Printer Friendly

POINT ARENA MOUNTAIN BEAVER (APLODONTIA RUFA NIGRA) SELECTS COOL CLIMATES AT FINE SPATIAL SCALES.

As the only living species from the once-diverse family Aplodontiidae, the Mountain Beaver (Aplodontia rufa) has endured through historical periods of climate change. The Mountain Beaver has primitive kidneys that are unable to concentrate urine (Nungesser and Pfieffer 1965). The species therefore requires access to plentiful fresh water. Because of this requirement, combined with a low heat tolerance, these fossorial rodents are currently limited to areas with cool, humid climates in California and the Pacific Northwest (Fig. 1; Nungesser and Pfeiffer 1965; Johnson 1971; Kinney 1971). Five of the 7 subspecies are considered species of concern or endangered in all or part of their range (USFWS 1998; Environment Canada 2015; CADFW 2017). Although the full extent of their Pleistocene range is unknown, fossil records suggest that since the Last Glacial Maximum, Mountain Beaver range has contracted significantly throughout the western United States, likely tracking cooler climates (Wake 2006; Blois and others 2010). Isolated relict populations, such as the federally endangered Point Arena Mountain Beaver (Aplodontia rufa nigra; hereafter PAMB), appear to exist in climate refugia (Wake 2006; Blois and others 2010).

The PAMB has a small, disjunct distribution of approximately 235 [km.sup.2] along the Mendocino Coast in California. Habitat loss and modification as a result of livestock grazing and localized urban development within this limited range is considered one of the species' primary threats (Fig. 1, USFWS 1998). Mountain Beavers are typically found in cool, dark forest understory near riparian areas (Beier 1989; Feldhamer and others 2003). However, PAMBs occupy areas on the coast that are notably different from the species' primary habitat, including coastal scrub and coastal dunes (Zielinski and Mazurek 2006). Although the temperature is moderated by summer fog, these areas have very little canopy cover and tend to be directly exposed to the sun. To persist in these warmer, drier habitats, PAMB may have developed local adaptations. Alternatively, PAMB on the coast may be selecting microclimates within their range that meet the typical needs of the species. Understanding habitat limitations at fine scales will be crucial in managing this species now and into the future.

Species distributions models (SDMs) have increasingly been used to understand the current geographic range and habitat requirements of taxa, particularly rare and sensitive species. Using a variety of statistical approaches, SDMs relate presence and either absence or background/pseudo-absence points to spatially explicit predictor variables to predict a species' distribution (Elith and others 2006). Although generally effective at broad scales, SDMs are often unable to accurately assess fine scale details of distribution (Pearson and Dawson 2003; Guisan and Thuiller 2005). This is likely a result of the spatial mismatch between readily available predictor variables and the scale at which most species experience their environment. This mismatch in scale may be especially important to consider for species with limited mobility (Pearson and Dawson 2003; Guisan and Thuiller 2005, Potter and others 2013). For a species like the Mountain Beaver, whose average home range estimates vary from 0.058 to 4.16 ha, climatic data at the 1-[km.sup.2] or even 900-[m.sup.2] resolution may not adequately capture variation in microhabitats available to them (Zielinski and Mazurek 2006; Arjo and others 2007).

Recent advances in downscaling climate data have allowed for the creation of fine-resolution environmental predictors. Fine-scale (often sub-meter) topographic data measured with aerial Light Detection and Ranging (LiDAR) sensors combined with microclimate data loggers have successfully been used to create temperature models that better correspond to the scale at which a species experiences the environment (George and others 2015; Frey and others 2016). Incorporating climate variables at an appropriate scale can improve predictive ability and provide insight into direct mechanisms by which Mountain Beavers may respond to climate change. For example, in coastal Mendocino County, a 1-[km.sup.2] area may appear to have quite hot maximum summer temperatures. However, within this area, Mountain Beavers will often have access to cooler ravines shaded by a dense redwood canopy. The difference in temperature between a south-facing, open meadow and a north-facing, old riparian forest could be up to 20[degrees]C in summer. Down-scaling topographic-driven climate data from 1-[km.sup.2] to 1-[m.sup.2] scale can provide much more accurate estimates of the climate actually experienced by a Mountain Beaver within the few hectares of its home range. However, owing to the amount of data required in fine scale models, they are generally not feasible for range-wide modeling for most species. As a climatically sensitive subspecies with a limited range, PAMB are an ideal organism for studying range restrictions based on microclimate variables.

