Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Screening for species potentially sensitive to habitat fragmentation

Screening for species potentially sensitive to habitat fragmentation The increasing fragmentation of the world's hiibitats has important consequences for biodiversity (Heywood 1995, Dudley et al. 1996). For conservation purposes there is thus a strong need for methods that can determine which speeies that are sensitive to fragmentation. A decrease in available habitats due to fragmentation may have different effects on the population sizes of species. In the simplest case, a reduction in habitat only results in a proportionally smaller population size. This means that small fragments can be considered to be random samples of larger ones (Connor and McCoy 1979, Haila 1983), However, in some cases the reduction in population size tnay be larger than predicted from habitat area alone, and these species may be described as experiencing "true" fragmentation effects (Andren 1994), These species are the most sensitive to fragmentation and may be of high priority for conservation efforts. To analyze whether a species" response to fragmentation only relates to habitat loss or also includes "true" fragmentation effects. Andren et al. (1997) developed a model where population size is related to the proportion of habitat in the landscape. They discuss the general situation where an organism may utilize more than one habitat type, e,g. where both the fragment and the matrix are used, and they describe different types of numerical responses to fragmentation. In the case of the random sample hypothesis, a halving of the proportion of preferred habitat would result in a halving of the population size for a species that only uses that habitat. This means that the slope in a log-log plot of population size and proportion of habitat in the landscape will be 1, However, if the species is sensitive to fragmentation, the slope will be > 1 since the relative change in population size will be larger than the relative change in preferred habitat (see Andren et al. 1997 for details). ECOGRAPHY 21:6 (1998) We have applied this model to empirical data on wood-inhabiting lichens, hepatics and fungi from a natural forest-wetland mosaic in northern Sweden, We want to evaluate the usefulness of the model to differentiate between species that are only sensitive to habitat loss from species that also are sensitive to fragmentation effects. In addition, we also discuss the model in relation to sessile organisms (plants) since the Andren et a!, model was based on mobile organims. It should be remembered that our data set was not collected specifically for this purpose and thus we do not aim at evaluating the response of particular species to fragmentation. Methods The data were collected in the nature reserve Granlandet as part of other studies (Kruys and Jonsson 1997. Gustafsson unpubl.. Berglund unpubl.. Moen and Jonsson unpubl.). Granlandet is a forest-wetland mosaic in the northern boreal zone of Sweden located ca 30 km southeast of Nattavara in Giillivarecounty (66°28'-44'N, 21°I5'-5O'E; see Edenius and Sjoberg 1997 and Kruys and Jonsson 1997 for details). The area is 28400 ha as defined by Grundsten et al. (1991) and consists of old-growth Norway spruce Pia-a ahies ssp. obovata moraine patches interspersed in a Sphagnum-dommaitiX wetland matrix. In total, Granlandet consists of > 1000 of these moraine patches (hereafter referred to as islands) that vary greatly in size and shape. There is a large variation in landscape mosaic structure in the area, which we use as a substitute for different degrees of fragmentation. The dominant trees are estimated to be 150-200 yr old. although individual spruces may be up to 300 yr old (Lovgren 1986). Most islands represent late suecessional stands where fire has been very rare or absent. Data on landscape composition were obtained by calculating the forested area within a circle with a 649 200-ni radius around the eentroids (i,c. the center of a N-S E-W rectangle covering the entire island) of the islands used in this study. This was done with a Geographic Information System (ESRl 1997). In the absence of data on dispersal distances in the area, the 200-m radius was chosen for two reasons: a) calculations of the fractal dimension of forested islands for ditTerent pixel sizes (20 x 20 m to 1000 x 1000 m) show a major change between 200 x 200 m and 300 x 300 m {cf. Obeysekera and Rutchey 1997). and b) 200 