Core terrestrial habitat for conservation of local populations of salamanders and wood frogs in agricultural landscapes

BIOLOGICAL CONSERVATION

Biological Conservation 120 (2004) 399–409 www.elsevier.com/locate/biocon

Core terrestrial habitat for conservation of local populations of salamanders and wood frogs in agricultural landscapes Deni Porej a

a,b,* ,

Mick Micacchion c, Thomas E. Hetherington

a

Department of Evolution, Ecology and Organismal Biology, 318 W. 12th Avenue, The Ohio State University, Columbus, OH 43210, USA b The Nature Conservancy Ohio Chapter, 6375 Riverside Drive, Suite 50, Dublin, OH 43221, USA c Division of Surface Water, The Ohio Environmental Protection Agency, 122 S. Front Street, Columbus, OH 43215, USA Received 12 November 2003; received in revised form 5 March 2004; accepted 25 March 2004

Abstract Pond-breeding amphibians require aquatic and terrestrial habitats to complete their lifecycles, and preservation of both habitats is necessary for maintaining local populations. Current wetland regulations focus primarily on aquatic habitats, and criteria to define critical upland habitats and regulations to protect them are often ambiguous or lacking. We examined the association between the presence of seven pond-breeding amphibian species and the landscape composition surrounding 54 wetlands located within the Till Plains and the Glaciated Plateau ecoregions of Ohio, USA. We quantified landscape composition within 200 m of the wetland (‘‘core terrestrial zone’’) and the area extending from 200 m to 1 km from the wetland (‘‘broader landscape context zone’’). We constructed binary logistic regression models for each species, and evaluated them using Akaike Information Criterion. Presence of spotted salamanders (Ambystoma maculatum), Jefferson’s salamander complex (A. jeffersonianum) and smallmouth salamanders (A. texanum) was positively associated with the amount of forest within the core zone. Presence of wood frogs (Rana sylvatica) was positively associated with the amount of forest within the core zone and the amount of forest within the broader landscape context zone. Presence of tiger salamanders (A. tigrinum tigrinum) was negatively associated with the cumulative length of paved roads within 1 km of the site, and presence of red-spotted newts (Notophthalmus v. viridescens) was negatively associated with the average linear distance to the five nearest wetlands. Overall salamander diversity was positively associated with the amount of forest within the core zone, and negatively associated with the presence of predatory fish and cumulative length of paved roads within 1 km of the site. Our results confirm the strong association between the structure of surrounding upland areas and amphibian diversity at breeding ponds, and stress the importance of preserving core terrestrial habitat around wetlands for maintaining amphibian diversity.
2004 Elsevier Ltd. All rights reserved. Keywords: Ambystoma; Forested wetlands; Landscape; Predatory fish; Salamander; Spotted salamander; Tiger salamander; Wood frog

1. Introduction Throughout the Midwestern United States, amphibian numbers have declined in the post-European settlement era with the conversion to an agriculturally dominated landscape (Lannoo, 1998). In addition to changes to terrestrial habitats brought about by agriculture, estimates of wetland losses within Midwestern states over the past 200 years range from 85% to 90% (Dahl, 1990). *

Corresponding author. Tel.: +1-614-717-2770x38. E-mail address: [email protected] (D. Porej).

0006-3207/$ - see front matter
2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2004.03.015

While studies of amphibian communities demonstrate strong associations between local diversity and the composition of the surrounding landscape (Laan and Verboom, 1990; Koloszvary and Swihart, 1999; Knutson et al., 1999, 2000; Lehtinen et al., 2001), protection and restoration of wetlands is often decoupled from preservation or restoration of surrounding uplands. This issue is particularly important for numerous pondbreeding amphibians in the Midwest that use forest as their primary upland habitat (Smith, 1961; Minton, 1972; Pfingsten and Downs, 1989; Davis and Menze, 2002). The landscape of Ohio has been converted from nearly a continuous cover of mature, deciduous forest to

400

D. Porej et al. / Biological Conservation 120 (2004) 399–409

a mosaic of agriculture, suburban/urban landuse and fragmented woodlands (Dunning et al., 1992, 1998). Around the turn of the 20th century, less than 10% forest remained in Ohio, mainly in the southeastern, unglaciated portion of the state. The overall forest cover in the state has now increased to around 30%, but the agricultural landscape of western Ohio still consists of less than 10% forest cover (Lafferty, 1979), mainly in the form of isolated woodlots. Conservation plans for amphibians must consider both local and landscape dynamics (Semlitsch, 2000), and identification of minimum habitat requirements for local populations is a necessary first step towards the creation of spatially explicit population models needed for effective conservation planning (Dunning et al., 1995; Marsh and Trenham, 2001). Marsh and Trenham (2001) suggest that species translocations in disturbed environments may be necessary to promote local and regional amphibian population persistence. Whether translocations of amphibians are regarded as a proven management method or just as the experimental technique (Dodd and Seigel, 1991; Seigel and Dodd, 2002; Trenham and Marsh, 2002), identification of both aquatic and terrestrial habitat requirements for local populations is nevertheless a critical issue. An important advance in our understanding of the spatial scale at which local amphibian populations use the surrounding terrestrial habitat is the biological delineation of terrestrial ‘‘core zones’’ (Semlitsch, 1998, 2001; Semlitsch and Bodie, 2003). Core zone widths were derived from a review of emigration distances for various species, and are defined as terrestrial habitats used by the local breeding populations that are biologically necessary for the maintenance of amphibian and reptilian diversity. These studies suggest that a ‘‘core zone’’ used by a local population extends as far as 218 m from the wetland edge for pond-breeding salamanders, and up to 290 m for amphibians in general (Semlitsch, 1998; Semlitsch and Bodie, 2003). Our study examines the association between the amount of forest within the core zone and the presence of pond-breeding salamanders and wood frogs in nat-

