Abstract
Non-native plants can affect communities through direct competition, and by providing refuge to seed predators, creating apparent competition with native plants. Ammophila arenaria (European beachgrass) has been introduced to coastal dune habitats throughout the western United States, where it forms dense monocultures, stabilizes dunes, and alters abiotic and biotic conditions. The dominance of European beachgrass is linked to declines of Lupinus tidestromii (Tidestrom’s lupine), an herb endemic to coastal dune communities in central and northern California. Peromyscus sonoriensis (western deer mice), a native seed predator, use beachgrass as refuge from predators. Tidestrom’s lupine near European beachgrass stands experience greater predation pressure from deer mice. At Point Reyes National Seashore, California, USA (PRNS), mechanical removal, manual pulling, and herbicide treatment have been used to reduce the density of European beachgrass near Tidestrom’s lupine populations. We trapped deer mice at five sites in PRNS that experienced different management regimes and used spatially-explicit capture-recapture models to estimate deer mouse density as a function of site and habitat treatment. We found that deer mouse density was lowest in areas where European beachgrass was mechanically removed and in herbicide-treated foredunes, and highest in areas highly invaded by European beachgrass and Carpobrotus spp. (iceplant). Deer mouse density increased from 2021 to 2022 at every site except one that underwent extensive mechanical removal of European beachgrass from 2010–2011. This study shows enduring effects of European beachgrass removal on the density of a native seed predator and highlights the importance of habitat management for conservation of Tidestrom’s lupine.
- Ammophila arenaria
- dune restoration
- Lupinus tidestromii
- seed predation
- spatially-explicit capture-recapture
Restoration Recap
European beachgrass has heavily altered native dune communities in coastal California and threatens the endangered native Tidestrom’s lupine by providing refuge for deer mice, a seed predator that depredates lupine racemes.
We trapped deer mice at five sites within Point Reyes National Seashore that comprised a variety of habitats and underwent different habitat treatments designed to remove European beachgrass. Treatments included herbicide application and mechanical removal and burial; control areas were left untreated. We used spatially explicit capture-recapture (SECR) models to estimate mouse density as a function of habitat type.
Deer mouse density was highest in areas highly invaded by European beachgrass and iceplant, and lowest in areas where European beachgrass was mechanically removed and in foredunes where European beachgrass was treated with herbicide.
Treatments that reduce the cover of European beachgrass and promote open habitats lead to lower density of native seed predators, which likely benefits the endangered Tidestrom’s lupine.
The introduction of non-native plants has had dramatic effects on biodiversity worldwide (Pyšek et al. 2012). Although non-native plants might directly compete with native species for space and resources, indirect effects can also be important drivers of decline. Apparent competition occurs when one prey species reduces the abundance of another prey species by supporting greater abundance of a shared predator (Holt 1977). A related mechanism in plants is refuge-mediated apparent competition, which occurs when one plant provides refuge for a predator, thereby allowing that predator to consume more of the vulnerable co-occurring plant (Orrock et al. 2010). Invasive plants can limit the recruitment of native species through refuge-mediated apparent competition and the facilitation of seed predation by small mammals (Orrock et al. 2008). The success of native plant community restoration could depend on accounting for indirect interactions between native and non-native plants mediated by a predator.
Coastal dune ecosystems in California have been disturbed by development, recreation, and the introduction of non-native plants, resulting in declines of native plant species and altered communities (Pickart and Barbour 2007). Non-native plants can alter natural physical processes of sand movement by stabilizing dunes (Wiedemann and Pickart 2004) and can facilitate high density of native seed predators by providing refuge (Orrock et al. 2008, Dangremond et al. 2010). Non-native plants including Ammophila arenaria (European beachgrass) and Carpobrotus spp. (iceplant) have been introduced to coastal ecosystems in western North America to stabilize sand dunes and bluffs, with negative effects on native plant communities (Wiedemann and Pickart 1996, Vilà and D’Antonio 1998). European beachgrass in particular establishes dense monocultures and alters soil chemistry (Parsons and Becker 2021) and the soil microbial community (Parsons et al. 2020), reduces sand movement (Wiedemann and Pickart 1996), and provides habitat for rodent seed predators (Pitts and Barbour 1979). Restoration of native dune ecosystems and conservation of native plants will require management to reduce the effects of European beachgrass.
