency of detections of glyphosate and AMPA increased when the spraying application (volume and area) increased and the annual rainfall was elevated, specifically during 2015 and 2017 when monsoon-related rainfall—typically intense with high runoff—was high (fig. 16A). The 2014 and 2016 years had greater winter precipitation—typically light to moderate intensity with low runoff—than 2015 and 2017. The driest of the 4 years analyzed was 2016 which coincided with the year that had the lowest number of detections in water and aquatic sediment samples within the Box Canyon watershed (fig. 16B).

The Madrona Canyon treatment area was also located on this soil type. The Madrona Canyon watershed was the least-sprayed watershed overall; in terms of volume and area, the treatment was an order of magnitude smaller than that of the Box Canyon watershed (fig. 17A) and the Madrona Canyon watershed had very few detections in the aquatic sediment and water (fig. 17B). The erodible sediments associated with the Lampshire-Romero-Rock outcrop complex were evident in an upstream tributary that drained the treatment area and it appeared to have flowed multiple times during the study. One aquatic sediment sample from Pool 02 in Madrona Canyon had glyphosate and AMPA present in the slightly drier year of 2016, although this might be explained by a September storm that had a 1-day total of greater than 1 inch of rain.

Results indicate that the soils in the Rincon Mountain District sprayed with GBHs most likely dissipated or transported with runoff. The few compounds that were detected in the Box Canyon and Madrona Canyon watershed pools were not detected on subsequent visits, indicating the compounds were from recent runoff and likely metabolized in the aquatic environment, and not from deeper groundwater sources that might renew concentrations. Snowmelt does infiltrate from higher elevations, mostly in untreated areas. This water sustains base flow in the pools the spring months and is unlikely transporting GBHs along deeper flow paths. At lower elevations there is potential for infiltration along short flow paths through shallow bedrock fractures associated with Cellar-rock and Lampshire-Romero-Rock outcrop complexes. Stable isotopic signatures of the groundwater can help to characterize water source and timing of infiltration. The groundwater collected at the sites will be marked with a time-averaged isotopic signature of precipitation in the areas where it fell as precipitation, unless evaporation or exchange with rock oxygen have taken place (Eastoe, 2012).

Groundwater stable isotope samples from previous sampling in the Rincon Mountain District and surrounding areas (Eastoe, 2012) were determined to plot along an elevation gradient (0.16 per mil per 100 meters for δ¹⁸O and 1.1 per mil per 100 meters for δ²H). The stable isotopes were determined from volume-weighted precipitation samples from Tucson, Ariz. (2,444 ft), and the Santa Catalina Mountains at a higher elevation (7,946 ft) just north of Tucson. The stable isotope samples collected for the present study were from a small elevation range (2,948 to 3,815 ft) but span a large range of values, which supports the idea that the water in this study represents recent contributions to the groundwater from precipitation rather than from integrated groundwater sources (fig. 13). Pools thought to be affected by groundwater sources were sampled during base flow to generalize the age and source of the groundwater at such sites.

A range of stable isotope ratios were measured at Pool 00 in the Box Canyon watershed and differences in stable isotope ratios helped to distinguish waters influenced by recent rain and elevated GBH compounds from other pool sources. The 2015 treatment was followed by a wet September and this resulted in 0.17 µg/L of glyphosate and 0.63 µg/L of AMPA, and the stable-isotope ratios were low, at −121.7 and −15.16 per mil for δ²H and δ¹⁸O, respectively. The low values are likely related to recent runoff, because the other two samples from Box Canyon Pool 00 from August, 2016, and April, 2017, had higher stable-isotope values (by an order of magnitude), which could be associated with other sources or evaporation. Glyphosate was not detected in these August and April samples and AMPA concentrations were much lower than those measured in September. The stable isotopes support runoff as the primary source of elevated GBH compounds. The timing of the GBH treatment along with rainfall likely contributed to the higher concentrations in the September 2015 sample.