ural wetlands within the agricultural landscapes of Ohio. We predicted that species would differ in the strength of positive association with the amount of forest in the core zone based on their natural history characteristics (Table 1). We hypothesize that the amount of forest in the core zone is critical for the persistence of amphibian species such as spotted salamanders, marbled salamanders, Jefferson’s complex salamanders and wood frogs, which are less likely to persist in, and travel long distances through non-forest habitat (Thompson et al., 1980; Douglas and Monroe, 1981; Kleeberger and Werner, 1983; Whitford and Vinegar, 1966; Demaynadier and Hunter, 1998; Rothermel and Semlitsch, 2002). The second group of species comprises eastern redspotted newts and smallmouth salamanders, which are generally associated with forests but can persist for shorter periods in and/or migrate long distances across open (prairie and pasture) lands (Smith, 1961; Minton, 1972). Finally, tiger salamanders can persist in nonforested habitats (Duellman, 1954; Petranka, 1998), and may perceive their surroundings as a heterogeneous undivided landscape (Addicott et al., 1987). Therefore, we expected the presence of tiger salamanders not to be significantly associated with the amount of forest within the terrestrial core zone. We further investigate the possible thresholds in forest landcover and other landscape variables associated with individual species’ presence, and discuss our findings in the context of current agricultural and wetland management practices.

2. Data and methods 2.1. Study sites and field methods We selected 54 wetlands for this study, located within the Till Plains and the Glaciated Plateau Ecoregions of Ohio. The landscape is dominated by intensive row-crop agriculture, mostly soybeans and corn on a two-year rotation. To minimize potential differences between sites not accounted for by the study variables, we chose

Table 1 Seven pond-breeding amphibian species studied and predictions of the strength of association with the amount of forest within the core zone (200 m radius) surrounding the breeding ponds Species

Prediction of the strength of association with the amount of forest within the core zone surrounding the wetlanda

Spotted salamander (Ambystoma maculatum) Marbled salamander (A. opacum) Jefferson salamander complex (A. jeffersonianum complex) Wood frog (Rana sylvatica) Smallmouth salamander (A. texanum) Eastern red-spotted newt (Notophthalmus v. viridescens) Tiger salamander (Ambystoma tigrinum tigrinum)

High High High High Medium Medium No

a

Natural history characteristics on which these predictions were based are described in text.

D. Porej et al. / Biological Conservation 120 (2004) 399–409

401

eocids, salmonids and ictalurids) and non-predatory (cyprinids) categories. Study sites were grouped geographically to minimize travel distance, but the order of visits to different groups was random within each sampling period.

permanent or semi-permanent depressional wetlands located in landscapes with less than 40% forest landcover within 1 km, and without any recent significant hydrological alterations or impacts to the surrounding terrestrial habitat (e.g., forest removal, urban development). All wetlands had a shallow littoral zone, circumneutral pH, and none dried out before the conclusion of the amphibian surveys in any year. Amphibian surveys were conducted at each site during three periods: 15 March–15 April, 15 May–10 June, and 20 June–10 July from 1997 through 2000. The first survey period was timed to coincide with the onset of spring pond-breeding salamander migration in all years. Presence of amphibian species was assessed using aquatic funnel traps, visual surveys, egg mass identification and dipnetting. Traps were made of aluminum and fiberglass window screen and had funnels at both ends that tapered from a 20 cm diameter to a 4 cm entrance hole. Ten traps were placed equidistantly from each other in shallow water around the perimeter of a site during each visit and were retrieved 24 h after deployment. Visual surveys (two observers) were conducted during each trap deployment and retrieval, and included captures of adults and juveniles, as well as identification of egg masses. Dipnetting for amphibian larvae was done concurrently with funnel trapping at each wetland during the three sampling periods. Dip net sweeps were made in all habitat types for a minimum of 30 min per habitat type. Woody debris and other substrate materials were manually collected and searched for eggs and larvae. Captured adults were identified in the field and released, and larvae that could not be identified in the field were preserved for later identification in the lab. We followed Hecnar and M’Closkey’s (1997) classification of fish into predatory (centrarchids,

2.2. Landuse data Landuse data were obtained from National Land Cover Database and National Wetland Inventory maps using ARC GIS applications (Environmental Systems Research Institute, 1999). Landuse data were verified by field reconnaissance and review of aerial photographs of the wetlands and the surrounding areas taken in the late 1990s. We calculated landuse variables in two regions: core habitat (200 m core zone from the wetland edge) and ‘‘broader landscape context’’ (a buffer zone extending from 200 m to 1 km away from the wetland’s edge). Cumulative length of paved roads was recorded for the entire 1 km radius zone. We used the average Euclidean distance to the five nearest wetlands (of any type or size) as a measure of the relative isolation of the study site from other wetlands. Distance to the five nearest wetlands, length of paved roads within 1 km, and wetland sizes were log10 -transformed. 2.3. Statistical analyses We assessed the correlation between independent variables using the Spearman rank correlation test, and selected variables with pair-wise jrj 6 0:4 for inclusion in the analysis (Tables 2 and 3). Row-crop agriculture (corn and soy beans) was the dominant land use type. It was highly correlated with several other landuse variables and was therefore not included as an independent

Table 2 Description of local and surrounding landscape characteristics of 54 study wetlands Variable

Mean ± SD

F200: Percentage of forest cover within 200 m of the study site Size: log10 of wetland size (ha) Fish: Presence of predatory fish Wet: log10 of average distance to five nearest wetlands Roads: log10 of the length of paved roads within 1 km of the study site For 1 km: Percentage of forest cover in the area extending from 200 m to 1 km from the wetland’s edge Past 1 km: Percentage of pasture and oldfield cover in the area extending from 200 m to 1 km from the wetland’s edge

40.84 ± 26.31 0.07 ± 0.52 n ¼ 17 2.48 ± 0.30 3.46 ± 0.25 20.20 ± 12.07 16.50 ± 10.93

Table 3 Correlation coefficient matrix for independent variables used in the analysis

