Testing the role of patch openness as a causal …Testing the role of patch openness as a causal...
1
Testing the role of patch openness as a causal mechanism for apparent area sensitivity c. Openness Index (degrees) 50 55 60 65 70 75 80 85 90 Bobolink Density (#/ha) 0 1 2 3 4 5 6 Area (ha) 0 10 20 30 40 50 60 Bobolink Density (#/ha) 0 1 2 3 4 5 6 a. b. c. 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Bobolink Density (#/ha) 0 1 2 3 4 5 6 c. Bobolink Density (#/ha) 0 2 4 6 8 10 12 14 16 Bobolink Density (#/ha) 0 2 4 6 8 10 12 Area (ha) 0 10 20 30 40 50 60 Bobolink Density (#/ha) 0.0 0.2 0.4 0.6 0.8 1.0 Openness Index (degrees) 50 55 60 65 70 75 80 85 90 a. b. c. d. e. f. r 2 = 0.25 (June) r 2 = 0.41 (July) r 2 = 0.32 r 2 = 0.81 July August September INTRODUCTION Area sensitivity is defined as a species being absent or occurring in lower densities in smaller habitat patches as compared to larger habitat patches. Area sensitivity is widespread across animal taxa and ecosystems, but is sometimes not consistently re- ported within a species. This suggests that it may not be area per se that is the causal variable these species are responding to. Openness can affect patch occupancy, with some animals avoiding less open sites, and is affected by both patch area and edge type. Openness is a measure of how much of an animal’s visual field is not occluded by ground, vegetation structure, or human-made structures in abutting habitat. It is a function of topography, (e.g., the top of a hill is more open than the base), the height and structure of adjacent habitat, and the distance to the adjacent habitat. One bio- logical argument for why animals might respond to openness is to increase predator detection and avoidance: (1) an advancing predator may be easier to detect in open habitat, (2) predators may be able to better detect prey from tall perches, and open landscapes may deny predators these perches, and (3) species specific escape behav- ior may require open habitat (e.g., some aerial maneuvers by birds). We tested field openness as a proximate mechanism for apparent area sensitivity in Bobolinks (Dolichonyx oryzivorus), a grassland passerine We examined linear, non-linear, and threshold relationships between openness and Bobolink density and compared them to area and edge effects We examined the relationship between these variables and Bobolink density during the breeding period and three post-breeding times We tested whether body condition, as measured by body mass, wing chord, and stress responses, increased with either patch area or openness METHODS Research was conducted from June-October 2009 and May-June 2010 in late-cut and uncut hayfields (grass/forb mixtures) within a primarily wooded matrix in Eastern and Central Massachusetts, USA (42°12’ - 42°53’N, 71°58’ - 70°50’ W). Habitat patches were delineated by roads, woods, pasture, or other habitat types. Breeding male Bobolink density was measured by two observers in June 2009 via line transect surveys and distance sampling. After the breeding season Bobolinks form flocks (July -October), so line transect surveys with distance sampling are not an effective survey method. In addition, male Bobolinks molt into basic plumage after breeding and re- semble females. Consequently for July-September surveys all Bobolinks encountered were counted, regardless of age or sex. In September, when Bobolinks are more cryptic, birds were surveyed using a 50 m rope dragged between two observers in non -overlapping transects. The total number of birds was divided by site area to avoid detecting area sensitivity due to passive sampling. Area and ln(area) were obtained from digitized polygons using ArcGIS. As Bob- olinks have lower population densities near forest edges, we modeled the expected edge effects for each patch using an edge effects model (EEM) based on Sisk’s effec- tive area model. If openness is only capturing a response to wooded edges, then the EEM is expected to outperform the openness index. Openness (Fig. 1) was measured by taking measurements at points every 50 m along a transect set across the longest axis of a site plus a final point at the far edge of the field (Fig. 2a). At each point, an angle measurement to the horizon was taken (measured at eye level, ~1.7 m) in the di- rection of the axis that was being walked plus 