m is the most common maximum distance from randomly chosen points on mires to the nearest forest edge in Granlandet (Dettki unpubl,). Thus, the 200-m radius is supposed to reflect a scale region where significant structural changes occur in our specific landscape. The data for epiphytic Calicioid lichens (i,e crustose species with "pin""-like apothecia) are from two separate studies. In 1995. data were collected from 24 islands in a size range of 0,4 15,9 ha. In this study, stems of all trees with a diameter a! breast height > 20 cm were analyzed for occurrence of Calicioid lichens from the base up Lo 2 m. On islands < 1 ha. one circular sample plot of 0.1 ha was sampled, while on larger islands three 0,1 ha plots were sampled (Kruys and Jonsson 1997), Additional data on Calicioid lichens were collected in 1996 from 22 small islands (0.7-1.0 ha) and 23 medium-sized islands (3.8-5,6 ha). The procedure was similar to the 1995 study although only one 0,1 ha plot was sampled on each island. Thus, data from a total of 67 separate islands (two islands were included in both studies) were available. Data on woodinhabiting hepatics were also collected in 1996 on the same islands as the lichens. All fallen logs with a base diameter > 15 cm were included. Wood-inhabiting fungi were inventoried in 1997 on 34 small ( < 1 ha) islands in circular 0,1 ha plots. In this study, all coarse woody debris longer than 50 cm and with a base diameter of > 10 cm were included. Eor al! these three species groups, presence, absence data on individual trees and logs in the 0,1 ha plots were collected. Accordingly, "abundance"" estimates are the fraction of trees with occurrence of a particular species within these plots. To estimate the "population size"" in the landscape, this fraction was multiplied by the estimated total number of trees or logs in the 12.5 ha landscapes (200-m radius circles) that were analyzed. The estimated number of trees or logs was based on the proportion of forest within each landscape multiplied by the mean density of standing and fallen trees in Granlandet. which was obtained from the separate studies mentioned above. Linear regression models with logarithmic population size as the dependent variable and logarithmic percent of forested area as the independent variable were constructed for each species. Our H(, hypothesis was that the regression coefficient, p, equals I. relating to a situation were population size decreases proportional to habitat ioss. The t-statistic for significance testing of the regression coefficient is given by t = (b— I),SE,, (with DF = n — 2). where b is the fitted slope parameter and SE(, is the standard error of the slope parameter. Since the null-hypothesis is of biological importance (i.e, that the species is not sensitive to fragmentation), we calculated for the non-significant tests the power to detect an effect size of 0.3 (i,e. slopes of 1,3 or 0,7 respectively) using the observed vai'iance and sample size (Thomas 1997), Results and discussion Six of the 32 species had slopes that were significantly steeper than I (Table I). Five of these were Calicioid lichens (Caliciuni glauccdliini, Chacnolhccii pluwocephuhi. Ch. suhroscida. Ch. irichialis and Micracalicium dis,seminatuin). and one was a hepatic (Anasirophylhim hellerianum). Two of the species showing a slope > I are on the Swedish red-list and classified as caredemanding {Ch, phaeucephahi and A, hdlcrianum; Aronsson et al. 1995). These species are considered as negatively influenced by forestry although the actual causes of decline are not known. Previous data from Granlandet have indicated that some of these red-listed species are more abundant on large forested islands or show negative correlations with edges (Gustafsson unpubl, Moen and Jonsson unpubl.). It is thus likely that the actual cause for the decline of several of these is related to disproportionally smaller core areas in smaller islands. Somewhat surprisingly, two species also had a slope that was significantly lower than I: Caliciuni Irabiiwlhim and Clmcnoiheca xyloxcna ( p < 0 , ] in the latter case). Andren et al, (1997) interpret this to mean that such species are generalists that are able to utilize both habitat types to some extent. However, this could not be the case here since these species cannot use the mire matrix due to lack of trees. We thus suggest that the occurrence of these species is related to the increasing amount of edge present as landscapes become more