F200 Size Fish Wet Roads For 1 km

Size

Fish

Wet

Roads

For 1 km

Past 1 km

0.143

0.119 0.130

)0.389 0.062 0.119

)0.287 )0.001 0.086 0.234

0.400 0.099 )0.049 )0.356 )0.226

)0.127 0.067 )0.064 0.322 0.358 )0.081

402

D. Porej et al. / Biological Conservation 120 (2004) 399–409

variable in the analysis. We chose seven variables to represent local wetland characteristics (presence of predatory fish and wetland size), forest coverage within the core (% forest cover within 200 m), isolation (cumulative length of paved roads within 1 km and distance to the five nearest wetlands) and ‘‘broader landscape context’’ (% forest and % pasture). We constructed binary logistic regression models with individual species presence/absence as a response variable, and used Akaike Information Criterion (AIC, Burnham and Anderson, 1998) to select models that best fit our data. In order to avoid data dredging, Burnham and Anderson (1998) stress that AIC should not be used for evaluating all possible models, but a subset of a priori models that the investigator considers being meaningful. In our opinion, current knowledge warrants the use of all seven selected independent variables in the analysis as biologically significant. We therefore constructed a priori binary logistic regression models for each species consisting of a constant and a single explanatory variable (one independent variable models), all pair-wise combinations of seven explanatory variables (two independent variable models), and several models incorporating interaction terms (three variable models). Due to some data dispersion, we used Akaike’s QAICc in the analysis. Akaike distance between models (Di ) is calculated as QAICc ) min QAICc . All models with Di < 2 can be considered potentially equal in making inferences. Model weight ðWi Þ is interpreted as the probability of that model being the best one among the models evaluated. Individual variable weights ðwi Þ are obtained by summarizing the weights of all the models that contained that variable. This value can be interpreted as the relative importance of that variable compared to the other variables in the set. We evaluated final model fit using Wald’s statistics, Hosmer–Lemeshow statistics, and Naglekerke R square (pseudo-R2 for binary logistic regression). Finally, we plotted presence/absence data and the fitted probabilities obtained from binary logistic regression against the variable selected in the most parsimonious model. Analyses were performed using SPSS (1998) and Microsoft Excel. Some of the study species’ current ranges do not cover our entire study area (Pfingsten, 1998; Pfingsten and Downs, 1989; Davis and Menze, 2002). We excluded sites from the individual-species analyses if the species was not recorded in the county within the last 60 years. For the salamander community level analysis, we used these recent distribution maps to calculate the total number of pond-breeding salamander records for the county, and then calculated the %SALAMANDER (number of salamander species recorded at the site/ number of salamander records for the county within which the site was located · 100). %SALAMANDER was then used in the analysis as a dependent variable in

a multiple linear regression model. To determine whether model assumptions were met we performed normality tests on the residuals and examined residual plots. We used the Durbin–Watson test to check for possible autocorrelation between residuals, and calculated the variance inflation factors to check for possible problems with multicolinearity. We found no significant violations of parametric test assumptions.

3. Results 3.1. Spotted salamander Spotted salamanders were recorded at 27% of the sites. The probability of presence was positively associated with the amount of forest within the core zone (Table 4). Average percent of forest cover within the core zone at sites where the spotted salamander was present was 62 ± 4.1% (range 34–99%), and the average for sites where it was not recorded was 26 ± 4.2% (range 0–86%). The amount of forest cover in the core zone, beyond which the fitted probability of recording spotted salamanders was >0.5, is 53% (Fig. 1). AIC analysis indicated that the presence of predatory fish (negative association) and wetland size (negative association) were potential candidates for the best model. 3.2. Smallmouth salamander Smallmouth salamanders were recorded at 54% of the sites. The probability of presence was positively associated with the amount of forest within the core zone (Table 4). Average percent of forest cover within the core zone at sites where the smallmouth salamander was present was 48 ± 4.8% (range 0–99), and the average for sites where it was not recorded was 33 ± 5.1% (range 0– 87). The amount of forest cover in the core zone, beyond which the fitted probability of recording smallmouth salamanders was >0.5, is 33% (Fig. 1). AIC analysis indicated the presence of predatory fish (negative association) and the amount of pasture within the zone from 200 m to 1 km (positive association) were potential candidates for the best model. 3.3. Jefferson’s salamander complex Jefferson’s salamanders were recorded at 43% of the sites. The probability of presence was positively associated with the amount of forest cover within the core zone (Table 4), and negatively associated with presence of predatory fish (present at 50% of sites without predatory fish and 12% of sites with predatory fish). For fishfree sites, the average percent of forest cover within the core zone at sites where Jefferson’s complex salamanders were present was 55 ± 4.7% (range 25–99), and the av-

Table 4 Analysis of logistic binary regression models for amphibian species occurrence in wetlands using QAICc Species

Red-spotted newt

Jefferson’s salamander complexb

Smallmouth salamander

Tiger salamander

Wood frogc

AIC analysis

Binary logistic regression

Model ) 2log (likelihood)

Di

Model weights ðWi Þ

Individual variable weights ðwi Þ

Variables in the selected model

B (SE)

Wald’s statistic

Hosmer–Lemeshow test

Naglekerke R square

Wetð
Þ

24.56

0.00

0.26

Wet

0.98

Wet

)8.37 (3.28)

.011

0.765

0.47

Wetð
Þ + Fishþ Wetð
Þ + Roadsð
Þ Wetð
Þ + SizeðþÞ

22.48 23.44 23.70

0.41 1.31 1.61

0. 21 0.13 0.11

Fish Road Size

0.22 0.15 0.11

F200ðþÞ + Size

ð
Þ

32.53

0.00

0.27

F200

0.92

F200

0.086 (0.024)