3 additional measurements offset at 90º intervals (Fig. 1b). Inclination was taken to the horizon; generally this was the field edge, but in some instances this was to the top of a rise in the field or to a distant hill not connected to the site. The four angle measurements were averaged for each point, and then averaged across all points to give a single index value for each field. This fi- nal number was subtracted from 90º so that the openness index increased with in- creasing openness. We used mist-nets and playback to capture male Bobolinks on their territories at fields across a range of sizes and openness. We measured body mass, natural wing chord, and circulating corticosterone (CORT) at 3 time points prior to, and a fourth after, injection with adrenocorticotropic hormone (ACTH), a hormone that stimulates the release of CORT. CORT concentrations were determined by radioimmunoassay. RESULTS Bobolink density and occupancy showed significant linear and logistic relationships with openness (respectively), but logistic models based on openness occupancy thresholds identified by binomial change-point tests had greater explanatory power. Specifically, our analyses suggest that Bobolinks in our study area were absent below a certain value of openness, and present above that threshold; above the threshold there appeared to be no relationship between Bobolink density and degree of open- ness (Figs. 3, 4). Thresholds remained approximately consistent from June to Au- gust, and shifted to greater openness in September (Fig. 4). Area models were poorer predictors of Bobolink occupancy and density outside of the breeding season. While our openness index was significantly linearly correlated with other predictor varia- bles (all P < 0.001), including area (r 2 = 0.26) and ln(area) (r 2 = 0.60), these inde- pendent variables were not equivalent when examining Bobolink density and occu- pancy (Fig. 4). Even though openness and area are correlated, variance partitioning of the logistic regression models show that openness consistently explained more pure variation – i.e., variation explained uniquely by the variable of interest – than did area (June: 20% vs. 8%, July: 28% vs. 1%, August: 30% vs. 2%, September: 1% vs. 0%). Openness did not explain much pure variation in September because the full model was over-parameterized (only 3 fields where Bobolinks were present); however openness alone was sufficient to describe Bobolink occupancy. There was no relationship between either openness or area and body condition measures (Fig. 5). Our analyses of leverage coefficients suggest one reason why studies might report differences in observed area sensitivity. In our area and edge effects model (EEM) regressions, there were two outlying points on the high end of the x-axis, and the re- gression was highly sensitive (strong leverage coefficients) to them (Fig. 2); their re- moval changed the r 2 value for both area and for the EEM from 0.18 to 0.06. This was not corrected by ln-transforming area, as the r 2 was still reduced from 0.16 to 0.07. In contrast, removing the single point with a high leverage coefficient from our openness regression caused a trivial increase in the r 2 (by 0.001). We also identified two open sites with small areas; removal of these points also strongly affected our ar- ea regression, increasing r 2 from 0.18 to 0.27. DISCUSSION We evaluated a novel measure of openness Bobolinks’ response to openness is stronger than their response to area Bobolinks respond to openness via an occupancy threshold (present or absent) not as a linear response to increasing openness For occupied sites there is no change in body condition – further supports an occupancy threshold If managing for Bobolinks, only habitat changes that cross the threshold are likely to change bird occupancy. The pattern with openness persisted across time and across chang- es in behavior Open sensitive species will appear area sensitive in a closed land- scape Inconsistency in area sensitivity may be partially due to statistical issues – especially leverage Adult predation risk is a possible mechanism for open-sensitivity ACKNOWLEDGMENTS We thank D. DesRochers, B. Tavernia, members of our lab group, and S. Lewis for discussions, and E. Wier, G. Surya, D. Peck, C. Le, D. DesRochers, and B. Tavernia for field