Our objective was to determine whether PAMB have adapted to habitats with hotter, drier temperatures than elsewhere in the species' range, or if the sub-species is able to find microclimates within this hotter landscape that match conditions throughout their distribution. We developed fine-scale climate surfaces using temperatures recorded from data loggers and topographic variables calculated from LiDAR data. Occurrence locations for PAMB were obtained from personal surveys and a previous study (Zielinski and others 2015). The fine-scale climate layers were incorporated with PAMB occurrence data in Maxent to model fine-scale habitat suitability throughout their range. We then used an independent data set to evaluate the accuracy of these models. We expected that PAMB were able to persist in apparently hotter, drier areas by taking advantage of microhabitat features that mirrored the cooler, wetter conditions in which the species is normally found. If so, we expected that the fine-scale model of PAMB distribution would identify a negative relationship between PAMB presence and: (1) mean daily maximum temperature during the summer-autumn; (2) mean daily minimum temperature during the summer-autumn; and (3) mean hours above PAMB upper critical temperature in the summer-autumn.

METHODS

The range of PAMB is currently limited to 235-[km.sup.2] area along the Mendocino County coast in California (Zielinski and others 2015). The area is comprised of a variety of habitat types including coastal dunes, coastal scrub, riparian, and non-native grasslands along the coast, and forests and riparian areas inland (LANDF1RE 2017). Land use varies, with approximately 25% of the range owned by timber companies, with the rest divided among public land, ranches and dairy farms, and other private parcels (ENPLAN 2017). The weather is moderate, with annual mean temperatures between 10 and 12[degrees]C (Fick and Hijmans 2017). Warmer temperatures are typically observed May-November (NOAA 2012). Summers have relatively frequent fog and low-level clouds, particularly along the coast (Torregrosa and others 2016).

In the summer of 2017, we conducted burrow surveys following USFWS protocols to determine patterns of occupancy (USFWS 2017). We also deployed temperature data loggers to develop fine-scale, spatially explicit climate layers. Individual PAMB construct a burrow system with multiple burrow entrances that are distinct and easily identified (Camp 1918; Feldhamer and others 2003; Zielinski and others 2015). Visual surveys for burrows were targeted at riparian areas and in areas where previous surveys had located burrows. Active burrows were classified as having fresh dirt and being clean of vegetation and other debris. PAMB burrow areas tend to have multiple entrances, and we did not attempt to classify individual openings as active or inactive. Instead, if a majority of burrow openings appeared active, we classified the entire area as active. For each active burrow area detected, a data logger was placed near one of the burrow entrances. Using ArcMap (Environmental Systems Research Institute, Inc., Redlands, CA), a grid of points was placed over the range of PAMB, such that each point was 150 m from the nearest neighbor. From this grid, we selected 2 subsets, 1 inland and 1 coastal, of 100 points for PAMB surveys and data logger deployment (Fig. 2). Sites for the coastal grid were randomly selected in R (George and others 2015). Sites were reselected if they were unsuitable for data logger deployment (such as in a river or inaccessible owing to topography or vegetation). Because of accessibility constraints (rugged terrain and landowner permission), points for the inland grid were selected using a stratified approach (Fridley 2009; Frey and others 2016). Inland points were placed along transects that were selected based on accessibility, presence of PAMB, and environmental variables. Histograms of elevation, slope, and aspect for the PAMB range were visually compared to grid points to ensure the range of environmental variables was reflected in the deployment sites. Additionally, as data loggers were placed in the field, the surrounding area was surveyed for any Mountain Beaver sign.

Microclimate data were collected using iButton[R] (Maxim Integrated, San Jose, CA, USA) temperature loggers to create a downscaled microclimate model of PAMB habitat. Each data logger was placed in a shield constructed from PVC and fiberglass screen that allowed for airflow while limiting the effect of direct solar radiation and precipitation. All data loggers were deployed during the summer, checked once in late summer-early autumn, and collected in early winter. Each data logger recorded temperatures on the same 3.5-h cycle to maximize battery life and capture different times of day. This deployment from July-December captured most of the typical warm season in the PAMB range (NOAA 2012).