Habitat alteration and increased seed predation are two threats to the endangered Lupinus tidestromii (Tidestrom’s lupine), a native plant occurring in early- to midsuccessional dune habitats along the north coast of California (U.S. Fish and Wildlife Service 1998, Pardini et al. 2015). Non-native plants that form dense monocultures reduce the availability of bare ground needed for colonization by Tidestrom’s lupine. Furthermore, in some populations, including those at Point Reyes National Seashore (PRNS), Tidestrom’s lupine experiences high seed predation rates (82%) from Peromyscus sonoriensis (western deer mice), particularly for plants located close to patches of invasive European beachgrass, greatly reducing the potential for successful reproduction and recruitment (Dangremond et al. 2010, Pardini et al. 2017). Based on population viability analyses, most Tidestrom’s lupine populations studied at PRNS are likely to be extirpated under observed levels of seed predation by deer mice (Dangremond et al. 2010).
The low viability of Tidestrom’s lupine populations was one impetus for PRNS to initiate an ambitious coastal dune restoration program in 2001. As of 2018, approximately 110 hectares of invasive European beachgrass (4.7% of 2,383 ha of coastal dune and scrub habitat at PRNS) and 11 hectares of iceplant (0.5%) had been removed from invaded nearshore foredunes (young dune features immediately adjacent to the beach; Parsons and Becker 2021) and inland backdunes (older dunes located at a greater distance from the beach, characterized by late-successional plant communities; Parsons and Becker 2021) at PRNS (Parsons et al. 2023). Habitat restoration through mechanical removal and burial of European beachgrass near Abbotts Lagoon at PRNS resulted in lower seed predation on Tidestrom’s lupine compared to sites treated with herbicide or manual removal (hand pulling) (Pardini et al. 2018). An early capture-mark-recapture study at PRNS established that the density of deer mice was greater in stands of European beachgrass and densely rooted native species such as Juncus spp. (rush) compared to native dune vegetation (Pitts and Barbour 1979). Because mechanical removal was expensive and had extensive indirect impacts from excessive sand remobilization on sensitive habitats such as wetlands, PRNS shifted its removal approach from mechanical to herbicide treatment (Parsons et al. 2023). Unlike mechanical removal, herbicide treatment of European beachgrass has, at PRNS and in other locations in California, resulted in extremely delayed decomposition of standing dead European beachgrass biomass and litter (Parsons et al. 2023). Dead European beachgrass may therefore continue to offer refuge for deer mice long after herbicide treatment. In addition, mice might simply move to other non-treated habitats nearby such as moist meadows or dune scrub such that European beachgrass removal alone would not suffice to reduce mouse predation on Tidestrom’s lupine and improve viability of imperiled lupine populations. To date, no study has investigated how habitat and vegetation management affect the density of deer mice in coastal sand dune vegetation at PRNS.
We conducted a capture-mark-recapture study of deer mice at five sites with extant Tidestrom’s lupine populations in PRNS that contain a variety of habitat types and have undergone different management regimes. We used spatially-explicit capture-recapture (SECR) models to estimate mouse density at each site for two years. We sought to quantify how mouse density varied among sites, habitats, and years. We hypothesized that deer mouse density would be lower in areas where European beachgrass was mechanically removed than areas that were untreated or only treated with herbicide. Our study encompassed highly invaded areas dominated by European beachgrass and iceplant, sparsely invaded native dune mat habitats where Tidestrom’s lupine occurs, as well as uninvaded habitats such as moist meadows and dune scrub that could support deer mouse populations. Our results have implications for invasive plant management projects that are focused on restoring native vegetation communities and reducing indirect effects of invasive plant establishment on rare plants and wildlife at PRNS and other coastal dune habitats.