assistance. We thank B. Parmenter for GIS assis- tance, P. Florance for GPS equipment and rangefinder, A. Strong for the playback recording, E. Rexstad for dis- cussion, and S. Keyel for bird bags. Special thanks to the landowners: Towns of Bedford, Carlisle, Lincoln, and Amesbury, Walden Pond, Moore, and Brookwood Farm State Parks, Massachusetts Dept. of Conservation and Recreation Division of Water Supply, MassWildlife, Massachusetts Audubon Society, Trustees of Reservations, United States Fish and Wildlife Service Eastern Massachusetts Wildlife Refuge Complex, Tufts’ Cummings School of Veterinary Medicine, and Nichols Brook Conservation Area. Funding for this research was provided by a Tufts Institute for the Environment Fellowship and NSF Graduate Research Fellowship to ACK, Nuttall Or- nithological Club Blake Fund grants to ACK & JMR, NSF IOB-0542099 to LMR, and NSF REU. Fig. 5 Male Bobolink corticosterone (CORT) levels signifi- cantly increased over time (t, in min) and with injection of adrenocorticotropic hormone (ACTH) (a). Letters denote sig- nificant differences (P ≤ 0.01; repeated measures ANOVA), * indicates that ACTH resulted in a significant increase in CORT relative to t = 110 (P = 0.05). There was no significant difference between sites <5 ha and sites >5 ha (all P > 0.30) in CORT (a), body mass (b), or wing chord (c). Sites are grouped into < 5 ha (n = 5) and > 5 ha (n = 4) to ease visual interpretation of results, results are similar when analyzed continuously. Number of birds sampled given in the respec- tive bars. Fig. 4 Total Bobolink density was better explained by a binary (presence/absence) openness threshold (b,d,f) than by area (a,c,e), and remained relatively consistent across months (b vs. d vs. f). Thresholds are shown as solid vertical lines; the dashed vertical line (b) indicates the openness threshold for June (based on data shown in Fig. 3a); r 2 values refer to threshold models and were calculated by converting from 2 x 2 χ 2 contingency table corrected for continuity. Regression lines are shown for significant regressions. Note that sample sizes and observed Bobolink densities changed between months. Fig. 3 Openness (a) provided a better fit to male Bobolink density than did area (b) or the edge effects model (EEM) (c). Also, the openness regression was less sensitive to individual points than were the other regressions. The black diamond denotes a point with a high leverage coefficient for the open- ness regression (it is depicted in all 3 graphs for comparison), the black triangle and black circle denote points with high lev- erage coefficients for area, and circles (gray and black) denote points with high leverage coefficients for the EEM. Gray squares on each graph indicate two small open sites that when removed increase the r 2 of the area regression. The site indi- cated by the black triangle was not included in the EEM be- cause it was unsuitable for that analysis (the field abutted grazed habitat- this did not strongly affect the openness or the area, but made the edge definition ambiguous and inconsistent with the rest of the patches). Fig. 1 An example of how openness changes over a field. Openness was interpolated based on a grid of points 50 m apart. Fig. 2 a) Openness transect, showing the survey points (black dots) at 50 m intervals and at the starting and ending edges at the large field from Moore State Park, Paxton, MA. b) Angle to the horizon was measured from eye lev- el to the visual horizon. This measure was taken with an inclinometer in 4 directions 90° apart and as perpendicular to the field edges as possible. Edge Effects Model (standardized # ha -1 ) Alexander C. Keyel 1 , Carolyn M. Bauer, Christine R. Lattin, L. Michael Romero, and J. Michael Reed Dept. of Biology, Tufts University, 163 Packard Ave, Medford, MA 02155 1 [email protected] Treatment CORT (ng/ml) 0 20 40 60 80 100 120 140 <5 ha >5 ha Site Size Body mass (g) 30 31 32 33 34 35 Site Size Natural Wing Chord (mm) 90 92 94 96 98 100 a. b. c. baseline t = 20 t = 110 ACTH a a b b c c * <5 ha >5 ha <5 ha >5 ha 11 10 11 11 12 11 11 9 12 12 13 13 Value High: 85.9 Low: 66.1
Transcript of Testing the role of patch openness as a causal …Testing the role of patch openness as a causal...
Testing the role of patch openness as a causal mechanism for apparent area sensitivity
c.