We then developed a series of topographic predictor variables using LiDAR data and other available GIS spatial data layers to generate a suite of range-wide, fine-scale climatic variables. Specifically, a Digital Terrain Model (DTM), Digital Surface Model (DSM), and Canopy Height Model (CHM) were derived from LiDAR data from NASA's Northern San Andreas Fault Study (NASA 2003). Environmental predictor variables that are generally accepted to affect temperature were calculated using the DTM, DSM, and CHM at 2 spatial grains, 10 and 100 m, to account for the effects of these variables at a local and more widespread scale (Table 1; Frey and others 2016). As a proxy for fog, we calculated distance to coast from a national US shapefile at a 1-m scale (GADM.org). We summarized 3 response variables for each data logger during the entire 6-mo deployment period: minimum, mean, and maximum temperature. We calculated an additional response variable, upper critical temperature ([CT.sub.max]), as the average number of times per day a data logger recorded temperatures above 32[degrees]C, which is the lower temperature limit that causes mortality in Mountain Beavers with prolonged exposure (Johnson 1971).

We used boosted regression trees in R package 'dismo' (Hijmans and others 2017) to analyze the relationship between topographic data at the data logger locations and climate data recorded on the data logger; we projected these climate models onto topographic data throughout PAMB range. Boosted regression trees can identify nonlinear relationships and explore complex interactions, and they can be particularly useful for a large number of predictor variables. Models were fitted using a tree complexity of 3 to 5 and a bag fraction of 0.75 owing to the small sample size (n = 168). All models were initially run with a learning rate of 0.005 and decreased until at least 1000 trees were used to fit the model (Elith and others 2008). Models were evaluated using cross validation, and we then simplified the top model by dropping non-informative predictor variables using methods outlined in Elith and others (2008). As the purpose of simplification is to improve the performance of the model, when the simplified model did not outperform the un-simplified model, the unsimplified model was considered the top model (Elith and others 2008). The models were also tested for correlation between predicted and observed values using 10-fold cross validation. Low correlations between predicted and observed values indicate over-fitting of the model; although there are no accepted target values for correlations, Frey and others (2016) specified that values >0.6 were "high."

Finally, we combined the occurrence data collected for PAMB and points from a previous study of PAMB distribution (Zielinski and others 2015) with the fine-scale climate data to construct local distribution models throughout PAMB range using Maxent in package 'dismo' (Hijmans and others 2017). Methods for distinguishing active from inactive burrow areas and documenting locations from the earlier study were substantially the same as ours. The mean, maximum, minimum, and critical temperature layers were used as predictor variables. Initially we considered all possible combinations of these 4 variables in our candidate set. Next, we filtered the set so that only variables with Pearson correlation coefficient between -0.5 and 0.5 were included in the same model. We evaluated the role of model complexity by varying the "beta" parameter in Maxent from 0.5 to 3.0 in steps of 0.5. The beta multiplier varies the number of different features used to relate habitat suitability to the predictor variables. A higher beta parameter results in a simpler model, and thus fewer total features used. Because of Maxent's tendency to overfit, we used model selection based on AICc to balance model fit with complexity (Warren and Seifert 2011). Accuracy of this model was also assessed using the Continuous Boyce Index (CBI), following the methods used for the broad-scale models (Hirzel and others 2006). Additional presence points provided from a variety of sources and collated by USFWS were used as an independent testing data set but were not used as training data owing to uncertainty in accuracy of location data.

RESULTS

We documented 20 PAMB active burrow areas. We used an additional 35 burrow areas from Zielinski and others (2015) for a total of 55 PAMB occurrences for the suitability models. We deployed a total of 195 iButton data loggers during the summer of 2017 (Fig. 2). Because of theft or loss, we removed 27 data loggers from the analysis, leaving 168 locations for this analysis. Several of the data loggers recorded temperatures above a reasonable limit for the area, likely resulting from solar radiation and heat reflection from the soil. We did choose to include these for analysis, as they likely captured relative patterns of temperature, if not reliable absolute values (Terando and others 2017).