Methods
Study Sites
The study took place in 2021 and 2022 in coastal dune habitat at Point Reyes National Seashore, Marin County, California, USA (from approximately 38°06′55.1″ N, 122°57′32.5″ W, to 38°03′04.4″ N, 122°59′13.5″ W; Figure 1) a region characterized by a Mediterranean climate. Rainfall at PRNS near our study sites was much lower in 2021 (24.4 cm) than 2022 (48.5 cm), a more typical year (Western Regional Climate Center 2023). Our study incorporated a range of invasion status, restoration approaches, and habitat conditions at five sites: Abbotts Lagoon (Abbotts), AT&T, B Ranch North, North Beach, and Population 9. Dune restoration has been performed at Abbotts, large portions of AT&T, and B Ranch North, but not at North Beach or Population 9. Both North Beach and Population 9 support dense stands of European beachgrass or iceplant or intermixed stands. Herbicide treatment of dense European beachgrass has been performed at Abbotts, AT&T, and B Ranch North. Abbotts, B Ranch North, and AT&T support sparsely invaded native dune mat (non-native cover < 15%) in which sporadic non-natives have either been manually removed or spot-sprayed. Mechanical removal was employed at Abbotts and was later followed by herbicide treatment in backdune and foredune areas south of the mechanically restored foredunes and backdunes. Dunes at PRNS encompass early- and mid-successional communities such as sparsely vegetated dune mat and foredunes, and later successional communities such as dune scrub and backdunes (Parsons and Becker 2021). Based on field surveys and interpreting satellite imagery (see Data Collection section below), we classified habitats into ten classes: moist meadow, dune scrub, iceplant-dominated, mechanically treated backdunes (backdunes formerly dominated by European beachgrass that was mechanically excavated and removed), herbicide-treated foredune, herbicide-treated mid-dune (intermediate distance to the beach, transitional between early successional foredunes and late successional backdunes), herbicide-treated backdune, moderately invaded dune mat, sparsely invaded dune mat, and highly invaded European beachgrass (Supplementary Material Table S1).
Moist meadows are depressional features that support dense cover of mesic genera such as rushes (Juncus spp.) or sedges (Carex spp.) along with more xeric species such as Baccharis pilularis (coyote brush). Dune mat habitat consists of sparsely-vegetated areas in foredunes and low-elevation inland areas populated by low-growing forbs such as Abronia latifolia (sand verbena) and Calystegia soldanella (seashore false bindweed) and grasses such as Leymus mollis (American dunegrass) that tolerate sandy soil (Wiedemann and Pickart 2004, Parsons et al. 2023). Dune scrub, in contrast, is characterized by denser shrub cover, including Ericameria ericoides (California goldenbush), Lupinus chamissonis (chamisso bush lupine), and coyote brush in backdune and higher elevation inland areas (Parsons et al. 2023). All sparsely invaded native dune mat included populations of Tidestrom’s lupine. The highest abundance of Tidestrom’s lupine is found at Abbotts and lowest at North Beach; Tidestrom’s lupine has declined in recent years at B Ranch North (National Park Service, unpublished data).
