KNF_RemoteSensingVegetationMonitoringFinal

FOREST SERVICE
Land Cover Change on the
Kisatchie National Forest,
1999-2009
As measured using satellite imagery
Converse Griffith
5/10/2012

1
Abstract
The 1999 revised forest land and resource management plan calls for the monitoring of changes in land cover on
the Kisatchie National Forest (KNF), in order to determine if management actions such as prescribed fire and thinning
are bringing the KNF closer to the desired future conditions described in the plan for different portions of the KNF.
Budgetary and time constraints make on the ground monitoring difficult to accomplish. An alternative approach to
monitoring, using imagery from the Landsat 5 satellite is described in this document.
Landsat images from 1999 and 2009 were obtained from both the dormant and growing season for each year.
The images were processed in order to remove atmospheric effects, so that the light in the images corresponded to the
light actually reflected from the earth’s surface. Images from a particular date were combined into a single image that
covered the entire KNF. These images were then clipped to the administrative boundaries, and then further clipped to
remove private ownership within the administrative boundaries and to remove a 30 meter (about 100 foot) buffer
around the roads.
Landsat images have collect information in discrete bands, some of which are different colors of visible light,
others of which are infrared light. The original Landsat images also have bands that are not relevant to measuring
vegetation cover. For each year, the relevant visible light and infrared light bands from processed growing season and
dormant season images were combined into a single new image. These new images with the combined growing and
dormant season bands were the images used in classifying the KNF according to vegetation cover type.
A computer algorithm performed unsupervised classifications on the images, which were initially divided into 22
categories. Each category was then manually assigned a particular land cover class. The land cover classes were
hardwood forest, mixed hardwood-pine forest, pine forest, herbaceous vegetation, not vegetation (which was mainly
water), shrubs, and bare earth. To improve accuracy, an unsupervised classification was performed again on these land
cover classes (cluster busting); the resulting categories were again assigned to one of the land cover classes listed above.
Once the classification steps were complete, acres were calculated for each of the cover classes. The raster files
of the 1999 and 2009 land cover classes were also clipped by district and sub-management area, and acres calculated for
the clipped images. Changes in cover, such as increases or decreases in pine forest acreage, were calculated as the
change in the 2009 acres from the 1999 acres.
To assess accuracy of the classifications, Geographic Information System files of random points, stratified by
land cover class and district, were constructed. Ocular inspection of aerial photography from 1998, 2009, and 2010 was
used to determine if the area around a random point was correctly or incorrectly classified. On the Kisatchie National
Forest as a whole, user’s accuracies for hardwood forest ranged from 65 to 67%, for pine forest from 81 to 90%, and a
disappointing 35 to 37% for mixed hardwood-pine forest. Accuracies on particular districts could be noticeably different
from these figures.
On the KNF as whole, forest cover increased in the hardwood and mixed hardwood-pine categories. Pine forest
acres decreased. The non-forested land cover classes occupied a much smaller portion of the forest than did any of the
forested land cover classes. The overall changes for the entire forest can mask changes in the opposite direction on
particular districts. For example pine forest acreage increased on the Calcasieu and Kisatchie districts, mixed hardwood-
pine forest area declined on both, and hardwood forest acres declined on the Calcasieu district.
The changes in land cover classes, when organized by district and sub-management area, were not always in
accord with the desired future conditions for a particular-sub-management area. The results suggest that there may be
challenges regarding sub-management area 5CL, among others, on some districts.
Provided that its limitations are understood, remote sensing could play a useful role in monitoring. The coarse
analysis presented here has some utility and could be repeated regularly, perhaps at the time of forest plan revisions.
More resource-intensive options include the establishment of permanent field plots for a more robust accuracy
assessment, and also to allow for the use of other remote sensing techniques such as supervised classification. Other
monitoring options could include contracting for the more frequent data collection from Forest Inventory and Analysis
plots located on the KNF.

2
Table of Contents
Abstract...................................................................................................................................................................................1
List of Equations......................................................................................................................................................................2
List of Figures ..........................................................................................................................................................................3
List of Tables ...........................................................................................................................................................................5
Introduction ............................................................................................................................................................................7
Methods..................................................................................................................................................................................7
Data.....................................................................................................................................................................................7
Image Processing ................................................................................................................................................................7
Accuracy Assessment..........................................................................................................................................................8
Results...................................................................................................................................................................................10
Classification Areal extents...............................................................................................................................................10
Accuracy Assessment Results ...........................................................................................................................................12
Accuracy Assessment for the entire KNF......................................................................................................................12
Accuracy Assessments for individual districts ..............................................................................................................17
Discussion..............................................................................................................................................................................21
Acknowledgements...............................................................................................................................................................25
References ............................................................................................................................................................................25
Appendix: Discussion, figures, and data tables relating to sub-management areas............................................................26
List of Equations
Equation 1 Normalized Difference Vegetation Index (NDVI) ......................................................................................................8
Equation 2 Standard Error......................................................................................................................................................10

3
List of Figures
Figure 1 shows the classification results for the Evangeline Unit of the Calcasieu Ranger District. ..................................................12
Figure 2 below shows the classification Results for the Vernon Unit of the Calcasieu Ranger District. ............................................13
Figure 3 below shows the classification results for the Caney District. Note the change in the classification in the wetland area
northwest of Corney Lake in the easternmost area of the Caney. The rainfall data in Table 11 may explain the changes in
water cover.............................................................................................................................................................................14
Figure 4 below shows the classification results for the Catahoula District and also the southwestern portion of the Winn District.
Note the absence of areas classified as water in the 2009 map...............................................................................................15
Figure 5 below shows the classification results for the Kisatchie Ranger District.............................................................................16
Figure 6 below shows the classification results for the Winn Ranger District. Note the absence of areas classified as water in the
2009 map................................................................................................................................................................................17
Figure 7 below is one of the random points used to assess the accuracy of the herbaceous category on the Catahoula District. The
vegetation on this spot in the pipeline right of way was readily recorded as correctly classified as herbaceous vegetation...22
Figure 8 below is one of the random points used to assess the accuracy of the mixed hardwood-pine forest classification on the
Calcasieu District. Given the different colors of the tree canopies within the 15 meter radius buffer circle, it appears that this
particular location was correctly classified, but judgment had to be exercised regarding if the proportions of hardwoods and
pines were such that it was mixed hardwood-pine forest.......................................................................................................22
Figure 9 below plots the user’s accuracies for each combination of Ranger District and year by classification category. ................23
Figure 10 below shows monthly mean daily minimum and monthly mean daily maximum temperatures for 1999 and 2009 from
Shreveport Louisiana. The graph also displays normal (30 year monthly average) minimum and maximum temperatures for
each month. The data was downloaded from the National Climatic Data Center website. ....................................................24
Figure 11 below is a plot of site index against the change in hardwood acres. Each point is the data from a single ranger district.24
Figure 12 below shows the changes in pine forest acres in sub-management area 1C. The error bars were calculated by applying
the user’s accuracies for pine forest on the KNF from 1999 and 2009 respectively to the appropriate number of acres for a
particular sub-management area, district and year.................................................................................................................28
Figure 13 below shows the changes in mixed hardwood-pine forest acres in sub-management area 1C. The error bars were
calculated by applying the user’s accuracies for mixed hardwood-pine forest on the KNF from 1999 and 2009 respectively to
the appropriate number of acres for a particular sub-management area, district and year. ...................................................28
Figure 14 below shows the changes in hardwood forest within sub-management area 1C. The error bars were calculated by
applying the user’s accuracies for hardwood forest on the KNF from 1999 and 2009, respectively, to the appropriate number
of acres for a particular sub-management areas, district, and year.........................................................................................28
Figure 15 below shows the changes in pine forest area in sub-management area 3BL. The error bars were calculated by applying
Figure 16 below shows the changes in mixed hardwood-pine forest area in sub-management area 3BL. The error bars were
calculated by applying the user’s accuracies for mixed-hardwood pine forest on the KNF from 1999 and 2009 respectively to
Figure 17 below shows the changes in hardwood forest within sub-management area 3BL. The error bars were calculated by
Figure 18 below shows the changes in pine forest area in sub-management area 3BM. The error bars were calculated by applying

4
Figure 19 below shows the changes in mixed hardwood-pine forest area in sub-management area 3BM. The error bars were
Figure 20 below shows the changes in hardwood forest within sub-management area 3BM. The error bars were calculated by
Figure 21 below shows the changes in pine forest area in sub-management area 3BS. The error bars were calculated by applying
Figure 22 below shows the changes in mixed hardwood-pine forest area in sub-management area 3BS. The error bars were
Figure 23 below shows the changes in hardwood forest within sub-management area 3BS. The error bars were calculated by
Figure 24 below shows the changes in pine forest area in sub-management area 5CL. The error bars were calculated by applying
Figure 25 below shows the changes in mixed hardwood-pine forest area within sub-management area 5CL. The error bars were
calculated by applying the user’s accuracies for mixed hardwood-pine forest on the KNF from 1999 and 2009, respectively, to
the appropriate number of acres for a particular sub-management area, district, and year. ..................................................32
Figure 26 below shows the changes in hardwood forest within sub-management area 5CL. The error bars were calculated by
Figure 27 below shows changes in the area of pine forest located in sub-management areas 11DM. The error bars were calculated
by applying the user’s accuracies for the KNF from 1999 and 2009 respectively to the appropriate number of acres for a
Figure 28 below shows the changes in mixed hardwood-pine forest area located in sub-management area 11DM. The error bars
were calculated by applying the user’s accuracies for the KNF from 1999 and 2009 respectively to the appropriate number of
acres for a particular sub-management area, district and year. ..............................................................................................33
Figure 29 below shows the changes in hardwood forest within sub-management area 11DM. The error bars were calculated by