0.00

0.579

0.57

F200ðþÞ + Fishð
Þ F200ðþÞ

33.22 36.34

0.61 1.22

0.20 0.15

Size Fish

0.27 0.20

F200ðþÞ þ Fishð
Þ

37.71

0.00

0.55

F200

0.97

F200

0.002

0.226

0.50

F200ðþÞ

43.51

3.35

0.10

Fish

0.56

Fish

0.064 (0.021) )2.88 (1.43)

Past 1 kmþ + Fishð
Þ

67.07

0.00

0.14

F200

0.43

F200

0.023 (0.011)

0.048

0.300

0.10

F200ðþÞ + Fishð
Þ Past 1 kmþ + F200ðþÞ

67.58 67.59

0.51 0.52

0.11 0.11

Fish Past 1 km

0.41 0.40

F200ðþÞ Past 1 kmðþÞ

70.31 71.09

0.90 1.68

0.09 0.06

Roadsð
Þ + F200þ

32.70

0.00

0.28

Road

0.99

Road

)12.85 (3.68)

0.00

0.328

0.65

Roadsð
Þ Roadsð
Þ + Wetð
Þ

35.69 34.02

0.51 1.17

0.22 0.16

F200 Wet

0.28 0.16

F200ðþÞ þ For1kmðþÞ

24.69

0.00

0.27

F200

0.63

F200

0.059 (0.026)

0.23

0.971

0.60

For 1 km Size

0.39 0.18

For 1 km

0.123 (0.058)

0.34

Fish

N/A Fish

N/A

F200ðþÞ + For 1 kmðþÞ + F200*For 1 kmðþÞ

22.79

1.39

0.13

F200ðþÞ + Sizeð
Þ Fishð
Þ

27.22

1.91 N/A

0. 10 N/A

0.045

D. Porej et al. / Biological Conservation 120 (2004) 399–409

Spotted salamander

Modela

Only models with AIC model weights 6 2 are listed (except for Jefferson’s salamander complex), and the most parsimonious model within this group is circled. Positive coefficient values for a given variable in the model are indicated with a (+), negative with an ()). b F200 model included for comparison purposes only ðDi > 2Þ. c Wood frogs were absent from all wetlands with fish, analysis performed on fish-free wetlands only. a

403

404

D. Porej et al. / Biological Conservation 120 (2004) 399–409 Smallmouth salamander Fitted probability of occupancy

Fitted probability of occupancy

Spotted salamander 1

0.5

0 0

25

(a)

50 % Forest within 200m

75

0

Fitted probability of occupancy

Fitted probability of occupancy

25

0 50 % Forest within 200m

75

50 % Forest within 200m

75

100

Tiger salamander

0.5

(c)

0

(b)

Jefferson's salamander complex

25

0.5

100

1

0

1

100

1

0.5

0 3

(d)

3.2 3.4 3.6 3.8 Log (road length within 1km radius)

4

Fitted probability of occupancy

Red-spotted newt 1

0.5

0 2

(e)

2.1

2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 Log (avg. distance to 5 nearest wetladns)

3

3.1

Fig. 1. Fitted logistic-regression models displaying the probability of occurrence of spotted salamanders, smallmouth salamanders, and Jefferson’s complex salamanders as related to forest coverage within 200 m, tiger salamanders as related to the cumulative length of roads within 1 km of the wetland, and eastern red-spotted newts as related to log10 distance to five nearest wetlands. Circles represent individual study sites (0, absent; 1, present), and triangles represent observed proportions of occupied wetlands grouped into four categories based on ranks of forest coverage (spotted and smallmouth salamanders), road length within 1 km (tiger salamanders) and log10 distance to five nearest wetlands (red-spotted newt). For Jefferson’s complex salamanders the fitted logistic regression curve has been created based on fish-free wetlands only (see text).

erage for sites where they were not recorded was 23 ± 4.8% (range 0–66). The amount of forest cover in the core zone around fish-free wetlands, beyond which the fitted probability of recording Jefferson’s complex salamanders was >0.5, is 47% (Fig. 1). 3.4. Marbled salamander Presence of marbled salamanders was recorded at only 7% of the sites, thus our records are insufficient to conduct the analysis.

3.5. Tiger salamander Tiger salamanders were recorded at 46% of the sites. The probability of presence was negatively associated with the length of paved roads within 1 km of the site (Table 4, Fig. 1). Average length of paved roads around sites where tiger salamanders were present was 2091 ± 154m (range 1000–4600 m), and the average for sites where they were absent was 4625 ± 425 m (range 1700–10,000 m). AIC analysis indicated that models including the amount of forest cover within the core

% SALAMANDER is calculated as the number of salamander species recorded at the site divided by the number of salamander species in the county within which the study site was located. F200 + Roads model included for comparison purposes only ðDi > 2Þ.

0.423 0.004 (0.001) )0.043 (0.014) )0.259 (0.084) % Salamander

F200ðþÞ þ Roadsð
Þ þ Fishð
Þ F200ðþÞ + Roadsð
Þ

28.37 33.05

0.00 7.02

0.926 0.027

F200 Roads Fish

0.995 0.976 0.968

F200 Roads Fish

Adjusted R square B (SE) Variables in the selected model )2log (likelihood)

