Peer-Reviewed Articles
897 Land-cover Mapping in the Brazilian Amazon Using
SPOT-4 Vegetation Data and Machine Learning Classification Methods
João M. B. Carreiras, José M. C. Pereira, and Yosio E. Shimabukuro
Abstract Download Full Article
The main objective of this study is to evaluate the feasibility
of deriving a land-cover map of the state of Mato
Grosso, Brazil, for the year 2000, using data from the 1 km
SPOT-4 VEGETATION (VGT) sensor. For this purpose we
used a VGT temporal series of 12 monthly composite
images, which were further transformed to physicalmeaningful
fraction images of vegetation, soil, and shade.
Classification of fraction images was implemented using
several recent machine learning developments, namely,
filtering input training data and probability bagging in a
classification tree approach.
A 10-fold cross validation accuracy assessment indicates that filtering and probability bagging are effective at increasing overall and class-specific accuracy. Overall accuracy and mean probability of class membership were 0.88 and 0.80, respectively. The map of probability of class membership indicates that the larger errors are associated with cerrado savanna and semi-deciduous forest.
911 Socioeconomic-Vegetation Relationships in Urban,
Residential Land: The Case of Denver, Colorado
Jeremy Mennis
Abstract Download Full Article
This research investigates the relationship between socioeconomic
status and remotely sensed vegetation intensity in
residential land in the Denver, Colorado metropolitan area.
Land-cover data derived from aerial photography and normalized
difference vegetation index data (NDVI) derived from
Landsat ETM™ imagery were integrated with U.S. Bureau
of the Census tract-level data and analyzed using choropleth
mapping and multivariate statistics. Association rule mining,
a data mining technique, is used to explore nonlinear relationships
among variables. Results indicate that higher vegetation
intensity is associated with socioeconomic advantage
in both sparsely populated, large lot suburban developments,
as well as in older, urban neighborhoods. This pattern likely
reflects residents’ ability to pay for the cost of maintaining
high vegetation intensity, suburban lawn ecosystem vegetation
in a semi-arid grassland environment. Additionally,
residential choices may be limited by a home price structure
that is closely related to the concentration of vegetation in
the residential landscaping.
923 Isolating Individual Trees in a Savanna Woodland Using
Small Footprint Lidar Data
Qi Chen, Dennis Baldocchi, Peng Gong, and Maggi Kelly
Abstract Download Full Article
This study presents a new method of detecting individual
treetops from lidar data and applies marker-controlled
watershed segmentation into isolating individual trees in
savanna woodland. The treetops were detected by searching
local maxima in a canopy maxima model (CMM) with variable
window sizes. Different from previous methods, the
variable windows sizes were determined by the lower-limit
of the prediction intervals of the regression curve between
crown size and tree height. The canopy maxima model was
created to reduce the commission errors of treetop detection.
Treetops were also detected based on the fact that they
are typically located around the center of crowns. The
tree delineation accuracy was evaluated by a five-fold,
cross-validation method. Results showed that the absolute
accuracy of tree isolation was 64.1 percent, which was
much higher than the accuracy of the method, which only
searched local maxima within window sizes determined
by the regression curve (37.0 percent).
933 Error Assessment in Two Lidar-derived TIN Datasets
Miao-Hsiang Peng and Tian-Yuan Shih
Abstract Download Full Article
An accuracy assessment of two lidar-derived elevation
datasets was conducted in areas of rugged terrain (average
slope 26.6°). Data from 906 ground checkpoints in various
land-cover types were collected in situ as reference points.
Analysis of the accuracy of lidar-derived elevation as a
function of several factors including terrain slope, terrain
aspect, and land-cover types was conducted. This paper
attempts to characterize vegetation information derived from
lidar data based on variables such as canopy volume, local
roughness of point clouds, point spacing of lidar ground
returns, and vegetation angle. This information was used to
evaluate the accuracy of elevation as a function of vegetation
type. The experimental results revealed that the accuracy
of elevation was considerably correlated with five
factors: terrain slope, vegetation angle, canopy volume, local
roughness of point clouds, and point spacing of lidar ground
returns. The results show a linear relationship between
the elevation accuracy and the combination of vegetation
angle and the point spacing of ground returns (r2 > 0.9).
The combination of vegetation angle and point spacing of
ground returns explains a significant amount of the variability
in elevation accuracy. Elevation accuracy varied with
different vegetation types. The elevation accuracy was also
linearly correlated with the product of the point spacing of
ground returns and the tangent of the slope (r2 > 0.9). A
greater product value implies a greater elevation error. In
addition, with regard to terrain aspect, one dense dataset
with extra cross-flight data revealed a lesser impact of
aspect on elevation accuracy.
949 Automated Forest Area Estimation Using Iterative
Guided Spectral Class Rejection
Rebecca Musy, Randolph Wynne, Christine Blinn, John Scrivani,
and Ronald McRoberts
Abstract Download Full Article
USDA Forest Service Forest Inventory and Analysis (FIA) forest
area estimates were derived from 4 Landsat ETM+ images
in Virginia and Minnesota classified using an automated
hybrid classifier known as Iterative Guided Spectral Class
Rejection (IGSCR). Training data were collected using regiongrowing
initiated at random points within each image. The
classified images were spatially post-processed using five
different techniques. Image accuracy was assessed using the
center land-use of all available FIA plots and subsets containing
plots with 50, 75 and 100 percent homogeneity.
Overall accuracy (81.9 to 95.4 percent) increased with homogeneity of validation plots and decreased with fragmentation (estimated by percent edge; r2 = 0.932). Filtering effects were not consistently significant at the 95 percent level; however, the 3 x 3 majority filter significantly improved the accuracy of the most fragmented image. The now-automated IGSCR is a suitable candidate for operational forest area estimation, with strong potential for use in other application areas.
961 Accuracy Assessment of Lidar Saltmarsh Topographic
Data Using RTK GPS
Juana M. Montané and Raymond Torres
Abstract Download Full Article
An evaluation was completed to compare the accuracy of
lidar (Light Detection and Ranging) against a statistically
representative array of Real-Time Kinematic (RTK) GPS data
in a low gradient, vegetated Southeastern U.S. salt marsh. In
order to discern potential bias, analyses were carried out
separately on the platform-only data, the creek-only data
and then the combined datasets. Lidar data were found to
overestimate the RTK GPS topographic data by an overall
average of only 7 cm. Additionally, these data showed little
effect from the dominant macrophyte vegetation within the
lidar footprint. From this evaluation, 7 cm appears to be an
appropriate vertical adjustment factor for using lidar data in
low gradient salt marshes. However, local ground control
will continue to be crucial in studies of intertidal environments
incorporating airborne laser data collection.