fragmented. Both of these species are dependent on standing dead trees with hard and dried wood, which may be more common in wind- and sun-exposed edge habitats. This may cause their population sizes to decrease at a slower rate compared to the proportion of forest in the landscape since the proportion of edge actually increases in smaller fragments. We thus suggest that species with slopes > 1 may be seen as specialist species sensitive to fragmentation as was suggested by Andren et al. (1997), but also that species that have slopes < I may be seen either as generalist species that are able to utilize both habitats or as specialist species that are favored by edge habitats. ECOGRAPHV 21: The majority of the species did not show a slope significantly different from I. This could simply imply that fragmentation for these species is nothing but pure habitat loss, i.e, that the null-hypothesis of slope = 1 is true. This could be evaluated through a power analysis (e,g, Rotenberry and Wiens 1985. Peterman 1990), We assumed that a biological significant response to fragmentation includes slopes that deviate from 1 with at least 0.3 (i,e, slopes > 1.3 or <0.7). (This assumption is based on the mean response of the significant tests in our study ( + 0.39). but the reader could perform the power analyses with a different assumed effect size if he/she finds it unreasonable. A slope of 1,3 corresponds to 15'Mi additional loss in abundance compared to a slope of I.) The power gives the probability that the test would find a difference of at least the assumed effect size given the observed variance and sample size. It is not possible to statistically accept a null-hypothesis (Hilborn and Mangel 1997). but if power is high enough (0,95 if we want to assume the same precision as 7) we could say that the test was powerful enough to have picked up the assumed effect size and that the null hypothesis is in a sense "true". In our data set there is only one species with a high enough power to conclude that it is not sensitive to tragmentation {Pulidhiiii pulclicrn'/nuin),. and one species which is close {Caliciuni rlride). The rest of the species are inconclusive due to high variance and/or low sample size. The differences between the studied species groups make biological sense as regards their potential sensitivity to edges and fragmentation. The epiphytic Calicioid lichens are the most exposed group, compared to woodinhabiting hepatics. which grow close to the ground, and wood-inhabiting fungi that grow within logs. Thus, our analysis shows that the approach of Andren et al. (1997) is a simple and potentially useful tool to screen sets of species in order to find those that are sensitive to "true"" fragmentation effects, Ftirther application of the model in other systems is. however, largely dependent on the possibility to replicate landscapes which could be a severe restriction in most systems. The present system might be a special Table I. Statistics from linear rL'gressions of log population size iigainsl log proportion of habitat for 33 species from Granlandet, northern Sweden, Species Calicioid lichens Cdlicium glaucelliim C, lirlictioidc.s C, trabincllum C, viriik Chaenollu'ca clirv.wceplialu Ch, furfiiracca Ch, pluii'Ci'phala Ch, siemonea Ch, subroscida Ch. Irichialis Ch. xyloxena Ch. viridialbci Cypht'lium karcliciim Microcalicium ilisseininaium Wood-inhabiting fungi Anlrodiu .seriali.'; Cnlymnocyslis abiainu Foniiiopsis rosca Gloc'phylhinj scpianim Phcilimis chrysoloma P. JvrrugiiK'ofuscus P, nlgrolimlnam P. vil'icoki STcrcum .uinguilenniiii Trii'luipluin ubictimiiu Wood-inhabiting hepatics Anaslrophylluni hi'llcrictniiin Ci'pludozia bicuspidaia Ci'phulozia hiiiullfoliu Lophozia ascendens L, lougidc'ii'. L, longiflora !., .s-ilric(ilit Pliiiditun pidclwrrimum 0,72 0,64 0,2^ 0,70 0,47 0,21 0-51 0,16 0,66 0,68 0,42 0,77 0,43 0,66 0,61 0,70 0,42 0-30 0.07 0,69 0,-12 0,56 0,72 0,58 0,76 0,45 0.46 0.58 0,60 0,47 0,49 0,85 1,25 1,08 0,55 1,11 0,90 0,71 1,46 0,64 \35 1,30 0.67 1,21 0,86 1,52 1,02 1,03 0,85 0,85 0,24 0.87 0,71 1.06 0,96 0,90 1,49 0,77 0,92 0.99 1,18 1,00 i,OI 0,92 0,10 0.12 0.18 0.09 0,12 0,34 0,26 0,27 0,13 0,12 0,19 0,13 0,19 0,14 0,18 0-14 0,23 0,34 0,25 0,14 0,26 0,24 0,16 0,16 0,17 0,17 0,24 0,16 0,19 0,20 0,21 0,07 61 43 30 63 62 19 33 31 61 60 19 29 30 59 21 23 19 15 12 18 16 16 15 2? 