Creation of fine-scale climate layers from the data loggers and LiDAR data resulted in accurate models for mean, minimum, and [CT.sub.max] (Table 2, Fig. 3). Deviance for these 3 layers ranged from 0.023 ([CT.sub.max]) to 2.240 (minimum). However, the maximum temperature model had high deviance. The cross-validation correlation values were moderate (ranging from 0.586 to 0.794), indicating that there was some agreement between training and testing data.

The top model for PAMB distribution included mean temperature and [CT.sub.max] (Table 3, Fig. 4). Maximum temperature was too correlated with minimum temperature (0.71) and [CT.sub.max] (0.90), as were minimum temperature and [CT.sub.max] (0.88), therefore no models that included these variables together were considered. The top 5 models all included mean temperature, 4 included [CT.sub.max] with varying beta values, and 1 incorporated minimum temperature (Table 3). Maximum temperature was not included in any of the best-supported models. In the top model, mean temperature had a greater contribution (94.3%) than [CT.sub.max] (5.7%). Suitability decreased as mean temperature increased above 12.0[degrees]C, and decreased as daily readings above [CT.sub.max] increased (Fig. 5). The model had an AUC of 0.880. The mean CBI value for the top model was 0.86 when the original data were divided into training and testing datasets. The mean CBI value was 0.99 when the additional USFWS points were used as the testing dataset.

DISCUSSION

In this study, we created species distribution models for PAMB at a scale that reflects how they interact with their environment. A previous PAMB habitat modeling study pointed to the need to assess critical habitats for management and recovery of this taxon (Zielinski and others 2015), and our study presents a novel estimate of habitat suitability driven by climatic variables at a scale appropriate for the species. Such models have long been a tool for conservation and recovery of endangered species (Eads and others 2011; D'Elia and others 2015). Several studies have shown that increased grain size decreases the accuracy of SDMs (Gottschalk and others 2011; Song and others 2013). This does not discount the need for coarser models, as they can often be used in conjunction with fine-scale models to increase understanding of factors limiting a species distribution at multiple scales (Fournier and others 2017). These fine-scale models provide insight into how PAMB is distributed through its range, whereas coarser models may elucidate information on why PAMB is limited to such a small area.

Similar to other climatically sensitive species, PAMBs appear to select habitat at a microclimatic scale (Wilkening and others 2011; Shi and others 2014; Varner and Dearing 2014). The fine-scale SDM showed that PAMB inhabited cooler microclimates within their range. These microclimates may be particularly important at range margins such as the conditions experienced by PAMB (Ray and others 2016). As temperature and moisture are believed to be 2 of the primary environmental variables limiting Mountain Beaver distribution, incorporating fine-scale humidity, fog, and precipitation in future models may help to further elucidate how PAMB select habitats (Nungesser and Pfieffer 1965; Johnson 1971).

A previous study on habitat suitability for PAMB used 25-ha sample units, likely leading to an overestimation of suitability in some areas and an underestimation in others (Zielinski and others 2015). Both models showed a clear affinity for riverine corridors; however, our fine-scale model showed increased connectivity between suitable areas, particularly in the southern portion of their range. The pattern of suitability in the northern portion of the range is more strictly defined by riverine corridors, whereas in the south their distribution appears to be more diffuse. This may be a result of the increased fog in the southern portion of the range, but warrants further research (Torregrosa and others 2016). Maintaining and increasing connectivity is critical for the survival of this species, and these suitability studies provide insight into high-priority areas, such as riverine corridors and the southern portion of the range.

Microclimate appeared to be a driving factor in PAMB habitat selection; however, this does not exclude the possibility of local adaptations, and there are multiple mechanisms in which this adaptation can be realized. Behavioral adaptations, such as shifting foraging activities to cooler parts of the day, have been documented in other climatically sensitive species (for example, Pikas [Ochotona princeps]; Smith 1974). Activity patterns of Mountain Beavers have only been studied in the subspecies californica, which were active at various times throughout the day and night with increased activity and ranging at night. During summer, these Mountain Beavers shifted to more nocturnal activity, potentially to avoid increased temperatures (Ingles 1959). This shift may be more conspicuous in populations living in areas with hotter temperatures, as PAMB do, and may occur seasonally or daily as weather conditions, such as fog, shift. Future studies on differences in active periods between subspecies could provide insight into behavioral adaptations.

