Final Report: Effects of Forest Fragmentation on Community Structure and Metapopulation Dynamics of Amphibians

EPA Grant Number: R827642
Title: Effects of Forest Fragmentation on Community Structure and Metapopulation Dynamics of Amphibians
Investigators: Johnson, Lucinda , Boone, Randall , Breneman, Dan , Gross, John , Johnson, Catherine , Olker, Jennifer H.
Institution: University of Minnesota - Duluth , Colorado State University
EPA Project Officer: Packard, Benjamin H
Project Period: December 1, 1999 through November 30, 2002
Project Amount: $769,623
RFA: Ecological Indicators (1999) RFA Text |  Recipients Lists
Research Category: Ecological Indicators/Assessment/Restoration , Ecosystems

Objective:

The objective of this research project was to quantify the effects of forest fragmentation on amphibian community structure and population dynamics in vernal pool ecosystems. The specific objectives were to: (1) quantify the manner and extent to which forest fragmentation influences amphibian and invertebrate community structure; (2) assess the extent to which regional and local-scale indices reflect fundamental structural properties of vernal pool habitats and their biotic communities and, conversely, the extent to which amphibian and invertebrate community structure reflects local and landscape properties; and (3) develop predictive models to quantify the extent to which forest fragmentation influences the metapopulation dynamics of woodland amphibians and predict the consequences of landscape change on these metapopulations.

Summary/Accomplishments (Outputs/Outcomes):

In the forested regions of the Great Lakes, fragmentation is accelerating as timber supplies in western states diminish (Christian, et al., 1992). Forest fragmentation reduces woodland habitat, isolates habitat patches, changes surface water flow patterns, and alters microclimates, which can have concomitant effects on the light regime, hydrology, and nutrient dynamics. Such stressors may result in both population- and community-level responses from aquatic invertebrates and amphibians. Aquatic invertebrates have a long history of use as indicators of environmental conditions in streams and lakes (e.g., Plafkin, et al., 1989; Karr and Chu, 1997), but invertebrate communities of vernal pools are poorly known, and little is known about the responses of invertebrates to stressors impacting temporary wetlands. The impacts on amphibians include increased mortality and decreased recruitment (e.g., from direct habitat loss, road traffic, increased predation, and dessication) and more rapid extinctions of metapopulations resulting from patch isolation. Woodland amphibians requiring vernal pools for breeding and larval development may be especially susceptible to such effects (Gibbs, 1998). Because amphibians are known to be affected by many stressors (both direct and indirect) associated with fragmentation (Green, 1997; Ovaska, 1997; Waldick, 1997), these animals should be robust ecosystem indicators of the effects of forest fragmentation that can be applied across geographic regions.

Effective ecological monitoring requires specific knowledge of the target ecosystem and the factors that regulate ecosystem dynamics. Indicators that accurately characterize ecosystem integrity must integrate those aspects of landscapes that influence the movement of energy, materials, and organisms within and between ecosystems; the local availability of resources; the structure of biological communities; and the composition and juxtaposition of ecosystem elements. The identification and quantification of these factors is essential for developing robust indices that cross geographic regions. Although many useful databases and analytical methods now are available to assess ecological integrity at multiple spatial scales, the power of various indices and the usefulness of various data for capturing essential components of landscapes that influence woodland amphibian communities have not been quantified adequately.

Methods

Study Area. The study area is located in three counties in northern Minnesota near Duluth, Grand Rapids, and Cloquet and three ecoregions: the Northern Superior Uplands, the Western Superior Uplands, and the Northern Minnesota Drift and Lake Plains (see Figure 1). The region is dominated by the glacial landscapes of the Great Lakes region, including many lakes, wetlands, and peatlands. Land cover primarily is boreal hardwood forest, including birch, maple, and aspen. The natural landscapes are fragmented by areas of tree harvest, small-scale agricultural or pasture plots, and residences.

Figure 1. Project Study Area

Site Selection. Groups of small palustrine wetlands in the three regions were identified as using National Aerial Photography Program high altitude aerial photographs, U.S. Geological Survey 1:24,000 scale topographic maps, and U.S. Fish and Wildlife Service National Wetlands Inventory (NWI) maps. Two clusters of vernal pools, one located in a relatively unfragmented forest patch and one located in a more fragmented landscape, were identified in each region. A cluster was defined as a minimum of four vernal pools within 2 km of one another (the mean dispersal distance for wood frogs based on Berven and Grudzien, 1990) and located within the same type of landscape (topography, geology, soils, and dominant forest type).