24 24 17 30 26 29 28 32 2,44 0.66 -2.48 1,15 -0,82 -0,87 1,80 -1,30 2,72 2,60 -1,74 1,63 -0,78 3,56 0,14 0-21 -0.65 -0,43 -3,09 -0,94 -1,12 0,27 -0,23 -0,65 2,83 -1.32 -0.32 -0,08 0,96 0,03 0,03 - 1,22 0,018 NS 0,019 NS NS NS 0,081 NS 0,008 0,012 0,098 NS NS 0,001 NS NS NS NS _^NS NS NS NS NS 0.009 NS NS NS NS NS NS NS _ 0,68 _ 0,91 0,69 0.13 0,19 _ _ _ 0.61 0,33 0,36 0,54 0,24 0-13 _ 0.22 0.19 0.22 0.20 0.43 _ 0.40 0.22 0.44 0.33 0.30 0.28 0,99 RSlope Slope SE il t-value Slope p' Power' ' H(,: slope = I, - Model not significant (p = 0.35), ' Power to detect a difference of + 0.3 from ilope = I given obser\ed sample size and \ariance. F.rOGRAPHY 21:6 (199K) 0,00 0,40 0,90 1,90 Log percent forest Fig, I. Examples of linear regressions of log population si7e agamsl log percent forest in the landscape. Boxes and solid line: Caliciuni oirlde Islope not significantly different from 1: y = l,lO6x-h U,9U4). Crosses and dotted line: Microcaliciiiin disseniinatum (slope significantly > 1: y = L516x-I-0,008), and circles and dashed line: Caliciuni irabinelliim (slope significantly < 1 ; y = 0.554x +0,774), case due to the large number of isolated forest patches available, 1!' we plot the detectable difference in slope (from 1) against sample size (given a power of 80% and a standard deviation of 0.94 which is Ihe mean variation in our data set), we see that we need ca 80 replicates to detect a slope of 1.3 (Fig, 2), Regardless of the detectable effect size that we deem is biologically significant. Fig. 2 suggests that we need a large number of replieated landscapes to be able to detect slopes different from 1 if the variation in our data set is in any way representative of varialion in other data sets. If it is difficult to inerease sample size in a study beeause landseapes are limiting, it is imperative that the sampling design is such that estimations of abundance within landscapes are optimized. No. of cases Fig, 2, The detectiibic diffi^rcncc from a slope of 1. given a power of SO"Ai and a standard deviation of 0,94 (which is the mean standard deviiilion in our daia sets), as a function of the number of cases. Connor. E, F, and McCoy. E. D. 1979, The statistics and biology of the species-area relationship, - Am, Nat, 113: 791-833. Dudley, N,. Gilmour, D. and Jeanrenaud. J,-P, 1996, Forests for life, - WWF Int, and IUCN, Gland. Fdenius. L, and Sjoberg, K, 1997, Distribution of birds in natural landscape mosaics of old-growth forests in northern Sweden: relations to habitat area and landscape context. Ecography 20: 425 431, ESRI 1997, ArcView ver, 3,0a, Hnvironmental Systems Research Inst., Redknds, USA, Grundsten, C . Haggbom, J, and Lilja. T, 1991, Omraden av riksintresse for naturvSrd och friluftsliv. Report 3771, Swedish Environmental Protection Agency, Stockholm. Haila. Y. 1983, Land birds on northern islands: a sampling metaphor for insular colonization, Oikos 41: 334-351. Heywood. V. H, (ed,) 1995, Global biodiversity assessment, Cambridge Univ, Press, Hilborn. R. and Mangel. M, 1997, The ecological detective, Confronttng models with data, - Princeton Univ, Press, - We would like lo thank Hakan Rydin for Kruys. N, and Jonsson, B. G, 1997, Insular patterns of comments on an earlier version of the paper. calicioid lichens in a boreal old-growth forest-wetland mosaic, - Ecography 20: 605-613, Lovgren, R, 1986. Urskogar, lnventering av urskogsartade omraden i Sverige, Del 3-Norra Sverige, Report 1509, Swedish Environmental Protection Agency, Stockholm, References Obeysekera, J, and Rutchey, K. 1997, Selection of scale for Everglades landscape models, - Landscape Ecoi, 12: 7-18, Andren, H, 1994. Effects of habitat fragmentation on birds Peterman, R, M. 1990. Statistical power analysis can improve and mammals in landscapes with different proportion of fisheries research and management, - Can, J, Fish, Aq, Sci, suitable habitat: a review, Oikos 71: 355-366, 47: 2-15, Andren, H,. Dclin. A. and Seller, A, 1997, Population rcRotenberry. J, T, and Wiens. .1, A, 1985, Statistical power spotise to landscape charges depends on specialization to analysis and community-wide patterns, - Am, Nat, 125: different landscape elements, - Oikos 80; 193-196. 164-168, Aronsson, M,. Hallingback, T, and Mattson. J.-E, (eds) 1995, Thomas. L. 1997. Retrospective power analysis, - Conserv. Rodlistade vaxter i Sverige 1995 [Swedish Red Data Book Biol. 11: 276-280, of Plants 1995], - Artdatabanken, Uppsala. ECOGRAPHY 21:6 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecography Wiley

Screening for species potentially sensitive to habitat fragmentation

Download PDF
5 pages

Loading next page...
