Remote Sensing Classification: Once gross spatial and thematic errors were
corrected in the vector data, Erdas Imagine was employed to perform a series of
unsupervised and supervised classifications of each SPOT image with the
corrected polygon data as a guide.
Process Date: Not complete
Vector Data Review, Editing, and Classification: The Cooperative Land Cover Map
(CLC) served as the foundation dataset for revision efforts. Polygon vector data was
compared against high resolution Digital Ortho Quarter Quads (DOQQ) and Google
Earth imagery. Google Earth?s imagery was the most effective imagery available for
visual data editing, due to its clarity, resolution, loading speed and data age. Vector
data of individual land cover classes were converted to *.KML format for use in
Google Earth. Errors identified through visual review were manually corrected. The
most common errors encountered included incorrect boundaries, mislabeling of
classes, hard edges between classes, and features containing multiple classes.
Sliver polygons were eliminated based on two criteria: 1) Area 2) Preimeter to Area
Ratio 1) Polygons with an area greater than or equal to 110 square meters were
selected and dissolved into the largest adjacent polygon utilizing the Eliminate tool.
110 square meters was selected because the final raster will have a 10 meter cell
size (100 square meters per cell) and the additional 10 square meters help to further
remove sliver polygons that would require significant time to remove manually during
the editing process. 2) Polygons with large ratios are less likely to accurately
describe a significant landcover at the statewide scale. Therefore, polygons with a
perimeter to edge ratio greater than or equal to 0.5 were eliminated into adjacent
polygons sharing the longest edge.
Process Date: not complete
Assemblage of Data Sources into Final Map: Tthe data was separated into 3
components for assembly into statewide land cover: 1) Local Source data, which
consisted of all local sources with confidence category 1 through 3; 2) FNAI
Review data, which consisted of all datasets that were inspected and classified
through aerial photo review; and 3) FLUCCS. The SWFWMD published a new
version of FLUCCS based on 2008 photography in spring 2010. Although we used
2007 FLUCCS for aerial photo review and comparative analyses in that district, we
incorporated the 2008 data in the final land cover map. We converted all datasets
into 15 m ESRI grids and combined them based on the following rules: 1) Local
Source data with confidence category 1 and 2 superseded FNAI Review data; 2)
FNAI Review data superseded Local Source data with confidence category 3; 3) all
Local Source 1 through 3 and FNAI Review data superseded FLUCCS.
Process Date: 20100101
Aerial Photography Review of Focal Communities: Areas within existing source data
in categories 1 through 3 (non-ancillary sources) were excluded from the set of
polygons to be reviewed. Scrub, scrubby flatwoods, sandhill, dry prairie and mesic
flatwoods (in SWF and SF only) were reviewed simultaneously as a single set of
review polygons. Review polygons as well as proximal areas were inspected with
the latest high resolution aerial photography (2006 - 2009) and other ancillary data
sources including aerial photography from 2004, 1999 and 1995, topographic maps,
county soils maps and other land cover datasets. Areas were reviewed at a scale of
1:5000 with a minimum mapping unit of 0.5 acres with exception to include smaller
polygons for scrub and pine rockland. Polygons were spatially edited and new
polygons were delineated where necessary to identify focal communities and then
assigned the polygon a land cover type. Polygons were deleted from the set of
review polygons that did not represent priority communities and were otherwise
correctly classified. A land cover type was assigned to polygons classified as
FLUCCS Coastal Scrub, Xeric Oak, Sand Pine, or Longleaf Pine - Xeric Oak; in
addition almost all review polygons in the SWF and SF districts were assigned a
land cover type. Any deleted polygon will default to its FLUCCS class in the final
land cover map. FNAI biologists familiar with the focal communities both on the
ground and through aerial photo interpretation performed the initial polygon
inspections. A second reviewer then re-inspected the polygons that were assigned
as one of the focal communities. A locations were checked from the FNAI element
occurrence database that reference scrub, scrubby flatwoods, sandhill or dry prairie.
Areas were identified that appeared to be functioning as viable natural communities.
