This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
RESEARCHPAPER
At the limits: habitat suitabilitymodelling of northern 17-yearperiodical cicada extinctions(Hemiptera: Magicicada spp.)John R. Cooley1,2,3, David C. Marshall1, Chris Simon1,
Michael L. Neckermann,3 and Gerry Bunker3
1Department of Ecology and Evolutionary
Biology, University of Connecticut, Storrs, CT,
USA, 2Department of Systems Engineering,
Shizuoka University, Hamamatsu, Japan,3Cicada Research Consulting, Storrs, CT, USA
ABSTRACT
Aim Adult periodical cicadas emerge as temporally isolated, synchronized multi-species communities (‘broods’) that are for the most part geographically contigu-ous and that fit together in jigsaw-puzzle-like fashion. Some year-classes of 17-yearcicadas have become extinct within historical times. We investigate two generalcauses for these extinctions – anthropogenic habitat destruction and post-glacialclimate change.
Location Periodical cicadas are confined to the eastern United States, east of theGreat Plains. We document the locations of known periodical cicada extinctions intwo broods of 17-year cicadas in Connecticut, Rhode Island, and upstate New York,USA.
Methods Using additional distributional records of 17-year cicadas, we develophabitat suitability models for all 17-year periodical cicadas, using data layers thatreflect both ecological and anthropogenic factors.
Results Climatological data layers related specifically to annual mean temperatureand temperature during the warmest months make the greatest contributions toour models, and data layers most specifically related to deforestation and habitatfragmentation tend to make much smaller contributions. Two well-documentedextinct populations of periodical cicadas occurred in locations where these modelspredict relatively low habitat suitability for 17-year cicadas.
Main conclusions Our results and other circumstantial evidence discount theimportance of anthropogenic habitat destruction in explaining these particularextinctions.
KeywordsBrood, cicada, change, eastern USA habitat loss, hypsithermal, metapopulation.
*Corresponding: John Cooley, Department ofEcology and Evolutionary Biology, University ofConnecticut, Storrs, CT 06269-3048, USA.E-mail: [email protected]
INTRODUCTION
There has been increasing interest in ecological studies that use
modelling to predict and understand the environmental corre-
lates of species distributions (reviewed in Guisan & Zimmer-
databases.php). This dataset represents an extensive sampling
effort using methods similar to those described above and
designed to locate the perimeters of various broods. Each pres-
ence record used in this study was obtained or confirmed by
one or more of the authors. The overall range of 17-year cicadas
delineated by our records is similar to the range shown in Mar-
latt’s periodical cicada maps as edited by Simon (Marlatt, 1923;
Simon, 1988). No records of off-cycle or ‘straggler’ cicadas were
included, and all duplicate or geographically coincident records
were removed. The full dataset did include the records of veri-
fied extinct populations. To speed processing, a ‘thinned’ dataset
(Fig. 1) was created by randomly thinning clumped data points
resulting from the use of an automated GPS datalogger in the
most recent field seasons (Cooley et al., 2011). Thinning was
accomplished via construction of a point density contour map
of the ‘full’ dataset; areas that were most clumped on this map
were thinned by a factor of 16. The final thinned dataset con-
sisted of 1737 positive records.
Because some modelling methods make use of both presence
and absence data, we also constructed a ‘known absence’ dataset.
Although the Cicada Central database contains many absence
(‘negative’) records, these records are organized by brood mem-
bership, so they are not necessarily absence records for 17-year
cicadas in general. For instance, many of the absence records for
Brood X (Cooley et al., 2009) fall within the territory of Brood
XIV (Cooley et al., 2011) and vice versa. To select only records
that represent the complete absence of any 17-year Magicicada
brood, we constructed a detailed hull around all positive records
of 17-year cicadas and buffered it to 1 km. We discarded all
absence records falling within the periphery of the 1-km buff-
ered hull, leaving 1388 records in which 17-year Magicicada
were known to be absent. Because these absence records had the
same datalogger biases as the positive records, plus the addi-
tional bias that our absence records were concentrated on areas
in the immediate periphery of our positive records, we thinned
this dataset to 706 observed negative records, using the data
thinning techniques described above. We then constructed a
raster of inferred absences, using a c. 10-km2 grid cell size con-
tained within the bounding box of this study (see above). This
raster was converted to a point shapefile with geocoordinates
placed at the centre of each raster grid. All such points falling
within a 5-km buffered hull of positive data points, or within
Figure 1 Periodical cicada distributionrecords used to create habitat suitabilitymodels. Hatched areas are estimates ofthe placement of Marlatt’s 17-yearMagicicada records (Marlatt, 1923) asrevised by Simon (1988).
Extinct populations of Broods VII and XI fall in or near areas
where our models predict lower suitability for 17-year cicadas
(Fig. 4). A plot of model values for extant and extinct popula-
tions shows that extinct populations of Broods VII and XI are
low-value outliers in all models (Fig. 5). These differences are
statistically significant; when model scores for extinct and extant
populations are extracted from the thinned-dataset model,
scores for extinct populations are statistically lower for the
MaxEnt model (Mann–Whitney–Wilcoxon test; W = 2405.5, P <0.001) and the summary statistic for the BIOMOD ensemble
models (Mann–Whitney–Wilcoxon test; W = 801, P < 0.001).
Thus, extinct populations of Broods VII and XI fall into areas
where our models predict a lower probability of conditions
appropriate for 17-year Magicicada.
DISCUSSION
In this paper we investigate factors that affect habitat suitability
for 17-year cicadas in the genus Magicicada. It would be inac-
curate for us to use the term ‘species distribution modelling’ to
describe our efforts, because of the quirky biology of periodical
cicadas: Whereas the 17-year cicadas are a group of closely
related and largely co-distributed species that in some respects
function as a single ecological unit, they do not form a mono-
phyletic clade. Even so, these species are locked together in space
and time, and it is the manner in which their shared ecology
binds them together and limits their distribution that we wished
to capture in our models. While the accuracy of HSMs tends to
be greatest for taxa with small geographic ranges, limited eco-
logical tolerance (Hernandez et al., 2006) and data sets of inter-
mediate prevalence (McPherson et al., 2004), we generated
HSMs that accurately represent the distribution of 17-year peri-
odical cicadas, even though their range is large and ecologically
diverse.
Two aspects of our models are relevant to understanding
extinct populations and broods. First, documented extinct
populations of Broods VII and XI were found in areas where our
models predict a comparatively low probability of conditions
appropriate for 17-year cicadas. Second, climatological variables
(e.g. temperature and precipitation) seem to have the greatest
importance in our models, while non-climatological variables
(e.g. land cover and fragmentation) tend to make smaller model
contributions. Taken at face value, these results suggest that
climatological explanations for the increased susceptibility of
periodical cicadas to extinction along the northern edge of their
distribution are more plausible than non-climatological expla-
nations like ‘ecoregions’, anthropogenic fragmentation, forest
connectivity and land-cover classification (see Appendix S2).
The low contribution of the forest fragmentation data layers
to our models was unexpected. To the extent that fragmentation
involves the reduction of existing forests into smaller and
smaller patches, we expect the probability of local extinction to
increase with increasing fragmentation, because periodical
cicadas depend on high community densities of at least tens of
thousands per hectare that decrease per capita predation risk
(Beamer, 1931; Dybas, 1969; Karban, 1982a; Williams & Simon,
1995; Marshall et al., 2011). In addition, forest patchiness has
been found to affect the spatial distributions of periodical
cicadas (Rodenhouse et al., 1997), and Magicicada fecundity
may be lowered in sparse populations (Karban, 1982b). Thus,
scarcity tends to magnify conditions that further depress popu-
lation size leading to demographic time lags that make it diffi-
cult for declining populations to recover. It is possible that the
measures of fragmentation used in our study fail to capture the
importance of fragmentation to periodical cicadas, either
because the resolution of the data layers is too coarse or because
current levels of fragmentation are irrelevant; instead, informa-
tion about fragmentation that occurred in the past might be
Figure 2 Periodical cicada Brood XI2005 search area in Connecticut andRhode Island in 2005. Cross symbolsdenote the absence of periodical cicadas;no positive records of Brood XI werefound in 2005. Black circles are verifiedpositive records of 17-year periodicalcicadas belonging to other broods.Hatched areas are estimates of theplacement of Marlatt’s Brood XI records(Marlatt, 1923) as revised by Simon(1988).
more informative. Unfragmented habitat that has regenerated
during the past century following agricultural abandonment
(such as habitat found broadly throughout the north-eastern
states today) may contain few Magicicada populations, as a
legacy of widespread extinction that occurred long ago during
extensive deforestation. Indeed, the last known location of
Brood XI and some of the areas formerly inhabited by VII are
currently forested and classified as intact and interconnected,
while an early 20th century photograph of this area, taken when
the brood was still in existence, shows patchy woodlots and
fencerows (Appendix S3; Favretti, 2003).
One issue raised by our study is why extinction has been well
documented only for Broods VII and XI, when portions of
Broods VIII, X, XIII and XIV occupy similar latitudes. In fact,
similar declines have been reported on the basis of county-level
maps in Indiana (Brood X; Kritsky, 1987) in areas where our
models predict lower suitability for 17-year Magicicada. It is
possible that 17-year populations are in retreat along the entire
northern edge of their range, and the lack of documentation
may be simply an artefact of missing or ambiguous historical
records. Indeed, along the northern periphery of the general
periodical cicada distribution, recent emergences of Broods X,
XIII, XIV and several disjunct populations of various broods on
Long Island (Simon & Lloyd, 1982) have fallen short of their
historically reported ranges (J.R.C., D.C.M. and C.S. unpub
lished data). Reporting biases may also explain the lack of
well-documented declines; for example, the complete extinction
of a peripheral brood such as VII or XI may be far more likely to
be noticed than a local extinction near the centre of a large
brood.
Figure 3 (a), (b). Species distributionmodels based on a thinned 17-yearMagicicada dataset and six variables.White indicates areas where probabilityof conditions appropriate for 17-yearcicadas approaches zero. Shadingindicates areas where conditions have thehighest probability of being appropriatefor 17-year cicadas, with darker shadingindicating higher probabilities. (a)BIOMOD ensemble forecast. (b) Averageof 1000 replicate MaxEnt models.
Figure 4 (a) Verified extinct populations (crosses) of periodical cicada Broods VII and XI. Dark shading is the MaxEnt model from Fig. 3.White indicates areas where the probability of conditions appropriate for 17-year cicadas approaches zero. Dark shading indicates areaswhere conditions have the highest probability of being appropriate for 17-year cicadas. Hatched areas are estimates of the placement ofMarlatt’s 17-year Magicicada records (Marlatt, 1923) as revised by Simon (1988). (Inset A) Verified extinct populations (crosses) ofperiodical cicada Brood XI in Connecticut and Rhode Island. The background is the MaxEnt habitat suitability model based on thinned17-year periodical cicada dataset from Fig. 3. Hatched areas are estimates of the placement of Marlatt’s 17-year Magicicada records (Marlatt,1923) as revised by Simon (1988). (Inset B) Verified extinct populations (crosses) of 17-year periodical cicada Brood VII in New York. Thebackground is the MaxEnt habitat suitability model based on thinned 17-year periodical cicada dataset from Fig. 3. Black circles are verifiedpositive records of 17-year periodical cicadas belonging to other broods. Hatched areas are estimates of the placement of Marlatt’s 17-yearMagicicada records (Marlatt, 1923) as revised by Simon (1988).
Another question raised by our models is how the now-extinct
periodical cicada communities became established in the first
place, if indeed the locations of these populations are of low
suitability for 17-year periodical cicadas. It is possible that
improved sampling of periodical cicadas in areas where our
sampling is weak (parts of Indiana, Ohio, New Jersey and Con-
necticut) would have changed habitat suitability estimates for the
extinct populations, though it seems doubtful that improved
sampling would significantly decrease the importance of the
climatological variables. It is also an intriguing possibility that
northern marginal populations of periodical cicadas are relicts
left behind by prior range extension during the drier, warmer
conditions of the Hypsithermal Interval (c. 9000–5000 bp). If
17-year periodical cicadas expanded their range northward
(c)
Figure 4 (c) An enlargement of Inset B.
Figure 5 Scatterplot of habitat suitability model scores for locations of positive records in thinned 17-year Magicicada dataset (blackcircles), verified extinct populations (black crosses), verified absence records (open triangles) and inferred absences (open diamonds). Thehorizontal axis is the score from the MaxEnt average model; the vertical axis is the score from the BIOMOD ensemble forecast.