Tritium isotopes were also measured at sites thought to have more groundwater contributions. Nuclear bomb tritium is still present in aquifers recharged with rainwater since about 1953 and therefore tritium can be used to distinguish between water containing tritium below detection limit, which must have recharged before about 1953, and water containing tritium above detection limit, which must contain some water that fell as rain since 1953. Values between 11.3 and 20.9 pCi/L are considered to represent water with residence time on the order of a few decades (Eastoe and others, 2012). The water samples ranged from 9.9 to 18.9 pCi/L, which represent groundwater with residence times of at the most a few decades and may represent more recently recharged waters on the order of years. Box Canyon watershed waters were all greater than 11 pCi/L. The two samples from Madrona Canyon (10.96 pCi/L) and Loma Verde canyon (9.87 pCi/L) were slightly below the 11.3 pCi/L threshold and may represent water that has a small component of water recharged prior to the initiation of above ground nuclear testing starting in 1953. A sample collected from Loma Verde Pool 03 in June 2014 had a tritium value of 2.9 pCi/L with an associated δ¹⁸O and δ²H of −7.5 and −59 per mil respectively, indicating a minimal evaporative signature. The low tritium value from Loma Verde may suggest a component of older water in at least a portion of this watershed.

The tritium and stable isotope data provide supporting information about the waters sampled and help characterize the infiltration from recent runoff. Based on these data, the waters sampled for GBHs were young in age and most originated from recent runoff or nearby infiltration. Overall the tritium results were more like the tritium signature of rain in the region (Eastoe and others, 2012) and there was minimal evaporation based on most of the stable isotope samples, which indicates these waters are most commonly sourced from localized storm events rather than integrated during several storm seasons.

Ratios of concentrations of glyphosate to AMPA can provide a generalized interpretation of the GBH source and stage of dissipation in the watersheds. Because the soil half-life of AMPA (metabolite) is several times longer than glyphosate (parent compound), the increased accumulation of AMPA over time can provide information about the sources transported and the residence times of compounds. Larger ratios indicate soils were recently treated and transported, whereas smaller ratios indicate a longer period of transformation (dissipation) prior to transport from source to stream (Battaglin and others, 2005; Ramwell and others, 2014; Medalie and others, 2020). In the current study, ratios of one indicate neither glyphosate nor AMPA were detected. Few sample ratios were above one and these samples were water samples collected during runoff or soil collected shortly after GBH treatment. Ratios of 0.5 or less indicate AMPA concentrations are double or more than glyphosate concentrations. Using only ratios below one, the average ratio for soil samples (0.32) was very similar to the average ratio of aquatic sediment samples (0.36), which indicates the sediments found in pools were probably recently deposited from a terrestrial soil source where past GBH application has enough time for glyphosate to degrade AMPA. Otherwise, aquatic sediments affected by persistent GBHs would likely manifest as smaller ratios than soil. The Box Canyon watershed had the most samples (all medium types) that were below a ratio of 0.5, which is consistent with most years of treatment (average ratio of 0.34). The one soil sample ratio from the 2014 treated Loma Verde canyon was very low at 0.072 whereas the most recent treatment area of Madrona Canyon had an average a ratio of 0.46.

Because glyphosate and AMPA often bind with soil particle surfaces, high intensity precipitation is needed to mobilize such particles and thus the GBHs. Studies have shown a peak in concentrations following heavy rain (Borggaard and Gimsing, 2008) and in one setting AMPA residues peaked as many as 3 weeks later in stream sediments after nearby application and then persisted for more than 1 year (Wan, 1986). Although no discharge was measured during runoff events, maximum flows can be estimated for the watershed from peak-flow recurrence interval regression equations to estimate instantaneous maximum loads (Paretti and others, 2014). The estimated peak discharges listed in table 2 for the Box Canyon watershed were used to compute an instantaneous load calculation. The 100-year recurrence interval flood for peak discharge is estimated between 1,800 and 6,950 cubic feet per second (95-percent confidence interval) (U.S. Geological Survey, 2018). The maximum glyphosate concentration observed in flood waters during the study was 0.20 µg/L in the Box Canyon watershed. Using these flow estimates and maximum glyphosate concentration in an instantaneous load calculation, the maximum peak load estimates range between 9.84 x 10^-3 and 3.79 x 10^-2 grams per second for the 95-percent lower and upper confidence interval, respectively.