5
List of Tables
Table 1 indicates the categories used in the classifications and the colors given to the different categories in the resulting raster
files. ........................................................................................................................................................................................12
Table 2 below indicates the total acres (rounded to the nearest acre) and pixels in each category for the entire Kisatchie National
Forest (KNF) for 1999 and 2009, and the change in acres. The figures in Table 2 are the initial figures before any adjustments
were done for the accuracy assessment..................................................................................................................................15
Table 3 uses the results given in Table 1, but groups the categories in Table 1 into larger, less precise sets. Once again no
weighting adjustments for accuracy have been made.............................................................................................................16
Table 4 below gives the areas for the different classification categories for each district for both 1999 and 2009. The “difference”
column is the result of subtracting the 1999 acres from a particular category from the appropriate 2009 category. The values
in the difference columns are rounded to the nearest acre. Once again no weighting adjustments for accuracy have been
made.......................................................................................................................................................................................18
Table 5 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Kisatchie
National Forest as a whole for the 1999 and 2009 classifications. The overall classification accuracy was 70% in 1999 and
69% in 2009. The kappa value was approximately 0.51 for 1999 and about 0.47 for 2009. The not applicable result for the
2009 producer’s accuracy for shrubs is due to the impossibility of carrying out the calculation, which would require dividing
by zero. No areas were classified as shrubs in 2009, but the accuracy assessment found some locations that appeared to be
covered with shrubs................................................................................................................................................................19
Table 6 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Calcasieu
Ranger District for the 1999 and 2009 classifications. The overall classification accuracy was 75% for both 1999 and 2009.
The kappa value was approximately 0.55 for 1999 and about 0.45 for 2009. The not applicable result for the 2009 producer’s
accuracy for shrubs is due to the impossibility of carrying out the calculation, which would require dividing by zero. No
areas were classified as shrubs in 2009, but the accuracy assessment found some locations that appeared to be covered with
shrubs.....................................................................................................................................................................................19
Table 7 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Caney Ranger
District for the 1999 and 2009 classifications. The overall classification accuracies were 70 % for 1999 and 55% in 2009. The
kappa value was approximately 0.58 for 1999 and about 0.37 for 2009. The not applicable result for the 2009 producer’s
areas were classified as shrubs on the Caney District in 2009, but the accuracy assessment found some locations that
appeared to be covered with shrubs.......................................................................................................................................20
Table 8 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Catahoula
Ranger District for the 1999 and 2009 classifications. The overall classification accuracies were 67% for 1999 and 68% in
2009. The kappa value was approximately 0.45 for 1999 and about 0.50 for 2009. The not applicable result for the 2009
producer’s accuracy for shrubs is due to the impossibility of carrying out the calculation, which would require dividing by
zero. No areas were classified as shrubs in 2009, but the accuracy assessment found some locations that appeared to be
Table 9 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Kisatchie
Ranger District for the 1999 and 2009 classifications. The overall classification accuracy was 62% for 1999 and 69% in 2009.
The kappa value was approximately 0.41 for 1999 and about 0.45 for 2009. The not applicable result for the 2009 producer’s
areas were classified as shrubs in 2009, but the accuracy assessment found some locations that appeared to be covered with
shrubs.....................................................................................................................................................................................21
Table 10 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Winn
Ranger District for the 1999 and 2009 classifications. The overall classification accuracies were 79 % for 1999 and 62% in
2009. The kappa value was approximately 0.65 for 1999 and about 0.42 for 2009. The not applicable result for the 2009

6
producer’s accuracy for shrubs is due to the impossibility of carrying out the calculation, which would require dividing by
zero. No areas were classified as shrubs in 2009, but the accuracy assessment found some locations that appeared to be
Table 11 below provides precipitation data (total water equivalent, in inches) from Shreveport Louisiana (downloaded from the
National Climatic Data Center web site) for 1998, 1999, 2008, and 2009. The precipitation data highlighted in yellow are the
same months the 1999 Landsat scenes were taken, and the green highlighted precipitation data are the same months that
the 2009 Landsat scenes were imaged....................................................................................................................................24
Table 12 below shows the areas (values are acres) occupied by each classification category within each of the 12 sub-management
areas of the Calcasieu Ranger District, and the changes in area between 1999 and 2009. The percentages given for the
forested categories are the percent of total classified acres. Because no areas were classified as shrubs in 2009, zero acres
were counted for this category in 2009 in order to calculate the change in area during the decade. A similar procedure was
used when a particular classification category was not found in a particular combination of sub-management area and year.
................................................................................................................................................................................................34
Table 13 below shows the areas occupied by each classification category within the two sub-management areas of the Caney
Ranger District, and the changes in area between 1999 and 2009. The percentages given for the forested categories are the
percent of total classified acres. Because no areas were classified as shrubs in 2009, zero acres were counted for this
category in 2009 in order to calculate the change in area during the decade. A similar procedure was used when a particular
classification category was not found in a particular combination of sub-management area and year...................................35
Table 14 below shows the areas occupied by each classification category within the sub-management areas of the Catahoula
Ranger District, and the changes in area between 1999 and 2009. Because no areas were classified as shrubs in 2009, zero
acres were counted for this category in 2009 in order to calculate the change in area during the decade. A similar procedure
was used when a particular classification category was not found in a particular combination of sub-management area and
year. The percentages given for the forested categories are the percent of total classified acres. Because the sub-
management areas that coincide with the National Catahoula Wildlife Preserve overlap the boundary between the
Catahoula and Winn Ranger Districts, the GIS polygons representing these sub-management areas were split at the district
boundary and the acres attributed to the appropriate district................................................................................................36
Table 15 below shows the areas occupied by each classification category within the sub-management areas of the Kisatchie
Ranger District, and the changes in area between 1999 and 2009. Because no areas were classified as shrubs in 2009, zero
acres were counted for this category in 2009 in order to calculate the change in area during the decade. A similar procedure
was used when a particular classification category was not found in a particular combination of sub-management area and
year. The percentages given for the forested categories are the percent of total classified acres. .........................................37
Table 16 below (and continued on the next page) shows the areas occupied by each classification category within the sub-
management areas of the Winn Ranger District, and the changes in area between 1999 and 2009. Because no areas were
classified as shrubs in 2009, zero acres were counted for this category in 2009 in order to calculate the change in area during
the decade. A similar procedure was used when a particular classification category was not found in a particular
combination of sub-management area and year. The percentages given for the forested categories are the percent of total
classified acres. Because the sub-management areas that coincide with the National Catahoula Wildlife Preserve overlap the
boundary between the Catahoula and Winn Ranger Districts, the GIS polygons representing these sub-management areas
were split at the district boundary and the acres attributed to the appropriate district. ........................................................38

7
Introduction
The Land and Resource Management Plan of the Kisatchie National Forest (hereafter usually referred to as the
“forest plan”) require the monitoring of vegetation on the Kisatchie National Forest (KNF). The forest plan indicates
desired future conditions for different areas, known as sub-management areas, on parts of the KNF. The object of the
management activities on the KNF (such as prescribed fire, thinning of pine trees, clear cuts of pine species not native to
the area, planting of longleaf pine, and mid-story removal) is to bring the vegetation of the KNF closer to the desired
future conditions. Different sub-management areas may have different desired future conditions. . The forest plan calls
for monitoring of particular aspects of the vegetation on the KNF. For example, measuring the change in hardwoods is
pertinent to forest plan monitoring task numbers 18, 19, 20, 76, and 78 on pages F-2 through F-13 of the forest plan.
Ideally a large set of permanently marked plots would be regularly surveyed in order to assess changes in the vegetation
on the KNF. Budgetary and time limitations have prevented the implementation of such a program.
Remote sensing offers the possibility that some of the grosser changes of the vegetation might be monitored in
a less costly and more time efficient manner. One aspect of vegetation change of particular interest is the change in
forest types within the categories of pine forest, hardwood forest, and mixed hardwood pine forest. Changes in any
direction (for example, from pine to hardwood or vice versa) are all of interest. Also of interest were the areas of
herbaceous vegetation, and changes in acreage of herbaceous vegetation. Finally, it was assumed that a “not
vegetation” category would be needed to create an exhaustive classification.
Within each of the five ranger districts, the forest plan categories different areas based upon the management
goals for each area. These categories are called sub-management areas. Management goals for various sub-
management areas include maintaining and improving Red-Cockaded Woodpecker (RCW) habitat, production of forest
products, or native community restoration (such as longleaf pine restoration). Different districts may have areas that
have the same management goals, and thus are categorized in the forest plan as belonging to the same sub-
management area. Remote sensing may be useful in monitoring how vegetation changes within sub-management areas
are in accord with the goals of the forest plan.
Methods
Data
Satellite scenes taken by the Landsat 5 satellite were used to classify the features of the Kisatchie National
Forest. The scenes have bands corresponding to both visible and infrared bands of light; both categories are relevant for
the classification of land. Scenes from 1999 (the start of the current forest plan) and 2009 (the latest year in which good
satellite imagery was available when the project began) were used. The scenes used were from path 24, rows 37, 38 (the
scenes from which contained most of the KNF), and 39. Images from both the dormant and growing season of each year
were used. For 1999, the scenes were from January 24 and April 30. June 6 and December 5 were the dates on which
the 2009 scenes were imaged. Each pixel in a Landsat scene is 30 meters (about 100 feet) on a side. The information
recorded in a pixel for a particular band is an average of all the electromagnetic radiation (i. e. “light”) reflected by the
different objects within that pixel. Images were downloaded from the United States Geological Survey (USGS) Earth
Explorer web site (http://earthexplorer.usgs.gov/).
Image Processing
A Perl script (computer program) from the Remote Sensing Applications Center (RSAC) was used to remove
atmospheric effects from the Landsat imagery, so that the information in the modified scenes more nearly corresponds
to the light reflected from the ground.
Most of the image processing was done on a HP laptop using the ERDAS Imagine 2011 software package. The
scenes for a particular date were mosaicked (combined) in order to obtain a seamless image of the KNF. These images
were then clipped, first to the administrative boundary of the KNF, and then clipped again to remove private in-holdings

8
so that the analysis could focus only on actual Forest Service property. Additionally, a 30 meter buffer was constructed
around roads within the administrative boundary of the KNF. This buffered road area was clipped out of the dataset to
avoid the complications that roadside features can add to the analysis. Information about the roads and land ownership
came from KNF Geographic Information System (GIS) database. During image processing these files were re-projected
to correspond to the projection (UTM, WGS84 datum) used in the Landsat scenes.
Before the classification process began, particular bands (corresponding to particular wavelength categories of the
electromagnetic spectrum – e.g. near-infrared) from the Landsat images were selected for the analysis, based upon the
experiences of the regional remote sensing coordinator and RSAC staff on which bands were most helpful for a project
of this type. In addition, a new “band” was created based on arithmetic manipulations of the Landsat bands. This
combination is known as the Normalized Difference Vegetation Index (NDVI). NDVI is a proxy for vegetation growth.
Equation 1 (Jones and Vaughn 2010) is used to calculated NDVI; рNIR is the near infrared reflectance, while рR is the red
light reflectance.
Equation 1 𝑵𝑵𝑵𝑵𝑵𝑵𝑵𝑵 =
(р𝐍𝐍𝐍𝐍 𝐍𝐍 − р𝐑𝐑)
(р𝐍𝐍𝐍𝐍 𝐍𝐍+ р𝐑𝐑)
A single image for each year was created by layer stacking (combining) Landsat Thematic Mapper bands 3
(visible light, 0.63-0.69 micrometers (µm)), 4 (near-infrared, 0.76-0.90 µm) 5 (near-infrared, 1.55-1.75 µm) and 7 (the
NDVI band) into one image file, for both the dormant and growing season scenes. The combination of dormant and
growing season data is intended to improve the accuracy at the core of the process, the unsupervised classification of
the images.
After layer stacking, unsupervised classifications were done separately on the 1999 and 2009 images. In
unsupervised classification, the ERDAS imagine software applies a statistical algorithm to the images to divide the pixels
into separate categories. This is done by grouping pixels with similar spectral characteristics into unique clusters
according to the statistically determined criteria. Initially the number of categories is greater than the categories used in
the classification (which are listed in Table 1). After some trial and error, 22 categories (using 12 iterations and a 95%
confidence threshold – settings in the unsupervised classification tool) were used for the initial classifications. It is up to
the analyst to then interpret the unsupervised classification and apply the appropriate classification category to the
categories created by the computer algorithm. In order to do so, 1998 NAPP false color infrared photography and 2010
National Agriculture Imagery Program (NAIP) photography (which also includes an infrared band) were used from image
files provided by RSAC. Once an initial classification has been carried out, it was usually necessary to re-apply the
unsupervised classification procedure to a particular category (for example, pine forest) to refine the classification and
reduce errors. The term for this process is “cluster busting”. Cluster busting generally requires more categories (80 for
pine forest) and typically a greater number of iterations and a higher confidence threshold.
Although initially it was assumed that categories for hardwood forest, mixed hardwood-pine forest, pine forest,
herbaceous vegetation, and “not vegetation” would suffice, during the process of interpreting the results from the
unsupervised classification it appeared that two additional categories were desirable. Several areas of the intensive use
area of the Vernon Unit of the Calcasieu Ranger District appeared to have bare soil or rock exposed, and portions of the
Kisatchie Hills Wilderness area on the Kisatchie District appeared to be covered by shrubs in 1998 aerial photography.
Consequently, “bare earth” and “shrub” categories were added to the possible categories and included in Table 1.
For each classification category (pine forest, etc.) areas were calculated (in acres) for both the KNF as a whole
and for particular ranger districts. Within the five ranger districts, classification categories were further subdivided by
sub-management area, using files from the KNF GIS data.
Accuracy Assessment
Once the final classifications for each year were completed, the accuracy of the classifications was assessed
using 1998 National Aerial Photography Program (NAPP), true color 2009 National Agriculture Imagery Program (NAIP),
and 2010 NAIP imagery (which includes an infrared band) in ArcMap, as references for photointerpretation of accuracy

