Population dynamics and resource availability drive seasonal shifts in the consumptive and competitive impacts of introduced house mice (Mus musculus) on an island ecosystem

Study site

As described by Chandler et al. (2016), SEFI is the largest of the Farallon Islands, which is one of several islands that compose the Farallon Islands National Wildlife Refuge. Topographically, it is characterized by a 90 m high hill that rises from the center of the island, and a wide, flat marine terrace that extends outward from the base of the hill (Fig. 1). The terrace is widest from the southeastern to the western portions of the island, and seabirds have excavated numerous burrows within its friable soil. The hill is composed of crumbling granite cliffs that contain fissures, crevices, caves, and rocky scree fields at its base. SEFI has a temperate, maritime climate, with relatively wet winters and dry summers (average annual rainfall = 51 cm) (Kim, Nelson & Seo, 2009). The air temperature is warmest in October (average: 16.1  °C) and coldest in January (average: 11.4  °C; Kim, Nelson & Seo, 2009).

Over 25% of California’s breeding marine birds, with nearly 400,000 individuals of 13 species, are found on the South Farallon Islands (Carter et al., 1992; Johns & Warzybok, 2019; Karl et al., 2001). Most areas on SEFI are occupied continually by breeding seabirds between late March and mid-August with cormorants (Urile penicillatus, Urile pelagicus, Nannopterum auritum) and common murres (Uria aalge) inhabiting rocky slopes and cliffs, storm-petrels (Hydrobates homochroa, Hydrobates leucorhous), auklets (Ptychoramphus aleuticus, Cerorhinca monocerata), pigeon guillemots (Cepphus columba), and tufted puffins (Fratercula cirrhata) nesting in rock crevices and burrows, black oystercatchers (Haematopus bachmani) nesting along the rocky shoreline, and western gulls (Larus occidentalis) nesting across the island and most common on the flatter or more gently sloped areas (Ainley & Boekelheide, 1990; Johns & Warzybok, 2019; Karl et al., 2001). Five species of pinniped visit and/or breed on SEFI including the northern elephant seal (Mirounga angustirostris), harbor seal (Phoca vitulina), Steller sea lion (Eumetopias jubatus), California sea lion (Zalophus californianus), and the northern fur seal (Callorhinus ursinus) (Karl et al., 2001; USFWS, 2009). While hoary bats (Aeorestes cinereus), Mexican free-tail bats (Tadarida brasiliensis), and Western red bat (Lasirurs blossevillii), have been recorded visiting the island, house mice are the only breeding terrestrial mammals present on SEFI (Karl et al., 2001; Tenaza, 1966; Point Blue Conservation Science, unpublished data, 2022). The islands flora includes at least 44 species, 26 of which are non-native (Coulter & Irwin, 2005). Honda et al. (2017) reported 11 orders of terrestrial arthropods representing 60 families, 107 genera and 112 species identified on SEFI. Two endemic taxa are found on SEFI: the Farallon camel cricket (Rentz, 1972) and the Farallon arboreal salamander (Van Denburgh, 1905). Domestic cats (Felis catus) and European rabbits (Oryctolagus cuniculus) were intentionally introduced to SEFI in the late 1800s and successfully removed from the island in the early 1970s (Ainley & Lewis, 1974).

Ethics statement

Sampling was approved by the United States Fish and Wildlife Service under a cooperative agreement with Point Blue Conservation Science (no. 81640AJ008) and the California Department of Fish and Game scientific permit (no. SC–008556). All vertebrate sampling protocols were approved by and adhered to statutes of the Institutional Animal Care and Use Committee of the Woods Hole Oceanographic Institution (no. 4855-001).

Mouse abundance and resource availability

We compiled data over a 17-year period (December 2000 to January 2018; hereafter 2001–2017) to quantify seasonal trends in mouse abundance and resource availability on SEFI. This included an index of mouse abundance based on monthly trapping, precipitation data as a proxy for vegetation phenology, and measures of arthropod, seabird, and salamander abundances. For all metrics, monthly values were compiled and averaged to obtain seasonal values for the spring (March, April, May), summer (June, July, August), fall (September, October, November), and winter (December, January, February) for each season and year that data were available (Table S1).

