MAPPING OAK WILT IN TEXAS

1
MAPPING OAK WILT IN TEXAS

• Amuche Ezeilo
• Wendy Cooley

2
OAK WILT (Ceratocystis fagacearum)

• Oak wilt is an arboreal disease that affects oaks
  in Texas and the Northeastern part of the U.S.
• Central Texas has been the hardest hit-thousands
  of oak trees have died over the past 20 years

3
DISTRIBUTION IN THE U.S.
Figure 1.  2005 Oak wilt distribution map in the
United States (USDA Forest Service)
4
DISTRIBUTION IN TEXAS
Fort Worth
Dallas
College Station
Austin
Houston
San Antonio
Figure 2. Oak wilt coverage in Texas (The Texas
Forest Service)
5
WHAT IS OAK WILT

• Oak wilt is a vascular fungal disease that
  develops in the water conducting vessels (xylem)
• The fungus plugs up the vessels, reducing water
  flow in trees
• Due to a lack of water, the tree begins to wilt
  and often times die
• All oaks are vulnerable but red oaks are more
  susceptible than white oaks

6
TRANSMISSION ROUTE 1

• One method of transmission is through root grafts
• Oak trees, esp. live oaks, tend to grow in large
  groups
• Roots in these groups are all interconnected
  through root grafting
• Therefore, it is easy for an infected oak to pass
  the disease to healthy oaks
• Grafting can also be between live oaks and red
  oaks

7
TRANSMISSION ROUTE 2

• The other method of transmission is through an
  insect vector
• Fungal mats produced on red oak bark emit an odor
  that attracts sap feeding insects of the
  Nitidulidae family as well as the Oak Bark Beetle
• Beetles carry fungal spores on their bodies from
  the spore mat of an infected tree to a fresh
  wound on a healthy oak
• The beetle feeds on the sap from a fresh wound of
  a healthy oak and, thus, spreads the infection to
  the healthy tree

8
CURE?

• There is no known cure for oak wilt
• Prevention is the key to fighting this disease
• Early detection and rapid removal of infected
  trees including breaking grafted roots
• Avoid wounding oak trees and when wounding cannot
  be avoided, paint immediately with pruning paint
• Cutting deep trenches around infection centers

9
OAK WILT SUPPRESSION PROJECT

• Created by the Texas Forest Service to detect oak
  wilt centers
• They conduct aerial survey flights annually over
  59 counties to locate possible centers
• These centers are then confirmed on ground
• Using remote sensing on current aerials will help
  TFS to classify these areas
• Data used were 1 meter orthophotos from 2004,
  Kerr County, after resizing

10
AIMS

• Detect areas of Oak Wilt in Kerr County
• Classify and map these areas
• Compare results of various classifications
• Thus enabling easier monitoring and control
• of the disease

11
METHODS

• Supervised and Unsupervised ENVI Methods
• Supervised makes use of researchers a priori
• knowledge.
• Training areas of gray/grayish magenta created,
  representing
• dead or severely affected forest.
• This training area spectral information is input
  to maximum
• likelihood technique
• Which determines probability of each image pixel
• belonging in the training areas, and therefore of
  each pixel
• being either healthy or diseased

12
METHODS contd

• Unsupervised These methods use only
• statistical techniques to classify the image
• Two techniques
• 1. K-Means Clustering
• 2. Isodata

13
METHODS_K-MEANS

• K-Means Clustering
• Clustering analysis, requiring analyst to
• select of clusters
• Technique then arbitrarily locates this
• and iteratively repositions them until optimum
• separability is achieved
• (Univ of Lethbridge)

14
METHODS_ ISODATA

• Iterative Self-Organizing Data Analysis Technique
• Iterative-repeatedly performs entire
  classification and recalculates statistics.
• Self-organizing refers to way in which it locates
  inherent data clusters.
• Minimum spectral distance formula is used to form
• clusters
• (Univ of Lethbridge)

15
ISODATA contd

• Means shift with each iteration
• Until either
• 1. Maximum of iterations achieved, OR
• 2. Maximum percentage of unchanged pixels has
• been reached between 2 iterations
• (Univ of Lethbridge)

16
K-Means15 Means Selected, 3 IterationsSample
Location
17
Same Area on Image
18
RESULTSIsodata3 Iterations, Sample Location
19
Same Area on Image
20
Supervised ClassificationMaximum Likelihood,
Sample Location
21
Same Area on Image
22
Discussion

• Comparisons made by observing linked images of
  each classification and orthophoto
• Then determining which classification best
• fit the affected orthophoto vegetation

23
Summary

• Supervised maximum likelihood classification
  seems to best classify the data
• Unsupervised Isodata classification was
• second best
• Thirdly, Unsupervised K-Means classification
• However, no methods could separate water from
  diseased vegetation