9
assessment points. The three types of imagery mentioned above were obtained from the image server. In order to
create a product convenient for users on the KNF, and also to avoid having to re-project a great number of GIS files in
this process, the final classifications were first re-projected to the Louisiana North State Plane system used on the KNF –
the image processing and classification steps had been done in the UTM projection of the Landsat images, in order avoid
the inadvertent creation of spatial inaccuracies during these previous steps There were still some problems in the
overlay of different map layers, which were typically not serious for categories of large area extent (such as pine forest)
but which were problematic for categories of small areal extent (such as water bodies on the Winn district in 2009. To
assess the accuracy, a GIS shapefile of random points were created with the Image Sampler tool available from RSAC.
For each combination of classification category (Table 1) and ranger district (Calcasieu, Caney, Catahoula, Kisatchie, and
Winn) 30 random points were created. The area within a 15 meter radius circular buffer around each point was
examined to see if it matched the classification; if it did not it was recorded as being in the most appropriate of the other
classification categories indicated in Table 1. A total of 1050 points (7 categories times 5 districts times 30 points) were
examined for the 1999 classifications. Because no areas were classified as “shrubs” in the 2009 final classification, only
900 points were examined for that accuracy assessment.
In general, the accuracy of the classification was based upon what category occupied the majority of the area
within the buffer circle associated with each random point. For the forest categories, locations with 75% or more pine or
hardwood cover were classified as either pine forest or hardwood forest respectively. To assess the accuracy of the
mixed hardwood-forest category, the rule used was that if an area had either more than one quarter but less than three
quarters pine cover, with the remainder hardwood forest, or if it had more than one quarter but less than three quarters
hardwood cover with the remainder pine forest, then it was considered mixed hardwood-pine forest. This rule is similar
to that used in the Silvicultural Examination and Prescription Field Book to delineate either Pine-Hardwood or
Hardwood-Pine forest types; thus the category in the remote sensing classifications for mixed hardwood-pine forest
overlaps two categories used in the field book.
For categories other than forest, areas were classified according to the classification category (herbaceous, bare
earth, and so forth) that appeared to occupy a majority of the buffer circle. Sometimes the area around a random point
had 3 or more different kinds of features within its associated buffer. Then in the accuracy the random point was
classified according to the feature type which occupied the largest area according to ocular assessment.
As mentioned above, sometimes map layers did not line up as precisely as expected. Sometimes this resulted in
a random point apparently falling outside of Forest Service ownership, usually by less than 15 meters. This circumstance
was generally a problem with classification categories of small areal extent. In these cases, the area nearest to the
random point that was Forest Service property was used in the accuracy assessment.
The raw results of the process described above was a table or matrix indicating how many of the random points
for a particular category actually “fit” into that category or were best described by one of the other categories. From
this matrix (a separate one was devised for each year) calculations of user’s and producer’s accuracy were made. Before
the calculations of user’s or producer’s accuracy were done, the entries in the matrix were weighted by the ratio of the
area of each category divided by the total classified area (Table 2) as computed from the final classifications for 1999 or
2009. This weighting process was used because a stratified (by district and class) random sampling process was used. In
addition, the weighted results were used to calculate the kappa statistic, which is a measure of how likely such results
could have resulted from chance alone. The values of the kappa statistic can vary between zero and one, with a value of
zero indicating results due only to chance, while a value of one indicating a perfect classification with no effect of
chance.
Because the accuracy assessment was stratified by district, it was possible to create separate accuracy
assessments for each district. In a process similar to that used for the accuracy assessment of the KNF as a whole, the
entries in the matrix appropriate for each district was weighted by the ratio of the area in each category for that district,
divided by the total classified area for that district (Table 4 has the area data).
The accuracy assessment results made possible the calculation of the error bars shown in Figures 12 through 29.
These error bars were calculated by multiplying the standard error by the appropriate area. Equation 2 below (taken
from Foody 2008) was used to calculate the standard error.

10
Equation 2 𝑺𝑺𝑺𝑺𝑺𝑺 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆 (𝑺𝑺𝑺𝑺) = �
𝒑𝒑(𝟏𝟏−𝒑𝒑)_
𝒏𝒏
In equation 2, p is the user’s accuracy (for the appropriate classification category) calculated from the KNF
results (Table 5) and n is the sample size.
Results
Classification Areal extents
Maps showing the classification results are given in Figure 1 for the Evangeline Unit of the Calcasieu District,
in Figure 2 for the Vernon Unit of the Calcasieu, in Figure 3 for the Caney District, in Figure 4 for the Catahoula District
and the southeastern Winn District, in Figure 5 for the Kisatchie District, and in Figure 6 for the western portion of the
Winn District. As the maps are small, those needing to closely examine details should obtain the GIS raster files of the
classification.
Approximately 577,520 acres were classified based on the 1999 imagery, and about 580,100 acres based on the
2009 imagery (see Table 2, which has more exact figures)). The difference in acreage is attributed to about 2570 acres
that were classified in 2009 but not in 1999. Unclassified areas include locations within the Landsat imagery but outside
the administrative boundary of the KNF (the largest proportion of the unclassified area), private property within the KNF
administrative boundary, and the 30 meter buffer around roads on the KNF. Based upon the classifications, in both
1999 and 2009 pine forest was the largest category within the KNF (about 355,000 and 331,000 acres respectively). In
1999 the next largest category was mixed hardwood-pine forest, but in 2009 hardwood forest appeared to have
overtaken mixed forest to become the second largest category. Areas occupied by herbaceous vegetation, water, and
bare ground were much smaller (at most a seventh of the smallest forested category, hardwood forest in 1999) than the
areas occupied by any of the three types of forest cover.
From the results in Table 2, pine forest cover appears to have declined between 1999 and 2009 by about 24,000
acres, while mixed hardwood-pine forest and hardwood forest cover has increased. Two of the smaller categories,
herbaceous vegetation and shrub cover, also decreased, while the acres covered by bare ground increased by about 800
acres.
Table 3 takes the information in Table 2, but groups it into less precise categories, roughly forested areas, other
types of vegetation, and non-vegetated areas. It appears that forest cover increased, while the area occupied by other
types of vegetation, and also non-vegetated areas, decreased during the decade from 1999 to 2009.
When the categorization is applied to particular districts, it appears that pine forest cover has increased
between 1999 and 2009 on the Calcasieu and Kisatchie districts, while declining on all the other districts (Table 4).
Mixed hardwood-pine forest cover increased on the Caney, Catahoula and Winn Districts, while declining on the
Calcasieu and Kisatchie district between 1999 and 2009 (Table 4). Hardwood forest increased on the Caney, Catahoula,
Kisatchie and Winn District, while declining on the Calcasieu district between 1999 and 2009 (Table 4).
The results were also categorized by sub-management area within each district; see Table 12 for the
Calcasieu, Table 13 for the Caney, Table 14 for the Catahoula, Table 15 for the Kisatchie and Table 16 for the Winn. As
mentioned in the introduction, some areas in different districts are categorized as belonging to the same sub-
management area. In order to facilitate comparisons between districts, graphs of the areas of pine forest, mixed
hardwood-pine forest, and hardwood forest are provided for sub-management area 1C (Figures 12, 13, and 14), sub-
management area 3BL (Figures 15, 16, and 17), sub-management area 3BM (Figures 18, 19, and 20), sub-management
area 3BS (Figures 21, 22, and 23), sub-management area 5CL (Figures 24, 25, and 26), and 11DM (Figures 27, 28,
and 29). The results from other sub-management areas were not displayed graphically either because that particular
sub-management area is found only on one district, or, in the case of sub-management area 11DS, because the total
acreage is rather small.
In sub-management area 1C (Figures 12, 13, and 14), pine forest increased in area between 1999 and 2009 on
the Calcasieu, while declining somewhat on the Catahoula and Winn districts. Mixed hardwood-pine forest declined