AIC Analysis Model

Table 5 Analysis of linear regression models for %SALAMANDER using QAICc

Di

Model weights ðWi Þ

Individual variable weights ðwi Þ

Linear regression

D. Porej et al. / Biological Conservation 120 (2004) 399–409

405

zone (positive association) and the average distance to the five nearest wetlands (negative association) were potential candidates for the best model. 3.6. Red-spotted newt Red-spotted newts were recorded at 13% of the sites. The probability of presence was negatively associated with the average distance to the five nearest wetlands (Table 4, Fig. 1). Average distance to the five nearest wetlands from sites where red-spotted newts were recorded was 158 ± 16 m (range 100–205 m), and average for sites where it was absent was 466 ± 50 m (range 100–1625 m). AIC analysis indicated that the length of paved roads within 1 km (negative) and presence of predatory fish (positive association) were potential candidates for the best model. Average percent of forest cover within the core zone at sites where newts were present was 56 ± 7.2% (range 36–83), and the average for sites where it was not recorded was 37 ± 4.1% (range 0–99). 3.7. Wood frog Wood frogs were recorded at 43% of the sites. Wood frogs were not present in any wetlands containing predatory fish. At fish-free wetlands the probability of presence was positively associated with the amount of forest within the core zone and the amount of forest within the zone extending from 200 to 1000 m from the site (Table 4). Average percent of forest cover within the core zone at sites where wood frogs were present was 58 ± 5.7% (range 11–99), and the average for sites where they were not recorded was 27 ± 5.1% (range 0–66). Average percent of forest cover within the broader landscape (200–1000 m from the wetland’s edge) at sites where wood frogs were present was 30 ± 3.3% (range 6– 40), and average for sites where they were not recorded was 15 ± 2.1% (range 1–35). AIC analysis indicates that a model including both measures of forest cover and the interaction term between these two variables, and a model including wetland size are also potential candidates for the best model. 3.8. Salamander community The average number of salamander species recorded at our sites was 2.0 ± 0.7 (range 0–4). Average %SALAMANDER (number of salamander species recorded at the site divided by the total number of salamander species recorded in the county) was 47 ± 7.9% (range 0– 100). %SALAMANDER was positively associated with the amount of forest cover within the core zone, and negatively associated with presence of predatory fish and the cumulative length of paved roads within 1 km (Table 5).

406

D. Porej et al. / Biological Conservation 120 (2004) 399–409

4. Discussion Our results confirm the strong association between the structure of surrounding upland areas and amphibian diversity at breeding wetlands. Our expectations about the positive association between the amount of forest within the 200 m core zone and presence of local populations at study sites were confirmed for a majority of species studied. Amount of forest cover within the core zone was included in the most parsimonious models for overall salamander diversity, and individual models for presence of spotted salamanders, Jefferson’s salamander complex, smallmouth salamanders and wood frogs. Presence of spotted salamanders and Jefferson’s salamanders was strongly associated with the amount of forest within the core zone. This result is consistent with their small home range, philopatry, and dispersal behavior of juveniles (Whitford and Vinegar, 1966; Shoop, 1965, 1968; Douglas and Monroe, 1981; Stenhouse, 1985; Madison, 1997; Rothermel and Semlitsch, 2002). Marbled salamanders exhibit some of the same natural history characteristics as the above-mentioned species (Shoop and Dotty, 1972; Douglas and Monroe, 1981; Stenhouse, 1985), but our limited data were insufficient to confirm the same association. The historical range of the marbled salamander encompasses our entire study region, but the distribution of recent records within this portion of the state is extremely patchy (Pfingsten, 1998). Although this species if fairly common within the unglaciated portion of the state (M. Micacchion, unpublished data), we suggest that the preservation of this species within the glaciated region of the state should be given special attention. Smallmouth salamanders were most common at sites with medium amounts of forest cover within the core zone (Fig. 1). Selection of the amount of pasture within the broader landscape context, and presence of predatory fish as potential candidates for inclusion in the best model for this species, suggests that this species can tolerate a wider range of conditions, as long as some pastures and oldfields remain in the vicinity of fish-free wetlands. This is consistent with what we currently know about smallmouth salamander life history in Ohio and Missouri (Pfingsten, 1998; Semlitsch, pers. commun.). The most parsimonious models for overall salamander diversity, and individual models for the presence of red-spotted newts, tiger salamanders and wood frogs, included landuse variables measured outside the core zone. This result indicates that persistence of local populations of these species, and consequently preservation of the amphibian diversity as a whole, depends on the composition of the broader landscape context as well. Distribution of wood frogs has been significantly impacted by historical loss of forests within our study

area (Davis and Menze, 2002). Presence of this species was positively associated with measures of forest cover at both local and broader landscape context, and AIC analysis indicates that the interaction between these two variables can be considered as a potential candidate for the best model. Wood frogs were absent from several sites with high amounts of forest cover within the core zone and low amounts of forest in the broader landscape context. Alternately, several wetlands with relatively low amounts of forest cover within the core zone and relatively high amounts of forest within the broader landscape context had a breeding population of wood frogs. Wood frogs can migrate up to 1.2 km through forested habitats (Berven and Grudzen, 1990). Although their movements are hindered by forest edges (Demaynadier and Hunter, 1998), wood frogs can persist within forested landscapes regardless of whether wetlands are isolated from forests or not (Guerry and Hunter, 2002). Preservation of the forest core zone might not be sufficient to preserve local populations of this species (several isolated natural areas in central Ohio with ample breeding sites and forest within the core zone have lost their populations of wood frogs within the last few decades, pers. observation). Guerry and Hunter (2002) demonstrated very high sensitivity of red-spotted newts to forest fragmentation. This species was also the first one to disappear from fragmented forests studied by Gibbs (1998). These results appear contrary to the natural history and seemingly high dispersal capabilities of this species. Redspotted newts have a terrestrial stage (eft) that may last for up to seven years (Forester and Lykens, 1991), during which they may travel long distances from their natal pond (Gill, 1978a). Gill (1978b) demonstrated the importance of eft-stage migration to the persistence of newts at a metapopulation level, and postulated that individuals colonizing new sites have higher fitness due to density-dependent reproductive success and survival in ponds long inhabited by newts. Gibbs (1998) postulates that this strategy may be detrimental in fragmented landscapes where a significant portion of the juveniles may be lost during the prolonged time they spend in the terrestrial environment. All of our study sites where redspotted newts were present had a relatively high forest landcover within the core zone with many wetlands nearby. However, the absence of red-spotted newts from other seemingly suitable sites (in terms of forest cover) remains somewhat puzzling. Our analysis on a limited dataset suggests that the density of remaining wetlands may be a factor in sustaining red-spotted newt populations, and that the trend in increased inter-pond distances with reduced amount of forest landcover may be a factor in reducing the probability of survival of redspotted newt populations within agricultural landscapes. This species may require a different management approach and deserves further study.