Selected fragmented treatment pools were significantly different from unfragmented pools based on land use at the scale of 500-m buffers surrounding the pools. Unfragmented sites had higher proportions of forest and lower proportions of agriculture and pasture. Forest patches were more highly interconnected and contained higher proportions of core areas than fragmented sites. Residential development did not differ between treatment areas, but was relatively low across the region. One cluster, located near Cloquet, MN, was the site of an intensive study to document population dynamics and dispersal across the landscape.

Spatial Data. Pertinent digital spatial data were acquired or digitized for the study region, including land use, land cover, hydrography, roads, digital elevation maps, NWI maps, and soil coverages. Low-level, infrared aerial photos were acquired for the area within 5 km of each wetland cluster, scanned, georeferenced, and used to visually interpret high-resolution land cover mapping within 1 km of each site. This mapping was field checked, and a formal accuracy assessment was performed, with 110 polygons randomly selected and field checked. Land cover mapping was summarized within concentric buffers ranging from a 10 m-1 km radius from each study site. Landscape structure attributes were quantified using Fragstats (McGarigal and Marks, 1995). Land cover (satellite derived), stream density, road density, population density, wetlands, and elevation also were summarized for 2-km and 5-km buffers around each pair of fragmented and unfragmented study areas, allowing for a broader regional analysis of the study areas.

Field Methods. Amphibians were sampled using nighttime calling surveys, daytime dipnet surveys, and minnow traps at 37 pools during 3 surveys in 2000-2001 and 2 surveys (April and June) in 2002. Egg mass surveys also were conducted in the spring of 2001 and 2002. For dispersal studies, eight pools were encircled with a drift fence of black plastic, and pitfall traps were installed on both sides of the fence. All captured amphibians were identified and examined for malformations prior to their release; wood frogs (Rana sylvatica) and blue-spotted salamanders (Ambystoma laterale) were weighed, measured, and given a unique mark to allow individual identification during subsequent recaptures. The macroinvertebrate community associated with the temporary pools was sampled in triplicate using a D-frame (500 µm) kick net during spring (mid-April to mid-May) and late summer (July-August). Macroinvertebrates were identified, enumerated, and the total number of individuals per sample was recorded. Quantitative vegetative sampling was conducted in, and surrounding each, study wetland. A stratified-random design was used to sample vegetation within each vernal pool. Percentage cover estimates were determined for the lowest possible taxonomic level. Aquatic vegetative cover was measured in 4 x 4 m plots nested with 1 m x 1 m subplots with percent cover estimates of emergent vegetation (percent emergents). Percent understory cover (percent shrubs) was measured by calculating the area covered by shrub crowns. Canopy cover was measured in each quadrant of a plot using a spherical (convex) densiometer (Forest Densiometers model-A). Large wood was sampled along six transects. Diameter, length, decay class, and species (if known) were recorded for all logs (> 5 cm diameter and > 0.5 m length) intersecting the transect.

Physical habitat parameters that were measured included: water level, ultraviolet (UV) and photosynthetically active radiation (and UV attenuation), temperature, pH, specific conductance, and dissolved oxygen; along with dissolved organic carbon, true color, total suspended solids (TSS), ash-free dry weight of TSS, chlorophyll-a (chl-a), phaeopigment-a, and turbidity (NTU).

The findings were as follows:

• Higher amphibian species richness was found at fragmented versus unfragmented sites.

Hyla versicolor, Hyla spp., and Pseudacris triseriata most often were found in fragmented versus unfragmented sites. Canopy cover was correlated negatively with the occurrence of these species.

• Invertebrate (primarily aquatic insect) taxa richness was greater, and insect predators in particular were the larger proportion of the community in fragmented versus unfragmented sites.

• Increased proportion of open canopy in the vicinity of the pool was associated with a number of pool habitat characteristics: higher median and coefficient of variation of temperature, larger accumulated temperature degree days, and higher diversity and species richness of vernal pool vegetation.

• There were no differences in the vernal pool vegetation communities at the scale of plots or pools in fragmented versus unfragmented clusters. This likely is because of the presence of largely intact fringing forest surrounding vernal pools, even in fragmented regions.