11
markedly on the Calcasieu in 1C, while increasing markedly on the Winn and somewhat on the Catahoula. The
Catahoula and Winn also experienced increases in hardwood forest area in 1C between 1999 and 2009; on the Calcasieu
hardwood area was essentially unchanged.
In sub-management area 3BL (Figures 15, 16, and 17), pine forest area increased on the Calcasieu and Kisatchie
districts, while declining somewhat on the Catahoula and markedly on the Winn. Mixed hardwood-pine forest within
3BL decreased on the Calcasieu and Kisatchie, remained nearly constant on the Catahoula, and increased on the Winn.
Hardwood forest area within 3BL increased on the Catahoula, Kisatchie, and (most strikingly) on the Winn.
Sub-management area 3BM (Figures 18, 19, and 20) experienced declines in pine forest area on Kisatchie and
Winn; area remained nearly constant on the Catahoula. Within this sub-management area, mixed hardwood-pine forest
increased on the Catahoula and Winn, and possibly slightly on the Kisatchie district. Hardwood forest area remained
constant on those areas of 3BM within the Catahoula, while increasing on the Kisatchie and Winn districts.
In sub-management area 3BS (Figures 21, 22, and 23), pine acreage declined markedly on the Caney and Winn
districts, while undergoing only modest declines on the Catahoula and Kisatchie. Mixed hardwood-pine forest within
3BS increased strikingly on the Caney, noticeably on the Winn, modestly on the Catahoula, but remained nearly
unchanged on the Kisatchie. Hardwood forest also increased on those areas of 3BS within the Caney and Winn districts,
while remaining nearly unchanged on the Catahoula and Kisatchie.
Sub-management area 5CL (Figures 24, 25, and 26) experienced increases in pine forest area on the Calcasieu
and Kisatchie, but decline on the Winn, while pine acres remained nearly unchanged on the Catahoula. Within 5CL,
mixed hardwood-pine forest area grew strongly on the Winn and modestly on the Catahoula, while declining on the
Calcasieu and Kisatchie districts. Hardwood forest acreage within 5CL increased on the Catahoula, Kisatchie and Winn,
while declining on the Calcasieu.
In sub-management area 11DM (Figures 27, 28, and 29), pine forest area declined noticeably on the Catahoula
and slightly on the Winn. Mixed hardwood-pine forest within 11DM increased on the Catahoula, but was essentially
unchanged on the Winn. Hardwood forest within 11DM increased on both the Catahoula and Winn districts.
Although other sub-management areas may not found on multiple districts, they can still cover substantial areas
of a particular district or have other unusual aspects that make it worthwhile to draw attention to them. Four of these
locally important sub-management areas are 2AS, 6BL, 7C, 9DL, and 13.
Sub-management area 2AS on the Caney District emphasizes amenity values. It is one of only two sub-
management areas on the Caney (the other is 3BS, which is also found on other districts) and is located around Caney
and Corney lakes. From 1999 to 2009, hardwood forest and mixed hardwood-pine forest area increased by about 1200
and 1000 acres respectively in 2AL, while pine forest acres decreased by almost 1400 acres (Table 13).
Sub-management area 6BL is found on the Vernon Unit of the Calcasieu and corresponds to the limited military
use area on that district. From 1999 to 2009, pine forest in 6DL increased by about 1000 acres, mixed hardwood-pine
forest increased by about 450 acres, while hardwood forest decreased by almost 200 acres (Table 12).
Sub-management area 7C on the Winn District is only hardwood sub-management area on the KNF. During the
decade from 1999 to 2009, both hardwood and mixed-hardwood forest increased in this area, while pine forest
decreased (Table 16, page 41).
Sub-management area 9DL on the Vernon Unit of the Calcasieu Ranger District corresponds to the intensive
military use area. Pine forest within 9EL increased by just over 4000 acres from 1999 to 2009, while both mixed
hardwood-pine forest and hardwood forest declined by about 1200 and 200 acres respectively (Table 12).
Sub-management area 13 on the Kisatchie Ranger District is the only wilderness area on the KNF, the Kisatchie
Hills Wilderness. Both hardwood and pine forest increased within the wilderness (by about 1200 and 1300 acres
respectively) from 1999 to 2009, while mixed hardwood-pine forest declined by about 1500 acres (Table 15).
The relevance of these apparent changes depends upon the accuracy of the classification.

12
Table 1 indicates the categories used in the classifications and the colors given to the different categories in the resulting raster files.
Type Type Description Color used in maps
1 Hardwood Forest green
2 Mixed Forest cyan
3 Pine Forest dark green
4 Herbaceous vegetation yellow
5 Not vegetation (mainly water) blue
6 Shrubs magenta
7 Bare Ground tan
Figure 1 shows the classification results for the Evangeline Unit of the Calcasieu Ranger District.
Accuracy Assessment Results
Accuracy Assessment for the entire KNF
The classification categories varied considerably in their accuracy. In addition categories mapped accurately in
one year might be mapped much less accurately in the other year. It may be helpful to keep in mind that with 7
categories, the chance that a point would randomly fall into one of them, if the categories were all equal in areal extent,
is a bit over 14% for each category. If the area of the different classification categories are taken into account, a
randomly chosen point would fall on an area mapped as pine forest about 60% of the time, into an area classified as
hardwood or mixed hardwood-pine forest close to 20% of the time, and no more than 2% of the time would it land into
area classified into any one of the other categories (Table 2).

13
Figure 2 below shows the classification Results for the Vernon Unit of the Calcasieu Ranger District.

14
Figure 3 below shows the classification results for the Caney District. Note the change in the classification in the wetland area northwest of
Corney Lake in the easternmost area of the Caney. The rainfall data in Table 11 may explain the changes in water cover.
As described above in the methods, it is possible to assess the accuracy of the classification using aerial
photography. For the users of information derived from remotely sensed data, the “user’s accuracy” is probably the
most important aspect of the accuracy assessment. The user’s accuracy represents the probability that if you were to go
an area classified in a particular way that you indeed find conditions matching the classification.

15
Figure 4 below shows the classification results for the Catahoula District and also the southwestern portion of the Winn District. Note the
absence of areas classified as water in the 2009 map.
Table 2 below indicates the total acres (rounded to the nearest acre) and pixels in each category for the entire Kisatchie National Forest (KNF)
for 1999 and 2009, and the change in acres. The figures in Table 2 are the initial figures before any adjustments were done for the accuracy
assessment.
Class Name 2009 Acres % of total
2009
classified
acres
1999 Acres % of
total
1999
classified
acres
Difference (2009
acres - 1999
acres)
2009 pixel
count
1999 pixel
count
Unclassified 4,977,055 4,979,627 -2,572 22070218 22081623
Hardwood Forest 129,925 22 94,044 16 35,881 576138 417026
Mixed Hardwood-Pine Forest 105,132 18 95,527 17 9,606 466199 423604
Pine Forest 331,207 57 355,463 62 -24,256 1468701 1576262
Herbaceous Vegetation 7,161.07 1 13,457 2 -6,296 31755 59674
Not Vegetation (mainly water) 5,210 1 8,490 1 -3,280 23103 37649
Shrubs 0 0.00 9,876 2 -9,876 0 43794
Bare Ground 1,457 0.00 663.90 0.00 793 6462 2944
Total Acres (including
unclassified)
5,557,147 5,557,147 0
Total Classified Acres 580,092.45 577,520.51 2,571.94

16
Figure 5 below shows the classification results for the Kisatchie Ranger District.
Table 3 uses the results given in Table 1, but groups the categories in Table 1 into larger, less precise sets. Once again no weighting adjustments
for accuracy have been made.
2009
Acres
% of total
2009
classified
acres
1999
Acres
% of
total
1999
classified
acres
Difference (2009
acres - 1999 acres)
Forest (hardwood, mixed, & pine) 566,264 98 545,033 94 21,230.86
Non-forest vegetation (shrub & herbaceous) 7,161 1 23,333 4 -16,172.00
Sum of not vegetation & bare ground 6,667 1 9,154 2 -2,486.92
Sum of Changes - equals change in unclassified 2,571.94
The producer’s accuracy represents the probability that an area that actually belongs in a particular category was
actually mapped appropriately.
Table 5 shows the user’s and producer’s accuracies for the KNF taken as a whole for the 1999 and 2009
classifications. For pine and hardwood forest, user’s accuracy was good for both years, with a minimum user’s accuracy
for hardwood forest of 65% and a minimum user’s accuracy for pine forest of 81%. The user’s accuracy for mixed
hardwood-pine forest was lower, ranging from 35 to 37%.
Using statistical techniques, it is possible to estimate how often such results could result from chance events
alone. The kappa statistic is one such method. A kappa value of zero indicates results which could have resulted from
chance events alone; a kappa value of 1 indicates no effects due to chance. The calculated kappa values were
approximately 0.51 for the 1999 classification and 0.47 for the 2009 classification.

17
Figure 6 below shows the classification results for the Winn Ranger District. Note the absence of areas classified as water in the 2009 map.
Accuracy Assessments for individual districts
User’s accuracies for the same classification category varied between districts (Figure 25). For pine forest, user’s
accuracies were usually 60% or higher (see Tables 6, 7, 8, 9, and 10 for user’s and producer’s accuracies by district).
However, it was only 31% for the Caney District in 2009 (Table 7). User’s accuracies were also often acceptable (e.g., at
least 40%) for hardwood forest, but were only 54% for the Calcasieu District in 2009 (Table 6) and 53% and 62% on the
Catahoula District in 1999 and 2009 respectively (Table 8).
User’s accuracies for the other categories were generally either lower, or more variable between years. The
user’s accuracy of the mixed pine-hardwood forest category varied from a low of 11% on the Kisatchie District in 2009 to
a high of 61% on the Winn in 1999. The user’s accuracy on the Winn District for mixed hardwood-pine was acceptable in
both years (Table 10).
The user’s accuracies for areas classified as not vegetation, mainly water, was extremely variable. An
examination of precipitation data from Shreveport (Table 11 ) suggests that months from which Landsat data was
collected in 1999 were unusually wet, perhaps causing water bodies to stand out better from other features. During the
accuracy assessment process for this classification category, I noticed that the areas incorrectly categorized as water on
the Catahoula and Winn districts were almost invariably edges between Forest Service ownership and private property.

18
Table 4 below gives the areas for the different classification categories for each district for both 1999 and 2009. The “difference” column is the
result of subtracting the 1999 acres from a particular category from the appropriate 2009 category. The values in the difference columns are
rounded to the nearest acre. Once again no weighting adjustments for accuracy have been made.
Calcasieu Kisatchie District
2009 acres 1999 acres Difference 2009 acres 1999 acres Difference
Unclassified 628679 629251 -572 143081 143501 -420
Hardwood Forest 21389.3 22919.1 -1530 25971.3 18407 7564
Mixed Hardwood Pine
Forest
17138.6 22170.6 -5032 5342.37 14939.8 -9597
Pine Forest 131093 117165 13928 65674.4 59755.5 5919
Herbaceous Vegetation 3273.21 6763.7 -3490 1031.91 2798.4 -1766
Not Vegetation (mainly
water)
1485.38 1580.78 -95 422.995 607.583 -185
Shrubs 0 3382.85 -3383 0 1617.92 -1618
Bare Earth 708.773 534.415 174 151.896 48.9269 103
Caney Winn
Unclassified 481249.00 481455 -206 617989 618841 -852
Hardwood Forest 14217.5 8137.43 6080 45587.4 26095.2 19492
Mixed Hardwood Pine
Forest
10026 5737.12 4289 48543.9 32089.2 16455
Pine Forest 4528.18 12707.9 -8180 62957.1 94358.2 -31401
Herbaceous Vegetation 348.938 619.592 -271 1266.76 1495.38 -229
water)
1611.47 3051.26 -1440 1031.47 2287.78 -1256
Shrubs 0 517.068 -517 0 2348.49 -2348
Bare Earth 249.305 4.89269 244 177.249 37.5848 140
Catahoula
Unclassified 211222 211859 -637
Hardwood Forest 22643.6 18391 4253
Mixed Hardwood Pine
Forest
24000.9 20523.1 3478
Pine Forest 66760.5 71246 -4486
Herbaceous Vegetation 1216.5 1767.15 -551
water)
649.171 833.536 -184
Shrubs 0 2005.34 -2005
Bare Earth 171.689 38.9191 133