Data Collection
We trapped deer mice at five sites at PRNS, California, USA, in 2021 and 2022 (Figure 1). At each trapping array, Sherman traps (model LFAHD, H. B. Sherman Traps, Tallahassee, FL, USA) were arranged in a square 7 × 7-point grid spaced 10 m apart, covering 60 m × 60 m (0.36 ha), with two traps at each point for a total of 98 traps per array. In 2021, we sampled each site using four arrays each. Most arrays were active for four nights in 2021, with one array at AT&T sampled for just three nights. In 2022, we sampled two arrays of 98 traps each at Abbotts, North Beach, and Population 9, and four arrays of 98 traps each at AT&T and B Ranch North. All trapping arrays were active for four nights in 2022 except for two arrays at B Ranch North, which were each active for 3 nights. The total sampling effort was greater in 2021 than 2022 (Table 1). Sites were sampled from 4–12 nights per year. Trapping and processing of deer mice followed National Ecological Observatory Network small mammal sampling protocol NEON.DOC.000481 (Meier et al. 2023) with minor modifications. We baited Sherman traps with rolled oats, and cotton balls were placed in traps for nesting. We timed trap setting so that the last trap was opened near sunset, and we began checking traps the following morning near sunrise. We processed captures at the trap to minimize stress and handling time. We marked deer mice with a unique numeric ear tag (style 1005-1L1, National Band and Tag Company, Newport, KY, USA). We assessed mouse age based on pelage coloration, determined sex and reproductive status by visible genitalia, and weighed each mouse to the nearest tenth of a gram upon first capture. During subsequent recaptures, we recorded the mouse’s ear tag and trap number and released it immediately. Non-target species other than deer mice were identified to species, if possible, then sexed, aged, and released without tagging. All mouse handling was performed under Scientific Collecting Permit number SC-10779 issued by the California Department of Fish and Wildlife. Our methods for handling mice were approved by the National Park Service Institutional Animal Care and Use Committee (Protocol PWR_PORE_Halstead_Peromyscus_2019, as amended).
Because each trapping grid often incorporated multiple types of habitats, vegetation was monitored by creating a systematic grid of linear transects within each trapping grid in ArcGIS version 10.8.1 (Environmental Systems Research Institute 2020). Transects were oriented parallel to the ocean/shoreline, with approximately six transects ranging from 23 m to 72 m in length per trapping grid with transects located 10 m apart. Along each transect, vegetation data was collected using the point intercept method every 3 m along the linear transect using a GPS for orientation. At each monitoring point, a pin flag was dropped, and every plant species intercepted by the pin flag was recorded: if a species was hit multiple times by the pin flag, it was only recorded once. We estimated percent cover by dividing the number of point intercept locations at which a particular species or habitat category was found by the total number of point intercept locations on that transect. The data were then summarized for the entire grid, as well as for the subset of habitats falling within that grid (Table 2). Species were classified according to nativity status (i.e., native, non-native) and life form (i.e., shrub, rush/sedge, grass, herb). Cover of live and dead European beachgrass and iceplant was calculated, along with detritus (i.e., litter) and bare ground. We then created a GIS layer (a polygon shapefile) on the distribution of coastal sand dune habitat at PRNS based on: 1) vegetation transects within our trapping grids, 2) field surveys of dominant vegetation in coastal dunes adjacent to our trapping grids, 3) interpretation of satellite imagery, and 4) knowledge of the habitat management actions performed on dune plant communities. Polygons were defined as belonging to one of the ten habitat classes defined above based on the species composition and cover and habitat treatment.
Spatially-explicit Capture-recapture Modeling
We fit SECR models using the “secr” package version 4.6.0 (Efford 2023) in R version 4.2.3 (R Core Team 2023). We created a habitat mask using the GIS layer of vegetation data with ten habitat classes defined above and each polygon assigned to the nearest sampled site, with a spacing (grain) of 15 m and buffered traps by 75 m. We tested if the spacing and trap buffer were sufficient using the “mask.check” function in secr; convergence of log-likelihood values with finer spacing and larger buffers indicated a trap buffer of 75 m was large enough and a 15 m spacing for the habitat mask had sufficient resolution. We fit separate models for the 2021 and 2022 capture data following a stepwise procedure for each. We used a half-normal detection function, which is defined by two parameters: g0, the capture probability for an individual if the distance between a detector (trap) and its activity center was zero, and σ, the scale of the half-normal detection function, or the rate at which capture probability declines with increasing distance from an individual’s activity center. We first compared alternative models for the detection component of the SECR model, by including covariates on g0 (Table 3). In addition to a null model for g0, we tested for a learned response to trapping after initial capture, because deer mice can exhibit a “traphappy” response whereby capture probability increases after the initial capture, as individuals learn that baited traps are a food source (Zwolak and Foresman 2008). We also tested for an effect of the fraction of the moon that was illuminated (moon fraction) on g0, because rodent activity is often influenced by moon phase, with greater activity when less lunar illumination reaches the ground (Prugh and Golden 2014, Guiden et al. 2023). For σ, we fit a null model because we had no a priori hypotheses for how the scale of the detection function might vary spatially. We initially tested for effects of mouse sex on g0 and σ but did not find support for sex affecting detection; therefore, we pooled mice of both sexes together for analysis. We compared models using AICc (Akaike Information Criterion corrected for small sample size) and chose the model with the lowest AICc for inference about detection (Akaike 1973, Burnham and Anderson 2002).
