Skip to main content
Dryad

High rates of anticoagulant rodenticide exposure in California Barred Owls

Cite this dataset

Hofstadter, Daniel (2021). High rates of anticoagulant rodenticide exposure in California Barred Owls [Dataset]. Dryad. https://doi.org/10.5061/dryad.47d7wm3c8

Abstract

Pesticide use is pervasive and the exposure of non-target wildlife has been well documented over the past half century. Among pesticides, anticoagulant rodenticides (AR) have emerged as a particularly important threat in forests of the western United States, with exposure and mortality reported for several species of conservation concern. To further quantify this threat, we collected specimens of Barred Owls (Strix varia) and Barred Owl x Spotted Owl hybrids from the Klamath and Cascade Mountains and Sierra Nevada in California, USA to use as indicator species for environmental contamination with AR and to infer exposure of closely related and ecologically similar Northern and California Spotted Owls (S. occidentalis caurina, and S. o. occidentalis, respectively). We tested 115 Barred Owl and 12 Barred Owl x Spotted Owl hybrid livers for eight AR compounds and found high rates of exposure (62%) across our study area, and greater than previous studies in the Pacific Northwest. In addition, we sampled seven ovaries from seven females and 100% tested positive for AR. Female Barred Owls were more likely than males to be exposed (78% and 50%, respectively). Unlike previous studies, we found no clear link between illegal cannabis cultivation and AR exposure. However, Barred Owls sampled in proximity to the wildland-urban interface (WUI) were more likely to be exposed to AR. Though the exact source (e.g., cannabis cultivation or application around human dwellings) and location is unknown, the association of AR exposure with the WUI was supported from GPS data from Barred Owls, Northern and California Spotted Owls, and hybrids using the WUI for foraging. The high rate of AR exposure in Barred Owls and hybrids provides mounting evidence of an additional stressor that ARs may pose to Spotted Owls – including the first evidence for California Spotted Owls – and fauna native to western forest ecosystems.

Methods

Study Area

We collected Barred Owls and hybrids from the southern Klamath and Cascade Mountains and from the Sierra Nevada in northern California (Figure 1) on national forest lands, national park lands, and private commercial timberlands primarily owned by Sierra Pacific Industries. There was considerable variation in climate, elevation, topography, and vegetation, though both sub-regions were predominantly composed of mixed coniferous forest, dominated by ponderosa pine (Pinus ponderosa), sugar pine (P. lambertiana), incense cedar (Calocedrus decurrens), Douglas fir (Pseudotsuga menziesii), and white fir (Abies concolor). Neither the U.S. Forest Service (S.C. Sawyer, personal communication) nor Sierra Pacific Industries (B.P. Dotters, personal communication) use ARs on lands they manage within our study area. However, there are houses in WUIs adjacent to lands where owls were collected, and it is not known whether ARs are used in these areas.

Tissue Collection and AR Screening

We lured territorial Barred Owls and hybrids by broadcasting digitally recorded Barred Owl vocalizations and collected them with a 12-gauge shotgun following methods described by Diller et al. (2014). We collected Barred Owls and hybrids under federal and state Scientific Collecting Permits (United States Fish and Wildlife Service permits MB24592D-0, MB53229B-0 and California Department of Fish and Wildlife permits SC-002114, SC-11963).  We froze owls immediately after collecting them and stored the specimens in a –20˚C freezer until we delivered them to the Museum of Vertebrate Zoology (University of California, Berkeley), where we extracted livers and ovaries. We were careful to avoid contamination between the two organs by separating them immediately after they were removed from the abdominal cavity and placing them in separate containers. We thawed all specimens for a similar amount of time to extract tissues, and we left no specimen thawed for over 24 hours. We shipped tissue samples to the California Animal Health and Food Safety Laboratory System (CAHFS; University of California, Davis) where they were screened for eight commonly used ARs: warfarin, diphacinone, chlorophacinone, coumachlor, brodifacoum, bromadiolone, difethialone, and difenacoum. The first four belong within less-acutely toxic first-generation ARs (FGAR); the latter four are more acutely toxic second-generation ARs (SGAR) that were created in the 1970s due to rodents developing resistance to first-generation ARs (FGARs; Buckle et al. 1994). High performance liquid chromatography-tandem mass spectrometry was used to screen tissue samples for AR exposure (whether or not any ARs were detected) and to quantify the concentration of ARs detected (Marek and Koskinen 2007). We classified AR exposure in livers and ovaries using the limit of detection (LOD), which allowed us to detect presence of AR in any sample with a concentration above 0.005 μg/g wet weight (ww). We quantified AR concentrations in liver and ovary samples using the limit of quantification (LOQ), which was 0.050 μg/g ww for brodifacoum and 0.020 μg/g ww for all other ARs in owl livers, and 0.200 μg/g ww for all ARs in owl ovaries (Riley et al. 2007). Any sample above these LOQs could have concentrations quantified. These concentrations all fall below the 0.1 μg/g ww threshold for mortality rate of 10% of individuals previously reported in Barred Owls (Thomas et al. 2011). When samples had concentrations greater than the LOD and below the LOQ, we designated those individuals as having “trace” exposure.