19
Table 5 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Kisatchie National Forest as
a whole for the 1999 and 2009 classifications. The overall classification accuracy was 70% in 1999 and 69% in 2009. The kappa value was
approximately 0.51 for 1999 and about 0.47 for 2009. The not applicable result for the 2009 producer’s accuracy for shrubs is due to the
impossibility of carrying out the calculation, which would require dividing by zero. No areas were classified as shrubs in 2009, but the accuracy
assessment found some locations that appeared to be covered with shrubs.
1999 Classification 2009 Classification
User’s
accuracy (%)
Producer’s
accuracy (%)
User’s
Accuracy
(%)
Producer’s
Accuracy (%)
Hardwood Forest 65 81 67 71
Mixed Hardwood-
Pine Forest 37 40 35 27
Herbaceous
vegetation 22 49 20 33
Not Vegetation
(mainly water) 42 76 33 33
Shrubs 74 58 0 Not applicable
Bare earth 7 38 35 36
Table 6 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Calcasieu Ranger District for
the 1999 and 2009 classifications. The overall classification accuracy was 75% for both 1999 and 2009. The kappa value was approximately 0.55
for 1999 and about 0.45 for 2009. The not applicable result for the 2009 producer’s accuracy for shrubs is due to the impossibility of carrying out
the calculation, which would require dividing by zero. No areas were classified as shrubs in 2009, but the accuracy assessment found some
locations that appeared to be covered with shrubs.
User’s
accuracy (%)
Producer’s
accuracy (%)
User’s
Accuracy
(%)
Producer’s
Accuracy (%)
Mixed Hardwood-
Herbaceous
Not Vegetation

20
Table 7 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Caney Ranger District for
the 1999 and 2009 classifications. The overall classification accuracies were 70 % for 1999 and 55% in 2009. The kappa value was approximately
0.58 for 1999 and about 0.37 for 2009. The not applicable result for the 2009 producer’s accuracy for shrubs is due to the impossibility of
carrying out the calculation, which would require dividing by zero. No areas were classified as shrubs on the Caney District in 2009, but the
accuracy assessment found some locations that appeared to be covered with shrubs.
User’s
accuracy (%)
Producer’s
accuracy (%)
User’s
Accuracy
(%)
Producer’s
Accuracy (%)
Mixed Hardwood-
Herbaceous
Not Vegetation
Table 8 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Catahoula Ranger District
for the 1999 and 2009 classifications. The overall classification accuracies were 67% for 1999 and 68% in 2009. The kappa value was
approximately 0.45 for 1999 and about 0.50 for 2009. The not applicable result for the 2009 producer’s accuracy for shrubs is due to the
impossibility of carrying out the calculation, which would require dividing by zero. No areas were classified as shrubs in 2009, but the accuracy
assessment found some locations that appeared to be covered with shrubs.
User’s
accuracy (%)
Producer’s
accuracy (%)
User’s
Accuracy
(%)
Producer’s
Accuracy (%)
Mixed Hardwood-
Herbaceous
Not Vegetation

21
Table 9 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Kisatchie Ranger District for
the 1999 and 2009 classifications. The overall classification accuracy was 62% for 1999 and 69% in 2009. The kappa value was approximately
carrying out the calculation, which would require dividing by zero. No areas were classified as shrubs in 2009, but the accuracy assessment
found some locations that appeared to be covered with shrubs.
User’s
accuracy (%)
Producer’s
accuracy (%)
User’s
Accuracy
(%)
Producer’s
Accuracy (%)
Mixed Hardwood-
Herbaceous
Not Vegetation
Table 10 below contains the results (rounded to the nearest digit) of the area-weighted accuracy assessments for the Winn Ranger District for
the 1999 and 2009 classifications. The overall classification accuracies were 79 % for 1999 and 62% in 2009. The kappa value was approximately
carrying out the calculation, which would require dividing by zero. No areas were classified as shrubs in 2009, but the accuracy assessment
found some locations that appeared to be covered with shrubs.
User’s
accuracy (%)
Producer’s
accuracy (%)
User’s
Accuracy
(%)
Producer’s
Accuracy (%)
Mixed Hardwood-
Herbaceous
Not Vegetation
Discussion
For remote sensing to be a useful monitoring tool in monitoring the consequences of management activities on
the KNF, the process must be capable of answering questions relevant to the forest plan, and sufficiently accurate for
uses to be confident in the results. Classification using satellite imagery is capable of meeting the first criterion, but is
sufficiently accurate only for some categories of land use. Some land cover categories were consistently mapped fairly
accurately, while others were not. For some categories accuracy was high in both years on a particular district, while

22
being very inaccurate in one or both years on other districts. Pine and hardwood forest was mapped at least fairly well
for all districts in both years, mixed hardwood-pine forest was mapped well on the Winn, and water was mapped well on
the Caney ranger district. Despite the limitations in the precision of the classifications, it would appear (see the figures
in the Appendix for the various sub-management areas) that some of the changes in the area of the different categories
of forest were large enough to be outside of the margin of error.
A more fundamental limitation of the accuracy assessment is the absence of field data. The error bars are a
measure of precision as opposed to accuracy. The accuracy assessment that was performed relies on the interpreter’s
ability to correctly classify an area based upon its appearance in an aerial photograph. For some points, such as that
show in Figure 7 this is a simple matter; for other locations, such as Figure 8, judgment must be exercised. In addition,
pixels from the Landsat scenes and the aerial photography that are supposed to represent the same locations did not
always perfectly overlay one another. When these imperfections were readily apparent, such as at a property boundary,
I could adjust my procedures to account for the difficulty; but such problems may not have always been readily
apparent. The consequence of the need to exercise judgment and the imperfect geographic overlay between the
different kinds of imagery is that the accuracy assessment itself is not perfectly accurate.
A set of permanent field plots, appropriately stratified by district and possibly also by sub-management area, so
as to provide statistically reliable information for decision makers, would improve the accuracy assessment, make
possible the use of other remote sensing methods such as supervised classification, and might have other uses for
monitoring herbaceous management indicator species and also for fuels monitoring. The time and expense needed to
establish such a system of plots would be substantial. Furthermore some initial trial plots would be required to
determine the number of plots needed for statistically valid results. Perhaps a more cost effective means of acquiring
useful plot data would be to contract for the more frequent data collection from the Forest Inventory and Analysis (FIA)
plots that are located on the KNF. Based upon downloaded database files (http://apps.fs.fed.us/fiadb-
downloads/datamart.html ), it appears that of the 546 FIA plots located within the ownership of the KNF, 54 appear to
have sampled to one degree or another after 2004.
Figure 7 below is one of the random points used to assess the
accuracy of the herbaceous category on the Catahoula District. The
vegetation on this spot in the pipeline right of way was readily
recorded as correctly classified as herbaceous vegetation.
Figure 8 below is one of the random points used to assess the
accuracy of the mixed hardwood-pine forest classification on the
Calcasieu District. Given the different colors of the tree canopies
within the 15 meter radius buffer circle, it appears that this
particular location was correctly classified, but judgment had to be
exercised regarding if the proportions of hardwoods and pines
were such that it was mixed hardwood-pine forest.

23
At first glance, some of the results related to hardwoods, water cover, and shrub cover are puzzling. The
increase in hardwood acres on most districts of the KNF (Table 4) cannot be explained as artifact of when the Landsat
images were taken; consultation with experienced local botanists indicates that deciduous trees even on the
northernmost district (the Caney) are leafed out by April 30, the earliest date of the Landsat scenes used in the analysis.
Furthermore, the temperatures in 1999 were not unusual (Figure 10). Therefore the increase in hardwoods cannot be
explained as an artifact resulting from a low baseline level due to using Landsat scenes taken before the seasonal leafing
out of the deciduous trees on the KNF. The apparent decrease in area occupied by water (Table 2) would appear to be
an artifact of the high levels of precipitation that occurred just before the 1999 Landsat scenes were imaged (Table 11).
The absence of any areas covered by shrubs in 2009 is probably a result of forest succession. The increase in hardwood
area does not seem to be caused by differences in soil fertility between the districts, as a plot average loblolly site index
(which is related to soil fertility) calculated from NRCS soil data (downloaded from http://soildatamart.nrcs.usda.gov/ )
against the changes in hardwood area did not reveal a positive association between site index and increasing hardwood
acres (Figure 11). Perhaps differences in prescribed fire intensity between districts explain the differences.
Figures, detailed tables and a discussion of the results for particular sub-management areas are provided in the
Appendix. Overall, there is a mixed picture regarding the movement of the KNF towards the desired future conditions
for various sub-management areas. On some districts the objectives are being approached or being met, while on other
districts there are challenges, for example in sub-management area 5CL. The cause of the differences between districts
is unclear.
Figure 9 below plots the user’s accuracies for each combination of Ranger District and year by classification category.
As long as its limitations are kept in mind, a regular program of remote sensing would be beneficial for the KNF.
Remote sensing allows for a quantitative assessment of large communities, such as pine forest or hardwood forest. The
information from remote sensing could be of value in the preparation of the annual monitoring and evaluation reports.
Its limitations include an inability to discern small inclusive plant communities, and also not being sensitive to make
qualitative differences within a large plant community. For example, remote sensing does not readily distinguish
between a recently established loblolly pine plantation and a mature, frequently burned, open longleaf pine forest.
0
10
20
30
40
50
60
70
80
90
100
KNF 1999
KNF 2009
Calcasieu 1999
Calcasieu 2009
Caney 1999
Caney 2009
Catahoula 1999
Catahoula 2009
Kisatchie 1999
Kisatchie 2009
Winn 1999
Winn 2009

24
Because some aspects of the desired future conditions described within the forest plan are gross changes that can be
detected with remote sensing, this technique offers a cost-effective complement to other forms of vegetation
monitoring. Remote sensing could also be useful in identifying management challenges, such as those described in the
Appendix. If remote sensing projects were scheduled according to likely timing of forest plan revisions, enough years
would have passed between measurements so that any major changes in vegetation would be noticeable. More
frequent remote sensing projects would probably require the establishment of permanent field plots to provide
meaningful information, as the changes between years would be more subtle and therefore more likely to fall within the
margin of error.
Table 11 below provides precipitation data (total water equivalent, in inches) from Shreveport Louisiana (downloaded from the National
Climatic Data Center web site) for 1998, 1999, 2008, and 2009. The precipitation data highlighted in yellow are the same months the 1999
Landsat scenes were taken, and the green highlighted precipitation data are the same months that the 2009 Landsat scenes were imaged.
January February March April May June July August September October November December Totals
1998 5.84 7.19 4.28 0.79 0.15 1.35 2.84 3.83 7.79 5.72 4.58 6.24 50.6
1999 12.96 0.42 5.1 7.88 3.96 7.98 2.8 1.47 4.9 3.21 0.52 3.82 55.02
2008 2.65 4.96 3.25 2.62 11.56 3.85 1.08 5.73 3.84 1.41 4.98 3.14 49.07
2009 2.13 1.63 6.48 3.97 7.44 1.22 6.49 1.69 2.58 20.35 1.42 4.64 60.04
Figure 10 below shows monthly mean daily minimum and monthly mean daily maximum temperatures for 1999 and 2009 from Shreveport
Louisiana. The graph also displays normal (30 year monthly average) minimum and maximum temperatures for each month. The data was
downloaded from the National Climatic Data Center website.
Figure 11 below is a plot of site index against the change in hardwood acres. Each point is the data from a single ranger district.
30
40
50
60
70
80
90
100
January
February
March
April
May
June
July
August
September
October
November
December
degreesFarenheit
1999 Mean daily
minimum
1999 Mean daily
Maximum
2009 Mean daily
minimum
-5000
0
5000
10000
15000
20000
25000
60 70 80 90
Chageinhardwoodacres
from1999to2009
Area-weighted average district loblolly pine
site index