We used the top-ranked detection model to compare alternative models for deer mouse density, D, using AICc. We fit four density models: 1) a null model for D, 2) a model with a fixed effect of site on D, 3) a model with a fixed effect of habitat type on D, and 4) a model that included fixed additive effects of habitat type and site on D (Table 4). We did not fit an interaction term for habitat type × site because not all habitat types were present in all sites; our model estimates a fixed site effect and fixed effects of habitat type on D, which we then used to estimate D within the habitats that occurred at each individual site. To estimate covariate effects on D, the “secr” package requires covariates to be defined for each point in the habitat mask. We defined habitat and site covariates on the habitat mask using the “addCovariates” function of the “secr” package, which extracted the site and habitat type for each point of the habitat mask using the GIS layer with habitat polygons described above. We included site as a fixed effect to account for variation in deer mouse density resulting from differences in vegetation cover and the history of treatment among sites that could not be explained by the habitat type. We fit separate models to estimate D using data from 2021 and 2022. We used the top-ranked density model by AICc to project density surfaces for deer mice within each habitat type at each site in each year, using the “predictDsurface” function in the secr package. Data to reproduce analyses are archived on USGS ScienceBase (Rose et al. 2024a) and R code to reproduce analyses are archived on GitLab (Rose et al. 2024b).
Results
Vegetation Cover
Vegetation cover varied among sampled sites. The highest cover of live European beachgrass was at Population 9 and North Beach, where restoration was not performed, with lower cover of predominantly dead European beachgrass at AT&T where herbicide treatment was performed. European beachgrass cover was lowest at B Ranch North and Abbotts, where herbicide treatment and a combination of herbicide treatment and mechanical removal was performed, respectively. Iceplant cover was highest at North Beach, followed by B Ranch North and Population 9, with minimal and no cover recorded at AT&T and Abbotts, respectively. The highest bare ground cover was present at Abbotts and B Ranch North, due either to mechanical burial of European beachgrass with clean sand or to beach overwash blanketing herbicide-treated stands of European beachgrass with sand (Parsons et al. 2023). Average bare ground cover was estimated at ≤ 7% for the remaining three sites (Table 2).
Deer Mouse Captures
We captured 401 individual deer mice 876 times over 7,742 trap-nights in 2021, with the greatest number of captures and individuals at B Ranch North and North Beach and the fewest at Abbotts (Table 1). In 2022, we captured 547 individual deer mice 1,055 times over 5,292 trap nights, with the greatest number of captures and individuals at AT&T and B Ranch North, and the fewest at Abbotts. The number of captures and individuals of each sex were similar in 2021, with 449 captures of 194 females and 427 captures of 207 males. In 2021, 146 mice were caught once, 121 mice were caught twice, 86 mice were caught three times, 30 mice were caught four times, and 18 mice were caught five or more times (mean = 2.2 captures per individual, SD = 1.3). The number of captures and mice of each sex was nearly equal in 2022; we made 519 captures of 264 females, 515 captures of 272 males, and 21 captures of 11 mice of unknown sex. In 2022, 239 individuals were caught once, 157 individuals were caught twice, 102 were caught three times, and 49 mice were caught four times (mean = 1.9, SD = 1.0). The longest distance an individual moved between captures was 208 m, and the average distance moved was 14.8 m (SD = 18.2) in 2021 and 13.1 m (SD = 14.9) in 2022.