Openness Index (degrees)50 55 60 65 70 75 80 85 90
Bo
bo
link
De
nsity
(#
/ha)
0
1
2
3
4
5
6
Area (ha)0 10 20 30 40 50 60
Bo
bo
link
De
nsity
(#
/ha)
0
1
2
3
4
5
6
a.
b.
c.
Predicted Average Effective Density (#/ha)0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Bo
bo
link
De
nsity
(#
/ha)
0
1
2
3
4
5
6c.
Bo
bo
link
De
nsity
(#
/ha
)
0
2
4
6
8
10
12
14
16
Bo
bo
link
De
nsity
(#
/ha
)
0
2
4
6
8
10
12
Area (ha)
0 10 20 30 40 50 60
Bo
bo
link
De
nsity
(#
/ha
)
0.0
0.2
0.4
0.6
0.8
1.0
Openness Index (degrees)
50 55 60 65 70 75 80 85 90
a. b.
c. d.
e. f.
r2 = 0.25 (June)r2 = 0.41 (July)
r2 = 0.32
r2 = 0.81
July
August
September
INTRODUCTION Area sensitivity is defined as a species being absent or occurring in lower densities in smaller habitat patches as compared to larger habitat patches. Area sensitivity is widespread across animal taxa and ecosystems, but is sometimes not consistently re-ported within a species. This suggests that it may not be area per se that is the causal variable these species are responding to. Openness can affect patch occupancy, with some animals avoiding less open sites, and is affected by both patch area and edge type. Openness is a measure of how much of an animal’s visual field is not occluded by ground, vegetation structure, or human-made structures in abutting habitat. It is a function of topography, (e.g., the top of a hill is more open than the base), the height and structure of adjacent habitat, and the distance to the adjacent habitat. One bio-logical argument for why animals might respond to openness is to increase predator detection and avoidance: (1) an advancing predator may be easier to detect in open habitat, (2) predators may be able to better detect prey from tall perches, and open landscapes may deny predators these perches, and (3) species specific escape behav-ior may require open habitat (e.g., some aerial maneuvers by birds).
We tested field openness as a proximate mechanism for apparent area sensitivity in Bobolinks (Dolichonyx oryzivorus), a grassland passerine
We examined linear, non-linear, and threshold relationships between openness and Bobolink density and compared them to area and edge effects
We examined the relationship between these variables and Bobolink density during the breeding period and three post-breeding times
We tested whether body condition, as measured by body mass, wing chord, and stress responses, increased with either patch area or openness
METHODS Research was conducted from June-October 2009 and May-June 2010 in late-cut and uncut hayfields (grass/forb mixtures) within a primarily wooded matrix in Eastern and Central Massachusetts, USA (42°12’ - 42°53’N, 71°58’ - 70°50’ W). Habitat patches were delineated by roads, woods, pasture, or other habitat types. Breeding male Bobolink density was measured by two observers in June 2009 via line transect surveys and distance sampling. After the breeding season Bobolinks form flocks (July-October), so line transect surveys with distance sampling are not an effective survey method. In addition, male Bobolinks molt into basic plumage after breeding and re-semble females. Consequently for July-September surveys all Bobolinks encountered were counted, regardless of age or sex. In September, when Bobolinks are more cryptic, birds were surveyed using a 50 m rope dragged between two observers in non-overlapping transects. The total number of birds was divided by site area to avoid detecting area sensitivity due to passive sampling. Area and ln(area) were obtained from digitized polygons using ArcGIS. As Bob-olinks have lower population densities near forest edges, we modeled the expected edge effects for each patch using an edge effects model (EEM) based on Sisk’s effec-tive area model. If openness is only capturing a response to wooded edges, then the EEM is expected to outperform the openness index. Openness (Fig. 1) was measured by taking measurements at points every 50 m along a transect set across the longest axis of a site plus a final point at the far edge of the field (Fig. 2a). At each point, an angle measurement to the horizon was taken (measured at eye level, ~1.7 m) in the di-rection of the axis that was being walked plus 3 additional measurements offset at 90º intervals (Fig. 1b). Inclination was taken to the horizon; generally this was the field edge, but in some instances this was to the top of a rise in the field or to a distant hill not connected to the site. The four angle measurements were averaged for each point, and then averaged across all points to give a single index value for each field. This fi-nal number was subtracted from 90º so that the openness index increased with in-creasing openness. We used mist-nets and playback to capture male Bobolinks on their territories at fields across a range of sizes and openness. We measured body mass, natural wing chord, and circulating corticosterone (CORT) at 3 time points prior to, and a fourth after, injection with adrenocorticotropic hormone (ACTH), a hormone that stimulates the release of CORT. CORT concentrations were determined by radioimmunoassay.