Genetic adaptations may also allow PAMB to occupy atypical habitats. Genetic analyses support grouping the 7 subspecies of Mountain Beaver into 5 distinct clades. The Point Arena Mountain Beaver (A. r. nigra), Point Reyes Mountain Beaver (A. r. phaea), and Humboldt Mountain Beaver (A. r. humboldtiana) form a closely related California coastal clade, whereas the remaining 4 subspecies (A. r. olympica, A. r. rufa, A. r. califomica, A. r. pacifica) remain in separate clades (Piaggio and others 2013; Fig. 1). PAMB can be further divided into 3 distinct genetic groupings, 2 of which exist only on the coast (Zielinski and others 2013). The 3rd grouping primarily occupies inland habitats, which more closely resemble other Mountain Beaver habitat. It is possible that the coastal groupings may be more adapted to hotter, drier climates and are therefore critical to preserve in the face of climate change. A study into the physiology of California coastal Mountain Beavers could provide valuable insight into whether this species has adapted to a hotter, drier climate or if microclimate selection is enough to sustain them.

Although this study focused on the impacts of abiotic limits on Mountain Beaver distribution, there are likely biotic variables that also influence this species' distribution. Biotic variables are recognized as important factors in predicting distribution shifts in response to climate change; however, there is debate over their relative importance at broad scales (Pearson and Dawson 2003; Wisz and others 2013). Incorporating biotic factors, such as predation, competition, and dispersal abilities, into SDMs, particularly fine-scale SDMs, for Mountain Beavers may improve our understanding of how this species will respond to climate change (Wisz and others 2013).

ACKNOWLEDGMENTS

This work was partially funded by cooperative agreements with the Bureau of Land Management and the United States Fish and Wildlife Service. We are grateful for the insight on PAMB survey techniques and study design from WJ Zielinski, K Fitts, R Douglas, H Ross, G Schmidt, J Hunter, M Szykman Gunther, and J Szewczak. Thanks also to our field technicians T Yanin and S Scherbinski. This project would not have been possible without access to lands owned or administered by R Douglas and H Ross (Mendocino Redwood Company), T Fuller (California State Parks), W Standley and K Barnitz (US Bureau of Land Management), N Houtz (Mendocino Land Trust), H Newberger (The Conservation Fund), B Morrison (Soper Wheeler Company), F Fedeli, T Shi, B Lotan, N Pinette, K Fusek, and B Heick. Additional thanks to S Cardimona for coordinating field housing.

LITERATURE CITED

ARJO WM, HUENEFELD RE, NOLTE DL. 2007. Mountain Beaver home ranges, habitat use, and population dynamics in Washington. Canadian Journal of Zoology 85:328-337.

BEIER J. 1989. Use of habitat by Mountain Beaver in the Sierra Nevada. Journal of Wildlife Management 53:649-654.

BLOIS JL, MCGUIRE JL, HADLY EA. 2010. Small mammal diversity loss in response to late-Pleistocene climatic change. Nature 465:771-775.

[CADFW] CALIFORNIA DEPARTMENT OF FISH AND WILDLIFE. 2017. Special Animals List. Sacramento, CA: California Department of Fish and Wildlife.

CAMP DL. 1918. Excavations of the burrows of the rodent Aplodontia, with observations on habitat of the animal. University of California Publications in Zoology 17:517-536.

D'ELIA J, HAIC; SM, JOHNSON M, MARCOT BG, YOUNG R. 2015. Activity-specific ecological niche models for planning reintroductions of California Condors (Gymnogyps californianus). Biological Conservation 184:90-99.

EADS DA, MILLSPAUGH JJ, BIGGINS DE, JACHOWSKI DS, LIVIERI TM. 2011. Evaluation of a Black-footed Ferret resource utilization model. Journal of Wildlife Management 75:1155-1163.

ELITH J, GRAHAM CH, ANDERSON RP, DUDIK M, FERRIER S, GUISAN A, HIJMANS RJ, HUETTMAN F, LEATHWICK JR, LEHMANN A, LI J, LOHMANN LG, LOISELL BA, MANION G, MORITZ C, NAKAMURA M, NAKAZAWA Y, OVERTON JMM, TOWNSEND PETERSON A, PHILLIPS S], RICHARDSON K, SCACHETTI-PEREIRA R, SCHAPRE RE, SOBER6.N J, WILLIAMS S, WISZ MS, ZIMMERMANN NE. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography 29:129-151.