SECR Analysis
The top-ranked model for detection in both 2021 and 2022 according to AICc included a learned response to trapping and an effect of moon fraction on g0 (Table 3). There was a positive or “trap-happy” learned response whereby the probability of recapture was higher than the probability of initial capture (Figure 2, Table 5). The relationship between g0 and moon fraction was negative; capture probability was lower on days when a larger fraction of the moon was visible (Figure 2, Table 5). The scaling parameter of the detection function (σ) was 14.9 m (95% confidence interval = 14.2–15.7 m) in 2021 and 12.2 m (11.6–12.9 m) in 2022. These estimates of σ were approximately 1/4 of the length and width of our trapping grids and 1/5 of the buffer used to create the habitat mask, indicating the buffer width was sufficient.
The top-ranked models for deer mouse density in 2021 and 2022 included additive effects of habitat and site on D (Table 4). Deer mouse density was highest in areas that were highly invaded by European beachgrass and iceplant (at AT&T, North Beach, and Population 9), followed by herbicide-treated backdunes (at AT&T, herbicide treatment applied in 2018) and moist meadows (at AT&T, B Ranch North, North Beach, and Population 9) (Table 6). Mouse density was lowest in areas at Abbotts where invasive European beachgrass was mechanically removed and that have remained sparsely vegetated ≥ 10 years after restoration due to sand remobilization. The lowest deer mouse density in 2022 was in foredunes that had been treated with herbicide (found at B Ranch North) and that have less standing dead European beachgrass and higher bareground cover due to burial by extensive sand overwash. We also estimated deer mouse density using fixed effect of site only, for a simpler comparison among sites without explicit habitat effects on D. In the site-only density model, deer mouse density was lowest at Abbotts in 2021 and 2022 (Supplementary Material Figure S1 and Table S2) and highest at AT&T in 2022 (Supplementary Material Figure S2 and Table S2). Deer mouse density increased from 2021 to 2022 at AT&T (Supplementary Material Figure S2), B Ranch North (Supplementary Material Figure S3), North Beach (Supplementary Material Figure S4), and Population 9 (Supplementary Material Figure S5).
Discussion
Our results showed that dune habitats in which non-native European beachgrass was mechanically buried using clean sand and herbicide-treated foredunes in which extensive sand overwash occurred had the lowest density of deer mice. In contrast, areas dominated by European beachgrass and iceplant, and moist meadows, supported higher densities of deer mice. Habitats with lower deer mouse density shared a common feature in having extensive bare ground. Rodents often avoid open habitats to minimize predation risk, and forage more in areas with greater ground cover (Manson and Stiles 1998). Deer mouse density was low and stable across years at Abbotts Lagoon, particularly so in habitat that underwent mechanical treatment that increased bare ground. In contrast, deer mouse density increased greatly from 2021 to 2022 at the nearby AT&T site, at which extensive standing dead European beachgrass cover remained 5–6 years after herbicide treatment in 2018. Likewise, deer mouse density increased from 2021 to 2022 at the other three sites at which non-native plants were left untreated (North Beach and Population 9) or only herbicide treatments were applied (B Ranch North). The apparent effect of habitat treatment on deer mouse density was enduring; mechanical removal of European beachgrass at Abbotts Lagoon took place a decade earlier than our study in 2010 and 2011. It is likely that standing dead European beachgrass continues to provide habitat for deer mice for several years after herbicide treatment as the vegetation decomposes very slowly (Parsons et al. 2023). Although density of deer mice was highest in habitat dominated by non-native plants, moist meadows with dense growth of native sedges and rushes also supported a high density of deer mice, indicating that high ground cover and vegetation structure are key for coastal deer mouse populations at PRNS.