• The peak dispersal date for wood frogs from pools in the unfragmented sites at the intensively studied Cloquet clusters was approximately 2-3 weeks later than those in the more fragmented landscape, despite a relatively synchronous peak in breeding across all pools. No difference was observed for blue spotted salamanders from the limited number of pools where successful recruits were produced.

• Early pool drying likely was responsible for the lack of recruitment of blue spotted salamanders during 2001.

• Emigration patterns of R. sylvatica differed among pools, but there was no evidence that initial dispersal of emerging metamorphs was focused in a particular compass direction across sites. The direction of dispersal from individual pools was not significantly different from random in the two unfragmented pools or another pool whose perimeter is surrounded by woodland. The preferred direction of dispersal, however, was away from adjacent open areas (fields) and toward adjacent woodland in two pools in the fragmented cluster.

• Slope was useful in describing wood frog movement within the Cloquet pools, as was the satellite image for spring wetness. Our final model described 77 percent (i.e., r2 = 0.77; p < 0.001) of the variation in observed wood frog movements.

• On sunny days, canopy vegetation reduced the total amount of UV-B radiation penetrating through the top few centimeters of the water column almost as much as dissolved and particulate substances in the water column, thereby protecting amphibians from exposure to potentially dangerous levels of UV-B radiation.

Amphibian Community Structure

A total of 13 amphibian species were collected across the study region during calling, dipnet, and trapping surveys (see Table 1). One additional species, Plethodon cinereus, was captured in pitfall traps in Cloquet. We found that both region and treatment effects were highly significant (p < 0.000; p < 0.007), but there was no interaction between treatment and region. We also tested for differences in cumulative total species richness and again found significant differences from both treatment and region, but no interaction. Post hoc tests show strong differences in total species richness between Grand Rapids and the other two regions; Duluth was not different from Cloquet. Higher amphibian species richness was found at fragmented versus unfragmented sites.

Table 1. Summary Statistics for Amphibian Species Richness by Year and Region

2000
2001
2002
Treatment
Cloquet
Duluth
Gr. Rapids
Cloquet
Duluth
Gr. Rapids
Cloquet
Duluth
Gr. Rapids
Frag
Mean
1.8
3.0
3.0
2.0
3.0
5.7
1.8
2.5
3.9
SD
1.7
1.5
1.5
2.2
1.7
0.8
0.5
1.2
0.9
N
4
6
7
4
6
7
4
6
7
Unfrag
Mean
1.8
1.8
2.8
0.3
2.3
3.0
0.8
2.5
2.2
SD
1.3
1.2
0.8
0.5
1.1
1.2
0.5
1.4
0.8
N
4
11
5
4
11
5
4
11
5

Hyla sp., H. versicolor, and P. crucifer explained the majority of the variation in the amphibian community data set based on principal components analysis. Their presence was highly negatively correlated with percent canopy cover and percent litter, and was positively correlated with temperature variables such as the coefficient of variation and thermal degree days. Hyla sp., H. versicolor, and P. triseriata were strongly associated with the fragmented sites, but P. crucifera displayed no clear pattern in its distribution. The remaining species were either very rare or occurred at almost all sites (R. sylvatica).

Invertebrates

The macroinvertebrate community consisted of 114 taxa identified to the generic level, with the remainder identified to family and order. Aquatic insects dominated the macroinvertebrate community, although Gastropoda and Crustacea formed a significant portion of overall abundance. Total taxa plotted by percent forested land cover (1-km radius) showed a strong negative correlation (see Figure 2a). As percent wooded land cover increased, the total number of macroinvertebrate taxa steadily declined. The relationships between total number of predator taxa and percent wooded upland (500-m radius) were distinct and similar to total taxa (see Figure 2b). A steady decline in predator taxa occurred as percent wooded land cover increased in the 500-m buffer surrounding each pool. The open canopy and smaller wood accumulations, in addition to deeper water depths of the fragmented pools, provide a different habitat than the smaller, more shaded confines of the unfragmented temporary pools. The additional niches associated with the fragmented treatments provide a resource that potentially attracts a greater diversity of taxa than the unfragmented treatments.

Figure 2a. Distribution of the Total Number of Macroinvertebrate Taxa Plotted Against the Percent of Forested Landscape at a 1-km Radius

Figure 2b. Distribution of the Mean Number of Macroinvertebrate Predator Taxa Plotted Against the Percent of Wooded Upland Buffered at a 500-m Radius. Taxa richness of both categories is negatively related (slope = - 0.2, p < 0.001) to increasin percent cover.