ELITH J, LEATHWICK JR, HASTIE T. 2008. A working guide to boosted regression trees. Journal of Animal Ecology 77:802-813.

ENPLAN. 2017. ENPLAN Parcel Viewer Mendocino County. Accessed 5 May 2017.

ENVIRONMENT CANADA. 2015. Management Plan for the Mountain Beaver (Aplodontia rufa) in Canada. Species at Risk Act Management Plan Series. Ottawa, ON: Environment Canada.

FELDHAMER GA, ROCHELLE JA, RUSHTON CD. 2003. Mountain Beaver (Aplodontia rufa). In: Feldhamer GA, Thompson BC, Chapman JA (editors). Wild mammals of North America: Biology, management, and conservation. Baltimore, MD: Johns Hopkins University Press, p 179-187.

FICK SE, HIJMANS RJ. 2017. Worldclim 2: new 1-km spatial resolution climate surfaces for global land areas. International Journal of Climatology 37:4302-4315.

FOURNIER A, BARBET-MASSIN M, ROME Q, COURCHAMP F. 2017. Predicting species distribution combining multi-scale drivers. Global Ecology and Conservation 12:215-226.

FREY SJ, HADLEV AS, JOHNSON SL, SCHULZE M, JONES JA, BETTS MG. 2016. Spatial models reveal the microclimatic buffering capacity of old-growth forests. Science Advances 2: e1501392.

FRIDLEY JD. 2009. Downscaling climate over complex terrain. High finescale (<1000 m) spatial variation of near-ground temperatures in a montane forested landscape (Great Smoky Mountains). Journal of Applied Meteorology and Climatology 48:1033-1049.

GEORGE AD, THOMPSON FR, FAABORG J. 2015. Using LiDAR and remote microclimate loggers to down-scale near-surface air temperatures for site-level studies. Remote Sensing Letters 6:924-932.

GOTTSHALK TK, AUE B, HOTES S, EKSCHMITT K. 2011. Influence of grain size on species-habitat models. Ecological Modelling 222:3403-3412.

GUISAN A, THUILLER W. 2005. Predicting species distribution: Offering more than simple habitat models. Ecology Letters 8:993-1009.

HUMANS RJ, PHILLIPS S, LEATHWICK J, ELITH J. 2017. Package 'dismo': Species distribution modeling. R package version 1.1-4.

HIRZEL A H, LE LAY G, HELPER V, RANDIN C, GUISAN A. 2006. Evaluating the ability of habitat suitability models to predict species presences. Ecological Modelling 199:142-152.

INGLES LG. 1959. A quantitative study of Mountain Beaver activity. American Midland Naturalist 61:419-423.

JOHNSON SR. 1971. The thermal regulation, microclimate, and distribution of the Mountain Beaver, Aplodontia rufa pacifica Merriam [dissertation]. Corvallis, OR: Oregon State University.

KINNEY JL. 1971. Environmental physiology of a primitive rodent [dissertation]. Eugene, OR: University of Oregon.

LANDEIRE. 2017. Existing vegetation type layer. LAND-FIRE 1.1.0. US Department of the Interior, Geological Survey. Accessed 23 February 2017.

[NASA] NATIONAL AERONAUTICS AND SPACE ADMINISTRATION. 2003. Northern San Andreas Fault Study. Distributed by OpenTopography.org. DOI: 10.5069/G9MW2F2H.

[NOAA] NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION. 2012. National Data Buoy Center. Station ANVC1 - Climatic Summary Plots for air temperature. http://www.ndbcnoaa.gov/view_climplot.php?station=anvc1&meas=at. Accessed 31 Aug 2017.

NUNCESSER W, PPEIEFER E. 1965. Water balance and maximum concentrating capacity in the primitive rodent, Aplodontia rufa. Comparative Biochemistry and Physiology 14:289-297.

PEARSON RG, DAWSON TP. 2003. Predicting the impacts of climate change on the distribution of species: Are bioclimate envelopes useful? Global Ecology and Biogeography 12:361-371.