 
/lp/wiley/screening-for-species-potentially-sensitive-to-habitat-fragmentation-Y0cAtqsdyb

References (13)

Publisher
Wiley
Copyright
Copyright © 1998 Wiley Subscription Services, Inc., A Wiley Company
ISSN
0906-7590
eISSN
1600-0587
DOI
10.1111/j.1600-0587.1998.tb00559.x
Publisher site
See Article on Publisher Site

Abstract

The increasing fragmentation of the world's hiibitats has important consequences for biodiversity (Heywood 1995, Dudley et al. 1996). For conservation purposes there is thus a strong need for methods that can determine which speeies that are sensitive to fragmentation. A decrease in available habitats due to fragmentation may have different effects on the population sizes of species. In the simplest case, a reduction in habitat only results in a proportionally smaller population size. This means that small fragments can be considered to be random samples of larger ones (Connor and McCoy 1979, Haila 1983), However, in some cases the reduction in population size tnay be larger than predicted from habitat area alone, and these species may be described as experiencing "true" fragmentation effects (Andren 1994), These species are the most sensitive to fragmentation and may be of high priority for conservation efforts. To analyze whether a species" response to fragmentation only relates to habitat loss or also includes "true" fragmentation effects. Andren et al. (1997) developed a model where population size is related to the proportion of habitat in the landscape. They discuss the general situation where an organism may utilize more than one habitat type, e,g. where both the fragment and the matrix are used, and they describe different types of numerical responses to fragmentation. In the case of the random sample hypothesis, a halving of the proportion of preferred habitat would result in a halving of the population size for a species that only uses that habitat. This means that the slope in a log-log plot of population size and proportion of habitat in the landscape will be 1, However, if the species is sensitive to fragmentation, the slope will be > 1 since the relative change in population size will be larger than the relative change in preferred habitat (see Andren et al. 1997 for details). ECOGRAPHY 21:6 (1998) We have applied this model to empirical data on wood-inhabiting lichens, hepatics and fungi from a natural forest-wetland mosaic in northern Sweden, We want to evaluate the usefulness of the model to differentiate between species that are only sensitive to habitat loss from species that also are sensitive to fragmentation effects. In addition, we also discuss the model in relation to sessile organisms (plants) since the Andren et a!, model was based on mobile organims. It should be remembered that our data set was not collected specifically for this purpose and thus we do not aim at evaluating the response of particular species to fragmentation. Methods The data were collected in the nature reserve Granlandet as part of other studies (Kruys and Jonsson 1997. Gustafsson unpubl.. Berglund unpubl.. Moen and Jonsson unpubl.). Granlandet is a forest-wetland mosaic in the northern boreal zone of Sweden located ca 30 km southeast of Nattavara in Giillivarecounty (66°28'-44'N, 21°I5'-5O'E; see Edenius and Sjoberg 1997 and Kruys and Jonsson 1997 for details). The area is 28400 ha as defined by Grundsten et al. (1991) and consists of old-growth Norway spruce Pia-a ahies ssp. obovata moraine patches interspersed in a Sphagnum-dommaitiX wetland matrix. In total, Granlandet consists of > 1000 of these moraine patches (hereafter referred to as islands) that vary greatly in size and shape. There is a large variation in landscape mosaic structure in the area, which we use as a substitute for different degrees of fragmentation. The dominant trees are estimated to be 150-200 yr old. although individual spruces may be up to 300 yr old (Lovgren 1986). Most islands represent late suecessional stands where fire has been very rare or absent. Data on landscape composition were obtained by calculating the forested area within a circle with a 649 200-ni radius around the eentroids (i,c. the center of a N-S E-W rectangle covering the entire island) of the islands used in this study. This was done with a Geographic Information System (ESRl 1997). In the absence of data on dispersal distances in the area, the 200-m radius was chosen for two reasons: a) calculations of the fractal dimension of forested islands for ditTerent pixel sizes (20 