Our index of mice abundance was based on monthly trapping success on 4, ∼300 m transect lines (L, M, N, and S; Fig. 1) spread across available habitats at SEFI (Irwin, 2004; Chandler et al., 2016; Nur et al., 2019). Trapping was conducted for three consecutive nights each month between March 2001 and March 2004, and again from December 2010 to March 2012, and finally September 2016 through January 2018. All sampling periods used the same transects, each with seven traps per transect. For the 2010–2012 and 2016–2018 effort, five additional traps were added to transect L; these incorporated more of the vertical aspect of the island topography (Fig. 1). Trapping efforts used D-Con^® Ultra Set^® covered snap traps baited with peanut butter and oats. Trapping success was determined as the proportion of trap-nights set per monthly session (either 84 (2001–2004) or 99 (2010–2012 and 2016–2018)) in which house mice were captured.

Precipitation data (cm of rain) was collected on SEFI using a standard US National Weather Service rain gauge (∼1 m elevation) at the same location at noon Pacific Standard Time every day throughout 2001–2017 (Fig. 1). Precipitation data was used as a proxy for both terrestrial environmental conditions and as a leading proxy for vegetation phenology as the majority of vegetation on SEFI senesces or dies during the summer and recovers in the late winter and spring when seasonal rainfall begins (Coulter & Irwin, 2005).

Arthropod density (indiv./m²) on SEFI was evaluated during four collecting trips during 2014. Each trip lasted roughly 12 continuous days and fell within a season (winter, spring, summer, fall). Arthropod densities were enumerated by systematically surveying standardized (0.3 m²), untreated wooden cover boards placed throughout the island in a previous study to create salamander refuges (Lee et al., 2012). Complete methods of arthropod collection, species identifications, and density estimates are detailed in Honda et al. (2017).

We used daily year-round assessments of adult seabird carcasses found during routine island operations as proxy of seabird resource availability to house mice on SEFI between 2001 and 2017. As operations and research protocols have remained standard throughout the time series, effort is relatively consistent throughout the period. Adult birds had their primary tips clipped to prevent double counting. For the purpose of this analysis, we used carcass count numbers from one common burrow nesting species whose breeding habitat overlapped mouse sampling areas, the Cassin’s auklet (Ptychoramphus aleuticus), as a leading indicator of overall seabird resource abundance (Fig. 1). Cassin’s auklets on SEFI initiate breeding in early to middle of April on average (Ainley & Boekelheide, 1990). Other seabirds also breed in the areas sampled for mice on SEFI, especially western gulls which are ubiquitous and breed throughout the island (Ainley & Boekelheide, 1990). These other seabirds may also act as food resources to house mice, though they initiate breeding slightly later than Cassin’s auklets on average, in early to middle of May (Ainley & Boekelheide, 1990).

We quantified seasonal variability in the abundance (indiv./month) of Farallon arboreal salamanders on SEFI using data from a long-term monitoring program of this species (Lee et al., 2012). A proxy for salamander abundance was estimated by systematically surveying standardized (0.3 m²), wooden cover boards placed throughout the island between 2008 and 2017 (Fig. 1). One hundred cover boards along a path from the north to the south side of the island were consistently monitored in all years and additional 148 cover boards spread around the island were also monitored from 2013 to 2017 (Fig. 1). Complete methods for salamander surveys are detailed in Lee et al. (2012).

Prior to statistical analyses average monthly values were used to calculated seasonal averages for salamander abundance, mouse trapping success, and arthropod density. Total monthly values were compiled to calculate seasonal averages for rainfall and bird carcasses. We then used the Kruskal-Wallis Test with non-parametric post-hoc comparisons to test for differences among seasons in mouse trapping success and the availability of food resources.