D. Porej et al. / Biological Conservation 120 (2004) 399–409

The probability of presence of tiger salamanders was negatively associated with the length of paved roads within 1 km of the site. Although they will cross paved roads during migration (Porej, pers. observation), some data suggest that tiger salamanders may avoid paved roads if possible (Madison and Farrand, 1998). Roads have been documented to negatively affect amphibian communities in numerous studies (Fahrig et al., 1995; Ashley et al., 1996; Gibbs, 1998). As this is a correlative analysis, unmeasured variables potentially correlated with road density could have influenced the results (e.g., human population density, traffic intensity, feral predators, increase in other impervious surfaces). We suggest that urbanization of agricultural landscape creates terrestrial habitats less conducive to tiger salamander movements and consequently renders local populations more susceptible to extinction. Adult tiger salamanders are flexible in their terrestrial habitat use and have been recorded at sites surrounded almost exclusively by agricultural lands within a 1 km radius (Koloszvary and Swihart, 1999; this study). Although it was not included in the best model for our data, AIC analysis indicates that a higher amount of forest cover within the core zone increases the probability of the presence of tiger salamanders. Little is known about the habitat selection and habitat requirements for juvenile tiger salamanders, and it is possible that increased forest cover within the core zone increases their chances of survival. Knowledge of habitat requirements at different life-history stages for all species would add to our understanding of core terrestrial zones needed to maintain local populations of amphibians. 4.1. Thresholds Our study indicates that there are thresholds of forest cover within the core zone below which local populations of some study species were not present. Whereas smallmouth salamanders were present across the entire spectrum of forest cover within the 200 m core zone, spotted salamanders and Jefferson’s complex salamanders these thresholds had thresholds of 35% and 25%, respectively (36% for a limited number of red-spotted newt records). The interpretation of wood frog data in terms of critical forest cover thresholds is complicated by the association between this species’ presence and the amount of forest within the ‘‘broader landscape context’’. This highlights the importance of considering scale, composition of the matrix, and the overall amount of suitable habitat on the landscape when investigating critical thresholds and minimum patch sizes. The analysis of landscape composition effects on amphibian communities has varied in scale from 500–2500 m radii from the wetland’s edge (Hecnar and M’Closkey, 1997; Knutson et al., 1999; Lehtinen et al., 2001; Guerry and Hunter, 2002). Examination of spatial scales at which

407

amphibian communities respond to landscape composition is complicated by differences in habitat preferences and dispersal abilities between species, as well as differences in the importance of patch size in fragmented versus more continuous suitable habitats. Studies of artificial landscapes based on percolation theory demonstrated non-linear relationships between the proportion of remaining suitable habitat in the landscape and patch size and isolation (Turner, 1989; Gardner and O’Neil, 1991). As the amount of suitable habitat disappears, the size and configuration of remaining patches becomes more important in predicting community structure and diversity (Andren, 1996). Andren (1994, 1996) found that studies on small mammal and bird communities that documented the effect of patch area and/or isolation on species diversity were conducted in landscapes with low percentages of suitable habitat remaining. In contrast, in landscapes with a higher proportion of suitable habitat, small patches were likely to have large patches close by, and hence would not suffer severe species loss. The agricultural landscape of Ohio has less than 8% forest cover, and clearly falls within the category of landscapes with low amount of suitable habitat. In addition, recent changes in agricultural practices in Ohio such as elimination of fencerows, adoption of two-year rotations between corn and soy-beans, and use of feedlots instead of pastures (Lafferty, 1979) have created a very inhospitable matrix for amphibian migration between remnant woodlots. The amount of forest within the core zone within this type of landscape may be more critical for preservation of amphibian diversity than in landscapes with a higher proportion of forest cover or a more hospitable matrix (e.g., Maine, up to 100% forest cover within 1 km radius; Guerry and Hunter, 2002). This issue warrants further investigation.

5. Conservation implications Results of our study stress the importance of perceiving and managing wetlands and the surrounding terrestrial habitat as one ecological unit (Semlitsch, 2000). Data from studies that demonstrate the importance of surrounding upland habitat should be used to make more informed decisions in the process of preserving, creating, or replacing wetlands. The importance of understanding the effects of landscape composition on pond-breeding amphibian communities has been highlighted by the urbanization of former agricultural lands, the increase in the amount of wetland habitat restored/ created through initiatives such as the NRCS Wetland Reserve Program, and wetland replacement (‘‘mitigation’’) under Section 404 and Section 401 of the Clean Water Act. Generally, little

408

D. Porej et al. / Biological Conservation 120 (2004) 399–409

attention is given to the wetland design and evaluation criteria for animals other than waterfowl and other threatened/endangered species when wetlands are created (National Research Council, 1995, 2001). A recent study by Porej (2003) documents that, although forested wetlands form a substantial portion of impacted wetlands, salamanders and wood frogs are very rare or completely absent from wetlands constructed for mitigation purposes in Ohio. Over 85% of mitigation wetlands in that study were placed at locations containing less than 25% forest landcover within the core zone. Based on the results of our study on natural wetlands, we would expect that these constructed wetlands are unlikely to support forest-associated amphibian fauna, even if re-introductions were attempted. New housing and development projects often retain some of the wetlands on the property for aesthetic, biological, or recreational purposes, while substantially modifying the surrounding terrestrial habitat (‘‘on-site’’ wetland preservation). Since hydrology is retained, the functioning of these wetlands is often considered to be intact, but our results suggest that this may not always be the case. Our data contribute to a better understanding of the amount and type of upland habitat necessary to maintain local populations of amphibians, and these findings should be incorporated into plans for ‘‘on-site’’ preservation. Wetlands whose terrestrial surroundings are significantly modified by construction of roads, clearing of forests or other activities should be considered ‘‘impacted’’. These impacts can only be mitigated by constructing replacement wetlands at a more suitable location. Private landowners own over 90% of all the trees in Ohio, mostly in the form of isolated woodlots (Lafferty, 1979). Many of these woodlots contain wetlands, the presence of which may have discouraged their clearing for agriculture or other uses. Although they may harbor significant amphibian diversity, it is unlikely that we will have an opportunity for an on-site identification of all high-quality sites before they are irreversibly lost. We suggest that landscape-level models such as ours, which rely on remote-sensing data (e.g., NALCD, NWI, hydric soils maps, land ownership data), should be used as rapid-assessment tools to: (a) identify remnant woodlots on public and private lands that may potentially harbor a significant biodiversity of organisms characteristic of forested wetland habitats within the Midwestern agricultural landscape, (b) identify sites where forested wetland restoration, creation, or habitat enhancement (possibly through the wetland mitigation process) can be recommended in order to assure the future integrity of existing high-quality woodlots, and (c) identify critical gaps in the distribution of this type of habitat where restoration and/or preservation of remnant habitats should be a priority in order to maintain basic habitat connectivity.