We demonstrated that aquatic invertebrates, especially insects, respond to changes in forest fragmentation by increasing diversity and abundance, particularly in predator taxa. Increased proportion of open canopy in the vicinity of the pool was associated with a number of pool habitat characteristics: higher median and coefficient of variation of temperature, larger accumulated temperature degree days, and higher diversity and species richness of vernal pool vegetation. Interestingly, water column chlorophyll a was slightly lower with greater open canopy, probably because of the shading effects of emergent vegetation. In addition, no significant relation was observed between canopy cover and water volume change. Aquatic insects were more abundant, and the communities were more taxonomically diverse in these pools. The increases in predator taxa observed in the temporary pools in fragmented versus unfragmented pools suggest that this group is taking advantage of the increased food supply. Increased temperatures, particularly earlier in the season, may partially account for this response.

Vegetation Communities

Diversity, evenness, and diversity measures were calculated for vegetation within the vernal pool plots and summarized across sites. A two-way analysis of variance was conducted to examine the effects of treatment and region on these vegetation metrics. There was no significant effect as a result of treatment, but a small effect was noted because of the region for species richness (p < 0.02) and diversity (P < 0.02), but not evenness. No interaction effects were noted. The lack of treatment effect likely is because of the presence of largely intact fringing forest surrounding vernal pools across both fragmented and unfragmented regions (i.e., fragmentation occurs at the scale of the landscape rather than at the scale of individual vernal pools).

Amphibian Movement and Dispersal

Blue-spotted salamanders were captured at 16 of the 36 sites in the overall study, including 1 site in Duluth, 7 of 12 sites in Grand Rapids, and all of the 8 pools receiving more intensive study in Cloquet. Sixty percent of the captures in Cloquet were accounted for by one pool, CU5, with two other pools accounting for another 30 percent. The age ratio of captures was highly skewed, with more than one-half of the captures comprised of adults; very few juveniles were captured.

Of the eight Cloquet pools, at least five showed evidence of breeding during all 3 years (i.e., capture of breeding adults or larval salamanders). Migration to breeding pools took place very early in the season, well prior to ice out, and lasted for approximately 1 month. Breeding chronology did not appear to differ substantially across sites. Unlike the wood frog (see below), dispersal chronology for these salamanders was similar for both the fragmented and unfragmented sites. Approximately 13 percent of the blue-spotted salamanders encountered in this study were recaptured within or among sites or years. Eighty-four percent of those recaptures were comprised of adults; however, 20 individuals originally captured as metamorphs were recaptured, 2 as juveniles and 18 as adults; the vast majority of these were recaptured at their natal site (CF3 or CU5).

We examined population dynamics and juvenile dispersal and movement patterns of R. sylvatica at eight pools in Cloquet from 2000-2002 to assess whether the movements of dispersing wood frogs are random across the landscape, if a directional component of dispersal may be influenced by fragmentation of woodland habitat, and whether fragmented habitat could effect the potential for migration of individuals among local populations. More than 14,700 individual wood frogs were captured during the study, with the majority (60 percent) initially caught as metamorphs; adults made up only 14 percent of the total. Five sites had evidence of breeding (i.e., egg masses) in both 2000 and 2001, and wood frogs bred in an additional pool in 2001. Wood frogs successfully metamorphosed and dispersed from three pools in both years, and two pools during 2001 only. Approximately 80 percent of the captured metamorphs were considered to be dispersing from their natal pools, and the remaining individuals were considered transients from wetlands outside the study pools.

Metamorph dispersal chronology varied across sites and years, with dispersal generally beginning earlier in the fragmented wetlands. The peak dispersal date for pools in the unfragmented group was approximately 2-3 weeks later than those in the more fragmented landscape, despite a relatively synchronous peak in breeding across all pools. The number of dispersing metamorphs also varied across sites and years.

More than 6,000 unique dispersing metamorphs were captured during this study. Only 2.3 percent of those metamorphs, however, were recaptured in subsequent years, and only 1.6 percent were recaptured at a different site than that from which they emerged (either during the same or subsequent years, Table 2). The percentage of recaptures varied across sites, with recaptures at different sites ranging from 0.1 percent for dispersers from CF3 to 2.8 and 2.9 percent for dispersers from CU5 and CU3, respectively.