Areas that were historically scrub or sandhill but are now disturbed so that they likely
no longer support their characteristic ecological elements or that have succeeded
to another natural community type were excluded or classified as another land
cover type. Many former sandhills were reclassified as successional hardwood
forest. Pine plantation was reclassified as scrub or sandhill where it appeared to
function ecologically as a natural community. This was especially true of planted
sand pine scrub which can tolerate a high degree of disturbance. Aerial
photographs from 1995 and 1999 were examined to help determine the level of past
ground disturbance. Small patches within residential areas were not included,
although if there appeared to be functional large patches within low density or rural
residential areas we included them. Only obvious patches of scrubby flatwoods
were mapped. This community was sometimes difficult to distinguish from scrub and
we did not follow strict criteria for distinguishing the two. For dry prairie we strictly
followed the FNAI definition of treeless areas of low shrubs and grasses within the
buffered historic dry prairie extent. Many prairie-like areas are pine flatwoods in
which trees have been removed. To determine dry prairie from flatwoods we
considered geographic position, shrub patterns, proximity of wetlands and overall
landscape context.
Process Date: 20100101
The Cooperative Land Cover map integrates data from multiple sources. All sources
were crosswalked into the Florida Land Cover Classification System (Kawula 2014)
All source datasets were received and processed as vector data. A set of standard
geoprocessing and topology operations were employed in ArcGIS 9.3, 10.1 to
ensure no overlapping features within or among datasets. All data were projected
into the Florida Albers custom coordinate system with NAD 1983 HARN datum. A
minimum mapping unit of 0.5 acres was applied, and each polygon < 0.5 acres was
dissolved into its largest neighboring polygon except for scrub, pine rockland and
upland glade polygons for which we applied a minimum mapping unit of 0.1 acres.
Finally, lines between neighboring polygons with the same classification were
dissolved. Based on the review of each dataset, other modifications were made as
described in: Florida Natural Areas Inventory, 2010. Development of a Cooperative
Land Cover Map: Final Report (available from FNAI upon request). Explanation of
confidence categories: Datasets were evaluated based on metadata, discussions
with data providers and a general review of the spatial accuracy and classification.
Based on this review, a confidence category was assigned to each dataset that
indicated how or if the dataset, or certain classes within the dataset, would be
integrated into the final land cover map. A confidence category of 1 indicates the
highest level of confidence; these data spatially superseded all other intersecting
sources. Category 2 data took precedence over statewide datasets (FLUCCS,
FLVEG) but did not supersede category 1. Category 3 data were used with review
and revision. Category 4 data were used to identify additional areas for aerial photo
review and help interpret classification during the review process; these data,
however, were not directly integrated into the final map.
Process Date: 20100101
Edge Matching: Following map classification, we conducted further visual
inspection of classified areas for consistency, errors, and edge matching between
assembled data sets.
Process Date: 20140101
Topology: Rules were established to ensure that: 1) all areas within the mapping
area (i.e. Florida) are covered by a land class (polygon); and 2) that there is only one
land class (polygon) defining a given area. Two topological rules were created to
ensure these requirements are met: 1) Land Classes Must Not Have Gaps 2) Land
Classes Must Not Overlap For errors that result from ?Land Classes Must Not Have
Gaps?, a polygon was created to fill that gap. This new polygon created NULL
information in the attributes and must either: 1) be merged with an appropriate
adjacent polygon sharing the same land classification characteristics or 2) be given
a land classification indicating its uniqueness in comparison with adjacent polygons.
For errors that resulted from ?Land Classes Must Not Overlap?, areas were
identified that have more than one polygon. The overlapping area was merged with
the most appropriate land class, thereby removing the overlap.
Process Date: 20140101
GeoPlan downloaded this data in June 2017 from:
http://fnai.org/LandCover.cfm
-There are no new values or descripts.
-A reclass (matching v3.1) was done in order to create an 8 bit version instead of
the original 32 bit.
-Thre was not a NoData or 0 value present, so one was added (and calculated to
0) for analysis purposes.
Process Date: 20170616
Metadata imported.
Process Date: 20170616
|