x 20 m to 1000 x 1000 m) show a major change between 200 x 200 m and 300 x 300 m {cf. Obeysekera and Rutchey 1997). and b) 200 m is the most common maximum distance from randomly chosen points on mires to the nearest forest edge in Granlandet (Dettki unpubl,). Thus, the 200-m radius is supposed to reflect a scale region where significant structural changes occur in our specific landscape. The data for epiphytic Calicioid lichens (i,e crustose species with "pin""-like apothecia) are from two separate studies. In 1995. data were collected from 24 islands in a size range of 0,4 15,9 ha. In this study, stems of all trees with a diameter a! breast height > 20 cm were analyzed for occurrence of Calicioid lichens from the base up Lo 2 m. On islands < 1 ha. one circular sample plot of 0.1 ha was sampled, while on larger islands three 0,1 ha plots were sampled (Kruys and Jonsson 1997), Additional data on Calicioid lichens were collected in 1996 from 22 small islands (0.7-1.0 ha) and 23 medium-sized islands (3.8-5,6 ha). The procedure was similar to the 1995 study although only one 0,1 ha plot was sampled on each island. Thus, data from a total of 67 separate islands (two islands were included in both studies) were available. Data on woodinhabiting hepatics were also collected in 1996 on the same islands as the lichens. All fallen logs with a base diameter > 15 cm were included. Wood-inhabiting fungi were inventoried in 1997 on 34 small ( < 1 ha) islands in circular 0,1 ha plots. In this study, all coarse woody debris longer than 50 cm and with a base diameter of > 10 cm were included. Eor al! these three species groups, presence, absence data on individual trees and logs in the 0,1 ha plots were collected. Accordingly, "abundance"" estimates are the fraction of trees with occurrence of a particular species within these plots. To estimate the "population size"" in the landscape, this fraction was multiplied by the estimated total number of trees or logs in the 12.5 ha landscapes (200-m radius circles) that were analyzed. The estimated number of trees or logs was based on the proportion of forest within each landscape multiplied by the mean density of standing and fallen trees in Granlandet. which was obtained from the separate studies mentioned above. Linear regression models with logarithmic population size as the dependent variable and logarithmic percent of forested area as the independent variable were constructed for each species. Our H(, hypothesis was that the regression coefficient, p, equals I. relating to a situation were population size decreases proportional to habitat ioss. The t-statistic for significance testing of the regression coefficient is given by t = (b— I),SE,, (with DF = n — 2). where b is the fitted slope parameter and SE(, is the standard error of the slope parameter. Since the null-hypothesis is of biological importance (i.e, that the species is not sensitive to fragmentation), we calculated for the non-significant tests the power to detect an effect size of 0.3 (i,e. slopes of 1,3 or 0,7 respectively) using the observed vai'iance and sample size (Thomas 1997), Results and discussion Six of the 32 species had slopes that were significantly steeper than I (Table I). Five of these were Calicioid lichens (Caliciuni glauccdliini, Chacnolhccii pluwocephuhi. Ch. suhroscida. Ch. irichialis and Micracalicium dis,seminatuin). and one was a hepatic (Anasirophylhim hellerianum). Two of the species showing a slope > I are on the Swedish red-list and classified as caredemanding {Ch, phaeucephahi and A, hdlcrianum; Aronsson et al. 1995). These species are considered as negatively influenced by forestry although the actual causes of decline are not known. Previous data from Granlandet have indicated that some of these red-listed species are more abundant on large forested islands or show negative correlations with edges (Gustafsson unpubl, Moen and Jonsson unpubl.). It is thus likely that the actual cause for the decline of several of these is related to disproportionally smaller core areas in smaller islands. Somewhat surprisingly, two species also had a slope that was significantly lower than I: Caliciuni Irabiiwlhim and Clmcnoiheca xyloxcna ( p < 0 , ] in the latter case). Andren et al, (1997) interpret this to mean that such species are generalists that are able to utilize both habitat types to some extent. However, this could not be the case here since these species cannot use the mire matrix due to lack of trees. We thus suggest that