Table 2. Movements of Dispersing R. sylvatica Metamorphs Between Years and Sites

Site of Origin
No. of Dispersers
No. (%) Recaptured at a Different Site
No. Recaptured in Subsequent Year(s)
2000
2001
Total
CF3
764
1,903
2,667
3 (0.1%)
40 (1.5%)
CF4
0
279
279
6 (2.2%)
15 (5.4%)
CF5
0
129
129
2 (1.5%)
2 (1.5%)
CU3
286
2,445
2,731
76 (2.8%)
54 (2.0%)
CU5
474
171
645
19 (2.9%)
37 (5.7%)
All Sites
1,524
4,927
6,451
106 (1.6%)
148 (2.3%)

R. sylvatica emigration patterns also differed among pools, with no evidence that initial dispersal of emerging metamorphs was focused in a particular compass direction across sites. The direction of dispersal from individual pools was not significantly different from random dispersal in the two unfragmented pools (CU3 and CU5) or in CF4, a pool whose perimeter is surrounded by woodland for a radius greater than 100 m. Dispersal from both CF3 and CF5, however, did show a significant directional component (CF3: r = 0.81, p = 0.02; CF4: r = 0.78, p < 0.03); in both cases, the preferred direction was away from adjacent open areas (fields) and toward adjacent woodland.

Amphibian Movement Model. Changes in land cover are hypothesized to affect the rates of dispersal of juvenile amphibians and movements of adults between pools. Quantifying the importance of the hypothesized causes of population declines, and policies meant to ameliorate losses, would be made more complete using metapopulation models. We developed a metapopulation model linking wood frog (R. sylvatica) production rates with movement probabilities using diffusion modeling. Diffusion models, and ecosystem models in general, often require land cover or vegetation maps. Creating land cover maps can be costly, error prone, and subjective, requiring the selection of a classification system. Spectral signatures in satellite images contain more information than land cover maps, and they are relatively inexpensive, updated frequently, and objective. We sought to create a diffusion model for wood frogs in Minnesota using satellite images as a primary input layer. Our model was constructed in Visual Basic (Microsoft Corporation, Redmond, WA) and described a correlated random walk diffusion model. The program, entitled Individual-Based Movement, was written in a flexible way, and has been used to model movements of other vertebrates such as mule deer (Odocoileus hemionus) in Colorado. Amphibian movements are represented using correlated random walk diffusion. Animals disperse from randomly selected locations within vernal pools and then move in a random walk (i.e., move in any of eight directions [queen's move] or remain at the current location) an adjustable number of moves, based on habitat attributes as described below. Another parameter controlled the strength of correlation in the movements, which is the likelihood that an amphibian traveling in a given direction continues moving in that direction. A final parameter added a uniform random noise to attributes determining movements. Euclidean distance between pools explained a large proportion of variation in frog movements (Poisson regression, r2 = 0.46, p < 0.001, df = 54 in all analyses). As expected, increasing distance decreased the likelihood that frogs would move between pools (i.e., y = 4.6054 - 0.0056 x), where x represents the distance in m and y represents the number of wood frogs. Wood frog movement was modeled separately using the tasseled cap images for spring brightness, leaf-on greenness, and fall brightness, as well as a linear relationship between image values and suitabilities. In each case, the results were not as robust as using Euclidean distance alone.

Slope was useful in describing wood frog movement within the Cloquet pools, as was the tasseled cap satellite image for spring wetness. A series of simulations were made to progressively improve model fit. Our final model described 77 percent (i.e., r2 = 0.77; p < 0.001) of the variation in observed wood frog movements. Regardless, when the two surface data components in the final model were modeled separately, slope described 68 percent of the variation, and spring wetness described 33 percent (this is less variation explained than when surfaces were ignored, so wood frog movements were best described using a true interaction involving two surfaces). In this model, slope was related to movements in a complex way. Landscape patches became gradually less suitable to wood frogs as slope declined, with a discontinuity at slopes that approached zero or flat. In contrast, spring wetness did not play a role unless sites were quite dry.

We conclude that a simple correlated random walk model using two data surfaces explained a great deal of variation in wood frog movements between eight pools in Cloquet; only 23 percent of the observed movements remained unexplained. The explanatory power is more surprising considering that in the model the number of animals dispersing from each of the pools was equal (i.e., 10,000), but the number actually dispersing from pools was almost certainly unequal.