PIAGGIO AJ, COGHLAN BA, MISCAMPBELL AE, ARJO WM, RANSOME DB, RITLAND CE. 2013. Molecular phytogeny of an ancient rodent family (Aplodontiidae). Journal of Mammalogy 94:529-543.

POTTER KA, WOODS HA, PINCEBOURDE S. 2013. Microclimatic challenges in global change biology. Global Change Biology 19:2932-2939.

RAY C, BEEVER EA, ROADHOUSE TJ. 2016. Distribution of a climate-sensitive species at an interior range margin. EcoSphere 7:e01379.

SHI H, PAULL D, BROOME L, BATES H. 2014. Microhabitat use by mountain pygmy-possum (Burramys parvus): Implications for the conservation of small mammals in alpine environments. Austral Ecology 10.1111/aec.l2220.

SMITH AT. 1974. The distribution and dispersal of pikas: influences of behavior and climate. Ecology 55:1368-1376.

SONG W, KIM E, LEE D, LEE M, JEON SW. 2013. The sensitivity of species distribution modeling to scale differences. Ecological Modelling 248:113-118.

TERANDO AJ, YOUNGSTEADT E, MEINKE EK, PRADO SG. 2017. Ad hoc instrumentation methods in ecological studies produce highly biased temperature measurements. Ecology and Evolution 7:9890-9904.

TORREGROSA A, COMBS C, PETERS J. 2016. GEOS-derived fog and low cloud indices for coastal north and central California ecological analyses. Earth and Space Science 3:46-67.

[USFWS] US FISH AND WILDLIFE SERVICE. 1998. Point Arena Mountain Beaver (Aplodontia rufa nigra) (Rafinesque)) recovery plan. Region 1, Portland, OR.

[USFWS] US FISH AND WILDLIFE SERVICE. 2017. Draft guidelines for project-related habitat assessments and presence-absence surveys for the Point Arena Mountain Beaver (Aplodontia rufa nigra). Areata, CA: Areata Fish and Wildlife Office.

VARNER J, DEARING MD. 2014. The importance of biologically relevant microclimates in habitat suitability assessments. PLoS ONE 9:e104648.

WAKE TA. 2006. Archaeological sewellel (Aplodontia rufa) remains from Duncan's Point Cave, Sonoma County, California. Journal of Mammalogy 87:139-147.

WARREN DL, SEIFERT SN. 2011. Ecological niche modeling in Maxent: The importance of model complexity and the performance of model selection criteria. Ecological Applications 21:335-342.

WILKENINC JL, RAY C, BEEVER EA, BRUSSARD PF. 2011. Modeling contemporary range retraction in Great Basin Pikas (Ochotona princeps) using data on microclimate and microhabitat. Quaternary International 235:77-88.

WIsz, MS, POTTIER J, KISSLINC WD, PELLISSIER L, LENOIR J, DAMGAARD CF, DORMANN CF, FORCHHAMMER MC, GRYTNES JA, GUISAN A, HEIKKINEN RK, HOYE TT, KUHN I, LUOTO M, MAIORANO L, NILSSON MC, NORMAND S, OCKINGER E, SCHMIDT NM, TERMANSEN M, TIMMERMAN A, WARDLES DA, AASTRUF P, SVENNING JC. 2013. The role of biotic interactions in shaping distributions and realized assemblages of species: Implications for species distribution modelling. Biological Reviews 88:15-30.

(*) ZIELINSKI WJ, MAZUREK MJ. 2006. The reproductive ecology and home range of the Point Arena Mountain Beaver (Aplodontia rufa nigra). Unpublished report to the US Fish and Wildlife Service. Areata, CA: Areata Fish and Wildlife Office.

ZIELINSKI WJ, SCHLEXER FV, PARKS SA, PILGRIM KL, SCHWARTZ MK. 2013. Small geographic range but not panmictic: How forests structure the endangered Point Arena Mountain Beaver (Aplodontia rufa nigra). Conservation Genetics 14:369-383.

ZIELINSKI WJ, SCHLEXER FV, DUNK JR, LAU MJ, GRAHAM JJ. 2015. A range-wide occupancy estimate and habitat model for the endangered Point Arena Mountain Beaver (Aplodontia rufa nigra). Journal of Mammalogy 96:380-393.