Acknowledgements We thank J. Mack, M. Gray, M. Brady and numerous Ohio EPA interns for their help in fieldwork, and J. Bishop from the Penn State Cooperative Wetlands Research Center for his help in GIS analysis. We thank S. Nikolic, T. Waite, R. Semlitsch, and J. Porej for suggesting improvements to the earlier versions of this manuscript. This research project was funded by US EPA Region 5 Wetland Protection Grants Nos. CD985875 and CD985276 to Ohio EPA, and OSU Department of EEOB R. Osbourn Fellowship to D. Porej. Specimens were collected under permit from Ohio Division of Wildlife and are currently deposited at the Ohio EPA field facility in Groveport, Ohio.

References Addicott, J.F., Aho, J.M., Antolin, M.F., Padilla, D.K., Soluk, D.A., 1987. Ecological neighborhoods scaling environmental patterns. Oikos 49 (3), 340–346. Andren, H., 1994. Effects of habitat fragmentation on bird and mammals with different proportions of suitable habitat: a review. Oikos 71 (3), 355–366. Andren, H., 1996. Population responses to habitat fragmentation: statistical power and random sample hypothesis. Oikos 76 (2), 235– 242. Ashley, E., Robinson, E.P., Jeffrey, T., 1996. Road mortality of amphibians, reptiles and other wildlife on the long point causeway, Lake Erie, Ontario. Canadian Field Naturalist 100 (3), 403–412. Berven, K.A., Grudzen, T.A., 1990. Dispersal in the wood frog (Rana sylvatica), implications for genetic population structure. Evolution 44 (8), 2047–2056. Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference: A Practical Information-theoretic Approach. Springer, New York. Dahl, T.E., 1990. Wetland Losses in the United States 1780’s to 1980’s. US Fish and Wildlife Service, Washington, DC. Davis, J.G., Menze, S.A., 2002. Ohio Frog and Toad Atlas. Ohio Biological Survey Miscellaneous Contributions (6), Columbus, OH, USA. Demaynadier, P.G., Hunter Jr., M.L., 1998. Effects of silvicultural edges on the distribution and abundance of amphibians in Maine. Conservation Biology 12 (2), 340–352. Dodd Jr., C.K., Seigel, R.A., 1991. Relocation, repartiation and translocation of amphibians and reptiles. Are they conservation strategies that work? Herpetologica 47 (3), 336–350. Douglas, M.E., Monroe, B.L., 1981. A comparative study of topographical orientation in Ambystoma (Amphibia: Caudata). Copeia 1981, 460–463. Dunning Jr., J.B., Stewart, D.J., Danielson, B.J., Noon, B.R., Root, T.L., Lamberson, R.H., Stevens, E.E., 1995. Spatially explicit population models: current forms and future uses. Ecological Applications 5, 3–11. Dunning Jr., J.B, Danielson, B.J., Pulliam, H.R., 1992. Ecological processes that affect populations in complex landscapes. Oikos 65, 169–175. Duellman, W.E., 1954. Observation on autumn movements of the salamander Ambystoma tigrinum tigrinum in southeastern Michigan. Copeia, 156–157. Environmental Law Institute, 1998. Ohio’s Biological Diversity: Strategies and Tools for Conservation. Environmental Law Institute, Washington, DC.