Habitats with more bare ground generally had lower density of deer mice, but deer mouse density also varied among patches of the same habitat type at different sites. In particular, B Ranch North had very high densities of deer mice despite having bare ground comparable to Abbotts. We did not have resources to sample replicate grids in each habitat type for each site in each year, which means our projected density surfaces have some additional uncertainty in unsampled areas (e.g., moist meadow at North Beach and Population 9 in 2022). Closer inspection of the spatial distribution of habitat, deer mouse captures, and density at B Ranch North reveals further complexities in the habitatdeer mouse relationship. The lowest density of deer mice at B Ranch North was in herbicide-treated foredunes; density in these foredunes was comparable to density in herbicide treated mid-dunes at Abbotts. The foredunes at B Ranch North have been subject to sand overwash and burial of dead European beachgrass following herbicide application. As a result, these foredunes have less cover and forage than nearby dune mat and moist meadow further inland. Unlike Abbotts, however, deer mouse density was elevated in the sparsely invaded native dune mat that contains the remaining Tidestrom’s lupine at B Ranch North. The surrounding habitat at Abbotts and B Ranch North could in part explain the higher density of deer mice at the latter. Dune mat at Abbotts is surrounded by areas mechanically treated to remove European beachgrass and by freshwater grass-dominated wetlands associated with a large lagoon. Dune mat at B Ranch North is surrounded by dense moist meadow, dune scrub, and drier grasslands used for grazing, which might be more suitable for deer mice than wetlands (Pendleton 1984). Whatever the cause, the high density of deer mice (and associated high seed predation rate) in dune mat and surrounding moist meadow habitats poses a serious threat to the viability of Tidestrom’s lupine populations at B Ranch North.
Our finding that deer mouse density was lower where European beachgrass was mechanically removed or treated with herbicide and overwashed by sand strengthens known links between deer mouse populations and this invasive species. Predation on Tidestrom’s lupine seeds is higher near patches of European beachgrass (Dangremond et al. 2010, Pardini et al. 2017). The low density of deer mice at Abbotts aligns with a Before-After-Control-Impact (BACI) study conducted at PRNS from 2005–2017 that found greatly reduced rates of seed predation in mechanically treated habitats at Abbotts (Pardini et al. 2018). In contrast, only marginal decreases in seed predation were observed at other sites (including AT&T, North Beach, Population 9, and B Ranch North) during this same period regardless of restoration status (Pardini et al. 2018), indicating the decline at Abbotts was due to the mechanical removal of European beachgrass. Deer mouse density remaining elevated in herbicide-treated areas that had substantial cover of dead European beachgrass indicates that dead standing biomass continues to provide refuge to these seed predators. The persistence of dead European beachgrass after herbicide treatment could be a legacy effect of invasion (Parsons et al. 2023) with the remnant physical structure perpetuating refuge-mediated competition between European beachgrass and Tidestrom’s lupine (Dangremond et al. 2010, Orrock et al. 2010). This lingering effect of herbicide treatment could be exacerbated if killing European beachgrass reduces the availability of an alternative food source for deer mice, which will feed on European beachgrass fruit and seeds (Pitts and Barbour 1979).
The density of deer mice at PRNS is among the highest model-based estimates reported for the Peromyscus maniculatus species group. An earlier study at PRNS estimated densities between 15–40 mice/ha for deer mice based on trapping a 1 ha plot, with the highest number of captures in areas with greater European beachgrass cover (Pitts and Barbour 1979). Our estimates cover a wider range from 1.6 to 278.1 mice/ha depending on the site and habitat, but are concordant with Pitts and Barbour (1979) in the finding that deer mouse density was highest in areas where European beachgrass dominated. Our estimates of deer mouse density were higher than estimates derived from SECR models for a long-term study population inhabiting boreal forest in Yukon, Canada (1.2–5.4 mice/ha; Krebs et al. 2011) and from non-spatial capture-mark-recapture models for populations in montane forests in Montana (2–14 mice/ha; Zwolak and Foresman 2008) and coniferous forest in British Columbia, Canada (2–13 mice/ha; Sullivan et al. 2023). SECR-based density estimates for deer mouse populations in the winter (11.3–21.6 mice/ha; Berl et al. 2017) and white-footed mouse (P. leucopus) populations in the summer and autumn (9.9–19.6 mice/ha; Berl et al. 2018) in agricultural habitat in Indiana were closer to density estimates for deer mice in most habitats at PRNS. Deer mouse density can vary over time as well as space. A study of deer mice on Santa Barbara Island, California, found densities fluctuated between years, with an order of magnitude difference between peaks (> 370 mice/ha) and troughs (17 mice/ha) (Drost and Fellers 1991). Likewise population density of P. leucopus in Ohio (Lewellen and Vessey 1998) and deer mice in British Columbia (Sullivan et al. 2023) fluctuated dramatically on seasonal and annual time scales. The increase in deer mouse density observed from 2021 to 2022 at four of five sites in this study indicates that temporal variation in mouse populations could be an important factor for viability of Tidestrom’s lupine at PRNS.