Hydrology and Habitat

Although a few study sites retained water for only a few weeks, most temporary pools sampled in this study were categorized as "medium" size pools (retaining water for approximately 4 to 6 months) as described by Higgins, et al. (1996). On average, changes in water depth between treatments followed similar patterns, as all pools experienced a normal decrease in depth as the sampling season progressed. Unfragmented pools were smaller on average, and initial size varied less than fragmented pools. Fragmented pools, however, consistently lost more water volume, as a percent of total volume, than unfragmented pools as the sampling period progressed. By the end of the field season, more pools had dried in unfragmented areas than in fragmented areas. This likely is because of a bias towards larger, deeper pools in the fragmented areas.

Hydrologic Model. We demonstrated that a classic water budget model is able to model the depth profiles of boreal forest vernal pools of Minnesota. Modeling hydroperiods for vernal pools in this region, however, would be impractical. Adjustments to parameters allowed us to match well the observed depths of pools in 2000 and 2001 for simulations that began modeling depths in 1950. When modeled depths for 2002 were compared to the observed depths, however, the results were mixed. We believe that more years of depth profiles would allow us to adjust the model and add subtleties that would improve the agreement between modeled and observed depths. We also have demonstrated that extrapolating our hydroperiod modeling to regions (e.g., naively modeling the hydrology of vernal pools within the NWI) is not possible using this method because infiltration rates specific to each vernal pool had to be adjusted. Using the infiltration rates calculated from soil maps directly in a model such as this will not yield reliable depth estimates; some measured depths are required for parameterization.

Modeling hydroperiods for vernal pools within an entire region (e.g., several counties or larger) using the methods described would be impractical because some depth information is required. Reliable measures of infiltration rates and runoff coefficients would improve modeling, but a comparison to observed data still appears necessary. Given the difficulty in measuring those parameters (Pierce, 1993), water depth measures may remain the most useful data that can be collected. Automatic water depth gauges are available, but broad-scale deployment would be extremely expensive. Brief weekly visits by technicians to pools to measure water depth, from the first major snow melt event until the pool is dry or amphibian larvae disperse, may be sufficient to allow a water budget model such as was used here to be constructed for pools across a region.

Light Regime

UV Radiation in Vernal Pools. Amphibian malformations and worldwide population declines have led to many investigations and theories of causation. There currently are several hypotheses for these phenomena, including exposure to pesticides and other chemicals (Oullet, et al., 1997; Sparling, et al., 2001), natural cycles, parasitic infections (Johnson, et al., 1999; Sessions, et al., 1999), and increased exposure to UV radiation (UVR) (Blaustein, et al., 1997; Ankley, et al., 2000; Broomhall, et al., 2000). Numerous studies have been conducted to test the hypothesized effect of UV-B radiation (UV-B = 290-320 nm) on amphibian survival and the rate of malformation with conflicting results. We sought to quantify and model solar radiation attenuation in northern Minnesota vernal pools, and determine the UVR range and potential maximum dose to embryonic and premetamorphic R. sylvatica in northern Minnesota vernal pools.

Water chemistry variables were not significant in predicting diffuse UV-B attenuation (Kd B) for all pools. When pools divided into groups, chl-a and turbidity were significant predictors of Kd B for group 1 (R2adj = 0.98). A single variable model that included only chl-a was not as strong, but still was highly predictive of Kd B. Of all possible variable combinations for group 2, none were significant predictors of Kd B. As hypothesized, UVR was rapidly attenuated within the study vernal pools, with 99 percent of the UV-B attenuated in all pools within the top 35 cm, and 99 percent attenuated within the first 20 cm in most of the pools. Based on recent research, however, on UV-B dose and malformations (Ankley, et al., 2002), a more biologically significant value may be the depth to which 65 percent of surface irradiation can reach (Z65 percent; hereafter known as the "effects level"). Exposure to 63.5 percent (95 percent confidence interval = 58.6-67.6 percent) of full sunlight resulted in 50 percent hind-limb malformations (ED50) of northern leopard frog (R. pipiens) in outdoor microcosm experiments (Ankley, et al., 2002). Z65 percent for UV-B in this study ranged from 0.5 to 3.6 cm, suggesting a narrow range of water depths at which amphibians would experience a high risk of exposure to UV-B radiation.