RESULTS Bobolink density and occupancy showed significant linear and logistic relationships with openness (respectively), but logistic models based on openness occupancy thresholds identified by binomial change-point tests had greater explanatory power. Specifically, our analyses suggest that Bobolinks in our study area were absent below a certain value of openness, and present above that threshold; above the threshold there appeared to be no relationship between Bobolink density and degree of open-ness (Figs. 3, 4). Thresholds remained approximately consistent from June to Au-gust, and shifted to greater openness in September (Fig. 4). Area models were poorer predictors of Bobolink occupancy and density outside of the breeding season. While our openness index was significantly linearly correlated with other predictor varia-bles (all P < 0.001), including area (r2 = 0.26) and ln(area) (r2 = 0.60), these inde-pendent variables were not equivalent when examining Bobolink density and occu-pancy (Fig. 4). Even though openness and area are correlated, variance partitioning of the logistic regression models show that openness consistently explained more pure variation – i.e., variation explained uniquely by the variable of interest – than did area (June: 20% vs. 8%, July: 28% vs. 1%, August: 30% vs. 2%, September: 1% vs. 0%). Openness did not explain much pure variation in September because the full model was over-parameterized (only 3 fields where Bobolinks were present); however openness alone was sufficient to describe Bobolink occupancy. There was no relationship between either openness or area and body condition measures (Fig. 5). Our analyses of leverage coefficients suggest one reason why studies might report differences in observed area sensitivity. In our area and edge effects model (EEM) regressions, there were two outlying points on the high end of the x-axis, and the re-gression was highly sensitive (strong leverage coefficients) to them (Fig. 2); their re-moval changed the r2 value for both area and for the EEM from 0.18 to 0.06. This was not corrected by ln-transforming area, as the r2 was still reduced from 0.16 to 0.07. In contrast, removing the single point with a high leverage coefficient from our openness regression caused a trivial increase in the r2 (by 0.001). We also identified two open sites with small areas; removal of these points also strongly affected our ar-ea regression, increasing r2 from 0.18 to 0.27.
DISCUSSION We evaluated a novel measure of openness Bobolinks’ response to openness is stronger than their response to
area Bobolinks respond to openness via an occupancy threshold
(present or absent) not as a linear response to increasing openness For occupied sites there is no change in body condition – further
supports an occupancy threshold If managing for Bobolinks, only habitat changes that cross the
threshold are likely to change bird occupancy. The pattern with openness persisted across time and across chang-
es in behavior Open sensitive species will appear area sensitive in a closed land-
scape Inconsistency in area sensitivity may be partially due to statistical
issues – especially leverage Adult predation risk is a possible mechanism for open-sensitivity
ACKNOWLEDGMENTS We thank D. DesRochers, B. Tavernia, members of our lab group, and S. Lewis for discussions, and E. Wier, G. Surya, D. Peck, C. Le, D. DesRochers, and B. Tavernia for field assistance. We thank B. Parmenter for GIS assis-tance, P. Florance for GPS equipment and rangefinder, A. Strong for the playback recording, E. Rexstad for dis-cussion, and S. Keyel for bird bags. Special thanks to the landowners: Towns of Bedford, Carlisle, Lincoln, and Amesbury, Walden Pond, Moore, and Brookwood Farm State Parks, Massachusetts Dept. of Conservation and Recreation Division of Water Supply, MassWildlife, Massachusetts Audubon Society, Trustees of Reservations, United States Fish and Wildlife Service Eastern Massachusetts Wildlife Refuge Complex, Tufts’ Cummings School of Veterinary Medicine, and Nichols Brook Conservation Area. Funding for this research was provided by a Tufts Institute for the Environment Fellowship and NSF Graduate Research Fellowship to ACK, Nuttall Or-nithological Club Blake Fund grants to ACK & JMR, NSF IOB-0542099 to LMR, and NSF REU.