25
Acknowledgements
Among the employees of the Kisatchie National Forest who were especially helpful, forest botanist David Moore, forest
planner Carl Brevelle, staff officer David Byrd, former forest siliviculturist Jackie Duncan, GIS coordinator Joel Harrison,
district ranger Lisa W. Lewis and district biologist Steve Shively helped in winnowing the potential remote sensing
projects to a proposal which was both feasible and relevant to the goals of the KNF. I am grateful to acknowledge the
help of the southern region’s Remote Sensing Coordinator, Renee Jacokes-Mancini, and Brent Mitchell and Abigail
Schaaf (both of RedCastle Resources, working at RSAC) in project development and implementation. Abigail Schaaf
‘s editorial suggestions were most helpful. Michael Walterman of RSAC provided guidance on appropriate accuracy
assessment calculations for a stratified random sampling design. Michael MacRoberts and Dave Moore provided
information about the timing of the leafing-out of deciduous trees in Louisiana. Scott McClarin of RSAC was most
helpful obtaining the 1998 NAPP infrared photography for the area in a form which could be used in ERDAS Imagine.
References
FIA Data Mart. Plot data for Claiborne, Grant, Natchitoches, Rapides, Vernon, Webster and Winn parishes downloaded
from http://apps.fs.fed.us/fiadb-downloads/datamart.html on May 7, 2012.
Foody, Giles M. 2009. Sample size determination for image classification accuracy and comparison. Proceedings of the
8th
International Symposium on Spatial Accuracy Assessment in Natural Resources and Environmental Sciences, pp. 154-
162.
Jones, Hamyn, G., and Robin A. Vaughn. 2010. Remote sensing of vegetation: principles, techniques, and applications.
Oxford University Press, Oxford, Great Britain.
National Climatic Data Center. Shreveport annual data from 1998, 1999, 2008, and 2009 accessed on April 4,2012
from http://www7.ncdc.noaa.gov/IPS/lcd/lcd.html
Soil Survey Staff, Natural Resources Conservation Service, United States Department of Agriculture. Soil Survey
Geographic (SSURGO) Database for [Survey Area, State]. Available online at http://soildatamart.nrcs.usda.gov . Accessed
[5/4/2012]
USDA Forest Service. 1999. Revised Land and Resource Management Plan: Kisatchie National Forest. Pineville,
Louisiana.
USGS. Landsat 5 history. Accessed on April 16, 2012 from http://landsat.usgs.gov/about_landsat5.php.

26
Appendix: Discussion, figures, and data tables relating to sub-management areas
Many of the management objectives described in the forest plan are related to particular regions of the forest.
These areas are classified as sub-management areas. Some of the sub-management areas are found on multiple ranger
districts, while others are found on only a single district. The first several sub-management areas discussed, 1C, 3BL,
3BM, 3BS, 5CL, and 11DM, are some of the larger sub-management areas that are found on multiple districts. I then
discuss the results pertaining to five sub-management areas (2AS, 6BL, 7C, 9DL, and 13) that are locally important on
particular districts.
The management goals for sub-management area 1C include “producing and sustaining a high level of a mixture
of commodity outputs” (forest plan 3-3). Part of the desired future condition of this sub-management area is that “the
overstory vegetation on a large majority of the area consists primarily of pine stands which may contain up to 30
percent hardwoods (forest plan 3-3). On the Calcasieu, pine forest cover has increased in 1C (Figure 12), which would
appear to be consistent with the goals of the forest plan. The nearly unchanged area of pine forest in 1C on the
Catahoula (Figure 7) and the decrease in pine forest on the Winn District (Figure 7) are probably not consistent with the
goals of the forest plan. The proportion of hardwoods in 1C on the Catahoula (Table 13) may not be consistent with the
goals in the forest plan, considering the 20% of the areas (in 2009) that is mixed hardwood-pine forest, as well as the
22% that is hardwood forest (Table 14). The mixed hardwood-pine forest obviously contains hardwood trees, which
would add to the area of hardwoods in the hardwood forest category, so more than 22% of 1C on the Catahoula is
hardwood forest.
For sub-management area 3BL, the management goals are to “emphasize restoration of native, fire dependent
longleaf pine communities in an intermediate time period while providing a moderate level of protection of other
resources” (forest plan, 3-8). Part of the desired future condition of 3BL is that “the landscape is composed of relatively
open park-like pine stands eventually dominated by native, fire dependent longleaf pine communities (forest plan, 3-8).
On the Calcasieu and Kisatchie districts, the increases in pine forest within 3BL between 1999 and 2009 would appear to
be consistent with these objectives (Figure 15), if the increase in pine acres represents an increase in longleaf pine.
On the Catahoula District pine forest acreage scarcely changed within that decade, which may be inconsistent
with the goals of the forest plan, as the percentages of hardwood forest (22%) and mixed hardwood-pine (also 22%) in
2009 are not much different from what was observed for 1C on the Catahoula (Table 14). On the Winn Ranger District
pine forest declined by about 4100 acres between 1999 and 2009 (Table 16) a result which appears inconsistent with the
management objectives for this sub-management area.
The management goals for sub-management area 3BM are to “emphasize restoration of native mixed
hardwood-loblolly pine (MHL) communities in an intermediate time period while providing a moderate level of
protection of other resources” (forest plan 3-10). Part of the desired future conditions for 3BM is that “the forest is
dominated by communities composed primarily of various hardwoods, with loblolly pine as a major associate” (forest
plan 3-10). The increase in hardwood forest (Figure 20) and mixed-hardwood pine forest on the Winn (Figure 19)
suggest that management goals for 3BM are being met on the Winn. On the Catahoula and Kisatchie districts the
significance of the changes is less clear, as some appear to fall within the margin of error (see Figures 14 and 15).
However, it appears that for both the Catahoula and Kisatchie districts a large majority of the forested areas located
within 3BM are either hardwood forest or mixed hardwood-pine forest see (Tables 14 and 15). Therefore it appears that
the on all 3 districts the management objectives for 3BM have either been reached or that management is bringing the
area closer to the desired future conditions.
For sub-management area 3BS, the management goals are to “emphasize restoration of native shortleaf
pine/oak-hickory (SOH) communities in an intermediate time period while providing a moderate level of protection of
other resources” (forest plan 3-9). One aspect of the desired future condition of 3BS is that “the landscape is dominated
by mixtures of shortleaf pine, oaks, and hickories” (forest plan 3-9). The Caney district would appear to be near the
desired future condition (Table 13), assuming that the pine in the overstory is shortleaf pine rather than some other
species. The increases of mixed hardwood-pine forest and hardwood forest on the Winn (Figures 22 and 23) would
suggest that 3BS on the Winn is moving towards the desired future condition, once again on the assumption that the

27
pine areas have a large shortleaf pine component. On the Catahoula and Kisatchie districts, the high proportion of pine
forest (46% and 53% respectively in 2009; see Tables 14 and 15) and the lack of change (Figures 22 and 23) indicates that
the 3BS on these districts is not approaching the desired future conditions.
The management goals for 5CL are to “emphasize the management of RCW habitat and restoring native, fire
dependent longleaf pine communities in an extended time period” (forest plan 3-16). One part of the desired future
condition for 5CL is that “the landscape is composed of relatively open, park-like pine stands eventually dominated by
native, fire dependent longleaf communities” (forest plan 3-16). On the Calcasieu and Kisatchie districts, the increase in
pine forest (Figure 19) and the high proportion of 5CL classified as pine forest (Tables 12 and 15) suggest that the
management objectives are being met on these districts, assuming that the species composition of the pine forest is
appropriate. On the Winn district, the decrease in pine acres within 5CL (Figure 24) would appear inconsistent with
management objectives Additional indications of a management challenge regarding 5CL on the Winn is that more of
the forested area within 5CL on this district is hardwood or mixed hardwood-pine forest (see 2009 percentages in Table
16) rather than pine forest. On the Catahoula District, most of 5CL is pine forest (see Table 14), but the direction of
change (either a decrease in pine acres or at best no change; Figure 24) appears unfavorable.
For sub-management area 11DM, one of the management goals is to “emphasize management of RCW habitat
and production of high quality wildlife habitats within a mixed hardwood-pine landscape” (forest plan 3-42). One of
aspect of the desired future condition of 11DM is that “the forest is dominated by communities composed primarily of
various hardwoods, with loblolly pine as a major associate” (forest plan 3-42). The increase in hardwood area on the
Catahoula (Table 14) and Winn (Table 16) appears consistent with these objectives.
Turning now to the sub-management areas confined particular districts, the management objectives for sub-
management area 2AS on the Caney District include the “protection and enhancement of non-market resources and
values associated with shortleaf pine/oak-hickory (SOH) dominated landscapes” (forest plan 3-5). Part of the desired
future condition is that “mixed communities of shortleaf pine, oaks, and hickories dominate the landscape” (forest plan
3-6). Even if the decrease in water cover (Table 4) on the Caney all occurred in sub-management areas 2AS, and even if
all of the decrease in water cover was classified as hardwood forest in the 2009 Landsat scenes, there would still be a
4640 acre a net increase in hardwood forest area which could not be attributed to the change in water cover. The
changes observed (Table 13) would appear consistent with the objective in the forest plan, provided that the pine
species is actually shortleaf pine.
For sub-management area 6BL on the Vernon Unit of the Calcasieu District, the management goals
“emphasize management of RCW habitat and producing high quality wildlife habitats created with an open, frequently
burned landscape” (forest plan 3-22). One aspect of the desired future condition is that “the landscape is composed of
relatively open park-like stands eventually dominated by native, fire dependent longleaf pine communities” (forest plan
3-22). Thus the desired future condition for 6BL is similar to that for 5CL. The increase in pine area within 6BL (Table 12)
and the high proportion of the area covered by pine forest, would appear to be consistent with this objective, assuming
that the species composition of the pine component is appropriate.
The management goals for 7C on the Winn District include “providing high levels of hardwood composition, featuring
hardmast producers” (forest plan 3-25). Part of the desired future condition is that “the canopy consists of a variety of
hardwood trees, with pine being a major associate” (forest plan 3-25). The increase in hardwood and mixed hardwood-
pine species within 7C (Table 16) suggests that management objectives are being approached, provided that the
hardwood forest includes the desired hardmast producers.
For sub-management area 9DL on the Vernon Unit of the Calcasieu, the management goals are similar to those
for 5CL and 6BL, emphasizing the “management of RCW habitat and producing the highest quality wildlife habitats
crated with an open, frequently burned landscape” (forest plan 3-28). As with 5CL and 6BL, an aspect of the desired
future condition is that “the landscape is composed of relatively open park-like pine stands eventually dominated by
native, fire dependent longleaf pine communities” (forest plan 3-29). The increase in pine forest acres, and high
proportion of 9DL classified as pine forest (Table 12) suggest that the management objectives are being met, if the pine
component is of the desired species composition.