Tissue sampling

To complement our analysis of resource availability between 2001 and 2017, we collected tissue samples form mice and potential prey items in 2013 to quantify seasonal shifts in resource use by house mice and the potential for resource overlap with arboreal salamanders. Specifically, we sampled muscle and liver tissues from house mice (n = 63 individuals) during three different seasons in the spring (March), summer (August), and fall (October) in 2013 that reflect the natural oscillation of mouse abundance on SEFI (Irwin, 2006). Mouse liver tissue stable isotope values reflects short term diet on the scale of days to a week with an average half-life of 3.5 to 5.6 days, while muscle tissues stable isotope values reflect diets on the scale of weeks to months with an average half-life of 29.6 to 30.1 days (DeMots et al., 2010). Mice were not collected during the winter months due to logistical constraints. We also collected samples of possible food resources for mice based on those that were abundant on SEFI, have been identified in a prior analysis of mouse stomach contents (Jones & Golightly, 2006), and/or represented likely isotopically unique prey groups (Phillips et al., 2014). Specifically, during the spring (April) and fall (October) of 2013 we collected representative samples of prey items, vegetative tissue from common native (Lasthenia maritima and Spergularia sp) and non-native (Malva spp. and Plantago coronopus) plants and four different taxa of whole-bodied terrestrial arthropods (Coleoptera larvae, Farallonophilus cavernicolus, Oniscidea sp., Araneae spp.; Table S2). Five to six plant and arthropod samples were collected per taxa in both seasons. As prior studies suggest that invasive rodents can consume resources from the intertidal zone (Navarrete & Castilla, 1993; Newton et al., 2016), we sampled the muscle tissues from an intertidal snail (Nucella emarginata) during the spring (April; n = 6) and fall (October; n = 5) of 2013. We also collected seabird tissues during the summer (August; n = 16) once these resources became more readily available on the island (Table S2). This included Cassin’s auklet muscle, Cassin’s auklet and western gull egg membrane, and western gull guano which reflect seabird tissues that are highly abundant on SEFI and/or have been previously recovered from mouse stomachs (Jones & Golightly, 2006). Finally, we sampled tail clips from arboreal salamanders found under cover boards in the spring (April; n = 16) and fall (October; n = 16) when they are most active on the island. This allowed us to assess the likelihood of house mice consuming arboreal salamanders on SEFI. All samples were stored frozen (−20 °C) prior to processing for stable isotope analysis.

Stable isotope analysis

All samples were freeze-dried and then homogenized using a mortar and pestle. Dried homogenized muscle, liver and tail clip samples underwent lipid extraction using 2:1 chloroform:methanol. Samples were placed in a glass vial with a solvent volume 10 times greater than sample volume and sonicated in a water bath for 15 min and then decanted. This procedure was repeated for a total of three cycles and the samples were rinsed in DI water and oven dried at 60 °C for 24 h to remove any remaining solvent. We flash-combusted (PDZ Europa ANCA-GSL elemental analyzer) approximately 1.0 mg of each animal tissue sample and 3.0 mg of each plant tissue sample loaded into tin cups to analyze for carbon and nitrogen elemental composition (C:N ratio) and carbon and nitrogen isotopes (δ¹³C and δ¹⁵N) through interfaced PDZ Europa 20-20 continuous-flow stable isotope ratio mass spectrometers (CFIRMS). Raw δ values were normalized using glutamic acid (G-17; G-9/USGS-41), bovine liver (G-13), peach leaves (G-7) and nylon 5 (G-18) as standard reference materials. Sample precision based on repeated standard reference materials was 0.1‰for both δ¹³C and δ¹⁵N. Stable isotope ratios are expressed in δ notation in per mil units (‰), according to the following equation: ${\displaystyle \delta X=\left[\left(R_{\mathrm{sample}}/R_{\mathrm{standard}}\right)-1\right]\cdot 1000.}$ where X is ¹³C or ¹⁵N and R is the corresponding ratio ¹³C /¹²C or ¹⁵N /¹⁴N. The R_standard values are based on Vienna PeeDee Belemnite (VPDB) for δ¹³C and atmospheric N₂ for δ¹⁵N.