Fig. 5 Male Bobolink corticosterone (CORT) levels signifi-cantly increased over time (t, in min) and with injection of adrenocorticotropic hormone (ACTH) (a). Letters denote sig-nificant differences (P ≤ 0.01; repeated measures ANOVA), * indicates that ACTH resulted in a significant increase in CORT relative to t = 110 (P = 0.05). There was no significant difference between sites <5 ha and sites >5 ha (all P > 0.30) in CORT (a), body mass (b), or wing chord (c). Sites are grouped into < 5 ha (n = 5) and > 5 ha (n = 4) to ease visual interpretation of results, results are similar when analyzed continuously. Number of birds sampled given in the respec-tive bars.
Fig. 4 Total Bobolink density was better explained by a binary (presence/absence) openness threshold (b,d,f) than by area (a,c,e), and remained relatively consistent across months (b vs. d vs. f). Thresholds are shown as solid vertical lines; the dashed vertical line (b) indicates the openness threshold for June (based on data shown in Fig. 3a); r2 values refer to threshold models and were calculated by converting from 2 x 2 χ2 contingency table corrected for continuity. Regression lines are shown for significant regressions. Note that sample sizes and observed Bobolink densities changed between months.
Fig. 3 Openness (a) provided a better fit to male Bobolink density than did area (b) or the edge effects model (EEM) (c). Also, the openness regression was less sensitive to individual points than were the other regressions. The black diamond denotes a point with a high leverage coefficient for the open-ness regression (it is depicted in all 3 graphs for comparison), the black triangle and black circle denote points with high lev-erage coefficients for area, and circles (gray and black) denote points with high leverage coefficients for the EEM. Gray squares on each graph indicate two small open sites that when removed increase the r2 of the area regression. The site indi-cated by the black triangle was not included in the EEM be-cause it was unsuitable for that analysis (the field abutted grazed habitat- this did not strongly affect the openness or the area, but made the edge definition ambiguous and inconsistent with the rest of the patches).
Fig. 1 An example of how openness changes over a field. Openness was interpolated based on a grid of points 50 m apart.
Fig. 2 a) Openness transect, showing the survey points (black dots) at 50 m intervals and at the starting and ending edges at the large field from Moore State Park, Paxton, MA. b) Angle to the horizon was measured from eye lev-el to the visual horizon. This measure was taken with an inclinometer in 4 directions 90° apart and as perpendicular to the field edges as possible.
Edge Effects Model (standardized # ha-1)
Alexander C. Keyel1, Carolyn M. Bauer, Christine R. Lattin, L. Michael Romero, and J. Michael Reed Dept. of Biology, Tufts University, 163 Packard Ave, Medford, MA 02155 [email protected]
Treatment
CO
RT
(ng
/ml)
0
20
40
60
80
100
120
140
<5 ha >5 ha
Site Size
Bo
dy
ma
ss (
g)
30
31
32
33
34
35
Site Size
Na
tura
l Win
g C
hord
(m
m)
90
92
94
96
98
100
a.
b. c.
baseline t = 20 t = 110 ACTH
a a
b b
c c
*
<5 ha >5 ha <5 ha >5 ha
11 10 11 11 12 11 11 9
12 12 13 13
Value
High: 85.9
Low: 66.1