28
Figure 12 below shows the changes in pine forest acres in sub-management area 1C. The error bars were calculated by applying the user’s
accuracies for pine forest on the KNF from 1999 and 2009 respectively to the appropriate number of acres for a particular sub-management
area, district and year.
Figure 13 below shows the changes in mixed hardwood-pine forest acres in sub-management area 1C. The error bars were calculated by
applying the user’s accuracies for mixed hardwood-pine forest on the KNF from 1999 and 2009 respectively to the appropriate number of acres
for a particular sub-management area, district and year.
Figure 14 below shows the changes in hardwood forest within sub-management area 1C. The error bars were calculated by applying the user’s
accuracies for hardwood forest on the KNF from 1999 and 2009, respectively, to the appropriate number of acres for a particular sub-
management areas, district, and year.
2000
4000
6000
8000
10000
12000
1999 2009
Acres
Pine Forest in
Sub-Management
Area 1C
Calcasieu
Catahoula
Winn
750
1000
1250
1500
1750
2000
2250
1999 2009
Acres
Mixed Hardwood-Pine
Forest in
Sub-Management
Area 1C
Calcasieu
Catahoula
Winn
500
750
1000
1250
1500
1750
2000
2250
2500
1999 2009
Acres
Hardwood Forest
in Sub-Management
Area 1C
Calcasieu
Catahoula
Winn

29
Figure 15 below shows the changes in pine forest area in sub-management area 3BL. The error bars were calculated by applying the user’s
Figure 16 below shows the changes in mixed hardwood-pine forest area in sub-management area 3BL. The error bars were calculated by
applying the user’s accuracies for mixed-hardwood pine forest on the KNF from 1999 and 2009 respectively to the appropriate number of acres
Figure 17 below shows the changes in hardwood forest within sub-management area 3BL. The error bars were calculated by applying the user’s
2000
4000
6000
8000
10000
12000
14000
1999 2009
Acres
Pine Forest in
Sub-management
Area 3BL
Calcasieu
Catahoula
Kisatchie
Winn
0
1000
2000
3000
4000
5000
6000
1999 2009
Acres
Mixed Hardwood-Pine
Forest in
Sub-Management
Area 3BL
Calcasieu
Catahoula
Kisatchie
Winn
2500
3000
3500
4000
4500
5000
5500
6000
1999 2009
Acres
Hardwood Forest
in Sub-Management
Area 3BL
Calcasieu
Catahoula
Kisatchie
Winn

30
Figure 18 below shows the changes in pine forest area in sub-management area 3BM. The error bars were calculated by applying the user’s
Figure 19 below shows the changes in mixed hardwood-pine forest area in sub-management area 3BM. The error bars were calculated by
Figure 20 below shows the changes in hardwood forest within sub-management area 3BM. The error bars were calculated by applying the
user’s accuracies for hardwood forest on the KNF from 1999 and 2009, respectively, to the appropriate number of acres for a particular sub-
0
1000
2000
3000
4000
5000
1999 2009
Acres
Pine Forest in
Sub-management
Area 3BM
Catahoula
Kisatchie
Winn
0
500
1000
1500
2000
2500
3000
3500
4000
1999 2009
Acres
Mixed Hardwood-Pine
Forest in
Sub-Management
Area 3BM
Catahoula
Kisatchie
Winn
1000
2000
3000
4000
5000
6000
1999 2009
Acres
Hardwood Forest
in Sub-Management
Area 3BM
Catahoula
Kisatchie
Winn

31
Figure 21 below shows the changes in pine forest area in sub-management area 3BS. The error bars were calculated by applying the user’s
Figure 22 below shows the changes in mixed hardwood-pine forest area in sub-management area 3BS. The error bars were calculated by
Figure 23 below shows the changes in hardwood forest within sub-management area 3BS. The error bars were calculated by applying the user’s
2000
4000
6000
8000
10000
12000
1999 2009
Acres
Pine Forest in
Sub-Management
Area 3BS
Caney
Catahoula
Kisatchie
Winn
0
2000
4000
6000
8000
10000
1999 2009
Acres
Mixed Hardwood-Pine
Forest in
Sub-Management
Area 3BS
Caney
Catahoula
Kisatchie
Winn
0
2000
4000
6000
8000
10000
12000
1999 2009
Acres
Hardwood Forest in
Sub-Management
Area 3BS
Caney
Catahoula
Kisatchie
Winn

32
Figure 24 below shows the changes in pine forest area in sub-management area 5CL. The error bars were calculated by applying the user’s
Figure 25 below shows the changes in mixed hardwood-pine forest area within sub-management area 5CL. The error bars were calculated by
applying the user’s accuracies for mixed hardwood-pine forest on the KNF from 1999 and 2009, respectively, to the appropriate number of acres
for a particular sub-management area, district, and year.
Figure 26 below shows the changes in hardwood forest within sub-management area 5CL. The error bars were calculated by applying the user’s
15000
20000
25000
30000
35000
40000
1999 2009
Acres
Pine Area in
Sub-Management
Area 5CL
Calcasieu
Catahoula
Kisatchie
Winn
0
5000
10000
15000
20000
1999 2009
Acres
Mixed Hardwood-Pine
Forest in
Sub-Management
Area 5CL
Calcasieu
Catahoula
Kisatchie
Winn
4000
5000
6000
7000
8000
9000
10000
11000
12000
1999 2009
Acres
Hardwood Forest
in Sub-Management
Area 5CL
Calcasieu
Catahoula
Kisatchie
Winn

33
Figure 27 below shows changes in the area of pine forest located in sub-management areas 11DM. The error bars were calculated by applying
the user’s accuracies for the KNF from 1999 and 2009 respectively to the appropriate number of acres for a particular sub-management area,
district and year.
Figure 28 below shows the changes in mixed hardwood-pine forest area located in sub-management area 11DM. The error bars were calculated
by applying the user’s accuracies for the KNF from 1999 and 2009 respectively to the appropriate number of acres for a particular sub-
management area, district and year.
Figure 29 below shows the changes in hardwood forest within sub-management area 11DM. The error bars were calculated by applying the
user’s accuracies for hardwood forest on the KNF from 1999 and 2009, respectively, to the appropriate number of acres for a particular sub-
0
500
1000
1500
2000
2500
3000
1999 2009
Acres
Pine Forest in
Sub-Management
area 11 DM
Catahoula
Winn
0
200
400
600
800
1000
1200
1999 2009
Acres
Mixed Hardwood-Pine
Forest in
Sub-Management
Area 11 DM
Catahoula
Winn
500
600
700
800
900
1000
1100
1999 2009
Acres
Hardwood Forest
in Sub-Management
Area 11DM
Catahoula
Winn

34
Table 12 below shows the areas (values are acres) occupied by each classification category within each of the 12 sub-management areas of the
Calcasieu Ranger District, and the changes in area between 1999 and 2009. The percentages given for the forested categories are the percent of
total classified acres. Because no areas were classified as shrubs in 2009, zero acres were counted for this category in 2009 in order to calculate
the change in area during the decade. A similar procedure was used when a particular classification category was not found in a particular
combination of sub-management area and year.
Year Sub-
Management
Unclassified Hardwood
Forest
Mixed
Hardwood
Pine Forest
Pine Forest Herbaceous
Vegetation
Not Vegetation
(mainly water)
Shrubs Bare Earth
Acres Acres % Acres % Acres % Acres Acres Acres Acres
2009 2AL 2137 2356 33 1121 16 2757 38 66 788 0 75
1999 2AL 2154 2092 29 1432 20 2296 32 103 1071 148 4
Change -17 264 -311 461 -37 -283 -148 71
2009 2AM 202 496 93 26 5 8 2 0 3 0 0
1999 2AM 204 514 97 5 1 8 2 0 3 0 0
Change -2 -18 21 0 0 0 0 0
2009 1C 79854 1900 13 1324 9 10792 75 166 223 0 15
1999 1C 80076 1914 13 2154 15 9595 68 229 28 276 3
Change -222 -14 -830 1197 -63 195 -276 12
2009 3BL 11938 2790 28 1115 11 6037 60 74 75 0 5
1999 3BL 12014 3239 32 1432 14 5196 52 65 15 67 7
Change -76 -449 -317 841 9 60 -67 -2
2009 3CS 17 1 11 1 11 7 78 0 0 0 0
1999 3CS 17 4 36 3 27 4 36 0 0 0 0
Change 0 -3 -2 3 0 0 0 0
2009 5CL 6770 6045 13 3714 8 36347 78 270 161 0 82
1999 5CL 6831 6944 15 6195 13 31135 67 996 196 1051 40
Change -61 -899 -2481 5212 -726 -35 -1051 42
2009 6BL 16295 2787 6 5004 11 35296 81 470 189 0 53
1999 6BL 16466 2965 7 4551 10 34295 79 1170 98 524 25
Change -171 -178 453 1001 -700 91 -524 28
2009 9DL 2542 4681 10 4482 10 33033 74 2130 25 0 451
1999 9DL 2550 4892 11 5651 13 28962 65 3697 142 1007 444
Change -8 -211 -1169 4071 -1567 -117 -1007 7
2009 9E 13917 16 5 21 6 273 84 6 6 0 4
1999 9E 13923 8 3 39 12 242 76 28 0 2 0
Change -6 8 -18 31 -22 6 -2 4
2009 12D 186 183 7 132 5 2300 87 11 7 0 2
1999 12D 193 188 7 261 10 2042 78 91 5 40 0
Change -7 -5 -129 258 -80 2 -40 2
2009 12E 223 130 3 188 4 4158 91 71 2 0 12
1999 12E 225 177 4 408 9 3321 73 370 12 267 5
Change -2 -47 -220 - 837 - 299 -10 267 7

35
Table 13 below shows the areas occupied by each classification category within the two sub-management areas of the Caney Ranger District,
and the changes in area between 1999 and 2009. The percentages given for the forested categories are the percent of total classified acres.
Because no areas were classified as shrubs in 2009, zero acres were counted for this category in 2009 in order to calculate the change in area
during the decade. A similar procedure was used when a particular classification category was not found in a particular combination of sub-
management area and year.
Year Sub-
Management
Area
Forest
Mixed
Hardwood
Pine Forest
Pine
Forest
Herbaceous
Vegetation
Not Vegetation
(mainly water)
Shrubs Bare
Earth
2009 2AS 6450 3042 36 2272 27 1226 15 151 1466 0 180
1999 2AS 6508 1814 22 1320 16 2597 31 157 2235 153 3
Change -58 1228 952 -
1371
-6 -769 -153 177
2009 3BS 2207 1125 49 7735 34 3219 15 178 153 0 53
1999 3BS 2222 6314 28 4410 20 1009 45 448 814 355 1
Change -15 -5189 3325 2210 -270 -661 -355 52
The management goals for sub-management area 13, the Kisatchie Hills Wilderness on the Kisatchie Ranger
District include “maintaining and protecting the enduring resource of wilderness as one of the multiple uses of Kisatchie
National Forest while providing a wide range of suitable wildlife habitats for all native wildlife” (forest plan 3-48). Part of
the desired future condition is that “the landscape is composed of areas of relatively open pine stands eventually
dominated by native, fire dependent longleaf pine communities” (forest plan 3-48). The increase in pine forest area and
decrease in mixed hardwood-pine forest (Table 15) would appear consistent with moving towards the desired future
condition, but the increase in hardwood forest is not. Possibly mixed hardwood-pine forest is being replaced with pine
forest on the more upland areas, while being replaced with hardwood forest on areas near streams and drains. An
additional complication in determining if the area is approaching management objectives is that not all the pine in the
wilderness is longleaf pine, but remote sensing cannot distinguish longleaf pine from other southern pine species.
No explanation of the differences between portions of the same sub-management area located on different
districts is readily apparent. For example, no clear pattern emerged from an attempt to plot the average site index (as
calculated from NRCS soil data downloaded from http://soildatamart.nrcs.usda.gov/ ) for loblolly pine for the various
districts against the change in hardwood area on those districts.
As mentioned in the Discussion, the congruence between the observed changes and the management goals
varies between sub-management area and ranger district. The results suggest that there are management challenges
for some portions of sub-management areas 1C, 3BL, 3BS, and 5CL. The raster files of the 1999 and 2009 classifications
might be examined to determine the locations of the potential problem areas, with subsequent field visits to these areas
to determine the extent of the challenges.

