|
|
Bird conservation in the St. Lawrence Plain of New York: integrating geographic modeling techniques with multivariate habitat models to predict species distribution.
Daniel L. Wells
Introduction
Scientists and policy-makers concerned with bird conservation have recently adopted an ecosystem-level, habitat-based approach, with the goal of developing strategies of land management to enhance breeding habitat for multi-species bird communities. Partners in Flight (PIF), an international organization dedicated to the conservation of non-game landbird species, has been at the forefront of this new approach by developing a method of regional conservation planning based on habitat types and their corresponding bird species rather than an individual species approach.
Currently, PIF Bird Conservation Plans are being developed for fifty-eight regions across North America. Each plan offers recommendations of how to manage a particular region for multiple species, based on the habitat requirements of high
priority
species. But implementation of these plans is a challenge given that priority
species often have discordant habitat requirements. In the St. Lawrence
Plain Physiographic Area (SLP) , for example, many of the priority species
breed in and require completely different habitat types, making implementation
of the Bird Conservation Plan a difficult problem. Bobolinks inhabit grasslands,
golden-winged warblers are shrub-dependent, and eastern wood-pewees are
found only in forests, yet all three species are deemed as "high
priority" for conservation. Management prescriptions designed to
enhance breeding conditions for one will necessarily exclude the other
two. Decision-makers in the region are thus faced with an interesting
management dilemma: how do we simultaneously balance the conflicting habitat
needs of declining grassland, shrub and forest bird species in a dynamic
successional landscape? Objectives
The purpose of this study was to develop predictive bird-habitat models
for use as a tool in determining how to implement the Bird Conservation
Plan for the SLP, given the conflicting habitat requirements of its priority
bird species. I selected a portion of the SLP and used multiple logistic
regression models to predict bird presence/absence based on a set of environmental
and habitat variables. The objective was to determine if models developed
on such coarse grain information could provide acceptable predictability
of bird occurrence at the regional scale, and if the results of such a
process (maps of predicted distribution) would be useful for regional
conservation planning such as the PIF Bird Conservation Plans.
Study Area
The study was conducted in the St. Lawrence River Valley of northern New
York (Figure 2 - studyar.jpg), a representative portion of the SLP. It
consisted of 12,393 km2, mostly in St.
Lawrence and Jefferson Counties. The region is generally flat and low-elevation,
dominated by second-growth deciduous and mixed forest fragments scattered
among actively managed hayfields and croplands. The easternmost portions
of the study area gradually transition into the Adirondack highlands,
and therefore have higher elevations and greater forest cover. Recent
trends of farmland abandonment have allowed extensive shrub layers and
deciduous forest patches to develop, reducing the amount of habitat for
grassland nesting species while at the same time increasing habitat for
forest-dependent species. Added to this mosaic of different habitat types
are numerous scattered marshes and forested wetlands, which enhance further
the diversity of bird species in the region.
Methods
I collected bird presence / absence data from 1005 sample points, and
determined the value of various biotic and abiotic habitat variables at
each point using GIS. I then used multiple logistic regression to develop
a unique predictive model for each species, that predicts occurrence based
on a statistically significant set of the biotic and abiotic habitat variables.
For example, the model for wood thrush includes land cover type, mean
patch perimeter and precipitation, and can estimate probability of occurrence
at any location by inputting the value of those habitat variables. Finally
I used a geographic model to generate maps showing probability of occurrence
for each species at each location in the landscape.
Results
Four "high priority" species have been identified in the SLP
Bird Conservation Plan: the grassland-breeding bobolink and the forest
specialists eastern wood-pewee, veery, and wood thrush. For each species,
every 90 meter pixel in the GIS map was assigned a probability value by
applying the logistic regression equation to the corresponding values
of environmental variables. A map of predicted probability of occurrence
was generated for each of the pr
iority
species:
Figure 3 - Predicted distribution for bobolink (bobo.jpg)
Figure 4 - Predicted distribution for eastern wood-pewee (ewpe.jpg)
Figure 5 - Predicted distribution for veery (veer.jpg)
Figure 6 - Predicted distribution for wood thrush (woth.jpg)
Implementation of the Bird Conservation Plan
Management of the landscape for species with conflicting habitat requirements
may be easily achievable in the St. Lawrence Plain. A simple division
of grassland-focused and forest-focused management from west to east along
gradients of increasing elevation, precipitation and forest cover will
allow coexistence in the region of both forest and grassland species.
In the more agricultural areas, forest patches may be cut to increase
size of fields, and in the more forested areas, fields should be allowed
to grow up into forest. As these fields revert to forests, they will become
shrub habitat suitable for golden-winged warblers for a period of years.
There should thus be a continual balance of cut forests planted with grasses
and abandoned fields reverting to shrubs. Since different forest species
prefer different forest age classes and patch sizes, there should be representative
forest habitat along a continuous gradient of successional stages.
Printable Version (Adobe
Acrobat Reader) If you do not have pictures in the printable
document then you need to down load a more recent version of Acrobat
Reader.
About the
Author: Daniel L. Wells
Dan Wells - B.S. Biology from Gettysburg College, M.S. Environmental and
Forest Biology, SUNY-ESF. My research interests involve using GIS to model
bird-habitat relationships, and general wildlife habitat conservation
issues. I plan to ultimately pursue a Ph.D. in a related topic.
Contact info:
Daniel L. Wells
dlwells@syr.edu
|
|