Isotopic niche analysis

We assessed variation in isotopic niche (Newsome et al., 2007) position, width, and overlap across seasons (i.e., spring, summer, and fall) and tissue (muscle vs. liver) in house mice and between species (i.e., house mice and arboreal salamander) within each season using both multivariate and univariate techniques. We compared isotopic niche positions by computing the Euclidean distance (ED) between group centroids (δ¹³C and δ¹⁵N bivariate means) following the methods of Turner, Collyer & Krabbenhoft (2010). Isotopic niche positions were considered to be different if the ED between species examined was significantly greater than zero after comparison to null distributions generated by a residual permutation procedure. If niche positions differed, we then examined the results of univariate general linear models (GLM) and Tukey-Kramer multiple comparison tests to determine which axis (δ¹³C and/or δ¹⁵N) contributed to niche differences across seasons or between species (Hammerschlag-Peyer et al., 2011). To examine individual consistency in the isotopic niche of house mice we tested for relationships between individual’s liver (i.e., shorter-term dietary signal) and muscle (i.e., longer-term diet signal) stable isotope values using Pearson correlations.

In addition, we explored variation in niche area and overlaps using standard ellipse areas which can be interpreted as a measure of the core isotopic niche of a population (Jackson et al., 2011). We calculated the Bayesian estimation of standard ellipse area (SEA_b) for each group to compare two-dimensional niche areas of house mice and arboreal salamanders among species and seasons. We then used the resulting Bayesian posterior probability distributions of SEA_b estimates in pairwise-tests to identify significant differences in SEA_b at the p < 0.05 level. Specifically, we calculated the probability of whether posterior SEA_b values from one group are different than the posterior SEA_b value from the comparison group (Jackson et al., 2011). Lastly, we compared isotopic niche overlap between groups calculated as the proportion of standard ellipse area corrected for sample size (Jackson et al., 2011) for each group that overlap with a comparison group’s standard ellipse area (SEA_c overlap). Among other things, this isotopic niche overlap approach provided us with a quantitative estimate of the potential for competitive overlap between mice and salamanders.

Dietary mixing model analyses

We used univariate (δ¹³C or δ¹⁵N) GLM and Tukey-Kramer multiple comparison tests to examine isotopic differences across major prey resources (seabirds, intertidal snails, plants, and arthropods) collected in each season. As not all prey resources (i.e., seabirds) were collected in every season we grouped prey resources by taxa and season into a single factor when conducting GLM analyses. We then used the SIMMR Bayesian mixing model (Parnell & Inger, 2016; Parnell et al., 2010) in the R environment (Ver. 3.6.2) to quantify the relative use of these four major prey resources by house mice in each season. This model estimates the probability distributions of multiple source contributions to a mixture while accounting for the observed variability in source and mixture isotopic signatures, elemental concentration, and dietary isotopic fractionation. We focused our mixing model analyses on liver tissues given the strong correlation in isotopic values found between mouse tissues (see Results below) and the more constrained and shorter isotopic turnover time in liver (DeMots et al., 2010; MacAvoy, Macko & Arneson, 2005). Separate models were run in each season (spring, summer, summer). As the discriminatory power of Bayesian mixing models can decline markedly above six or seven prey sources (Phillips et al., 2014), we a priori defined five statistically, ecologically, and taxonomically relevant prey resource groups for incorporation into our mixing model analysis. As little to no differences in prey resource stable isotopes values were observed across seasons (see Results below), we used the same prey resource δ¹³C and δ¹⁵N values averaged across all seasons in each model analyses. As there is no existing evidence to support or refute the possibility that house mice may also consume arboreal salamanders on SEFI we also explored a separate subset of mixing models that include arboreal salamanders as a prey resource for house mice. We used diet to consumer isotopic discrimination factors for liver tissue (δ¹⁵N: +4.3 ± 0.2; δ¹³C: +0.7 ± 0.3) derived from captive studies of house mice on a controlled diet of wheat and corn (Arneson & MacAvoy, 2005). We incorporated elemental concentration dependence (Phillips & Koch, 2002) into the model, and ran 1 million iterations, thinned by 15, with an initial discard of the first 40,000 resulting in 64,000 posterior draws. Following Bicknell et al. (2020), model convergence and fit were checked using Gelman -Rubi diagnostic values (i.e., Gelman–Rubin statistics <1.1) and by plotting the posterior predictive distributions (Gelman et al., 2004).