36
Table 14 below shows the areas occupied by each classification category within the sub-management areas of the Catahoula Ranger District,
and the changes in area between 1999 and 2009. Because no areas were classified as shrubs in 2009, zero acres were counted for this category
in 2009 in order to calculate the change in area during the decade. A similar procedure was used when a particular classification category was
not found in a particular combination of sub-management area and year. The percentages given for the forested categories are the percent of
total classified acres. Because the sub-management areas that coincide with the National Catahoula Wildlife Preserve overlap the boundary
between the Catahoula and Winn Ranger Districts, the GIS polygons representing these sub-management areas were split at the district
boundary and the acres attributed to the appropriate district.
Year Sub-
Management
Area
Forest
Mixed
Hardwood
Pine Forest
Vegetation
Not Vegetation
(mainly water)
Shrubs Bare
Earth
2009 1C 19239 2134 22 1931 20 5348 55 109 112 0 18
1999 1C 19349 1642 17 1859 19 5649 59 127 99 165 1
Change -110 492 72 -301 -18 13 -165 17
2009 3BL 26953 4793 22 4869 22 11935 54 228 232 0 45
1999 3BL 27182 3833 18 4618 21 12272 56 461 69 611 10
Change -229 960 251 -337 -233 163 -611 35
2009 3BM 1689 1687 53 877 28 442 14 93 36 0 36
1999 3BM 1724 1614 51 505 16 491 16 19 464 39 3
Change -35 73 372 -49 74 -428 -39 33
2009 3BS 1097 1565 24 1836 28 2977 46 77 20 0 9
1999 3BS 1117 1247 19 1470 23 3591 56 52 17 86 0
Change -20 318 366 -614 25 3 -86 9
2009 5CL 21717 7589 14 9489 18 35653 67 542 205 0 55
1999 5CL 21920 6258 12 8045 15 36846 69 1017 130 1010 25
Change -203 1331 1444 -1193 -475 75 -1010 30
2009 5CS 528 651 18 1019 28 1982 54 24 8 0 0
1999 5CS 536 500 14 746 20 2355 64 18 2 54 0
Change -8 151 273 -373 6 6 -54 0 6
2009 11DM 394 966 24 997 24 2119 52 5 6 0 0
1999 11DM 400 688 17 801 20 2571 63 7 9 11 0
Change -6 278 196 -452 -2 -3 -11 0

37
Table 15 below shows the areas occupied by each classification category within the sub-management areas of the Kisatchie Ranger District, and
the changes in area between 1999 and 2009. Because no areas were classified as shrubs in 2009, zero acres were counted for this category in
2009 in order to calculate the change in area during the decade. A similar procedure was used when a particular classification category was not
found in a particular combination of sub-management area and year. The percentages given for the forested categories are the percent of total
classified acres.
Year Sub—
Management
Area
Forest
Mixed
Hardwood
Pine Forest
Vegetation
Not Vegetation
(mainly water)
Shrubs Bare
Earth
2009 3BL 28108 3928 39 588 6 5336 53 99 183 0 4
1999 3BL 28289 2818 28 1992 20 4734 48 204 53 155 1
Change -181 1110 -1404 602 -105 130 -155 3
2009 3BM 15010 5142 79 688 11 532 8 41 65 0 4
1999 3BM 15074 4677 73 560 9 873 14 16 268 13 0
Change -64 465 128 -341 25 -203 -13 4
2009 3BS 19613 1380 27 899 18 2702 53 63 56 0 1
1999 3BS 19669 1097 22 726 14 3039 60 36 6 139 0
Change -56 283 173 -337 27 50 -139 1
2009 5CL 7047 7924 21 1577 4 26926 73 336 111 0 90
1999 5CL 7158 5179 14 5526 15 24390 66 1135 200 377 46
Change -111 2745 -3949 2536 -799 -89 -377 44
2009 5CM 0 19 50 10 26 9 24 0 0 0 0
1999 5CM 0 10 28 15 42 11 31 0 0 0 0
Change 0 9 -5 -2 0 0 0 0
2009 5CS 10 64 31 42 20 103 49 0 0 0 0
1999 5CS 10 50 24 58 28 96 46 5 0 0 0
Change 0 14 -16 7 -5 0 0 0
2009 9E 5132 0 0 0 0 10 83 0 2 0 0
1999 9E 5134 0 0 1 10 8 80 1 0 0 0
Change -2 0 -1 2 -1 2 0 0
2009 11DL 1742 2796 10 768 3 23084 86 265 2 0 14
1999 11DL 1744 1525 6 3301 12 21069 78 691 31 309 1
Change -2 1271 -2533 2015 -426 -29 -309 13
2009 11DS 116 513 29 240 13 1030 58 8 0 0 0
1999 11DS 116 354 20 398 22 942 53 32 2 62 0
Change 0 159 -158 88 -24 -2 -62 0
2009 11E 156 1317 61 183 9 632 29 13 2 0 0
1999 11E 158 966 45 477 22 583 27 42 7 70 0
Change -2 351 -294 49 -29 -5 -70 0
2009 13 8 2844 33 333 4 5307 61 211 1 0 39
1999 13 9 1684 19 1874 21 4016 46 639 29 492 1
Change -1 1160 -1541 1291 -428 -28 -492 38

38
Table 16 below (and continued on the next page) shows the areas occupied by each classification category within the sub-management areas of
the Winn Ranger District, and the changes in area between 1999 and 2009. Because no areas were classified as shrubs in 2009, zero acres were
counted for this category in 2009 in order to calculate the change in area during the decade. A similar procedure was used when a particular
classification category was not found in a particular combination of sub-management area and year. The percentages given for the forested
categories are the percent of total classified acres. Because the sub-management areas that coincide with the National Catahoula Wildlife
Preserve overlap the boundary between the Catahoula and Winn Ranger Districts, the GIS polygons representing these sub-management areas
were split at the district boundary and the acres attributed to the appropriate district.
Year Sub-
Management
Area
Forest
Mixed Hardwood
Pine Forest
Vegetation
Not Vegetation
(mainly water)
Shrubs Bare
Earth
2009 1C 35612 1262 18 1852 27 3559 52 87 79 0 1
1999 1C 35689 674 10 1150 17 4519 67 142 19 256 4
Change -77 588 702 -960 -55 60 -256 -3
2009 3BL 11778 5642 35 5373 33 5082 31 66 120 0 16
1999 3BL 11897 2821 17 3708 23 9275 57 124 62 185 1
Change -119 2821 1665 -4193 -58 58 -185 15
2009 3BM 12680 3605 35 3312 32 2943 29 61 232 0 69
1999 3BM 12775 2476 24 2338 23 4410 44 51 672 182 0
Change -95 1129 974 -1467 10 -440 -182 69
2009 3BS 8863 3790 25 4128 27 7241 47 119 101 0 6
1999 3BS 8962 2515 16 3028 20 9318 61 124 23 275 0
Change -99 1275 1100 -2077 -5 78 -275 6
2009 3CL 10684 1785 42 1429 33 947 22 51 55 0 3
1999 3CL 10739 725 17 1083 26 2225 53 86 4 90 3
Change -55 1060 346 -1278 -35 51 -90 0
2009 3CM 1497 1282 58 749 34 134 6 10 41 0 1
1999 3CM 1537 548 25 619 28 934 43 30 38 8 0
Change -40 734 130 -800 -20 3 -8 1
2009 3CS 5187 620 28 832 38 734 33 9 22 0 0
1999 3CS 5209 259 12 370 17 1543 70 17 2 3 0
Change -22 361 462 -809 -8 20 -3 0
2009 5CL 12148 11155 25 15203 33 18425 40 540 180 0 24
1999 5CL 12313 4678 10 7831 17 31306 69 586 154 789 18
Change -165 6477 7372 -12881 -46 26 -789 6
2009 5CM 3530 4168 38 3355 31 3314 30 51 53 0 8
1999 5CM 3569 2310 21 2684 25 5579 51 55 184 94 5
Change -39 1858 671 -2265 -4 -131 -94 3

39
Year Sub-
Management
Area
Forest
Mixed
Hardwood Pine
Forest
Vegetation
Not Vegetation
(mainly water)
Shrubs Bare
Earth
2009 5CS 3949 1770 20 2478 28 4633 52 51 39 0 9
1999 5CS 3987 1081 12 1566 18 5953 67 98 63 181 0
Change -38 689 912 -1320 -47 -24 -181 9
2009 7C 26928 3261 32 3852 38 2786 28 63 100 0 7
1999 7C 27028 1853 19 2127 21 5676 57 84 102 126 1
Change -100 1408 1725 -2890 -21 -2 -126 6
2009 10DM 63 1816 71 574 22 130 5 37 1 0 5
1999 10DM 64 1400 55 490 19 372 15 1 298 1 0
Change -1 416 84 -242 36 -297 -1 5
2009 10EM 79 1677 63 699 26 220 8 29 7 0 16
1999 10EM 85 1223 46 436 16 369 14 5 607 3 0
Change -6 454 263 -149 24 -600 -3 16
2009 11DM 98 705 41 484 28 537 31 5 0 0 0
1999 11DM 98 610 35 515 30 600 35 0 5 0 0
Change 0 95 -31 -63 5 -5 0 0
2009 11DS 287 563 11 945 19 3540 70 34 0 0 5
1999 11DS 287 460 9 1042 20 3493 69 41 9 36 4
Change 0 103 -97 47 -7 -9 -36 1
2009 11E 1054 2429 17 3180 22 8653 60 41 0 0 6
1999 11E 1053 2437 17 3032 21 8689 61 36 20 97 0
Change 1 -8 148 -36 5 -20 -97 6

KNF_RemoteSensingVegetationMonitoringFinal

More Related Content

What's hot (19)

Viewers also liked (11)

Similar to KNF_RemoteSensingVegetationMonitoringFinal (20)

KNF_RemoteSensingVegetationMonitoringFinal