the occurrence of these species is related to the increasing amount of edge present as landscapes become more fragmented. Both of these species are dependent on standing dead trees with hard and dried wood, which may be more common in wind- and sun-exposed edge habitats. This may cause their population sizes to decrease at a slower rate compared to the proportion of forest in the landscape since the proportion of edge actually increases in smaller fragments. We thus suggest that species with slopes > 1 may be seen as specialist species sensitive to fragmentation as was suggested by Andren et al. (1997), but also that species that have slopes < I may be seen either as generalist species that are able to utilize both habitats or as specialist species that are favored by edge habitats. ECOGRAPHV 21: The majority of the species did not show a slope significantly different from I. This could simply imply that fragmentation for these species is nothing but pure habitat loss, i.e, that the null-hypothesis of slope = 1 is true. This could be evaluated through a power analysis (e,g, Rotenberry and Wiens 1985. Peterman 1990), We assumed that a biological significant response to fragmentation includes slopes that deviate from 1 with at least 0.3 (i,e, slopes > 1.3 or <0.7). (This assumption is based on the mean response of the significant tests in our study ( + 0.39). but the reader could perform the power analyses with a different assumed effect size if he/she finds it unreasonable. A slope of 1,3 corresponds to 15'Mi additional loss in abundance compared to a slope of I.) The power gives the probability that the test would find a difference of at least the assumed effect size given the observed variance and sample size. It is not possible to statistically accept a null-hypothesis (Hilborn and Mangel 1997). but if power is high enough (0,95 if we want to assume the same precision as 7) we could say that the test was powerful enough to have picked up the assumed effect size and that the null hypothesis is in a sense "true". In our data set there is only one species with a high enough power to conclude that it is not sensitive to tragmentation {Pulidhiiii pulclicrn'/nuin),. and one species which is close {Caliciuni rlride). The rest of the species are inconclusive due to high variance and/or low sample size. The differences between the studied species groups make biological sense as regards their potential sensitivity to edges and fragmentation. The epiphytic Calicioid lichens are the most exposed group, compared to woodinhabiting hepatics. which grow close to the ground, and wood-inhabiting fungi that grow within logs. Thus, our analysis shows that the approach of Andren et al. (1997) is a simple and potentially useful tool to screen sets of species in order to find those that are sensitive to "true"" fragmentation effects, Ftirther application of the model in other systems is. however, largely dependent on the possibility to replicate landscapes which could be a severe restriction in most systems. The present system might be a special Table I. Statistics from linear rL'gressions of log population size iigainsl log proportion of habitat for 33 species from Granlandet, northern Sweden, Species Calicioid lichens Cdlicium glaucelliim C, lirlictioidc.s C, trabincllum C, viriik Chaenollu'ca clirv.wceplialu Ch, furfiiracca Ch, pluii'Ci'phala Ch, siemonea Ch, subroscida Ch. Irichialis Ch. xyloxena Ch. viridialbci Cypht'lium karcliciim Microcalicium ilisseininaium Wood-inhabiting fungi Anlrodiu .seriali.'; Cnlymnocyslis abiainu Foniiiopsis rosca Gloc'phylhinj scpianim Phcilimis chrysoloma P. JvrrugiiK'ofuscus P, nlgrolimlnam P. vil'icoki STcrcum .uinguilenniiii Trii'luipluin ubictimiiu Wood-inhabiting hepatics Anaslrophylluni hi'llcrictniiin Ci'pludozia bicuspidaia Ci'phulozia hiiiullfoliu Lophozia ascendens L, lougidc'ii'. L, longiflora !., .s-ilric(ilit Pliiiditun pidclwrrimum 0,72 0,64 0,2^ 0,70 0,47 0,21 0-51 0,16 0,66 0,68 0,42 0,77 0,43 0,66 0,61 0,70 0,42 0-30 0.07 0,69 0,-12 0,56 0,72 0,58 0,76 0,45 0.46 0.58 0,60 0,47 0,49 0,85 1,25 1,08 0,55 1,11 0,90 0,71 1,46 0,64 \35 1,30 0.67 1,21 0,86 1,52 1,02 1,03 0,85 0,85 0,24 0.87 0,71 1.06 0,96 0,90 1,49 0,77 0,92 0.99 1,18 1,00 i,OI 0,92 0,10 0.12 0.18 0.09 0,12 0,34 0,26 0,27 0,13 0,12 0,19 0,13 0,19 0,14 0,18 0-14 0,23 0,34 0,25 0,14 0,26 0,24 0,16 0,16 0,17 0,17 0,24 0,16 0,19 0,20 0,21 0,07 61 43 30 63 62 19 33 31 61 60 19 29 30 59 21 23 19 15 12 18 16 16 15 2? 