In addition to the physical characteristics of the atmosphere and pool, actual UVR doses to amphibian embryos and larvae are affected by biological factors such as egg mass placement and larval behavior in pools. Because amphibian embryos and larvae have restricted potential habitats compared to adults, their exposure to UVR can be estimated by measuring UVR levels across a pool and through the water column. We found that the risk of exceeding the effects level was best predicted by water color when chemistry, vegetation, and landscape variables were included in the model.

Overall, we suggest that weather has the greatest individual effect on daily UV-B dose at 1-cm deep. The dissolved and particulate substances in the water column reduced the risk factor by 75 percent (with median water column attenuation rates) or 50 percent (with maximum attenuation rates). Canopy vegetation reduced the risk of exposure to potentially dangerous levels of UV-B on sunny days almost as much as water column components.

R. sylvatica embryos and larvae were at risk of UV-B-induced malformations in the study pools because the incident UV-B effects level was exceeded at the egg mass and other locations during embryonic and larval stages, as hypothesized. Incident daily UV-B dose in CF5 exceeded the effects level from Ankley, et al. (2002) in all of the plots on at least 1 day and on more than one-half of the days in the more open locations. An additional comparison of incident UV-B to maximum incident UV-B (full sun) revealed that nearly all of the pools (during both seasons) had at least one plot with incident UV-B levels greater than the effects level.

Although UV-B dose at the pool surface and at 1 cm deep reached or exceeded the effects level from Ankley, et al. (2002), UV-B dose at depths of 5 cm and greater was very low. Egg masses or larvae at 0 to 2 cm deep are expected to have the highest UV-B exposures, and those at 5 cm and deeper are expected to experience low exposures.

The following conclusions were made:

• Forest fragmentation had statistically significant effects on the physical properties (especially light and temperature) and biotic communities of vernal pools in northern Minnesota.

• Although vernal pools in Minnesota largely are embedded in a forest matrix, and may contain or be surrounded by dense vegetation, UV-B levels are sufficiently high to pose a potential risk to larval amphibians. Management practices that reduce canopy cover will increase potential UV-B exposure and could negatively affect larval survival, further impacting already stressed amphibian populations. Conservation and restoration of riparian vegetation within 100 m of vernal pools could be essential in mitigating potential impacts to these systems.

• Fragmentation effects were observed with respect to wood frog movement and dispersal. The timing of dispersal was advanced in fragmented pools; at a small scale, we observed avoidance of dispersal from pool margins that were not forested versus those that were forested. Furthermore, movement appeared to occur preferentially along wooded corridors linking the vernal pools. Conservation and restoration of wooded corridors between pools should be encouraged.

• Potential indicators of vernal pool condition include light regime, water temperature, and insect and frog community structure.

Journal Articles:

No journal articles submitted with this report: View all 6 publications for this project

Supplemental Keywords:

wetlands, frogs, salamanders, Rana sylvatica, Pseudacris crucifer, Ambystoma laterale, Ambystoma tigrinum, ecosystem protection, environmental exposure and risk, agronomy, ecological effects, ecological indicators, ecology, ecosystem assessment, ecosystem indicators, environmental chemistry, forestry, microbiology, exploratory research, environmental biology, ultraviolet, UV, UV effects, algae, amphibian, anthropogenic stresses, aquatic biota, community structure, ecological exposure, forest fragmentation, frogs, landscape indicator, multiple spatial scales, multiscale assessment, regional scale, risk assessment, vernal pool ecosystems., RFA, Scientific Discipline, Ecosystem Protection/Environmental Exposure & Risk, Ecology, exploratory research environmental biology, Environmental Chemistry, Ecosystem/Assessment/Indicators, Ecosystem Protection, Forestry, Ecological Effects - Environmental Exposure & Risk, Environmental Monitoring, Ecological Risk Assessment, Agronomy, Ecological Indicators, risk assessment, ecological exposure, anthropogenic stresses, wetlands, algae, aquatic biota , landscape indicator, UV effects, frogs, vernal pool ecosystems, amphibian, multiple spatial scales, ecosystem indicators, regional scale, multiscale assessment, forest fragmentation

Progress and Final Reports:

Original Abstract
  • 2000 Progress Report
  • 2001 Progress Report