D. Porej et al. / Biological Conservation 120 (2004) 399–409 Environmental System Research Institute, 1999. ArcView GIS version 3.2. Environmental System Research Institute, Redlands, CA. Fahrig, L.P., Pope, J.H., Taylor, S.E., Wegner, P.D., John, F., 1995. Effect of road traffic on amphibian density. Biological Conservation 73 (3), 177–182. Forester, D.C., Lykens, D.V., 1991. Age structure in a population of red-spotted newts from the Allegheny Plateau of Maryland. Journal of Herpetology 25, 373–376. Gardner, R.H., O’Neil, R.V., 1991. Pattern, process, and predictability: the use of neutral models in landscape analysis. In: Turner, M.G., Gardner, R.H. (Eds.), Quantitative Methods in Landscape Ecology. Springer, New York. Gibbs, J.P., 1998. Amphibian movements in response to forest edges, roads, and streambeds in southern New England. Journal of Wildlife Management 62 (2), 584–589. Gill, D.E., 1978a. The metapopulation ecology of the red-spotted newt, Nothphthalmus viridescens (Rafinesque). Ecological Monographs 48, 145–166. Gill, D.E., 1978b. Effective population size and inter-demic migration rates in a metapopulation of the red-spotted newt, Nothphthalmus viridescens (Rafinesque). Evolution 32, 839–849. Guerry, A.D., Hunter Jr., M.L., 2002. Amphibian distributions in landscape of forests and agriculture: an examination of landscape composition and configuration. Conservation Biology 16 (3), 745– 754. Hecnar, S.J., M’Closkey, R.T., 1997. The effects of predatory fish on amphibian species richness and distribution. Biological Conservation 79, 123–131. Kleeberger, S.R., Werner, J.K., 1983. Post-breeding migration and summer movement of Ambystoma maculatum. Journal of Herpetology 17, 176–177. Knutson, M.G., Sauer, J.R., Olsen, D.A., Mossman, M.J., Hemesath, L.M., Lannoo, M.J., 1999. Effects of landscape composition and wetland fragmentation on frog and toad abundance and species richness in Iowa and Wisconsin, USA. Conservation Biology 13, 1437–1446. Knutson, M.J., Sauer, J.R., Olsen, D.A., Mossman, M.J., Hemesath, L.M., Lanoo, M.J., 2000. Landscape associations of frog and toad species in Iowa and Wisconsin, USA. Journal of the Iowa Academy of Science 107 (3–4), 134–145. Koloszvary, M.B., Swihart, R.K., 1999. Habitat fragmentation and the distribution of amphibians, patch and landscape correlates in farmland. Canadian Journal of Zoology 77, 1288–1299. Laan, R., Verboom, B., 1990. Effects of pool size and isolation on amphibian communities. Biological Conservation 54, 251–262. Lafferty, M.B., 1979. Ohio’s Natural Heritage. The Ohio Academy of Science, Columbus, OH. Lannoo, M.J., 1998. Status and conservation of Midwestern amphibians. In: Lanoo, M.J. (Ed.), Status and Conservation of Midwestern Amphibians. University of Iowa Press, Iowa City, IA. Lehtinen, R.M., Galatowitsch, S.M., Tester, J.R., 2001. Consequences of habitat loss and fragmentation for wetland amphibian assemblages. Wetlands 19 (1), 1–12. Madison, D.M., 1997. The emigration of radio-implanted spotted salamanders, Ambystoma maculatum. Journal of Herpetology 31, 542–552. Madison, D.M., Farrand, L., 1998. Habitat use during breeding and emigration in radio-implanted tiger salamanders, Ambystoma tigrinum. Copeia 2, 402–410. Marsh, D.M., Trenham, P.C., 2001. Metapopulation dynamics and amphibian conservation. Conservation Biology 15, 40–49.

409

Minton Jr., S.A., 1972. Amphibians and Reptiles of Indiana. Indiana Academy of Sciences, Indianapolis, IN. National Research Council, 1995. Restoration of aquatic ecosystems: science, technology and public policy. National Academy Press, Washington, DC. National Research Council. 2001. Compensating for wetland losses under the Clean Water Act. National Academy Press, Washington, DC. Petranka, J.W., 1998. Salamanders of the United States and Canada. Smithsonian Institution Press, Washington, DC. Pfingsten, R.A., 1998. Distribution of Ohio Amphibians. In: Lanoo, M.J. (Ed.), In Status and Conservation of Midwestern Amphibians. University of Iowa Press, Iowa City, IA, pp. 221–255. Pfingsten, R.A., Downs, F.L., (Eds.), 1989. Salamanders of Ohio. Bulletin of the Ohio Biological Survey 7(2). College of Biological Sciences, The Ohio State University, Columbus, OH. Porej, D.P. 2003. Faunal aspects of wetlands restoration. Ph.D. Thesis, The Ohio State University, Columbus, OH. Rothermel, B.B., Semlitsch, R.D., 2002. An experimental investigation of landscape resistance of forest versus old-field habitats to emerging juvenile amphibians. Conservation Biology 16 (5), 1324–1332. Seigel, R.A., Dodd Jr., D.C., 2002. Translocations of amphibians: provem management method or an experimental technique. Conservation Biology 16 (2), 552–554. Semlitsch, R.D., 1998. Biological delineation of terrestrial buffer zones for pond-breeding salamanders. Conservation Biology 12, 1113– 1119. Semlitsch, R.D., 2000. Principles of management of aquatic-breeding amphibians. Journal of Wildlife Management 64 (3), 615–631. Semlitsch, R.D., 2001. Core habitat, not buffer zone. National Wetlands Newsletter 23 (4), 9–11. Semlitsch, R.D., Bodie, J.R., 2003. Biological criteria for buffer zones around wetlands and riparian habitats for amphibians and reptiles. Conservation Biology 17 (5), 1219–1228. Shoop, C.R., 1965. Orientation of Ambystoma maculatum: movements to and from breeding ponds. Science 149, 558–559. Shoop, C.R., 1968. Migratory orientation of Ambystoma maculatum movements near breeding ponds and displacement of migrating individuals. Ecology 55, 440–444. Shoop, C.R., Dotty, T.L., 1972. Migratory orientation by marbled salamanders (Ambystoma opacum) near a breeding area. Behavioral Ecology 7, 131–136. Smith, P.W., 1961. The amphibians and reptiles of Illinois. Illinois Natural History Bulletin 28 (1), 1–298. SPSS, 1998. SYSTAT, Version 8.0. SPSS, Chicago, IL. Stenhouse, S.L., 1985. Migratory orientation and homing in Ambystoma maculatum and Ambystoma opacum. Copeia 1985, 631– 637. Thompson, E.L., Gates, J.E., Taylor, G.J., 1980. Distribution and breeding habitat selection of the Jefferson salamander Ambystoma jeffersonianum in Maryland, USA. Journal of Herpetology 14 (2), 113–120. Trenham, P.C., Marsh, D.M., 2002. Amphibian translocation programs: reply to Seigel and Dodd. Conservation Biology 16 (2), 555– 556. Turner, M.G., 1989. Landscape ecology: the effect of pattern and process. Annual Review of Ecology and Systematics 20, 171–197. Whitford, W.G., Vinegar, A., 1966. Homing, survivorship, and overwintering of larvae in spotted salamanders, Ambystoma maculatum. Copeia 1966, 515–519.