Seasonal trends in mouse abundance and resource availability

Mouse trapping success varied by season (H₃ =19.1058, p < 0.001) with lower trapping success in the spring relative to the late summer and fall (Fig. 2). These long-term trends broadly concurred with the trapping success observed in 2013, when mouse tissue and food resource samples were collected for stable isotope analyses (Table S1, Fig. 2). Precipitation also differed across seasons (H₃ =43.8296, p < 0.001) being lowest in the summer, intermediate in the spring and fall, and highest in the winter (Fig. 2). Arthropod density varied significantly across seasons (H₃ =20.2933, p < 0.001) with lower values in the summer and higher values in the spring and winter (Fig. 2). Seabird carcass abundance also varied significantly across seasons (H₃ =47.3676, p < 0.001) with higher values in spring and summer relative to fall and winter (Fig. 2). However, seabird carcasses abundance in the spring and summer of 2013 was higher than the long time average. Finally, salamander abundances differed across seasons (H₃ =30.3015, p < 0.001) with higher average counts in spring and winter relative to summer (Fig. 2).

Isotopic niche of house mice and arboreal salamanders

The isotopic niche position of house mice differed significantly in all pairwise comparisons across seasons for both liver (ED = 4.01−5.37‰, p < 0.001) and muscle (ED = 2.00−5.02‰, p < 0.030) tissues. House mice stable isotope values differed among seasons (δ¹³C: F_2,126 = 66.61, p < 0.001; δ¹⁵N: F_2,126 = 26.60, p < 0.001) but not by tissue type (δ¹³C: F_1,126 = 66.61, p = 0.878; δ¹⁵N: F_1,126 = 0.92, p = 0.339) or the interactions between these two factors ( δ¹³C: F_1,126 = 2.32, p = 0.681; δ¹⁵N: F_1,126 = 1.22, p = 0.299). Seasonal differences in the isotopic niche position of house mice were due to lower tissue δ¹³C values in the spring relative to the summer and fall as well as lower tissue δ¹⁵N values in the summer relative to the spring (liver and muscle) and fall (liver only; Table 1). Muscle and liver tissue stable isotope values were positively correlated with one another for both δ¹³C (r = 0.908, p < 0.001) and δ¹⁵N (r = 0.946, p < 0.001) values.

The isotopic niche area of house mice differed significantly across seasons due to smaller SEA_b values in spring relative to summer and fall for liver tissues and higher SEA_bvalues in summer relative to spring and fall for muscle tissues (Table 2). While the isotopic niche of house mice did not overlap among seasons when examined using liver tissues, niche overlap based on muscle tissue was higher between spring and summer (7.6–18.9%) and spring and fall (11.0–12.4%) relative to fall and summer (0.0−0.1%; Fig. 3).

The isotopic niche position of house mice and arboreal salamanders differed significantly in all within-season, pairwise comparisons of arboreal salamander tail clips and mouse liver (spring: ED = 3.19‰, p < 0.001; fall: ED = 2.02‰, p = 0.004) and muscle (spring: ED = 3.05‰, p < 0.001; fall: ED = 1.71‰, p = 0.011) tissues. This was due to higher δ¹³C values in arboreal salamanders’ tail clips relative to mouse liver and muscle tissues in the spring (F_2,118 = 63.61, p < 0.001) and muscle tissues only in the fall (F_2,118 = 6.03, p = 0.004; Table 1). In contrast, arboreal salamanders tail clip δ¹⁵N values did not differ from house mouse liver and muscle tissues in the spring (F_2,118 = 2.23, p = 0.101) or the fall (F_2,118 = 3.40, p = 0.057; Table 1).