The causes of fluctuating deer mouse abundance at PRNS and consequences for Tidestrom’s lupine remain unknown. The 2021 water year was very dry, whereas total precipitation at PRNS in 2022 was approximately two times greater. Increased resource availability following a wetter winter in 2022 could explain the observed increase in deer mouse density. A study of the effects of deer mouse on viability of Tidestrom’s lupine populations at PRNS assumed constant levels of seed predation over time (Dangremond et al. 2010), but our results indicate predation pressure could vary from year to year. If deer mouse populations decline substantially in dry years resulting in a lower seed predation rate, it could allow for accumulation of a seed bank to sustain Tidestrom’s lupine populations in subsequent years. Alternatively, a high seed predation rate during wet years could have damaging effects on the viability of Tidestrom’s lupine populations, with higher seed predation leading to greater risk of extinction (Dangremond et al. 2010) if not compensated for by higher recruitment during dry years.
Conclusions
This study furthers the understanding of the relationships between deer mice, coastal dune habitats, and Tidestrom’s lupine (Dangremond et al. 2010, Pardini et al. 2017, 2018). Our results and previous work at PRNS (Pitts and Barbour 1979) indicate that areas with dense vegetation (living or dead), whether dominated by native or non-native species, support a greater abundance of deer mice. Given the susceptibility of Tidestrom’s lupine to seed predation by deer mice (Pardini et al. 2017), limiting the availability of densely vegetated refuges, including dead European beachgrass, near native dune mat could be essential to conserve this endangered plant. Wider implementation of treatments that eradicate European beachgrass and remove dead biomass could decrease local abundance of deer mice, with positive effects on endangered Tidestrom’s lupine populations. Mechanical removal and burial, while effective, can have unwanted side effects including remobilization of sand that drifts into nearby native plant communities and wetlands. Also, mechanical removal of European beachgrass is not an option for native moist meadow communities that can support high densities of deer mice. If one goal of dune restoration is to reduce densities of seed predators, herbicide treatment of non-native plants could be ineffective if it leaves dead standing vegetation as refuge for deer mice. Developing habitat management methods that reduce the density of live and dead non-native plants without negatively affecting native plant communities and wetlands would be valuable for reducing density of deer mice and conserving remaining Tidestrom’s lupine populations.
Acknowledgments
We thank Point Reyes National Seashore for allowing access and supporting our project. We thank the U.S. Fish and Wildlife Service for funding this project. Further salary support was provided by the National Park Service, Point Reyes National Seashore Association, and the U.S. Geological Survey Ecosystems Mission Area. We thank C. Cimmiyotti, C. Cardillo, E. Alvarado, and K. Kozlowski for assisting with field data collection. We thank L. D’Acunto for reviewing an earlier draft of this manuscript, and two anonymous reviewers for their helpful reviews. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
Footnotes
This open access article is distributed under the terms of the CC-BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) and is freely available online at: https://er.uwpress.org.
Supplementary materials are available online at: https://er.uwpress.org.
This open access article is distributed under the terms of the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0) and is freely available online at: http://er.uwpress.org
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.