To reframe these estimates using a daily load (total mass per day) calculation with a more common magnitude peak flow, like the 2-year to 5-year peak discharge, the daily loads would range between 132 and 335 grams per day for estimated flows between 278 and 709 cubic feet per second. To our knowledge, floods did not exceed a discharge of 300 cubic feet per second during the study. These observations were based on channel conditions (debris and water marks) during site visits while there was flooding. To put these daily load estimates in context, an over-the-counter common residential-use gallon of Roundup ready-to-use^® contains 45.4 grams (0.1 pound) a.e. of glyphosate and Roundup^® super concentration contains 1,633 grams (3.6 pounds) a.e. of glyphosate.

Studies of the household application measured maximum concentrations between 4.00 and 8.99 µg/L of glyphosate and 1.00 and 5.8 µg/L of AMPA in runoff waters collected in the downstream urban drainage (Ramwell and others, 2014; Tang and others, 2015). Rain total and intensity along with user application (frequency and amount applied) affected the runoff concentrations and caused high variability in the measured concentrations (Ramwell and others, 2014; Tang and others, 2015; Hanke and others, 2010). These findings indicate that only a small percentage of the applied GBHs was transported as glyphosate by rainfall runoff and flooding during this study. Relative to what has been reported with household usage, the concentrations transporting from the Rincon Mountain District are lower. It is also assumed that most of the applied GBH was retained in the watershed and (or) degraded. It should be noted that these daily load estimates make several assumptions and do not account for sources of variability. This calculation scenario uses a high-flow maximum discharge to demonstrate what a representative maximum load estimate might be during a high-flow event in the Rincon Mountain District. It should be noted that during the study many of the runoff events had no detections of glyphosate or AMPA.

National Comparison of Different Studies and Applications

Glyphosate-based pesticides are the most widely used in the United States for agricultural and non-agricultural use and have become the largest-selling pesticides in the world (Franz and others, 1997; Baylis, 2000; Veiga and others, 2001; Kolpin and others, 2006). As of 2014, non-agriculture glyphosate use was 26.5 million pounds (Aspelin, 1997; Kiely and others, 2004; Grube and others, 2011; Battaglin and others, 2014; Benbrook, 2016). Several studies have broadly summarized glyphosate and AMPA occurrence and concentration in surface water, groundwater, atmosphere, and sediment samples from diverse hydrologic settings and a wide geographic range of locations in the United States (Battaglin and others, 2005, 2014; Scribner and others, 2007; Struger and others, 2008; Medalie and others, 2020). The studies have also used the data to identify in which hydrologic settings, land-use activities, and medium (water, air, sediment) glyphosate and AMPA are likely to be present, and to a limited degree the temporal patterns of their occurrence or concentrations.