Calculating Biological Variables

We identified owls in the field as Barred Owls or hybrids based on both plumage and territorial vocalizations. Individuals with vertical barring on the breast feathers and horizontal barring around the nape that produced distinct two-phrase, eight-note calls (Odom and Mennill 2010) were identified as pure Barred Owls. Individuals with bars and spots on their breast feathers and that produced territorial calls that were not distinctly Spotted Owl or Barred Owl calls were identified as hybrids (Hamer et al. 1994). We classified age as either adult (≥3 years), sub-adult (1–2 years), or juvenile (0 years), based on adults having wider terminal bands than sub-adults on all flight feathers, and juveniles lacking most or all body contour feathers (Mazur and James 2000, J.D. Wiens, personal communication). We determined sex by examining gonads in the lab, and we assessed body condition by characterizing the amount of subcutaneous fat content into four categorical values, with no fat being our baseline (“0”), slight fat (“1”), moderate fat (“2”), and heavy fat (“3”). Because fat reserves in owls change throughout the year (Massemin et al. 1997, DeLong 2006), we obtained a corrected fat index by calculating the residuals of a linear regression of fat against the month of the year (Supplemental Material Figure S1).

Calculating Environmental Variables

We assigned owls that were collected north of the Pit River to the Klamath/Cascade sub-region, and owls sampled south of this river to the Sierra Nevada sub-region (Figure 1). We used this designation to differentiate Barred Owls collected within the range of the Northern Spotted Owl (Klamath/Cascade) or of the California Spotted Owl (Sierra Nevada; Barrowclough et al. 2005). We calculated remaining environmental variables within 2000ha circular buffers around collection locations that approximated Barred Owl home range size in the region that we measured using GPS-tagged individuals in a previous study (see Wood et al. 2020a). We used a combination of law enforcement databases (IERC 2019) to calculate the number of known cannabis cultivation sites detected from 2004 to 2019 within the circular buffers. We also related AR exposure to a measure of the probability of illegal cannabis cultivation within the buffers, estimated from a Maximum Entropy model (Wengert et al. in review) parameterized with variables indicative of the suitability of growing cannabis on California’s public and private lands. The important variables in this predictive model included elevation, slope, precipitation, canopy cover, stand age, and distances to disturbance, fresh water, roads, and private lands, and used a resolution of 90m for individual cells. From the MaxEnt model we obtained an averaged index of cannabis cultivation suitability (ranging from 0 to 1) for each buffer to assess whether owls were more likely to be exposed in areas with more suitable conditions for cannabis cultivation.

Additionally, we calculated the distance of each Barred Owl removal location to the WUI, based on 2010 census data (Radeloff et al. 2005, http://silvis.forest.wisc.edu/data/wui-change/, accessed Aug, 2020), where owls that occurred within the WUI were assigned a distance of 0 km. Both intermix (where housing and vegetation intermingle) and interface (where housing occurs in the vicinity of contiguous wildland vegetation) components of the WUI spatial dataset were used. Four thresholds are defined in the WUI data provided by Radeloff et al. (2005) based on the level of housing density: high, moderate, low, and very low. We chose to use the low density WUI threshold requiring at least 6.17 housing units/km2 because of concordance we observed with this threshold and buildings visible in a building footprint spatial layer developed from Microsoft (https://www.microsoft.com/en-us/maps/building-footprints). Finally, we calculated landownership as the proportion of the circular buffers that were composed of National Forest lands. Descriptive statistics of the environmental variables is listed in Table S1.

Characterizing Barred and Spotted Owl Foraging Activities

To characterize the distribution of Barred Owl foraging locations relative to environmental factors related to AR exposure (in this case WUIs, see below), we GPS-tagged seven Barred Owls and three hybrids between May and August of 2017 and 2018 in the northern Sierra Nevada. We used visual and vocal lures to attract Barred Owls and hybrids and captured them with dho-gaza nets, and applied Argos-enabled GPS backpack tags (Lotek Wireless, Newmarket, Ontario, Canada). We programmed tags to record 4–6 nighttime locations per week between April and August, and then to record one location per week between September and March.