The isotopic niche area of arboreal salamanders did not differ significantly between seasons, nor did they differ from estimates of house mice niche area calculated from liver or muscle tissues within each season (Table 2). Isotopic niche overlap between spring and fall for arboreal salamanders ranged from 16.0–24.9%. The isotopic niche of arboreal salamanders in the spring did not overlap with those of house mice estimated using either liver or muscle tissues (Fig. 3). In contrast, the isotopic niche of arboreal salamanders in the fall showed significant overlap with the isotopic niche area (SEA_c) of house mice estimated using liver tissue (16.3%) or muscle tissues (52.3%; Fig. 3).

Dietary mixing model analyses

Differences in prey resource stable isotope values between major taxonomic groupings were more common than differences within major taxonomic groupings collected in different seasons for both δ¹³C (F_4,152 = 132.60, p < 0.001) and δ¹⁵N (F_4,152 = 38.25, p < 0.001) values. For example, plant and intertidal resource δ¹³C and δ¹⁵N values, and arthropod and arboreal salamander δ¹⁵N values did not differ between seasons (Table S2). Arthropod and arboreal salamander δ¹³C values differed slightly between seasons, though these differences were small (1.5−1.9‰) relative to the differences observed between major taxonomic groupings (2.2–16.5‰; Table 1, Table S2). Because of these findings, and because not all prey resources were collected in each season due to logistical and financial constraints even though they were available to mice, prey resource δ¹³C and δ¹⁵N values were averaged across all seasons by major taxonomic groupings for subsequent analyses (Table S3). These averaged prey resource δ¹³C ( F_4,152 = 237.27, p < 0.001) and δ¹⁵N (F_4,152 = 75.35, p < 0.001) values differed significantly among major taxonomic groups thus providing isotopically unique endmembers for incorporation into the isotopic mixing model (Fig. 4).

When examined without arboreal salamanders as a possible prey resource, stable isotope mixing model analysis indicate seasonal shifts in the diet composition of house mice on SEFI. While there is overlap in 95% credibility intervals between some groups, plants dominated the diet in the spring, followed by a smaller proportion of arthropods, and very little intertidal and seabird resources (Table 3). The importance of plants decreased in the summer, while the relative importance of seabirds during this time increased to the highest observed across all seasons (Table 3). Fall diets were characterized by a relatively increased importance of arthropod and to a lesser extent intertidal resources, a continued decline in the importance of plants, and a decrease in the importance of seabirds. When prey resources were aggregated a posteriori, there were clear differences in the relative use of marine (i.e., seabird and intertidal) vs. terrestrial (i.e., plant and arthropod) resources by mice across seasons. While terrestrial resources were more important to mice in all three seasons, the relative importance of marine resources were higher in the summer and fall, relative to the spring (Table 3).

These same seasonal trends are also apparent when including salamanders as a possible prey resource. Plants are most important in the spring and decline in contribution to the diet throughout the summer and fall (Table 3). Seabird resources are consumed relatively more in the summer and arthropod resources are relatively more important in the fall. Arboreal salamanders are predicted to contribute between 10.1 and 21.6% of house mice diet (Table 3). Mixing models that include arboreal salamanders as a possible prey resource tend to predict lower arthropod contribution to house mice diets than those that do not include salamanders likely due to the fact that these two groups have the most similar stable isotope values of the possible prey resources examined in our study (Table 3, Table S3). When prey resources were aggregated a posteriori into marine vs. terrestrial resources, the results of models that included salamanders did not differ from those that did not (Table 3).