As a comparison to the Rincon Mountain District’s use of the GBH, the urban residential, commercial, and municipal uses of GBHs are far more common and frequent. Among other organic and inorganic compounds, stormwater runoff often contains glyphosate, AMPA, and related surfactants at concentrations far exceeding concentrations observed during the Rincon Mountain District study (Ramwell and others, 2002, 2014; Struger and others, 2008, 2015; Byer and others, 2008; Botta and others, 2009; Hanke and others, 2010; Tang and others, 2015). Nonetheless, the extent, type of GBHs used, and contribution to surface-water contamination in an urbanized setting have not been well quantified. Ramwell and others (2014) reported finding maximum concentrations of 8.99 µg/L of glyphosate and 1.15 µg/L of AMPA in runoff waters from localized applications but also noted a marked decrease with subsequent runoff events. Rain total and intensity along with the frequency and amount of GBHs applied affected the magnitude and variability of concentrations measured in runoff (Ramwell and others, 2002, 2014; Hanke and others, 2010; Tang and others, 2015). Differences between what was applied and what was measured as a load were large. Runoff from larger impervious surfaces, such as applications to weeds near sidewalks and roads, resulted in the highest concentrations in the urban setting (Ramwell and others, 2002; Botta and others, 2009; Struger and others, 2015). Concentrations from runoff of these surfaces were as high as 650 µg/L for glyphosate, but glyphosate measurements from subsequent events were 50 and 3.2 µg/L, 2 and 25 days later, respectively (Ramwell and others, 2014). The maximum glyphosate concentrations observed in the Rincon Mountain District runoff were 0.2 and 0.63 µg/L for standing base-flow samples. These measurements were an order to several orders of magnitude lower than concentrations observed in the urban environment where these herbicides are generally applied more frequently (multiple seasons) and where runoff may affect surface and groundwater resources (rivers and shallow wells).

During this study, water concentrations of glyphosate and AMPA were mostly below the laboratory reporting limit and the only samples with detections were collected during runoff and from Pool 00. The Rincon Mountain District median and maximum glyphosate and AMPA water concentrations were lower than median and maximum reported in two U.S. Geological Survey national assessments that compiled samples collected from various land-use settings and watershed sizes within the United States (Battaglin and others, 2014; Medalie and others, 2020; fig. 18A). Two of the runoff samples collected in the Box Canyon watershed had concentrations higher than many of the concentrations measured in the national studies, which is a compiled range of concentrations from rivers, streams, wetlands, ponds, and lakes. Medalie and others (2020) observed greater concentrations of glyphosate and AMPA in smaller watersheds or streams compared with larger watersheds and rivers, which is relevant to the small, steeply sloped watersheds commonly present in the Rincon Mountain District.

Similar to the water, median and maximum aquatic sediment glyphosate and AMPA concentrations were lower than the median and maximum soil and sediment sample concentrations compiled from the Midwest region in Battaglin and others (2014; data from 7 sites in Indiana and Mississippi; fig. 18B). The highest aquatic sediment concentrations measured in the Box Canyon Pool 00 were higher than a portion of those Midwest concentrations, but these concentrations were statistical outliers in the current study (22 and 76 µg/kg of glyphosate and AMPA, respectively). Aquatic sediments with elevated GBH compounds were mostly found in Pool 00. Pool 00 is below a small catchment—roughly 5 percent of the total watershed area—that had less sustained flow compared to the greater Box Canyon watershed; however, a relatively large percentage of the area of this small catchment was sprayed between 2014 and 2016 (fig. 17A). The infrequently flowing drainage feeding Pool 00 likely retains depositional materials sourced from upstream runoff, such as sediments and organic material, as well as potential for wind-blown materials owing to the near proximity of the treatment area. Most aquatic sediment samples had no glyphosate or AMPA, and the few samples from other pools with GBH detections were associated with specific runoff events. Outside of areas like Pool 00, aquatic sediments with GBHs present a low risk for the long-term exposure to aquatic organisms because fine sediments are so limited in the three watersheds.

Toxicological Review

Glyphosate is a weak acid (alone referred to as “technical grade glyphosate”) and the active ingredient in all glyphosate formulation products. The glyphosate in different products carry many chemical forms, such as isopropylamine, dimethylamine, trimesium, and potassium salts (Hanson and others, 2013). Each salt has a different molecular weight affecting the ratio of acid to salt and changing the acid equivalent and chemical properties of many of these formulations. The salts, as well as surfactants and other adjuvants, enhance the adsorption into the plant or change the delivery of the glyphosate into the plant membrane. Although most added surfactants or other adjuvants are not considered active ingredients, they nonetheless have known biological effects (Giesy, and others, 2000; Tsui and Chu, 2003; Edginton and others, 2004; Howe and others, 2004; Fuentes and others, 2011; Papoulias and others, 2013; Vincent and Davidson, 2015). Specifically, surfactants have the greatest negative effect on organisms of all formulation components. The various formulations will affect the physical and chemical properties, efficacy, and toxicity of the herbicide in different ways. Furthermore, these differences complicate comparisons between formulations and limit studies designed to understand the biological effects and biochemical pathways affecting plant and animal species. As a result, the primary studies of GBH toxicity focus on the active ingredient (glyphosate) or the most common salt (isopropylamine) rather than on the surfactants (which have the greatest negative effect on non-target organisms).

As previously described, when comparing glyphosate concentrations in the different forms (whether it be the technical grade, salts, or other formulations) the a.i. should be converted to the a.e. Glyphosate is an acid but generally comes in different types of salts mixed in solution. Labels contain percentage information as a weight of a.i. per weight of dry or liquid product. This percentage or concentration of glyphosate in a specific salt does not allow researchers to compare percent of glyphosate active ingredient among products. Using an a.e. measure (glyphosate acid per unit volume of product) instead standardizes the percent glyphosate regardless of salt formulation (Hanson and others, 2013).

All pesticides sold or distributed in the United States must be registered by the U.S. Environmental Protection Agency and this registration must be based on scientific studies showing the pesticides can be used without posing unreasonable risks to people or the environment (U.S. Environmental Protection Agency, 1993). Several studies and reviews have investigated various formulations for a wide range of concentrations under different conditions and for different species (Giesy, and others, 2000; Hurley and others, 2008; Székács and Darvas, 2012; Durkin, 2011; Relyea, 2011; Annett and others, 2014). Durkin (2011) provides an extensive literature review and comprehensive human health and ecological risk assessment of glyphosate and glyphosate formulations for the U.S. Department of Agriculture, Forest Service. From this risk analysis, tools were developed to perform many of the calculations used in the human health risk assessments and ecological risk assessments prepared for the different glyphosate applications used by the U.S. Forest Service. Annett and others (2014) also provide an extensive review of the chronic and acute effects.

The U.S. Environmental Protection Agency’s drinking water standards (2018) list maximum contaminant levels at 700 µg/L for glyphosate. The child health advisory of the 1- and 10-day dose that is not expected to cause any adverse noncarcinogenic effects during the given 1- or 10-day period is 20 mg/L. The daily oral exposure to the human population that is likely to be without an appreciable risk of deleterious effects during a lifetime is 2 milligrams per kilogram per day and the drinking water lifetime exposure level is 70,000 µg/L (U.S. Environmental Protection Agency, 2018). The aquatic life benchmarks for glyphosate and AMPA were all several orders of magnitude greater than any surface-water concentration observed during the study.

Acute toxicity testing determines the concentration of effluent or ambient waters that causes an adverse effect (usually death) on a group of test organisms during a short-term exposure (for example, 24, 48, or 96 hours; U.S. Environmental Protection Agency, 2019). Acute concentrations for fish are 21,500; 249,500; and 34,700 µg/L for glyphosate (technical grade weak acid), AMPA, and the glyphosate isopropylamine salt, respectively (table 8). The invertebrate acute benchmarks are 26,600 and 341,500 µg/L for glyphosate and AMPA, respectively. Chronic toxicity test is a short-term test, usually 96 hours or longer in duration, in which sublethal effects (for example, significantly reduced growth or reproduction) are usually measured in addition to lethality. Chronic levels of glyphosate for fish and invertebrates are 25,700 and 49,900 µg/L, respectively (U.S. Environmental Protection Agency, 2004). The Canadian freshwater-quality guideline is 800 µg/L (Canadian Council of Ministers of the Environment, 2012).

The U.S. Environmental Protection Agency Reregistration Eligibility Decision for glyphosate (U.S Environmental Protection Agency, 1993) states that glyphosate is slightly toxic to birds and is practically non-toxic to fish, aquatic invertebrates, and honeybees, but the inert ingredient—mostly surfactants—in some products must be labeled “toxic to fish” if used in aquatic environments. Although most of the available data indicate that the mammalian toxicity of glyphosate is low, and very few specific hazards can be identified, there are concerns about unknown, less obvious effects to biochemical parameters in blood as well as tissues, and inhibition of some enzymes (Jiraungkoorskul and others, 2003; Cavalcante and others, 2008; Modesto and Martinez, 2010; de Castilhos Ghisi and Cestari, 2013; Webster and others, 2014; Smith and others, 2019).

Acute toxicity testing determines what constitutes a lethal concentration of a substance for 50 percent (LC50) of a given animal population. Chronic toxicity studies generally feed different oral doses of the substance to determine a toxic dose for a specific animal. Chronic limits commonly use the no-observed-adverse-effect concentration (NOAEC) and the lowest-observed-adverse-effect concentration (LOAEC). The NOAEC is a concentration where there is no biologically or statistically significant increase in the frequency or severity of any adverse effects of the tested protocol. The LOAEC is the concentration or amount of a substance determined by experiment or observation that causes an adverse alteration of morphology, function, capacity, growth, development, or lifespan of a target organism distinguished from normal organisms of the same species under defined conditions of exposure. Studies that evaluate acute toxicity and chronic toxicity for glyphosate may not be realistically representative of environmental conditions or routes of exposure, especially for cases where herbicides will be applied in natural resource management settings. However, the LC50 does define an upper limit for organisms, which provides a means to compare various organisms’ sensitivities to a given glyphosate formulation, and the effect on a given organism of various glyphosate formulations. The 96-hour LC50 estimates are commonly used to compare this upper toxicity level (U.S. Environmental Protection Agency, 1993).

Durkin’s (2011) literature review and ecological risk assessment of glyphosate and glyphosate formulations presents a wide range of LC50 studies and toxicity testing performed by the U.S. Environmental Protection Agency and other researchers that includes data from the research studies mentioned earlier in the section (Giesy and others, 2000; Hurley and others, 2008; Relyea, 2011; Székács and Darvas, 2012). As a way to summarize Durkin’s (2011) data review in the context of the Rincon Mountain District study, LC50 a.e. concentrations were compiled for fish, aquatic-phase amphibians (tadpoles and frogs only), and aquatic invertebrates (appendixes 6–8 in Durkin, 2011), and broadly grouped by glyphosate formulation type as a means to generalize acute toxicity testing in the aquatic environment. This approach compiles single LC50 concentrations reported from multiple studies without consideration of uncertainty or study design and presented as a statistical distribution or boxplot. Because these LC50 distributions are presented for a broad group of organisms exposed to a range of glyphosate products, the values presented should by no means be used to define ecological criteria for aquatic resources in the Rincon Mountain District. The purpose of placing the Rincon Mountain District results within the context of the LC50 distributions is to provide a broader value range that could be considered the upper bounds of detrimental effects to aquatic organisms in the park. General comparisons among fish, amphibians, and aquatic invertebrates LC50 results were made for the technical grade glyphosate and glyphosate formulations presented by Durkin (2011). A higher LC50 response indicates a less toxic substance; therefore, a low LC50 response is related to a more toxic chemical response.

The technical grade form of glyphosate overall had a higher LC50 dose or required more chemical to get the LC50 response for all three groups of organisms (fish, amphibians, and invertebrates) when compared to other formulations. Median concentrations for fish and amphibians were near 100 mg/L a.e., whereas aquatic invertebrates were slightly higher at around 130 mg/L a.e. Aquatic macroinvertebrates overall appeared to be more tolerant of glyphosate than the other two groups. This tolerance was more obvious in the glyphosate formulation LC50 toxicity testing where LC50 responses for aquatic invertebrates were 5 to 10 times greater than for fish and amphibians, respectively. Median LC50 concentrations for the glyphosate formulations were around 8 mg/L a.e. for fish, 4 mg/L a.e. for frogs, and 44 mg/L a.e. for aquatic macroinvertebrates (Durkin, 2011). For reference these concentrations are 4 to 5 orders of magnitude greater than the highest concentration detected in water at the Rincon Mountain District (0.0002 mg/L a.e.).

The most common fish family tested was the salmonids (rainbow trout and other salmon species). Although no fish currently reside in the three the Rincon Mountain District watersheds studied, future introduction of native fish is a possibility. However, the glyphosate formulation LC50 species summarized here have little species relation to the possible options for native fish reintroduction and the LC50 response for native fish may be quite different (fig. 19A). The amphibian testing focused primarily on four species, but none appeared to have major differences in the glyphosate formulation LC50 testing. Life stage appeared to be an important tolerance factor; tadpoles and embryos overall had greater LC50 concentrations. There were species from the genus Lithobates, the same genus as the lowland leopard frog, used in many of the toxicity tests, but comparing this directly to native species response would require further testing (Durkin, 2011).

Distinguishing between the different formulations presented in Durkin (2011) would be difficult and would require individual analysis of each study, which was outside the scope of this review. Some general groupings were applied to the different glyphosate formulations (fig. 19B). Primarily, these were distinguished as the inclusion or non-inclusion of the surfactant POEA, or the inclusion of another unknown surfactant such as those in Roundup brand name formulations. General patterns indicated that lower LC50 (greater toxicity) concentrations were associated with the POEA surfactant formulations (median concentration of 7 mg/L a.e.) as well as the various unspecific Roundup formulations with surfactants (median concentration of 11 mg/L a.e.). Overall, the non-surfactant formulations had much higher LC50 (lower toxicity) concentrations. Although associated pH data were limited, there did appear to be a trend in the response to concentrations at different pHs (6 to 9.5) for technical grade glyphosate which increased 2 orders of magnitude with increased pH (less harmful) and decreased 1 order of magnitude with decreasing pH for glyphosate formulations (more harmful).

Research on more environmentally relevant concentrations, such as chronic toxicity, genotoxic effects, and endocrine disrupting effects were reviewed for this report. There is no literature specifically on the lowland leopard frog, but studies have been conducted on many aquatic invertebrate and vertebrate species. Physiological and neurological pathways can be affected by low-level concentrations of glyphosate and glyphosate formulations (Jiraungkoorskul and others, 2003; Cauble and Wagner, 2005; Cavalcante and others, 2008; Modesto and Martinez, 2010; de Castilhos Ghisi and Cestari, 2013; Webster and others, 2014). Published NOAECs and LOAECs are generally multiple orders of magnitude above the highest glyphosate concentrations measured in water resources during this study. More important considerations are the potential interactions among disease, pesticides, and water quality in amphibian habitats. It is possible low-level concentrations might not always have immediate or obvious effects, but the combination of exposure, timing, and emerging diseases will have the greatest effect on aquatic organisms. Sediments may also play a larger role in the potential effects to amphibians (Fuentes and others, 2011). Aquatic sediment detections in the Rincon Mountain District were infrequent, but they were observed in sensitive habitats used by lowland leopard frogs. Unfortunately, no studies provide sediment effect levels that could be compared to sediment levels in the Rincon Mountain District pools.

All measured glyphosate concentrations in the study were well below the Federal water-quality criteria and benchmarks for drinking water, chronic, and acute levels for surface-water and freshwater organisms (U.S. Environmental Protection Agency, 2018). Glyphosate from the application area is thus likely to have low ecotoxicological significance. Concentrations observed in the aquatic resources in the Rincon Mountain District were low relative to other published NOAEC and LOAEC criteria (U.S. Environmental Protection Agency, 1993). Additional studies could focus on relevant concentrations using environmental conditions consistent with the Rincon Mountain District, but in a laboratory setting. This would help control the variability of the levels observed in water and sediment and understand the exposure timing and potential effects this may have on the metamorphosis of lowland leopard frogs.