We also used locations from 24 GPS-tagged Northern Spotted Owls and 106 California Spotted Owls to characterize their use of areas associated with elevated AR exposure in Barred Owls – and thus the potential for Northern and California Spotted Owl exposure rates to mirror Barred Owl rates. Northern Spotted Owl locations were collected in the Klamath Mountains between March and August of 2017, and California Spotted Owl locations were collected in the Sierra Nevada between May and August of 2015 through 2020 as part of previous studies (Jones et al. 2016, Atuo et al. 2018, Kramer et al. 2020). We used vocal lures to locate Spotted Owls and captured them either by hand-grab, pan-trap, or snare-poles, and applied GPS backpack tags (Lotek Pinpoint VHF 120, Newmarket, Ontario, Canada). Spotted Owl tags were programmed to record 5 hourly nocturnal locations per night between March and August. From these data, we calculated the mean proportion of locations that occurred within the WUI for both Northern and California Spotted Owls, as well as the proportion of individuals of each subspecies with at least one location in the WUI. We assumed the majority of these locations were primarily foraging locations as owls are nocturnal predators, but we acknowledge that other behaviors such as territory defense, resting, and returns to roosts and nests may be included in these locations.

 Additionally, we calculated the proportion of all known Northern Spotted Owl activity centers and all California Spotted Owl activity centers in the Sierra Nevada whose home ranges at least partially overlapped with the WUI to assess the risk of Spotted Owl exposure to ARs via the possibility of foraging in the WUI. We used 2.1 km radius home ranges for Northern Spotted Owls and 1.6 km radius home ranges for California Spotted Owls (Wiens et al. 2014, Blakey et al. 2019). Activity centers were defined as nest locations or geometric centers of daytime roost locations and were obtained from California Department of Fish and Wildlife (https://www.wildlife.ca.gov/Data/CNDDB/Spotted-Owl-Info, accessed Nov, 2020). We also used both Northern and California Spotted Owl designated ranges (USFWS 2017) to calculate the proportion of WUI within each Spotted Owl subspecies’ range (only including the Sierra Nevada for California Spotted Owls).

Statistical Analysis

We used a set of generalized linear models (McCullagh and Nelder 1989) within an information-theoretic framework (Burnham and Anderson 2002) to test for associations between AR exposure and biological and environmental factors. Because most exposures were at the trace level, we modeled exposure as a binomial response (exposed = 1 and not exposed = 0). Biological factors consisted of species (pure Barred Owl versus hybrid), age, sex, and the index of body condition. Juvenile and un-aged owls were omitted from the generalized linear model because of small sample sizes. Environmental factors consisted of sub-region, proximity to the WUI, number of known cannabis cultivation sites within home ranges, the average index of predictive cultivation for each Barred Owl home range from the MaxEnt model, and landownership.

We used a multi-stage secondary candidate strategy to select top-ranked models (Morin et al. 2020). First, we ran all combinations of biological models and all combinations of environmental models separately. We then identified supported models as those within 5 AICc of the most supported model for each set of models. Second, we combined and evaluated support for variables in the top models from both the biological and environmental sets. In both model-selecting stages, models with uninformative variables (e.g., confidence intervals of variables overlap with zero) were not considered (Leroux 2019). We used package MuMIn (Bartón 2015) in R Studio 1.3.1073 (R Core Development Team 2020) for these analyses.

We also conducted a general Getis Ord-General G high/low cluster analysis (Getis and Ord 1992) to assess the degree to which AR exposure was more clustered than expected at random, less clustered than expected at random, or randomly distributed. We ran separate analyses for owls collected in the Klamath/Cascade sub-region and those collected in the northern Sierra Nevada (where the majority of Sierra Nevada removals were conducted), and only used locations for where owls were exposed, realizing that mates could be non-exposed. To reduce potential biases associated with sampling multiple owls from the same territory, owls collected within 2.52km (the radius of a 2000ha Barred Owl home range in the region, Wood et al. 2020a) of other owls were combined to single points based on the geometric centers of the points. We also conducted a Moran’s I spatial autocorrelation analysis with the same condensed points to assess the degree of concordance between different clustering procedures. All spatial analyses were conducted using ArcMap 10.6.1 (ESRI Inc., Redlands, California).