24 24 17 30 26 29 28 32 2,44 0.66 -2.48 1,15 -0,82 -0,87 1,80 -1,30 2,72 2,60 -1,74 1,63 -0,78 3,56 0,14 0-21 -0.65 -0,43 -3,09 -0,94 -1,12 0,27 -0,23 -0,65 2,83 -1.32 -0.32 -0,08 0,96 0,03 0,03 - 1,22 0,018 NS 0,019 NS NS NS 0,081 NS 0,008 0,012 0,098 NS NS 0,001 NS NS NS NS _^NS NS NS NS NS 0.009 NS NS NS NS NS NS NS _ 0,68 _ 0,91 0,69 0.13 0,19 _ _ _ 0.61 0,33 0,36 0,54 0,24 0-13 _ 0.22 0.19 0.22 0.20 0.43 _ 0.40 0.22 0.44 0.33 0.30 0.28 0,99 RSlope Slope SE il t-value Slope p' Power' ' H(,: slope = I, - Model not significant (p = 0.35), ' Power to detect a difference of + 0.3 from ilope = I given obser\ed sample size and \ariance. F.rOGRAPHY 21:6 (199K) 0,00 0,40 0,90 1,90 Log percent forest Fig, I. Examples of linear regressions of log population si7e agamsl log percent forest in the landscape. Boxes and solid line: Caliciuni oirlde Islope not significantly different from 1: y = l,lO6x-h U,9U4). Crosses and dotted line: Microcaliciiiin disseniinatum (slope significantly > 1: y = L516x-I-0,008), and circles and dashed line: Caliciuni irabinelliim (slope significantly < 1 ; y = 0.554x +0,774), case due to the large number of isolated forest patches available, 1!' we plot the detectable difference in slope (from 1) against sample size (given a power of 80% and a standard deviation of 0.94 which is Ihe mean variation in our data set), we see that we need ca 80 replicates to detect a slope of 1.3 (Fig, 2), Regardless of the detectable effect size that we deem is biologically significant. Fig. 2 suggests that we need a large number of replieated landscapes to be able to detect slopes different from 1 if the variation in our data set is in any way representative of varialion in other data sets. If it is difficult to inerease sample size in a study beeause landseapes are limiting, it is imperative that the sampling design is such that estimations of abundance within landscapes are optimized. No. of cases Fig, 2, The detectiibic diffi^rcncc from a slope of 1. given a power of SO"Ai and a standard deviation of 0,94 (which is the mean standard deviiilion in our daia sets), as a function of the number of cases. Connor. E, F, and McCoy. E. D. 1979, The statistics and biology of the species-area relationship, - Am, Nat, 113: 791-833. Dudley, N,. Gilmour, D. and Jeanrenaud. J,-P, 1996, Forests for life, - WWF Int, and IUCN, Gland. Fdenius. L, and Sjoberg, K, 1997, Distribution of birds in natural landscape mosaics of old-growth forests in northern Sweden: relations to habitat area and landscape context. Ecography 20: 425 431, ESRI 1997, ArcView ver, 3,0a, Hnvironmental Systems Research Inst., Redknds, USA, Grundsten, C . Haggbom, J, and Lilja. T, 1991, Omraden av riksintresse for naturvSrd och friluftsliv. Report 3771, Swedish Environmental Protection Agency, Stockholm. Haila. Y. 1983, Land birds on northern islands: a sampling metaphor for insular colonization, Oikos 41: 334-351. Heywood. V. H, (ed,) 1995, Global biodiversity assessment, Cambridge Univ, Press, Hilborn. R. and Mangel. M, 1997, The ecological detective, Confronttng models with data, - Princeton Univ, Press, - We would like lo thank Hakan Rydin for Kruys. N, and Jonsson, B. G, 1997, Insular patterns of comments on an earlier version of the paper. calicioid lichens in a boreal old-growth forest-wetland mosaic, - Ecography 20: 605-613, Lovgren, R, 1986. Urskogar, lnventering av urskogsartade omraden i Sverige, Del 3-Norra Sverige, Report 1509, Swedish Environmental Protection Agency, Stockholm, References Obeysekera, J, and Rutchey, K. 1997, Selection of scale for Everglades landscape models, - Landscape Ecoi, 12: 7-18, Andren, H, 1994. Effects of habitat fragmentation on birds Peterman, R, M. 1990. Statistical power analysis can improve and mammals in landscapes with different proportion of fisheries research and management, - Can, J, Fish, Aq, Sci, suitable habitat: a review, Oikos 71: 355-366, 47: 2-15, Andren, H,. Dclin. A. and Seller, A, 1997, Population rcRotenberry. J, T, and Wiens. .1, A, 1985, Statistical power spotise to landscape charges depends on specialization to analysis and community-wide patterns, - Am, Nat, 125: different landscape elements, - Oikos 80; 193-196. 164-168, Aronsson, M,. Hallingback, T, and Mattson. J.-E, (eds) 1995, Thomas. L. 1997. Retrospective power analysis, - Conserv. Rodlistade vaxter i Sverige 1995 [Swedish Red Data Book Biol. 11: 276-280, of Plants 1995], - Artdatabanken, Uppsala. ECOGRAPHY 21:6

Journal

Ecography – Wiley

Published: Dec 1, 1998

There are no references for this article.