JENNIE JONES SCHERBINSKI AND WILLIAM T BEAN

Humboldt State University, 1 Harpst St., Arcata, CA 95521 USA; jkj175@humboldt.edu; tim.bean@humboldt.edu

Submitted 2 January 2019, accepted 24 June 2019. Corresponding Editor: Robert Hoffman.

(*) Unpublished
TABLE 1. Predictor variables and the data scales used to calculate
fine-scale temperature models from the Digital Terrain Model (DTM),
Digital Surface Model (DSM), and Canopy Height Model (CHM). These
variables were selected as they likely affect air temperature (Frey and
others 2016)

Variable            Spatial scale   Description

Distance to coast    1 m            Euclidean distance from coast, proxy
                                    for fog
Mean cover          10 m/100 m      Mean percent cover from DTM and DSM
SD cover            10 m/100 m      Standard deviation of percent cover
                                    from DTM and DSM
Mean eastness       10 m/100 m      Mean sin (aspect *pi/180), between
                                    -1 and 1, DTM
Mean elevation      10 m/100 m      Mean elevation, DTM
SD elevation        10 m/100 m      Standard deviation of elevation, DTM
Elevation range     10 m/100 m      Range of elevation values, DTM
Mean northness      10 m/100 m      Mean cos(aspect*pi/180), between -1
                                    and 1, DTM
Mean slope          10 m/100 m      Mean slope (%) from DTM
SD slope            10 m/100 m      Standard deviation of slope (90 from
                                    DTM
Topo Index          10 m/100 m      Difference in elevation between the
                                    point and mean elevation, indicating
                                    local low and high points, DTM
Mean veg height     10 m/100 m      Mean vegetation height, CHM
SD veg height       10 m/100 m      Standard deviation of vegetation
                                    height, CHM

TABLE 2. Model settings and performance metrics for the top model for
each temperature variable, and whetherthe top model was simplified by
removing uninformative predictor variables.

Variable   Tree      Learning   No. trees   Deviance   Dev SE   CV corr
           complex   rate

Mean       5         0.005      1000         0.517     0.053    0.784
Max        3         0.002      1300        33.258     2.206    0.64
Min        5         0.002      1400         2.24      0.292    0.794
CTmax      5         0.002      1300         0.023     0.004    0.586

Variable   CV SE   Simplified

Mean       0.041   No
Max        0.034   No
Min        0.029   Yes
CTmax      0.101   Yes

TABLE 3. Model selection table for PAMB microclimate suitability. Only
the top 5 models are shown. Mean = mean daily temperature
(July-December); CTmax = mean daily temperature recordings above
32[degrees]C; Min = minimum temperature (July-December). (Note: the
beta multiplier determines the number of parameters Maxent uses to
relate the listed predictors to habitat suitability; therefore, 2
models with the same predictors will vary with different beta
multipliers. In this case, a simpler model IK = 4] was better supported
than a more complex model [K = 6] using the same predictors).

Model          Beta        Log         K   AICc     [DELTA]AICc   AICc
               multiplier  likelihood                             weight

Mean + CTmax   3.0         -442.2641    4  893.33    0.00          0.65
Mean + CTmax   2.0         -440.9083    6  895.57    2.23          0.21
Mean + CTmax   2.5         -441.5078    6  896.77    3.44          0.12
Mean + CTmax   1.5         -438.1371   10  901.27    7.95          0.01
Mean + Min     3.0         -445.2443    6  904.24   10.91         <0.01
COPYRIGHT 2019 Society for Northwestern Vertebrate Biology
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Scherbinski, Jennie Jones; Bean, William T.
Publication:Northwestern Naturalist: A Journal of Vertebrate Biology
Date:Dec 22, 2019
Words:5310
Previous Article:ABSTRACTS FROM THE 2019 ANNUAL MEETING OF THE SOCIETY FOR NORTHWESTERN VERTEBRATE BIOLOGY, HELD JOINTLY WITH THE WASHINGTON CHAPTER OF THE WILDLIFE...
Next Article:DIURNAL ACTIVITY PATTERN OF A PAIR OF BREEDING NORTHERN PYGMY-OWLS.
Topics:

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters