|
|
Project Sponsor
National Park
Service: Fire Island National Seashore
USGS National Park Service Vegetation Mapping
Program
Status
Complete
Cooperators
Fire Island National Seashore
NatureServe
New
York Department of Environmental Conservation - Division of Natural Heritage
USGS National Park Service Vegetation
Mapping Program
Abstract
|
In 1994, The United
States Geological Survey (USGS) began the National Park Service (NPS)
Vegetation Mapping Program in cooperation with the NPS. The vegetation mapping project at
Fire Island National Seashore (FIIS) was initiated in June of 1999. The Conservation Management Institute
at Virginia Tech was contracted to complete the photointerpretation, accuracy
assessment, and fieldwork stages of the project. The Nature Conservancy (now NatureServe) along with the
New York Natural Heritage Program (NYNHP), were subcontracted to complete
vegetation sampling and develop a classification system for the FIIS. The vegetation
communities and spatial patterns on Fire Island, as on all barrier islands,
is a direct result of dynamism.
The forces of sand deposition, storm-driven over wash, salt spray, and
surface water all play major roles in affecting vegetation distribution. More recently, disturbance from both
humans and white-tailed deer have impacted vegetation communities on Fire
Island. The vegetation-mapping
project at FIIS follows the standards and general procedures of other NPS
Vegetation Mapping Program projects.
The same standards, type of photography, vegetation classification
system, and field procedures were employed for FIIS. There were, however, some unique
conditions at FIIS that called for some changes in sampling as well as
techniques that may not be applicable in other regions. True-color photographs
were used to delineate and interpret vegetation polygons at Fire Island. The detailed resolution of the aerial
photography allowed us to identify and map many objects that would have been
difficult to identify from digital orthoquads (DOQs) or other smaller-scale
data. We opted to map at a
minimum mapping unit of 0.25-ha to address the needs of the park managers,
but could often discern objects and vegetation polygons well below this
threshold. NatureServe provided a
preliminary classification that was used in the initial delineation
phase. In October of 1999, CMI
field staff and photointerpreters visited Fire Island to familiarize
themselves with the dominant vegetative species on the island. The data collected during the initial
reconnaissance mission was used to better identify areas and vegetation types
that required comprehensive vegetation measurement. Field data collection was completed with two types of
plots. The first, which was
completed by the NatureServe ecologists, involved detailed data collection on
representative plots for each vegetation type encountered. The second type, completed by CMI
field staff, involved a more qualitative classification of vegetation types
observed on the ground. The plot
sampling methodology used by the NYNHP on Fire Island follow the methodology
developed by The Nature Conservancy and the network of Natural Heritage Programs. The second data point set was
selected in the field, and allowed us to maximize the contribution of each
point to the knowledge base of the photointerpreter. Polygon boundaries were
delineated from the georeferenced photos on-screen through heads-up
digitizing. Once the final
classification was available, all polygons were assigned a vegetation
association. A photointerpreter
key was employed for consistency.
Polygons were re-delineated and assigned as necessary. Polygons were also given values for
height, density, and distribution pattern. The final vegetation map layers were converted to ArcInfo
and all spatial errors were cleaned. An accuracy assessment
effort was completed for FIIS in accordance with the NPS vegetation mapping
specifications. The accuracy
assessment phase was carried out similar to other NPS vegetation mapping
projects. We used guidelines
from The Nature Conservancy (1994) to determine the number of accuracy
assessment points needed for Fire Island. A target number of 665 accuracy assessment points was
established, representing 579 polygons. A total of 5 broadly
defined vegetation groups were encountered on Fire Island and the William
Floyd Estate. These include salt
marshes, dune grasslands, dune shrublands, interdunal swales, and forests /
shrublands. These types were
further classified into 27 different associations. Six of the associations at Fire Island National Seashore are
broadly classified as forest types (“Forest Class” in the National Vegetation
Classification hierarchy), one as Woodland, five as Shrubland, two as
Dwarf-Shrubland, twelve as Herbaceous, and one as Sparse Vegetation. These
associations are representative of a typical middle and upper Atlantic
barrier island system. A total of 39 classes of
land cover were mapped on Fire Island and the William Floyd Estate. These are comprised of 24 types
mapped to NVCS association, 1 complex of 2 NVCS alliances, and 14 non-NVCS
classes. Four associations were
identified on Fire Island and the William Floyd Estate but do not appear on
the map due to their rarity, small relative size, and/or difficulties in
identifying them with aerial photography. Spatial accuracy was
assessed by collecting “map points” on the ground along with vegetation
fieldwork. A total of 47 points
were used to assess the spatial accuracy of the vegetation map. The mean error distance was found to
be 4.42 m (± 4.94 m). Errors
distances ranged from 0.00 m – 30.0 m.
When the single 30 m error point was removed (assumed to be an outlier),
the mean error distance was 3.86 m ((± 3.18 m) with a range of 0.00 m – 14.09 m. A total of 495 points
were used to assess the thematic accuracy of the vegetation map. Initial analysis showed a relatively
low overall accuracy of 57.6%.
Further examination revealed some fairly serious discrepancies between
vegetation types classified in the field and those depicted on the map. After review, 428 points were
available for use in the accuracy assessment. Of these 329 were located in polygons larger than 0.25 ha
and 99 were found in smaller polygons.
We used a fuzzy set matrix to evaluate the severity of error between
each class and every other class.
The fuzzy value was assigned based on the similarity between types
observed on Fire Island. We present
accuracy estimates for levels 5, 4 and 3 in the classification. The level 5 contains only those
points where the observed type matched the mapped type exactly. The level 4 assessment considers both
level 5 and 4 as being correct.
The level 3 assessment similarly considers levels 5,4, and 3. The overall accuracy (and Kappa
index) for the map at level 5 was 66.3% (64%). The level 4 and 3 accuracy was 78.1% (77%) and 87.5% (87%)
respectively. The vegetation of Fire
Island is not much different than types observed in similar NPS areas (e.g.,
Asseteague Island). There are
some unique associations on Fire Island not seen elsewhere. It was apparent early in the project
that a minimum mapping unit of 0.25 hectares was insufficient to adequately
capture the structural and vegetative diversity of the island. Attempts were made to delineate
smaller discrete polygons. The
final vegetation classification includes several types that were not
delineated on the vegetation map.
Many of these types are extremely rare on Fire Island (or the Floyd
Estate) or are indistinguishable from the more prevalent associations on the
island. This project has identified several potential points for improvement or additional study. The base information for this project was more than adequate for the task of mapping vegetation on a barrier island. The barrier island is constantly changing. Dynamism is the foundation on which these species and communities are built. Updates to this product should be completed on a regular basis to ensure these changes are included. |
Products
Vegetation Map
(ArcView shapefile plus metadata and legend) – FIIS_vegshape.zip
Photointerpretive Key (HTML document) - FIIS_Photo_Interp.htm
Final Report (PDF format) – FIIS_report.pdf
Contacts
Scott Klopfer, Conservation Management
Institute
Lesley
Sneddon, NatureServe