Short URL

Quantitative Change Detection & Analysis

 

Other Results & Analysis areas:

Loading, please wait...

Chapter 5 - Results & Analysis

1999 Land Use & Land Cover Accuracy Assessment  |  1984 & 1999 Land Use & Land Cover Statistics  |  Quantitative Change Detection & Analysis  |  Land Assessment

Quantitative Change Detection & Analysis

The areal extent of the forest to pasture and pasture to forest conversions are based on the statistical change between the two land use categories. The Land Use & Land Cover Change Map, 1984 - 1999, for Carroll County is presented in Map Foldout 5.3. The statistical land use change for Carroll County was computed from the loss and gain for each level I land use category between 1984 and 1999 and is further described in Table 5.4. These figures are based directly upon the data derived from the change detection techniques employed in this study.

The land use and land cover change areal statistics were also summarized to express the net forest lost and forest gain. This was accomplished by subtracting the total amount of forest gain from the total amount of forest loss.

Map Foldout 5.3

Table 5.4  Land Use & Land Cover Change 1984-1999 Area Statistics

According to the statistics calculated from the land use and land cover change data, a total of 12.43 squares miles (7,952 acres) of forest was lost, or converted to pasture lands in Carroll County, Arkansas from 1984 to 1999. Although it is less than 2% (1.94%) of the total area of the county, this is a considerable amount when considering the time span of the land use and land cover change, only 15 years. However, this amount may be somewhat misleading since this is the total or gross amount of forest conversion.

These statistics also indicate that 2.46 square miles (1,574.4 acres) of forest was regenerated from fallow pastures. However this is only roughly one-third of one percent (0.38%) of the total area of the county. By subtracting this amount from the total amount of forest conversion (12.43 square miles – 2.46 square miles) the net amount of forest conversion or loss can be calculated: 9.97 square miles (6,380.8 acres) or 1.56% of the total county area. Assuming these statistics to be accurate and precise, the net amount of forest conversion to pasture in Carroll County, Arkansas is still approximately 10 square miles (9.97 square miles or 6,380.8 acres) over a time period of only 15 years. When considered cumulatively, 10 square miles is a large expanse of land to have endured a clearcut.

The individual areas of clearcut were also analyzed based upon their contiguous area and categorized initially, into seven classes. (Table 5.5) A total of 1388 different contiguous areas were delineated as being cleared forest and new pastures. These figures were calculated in ArcView and are derived from the cleared forest vector data, (converted raster data) which is a direct result of the change detection technique discussed in the previous chapter, Chapter Four, Methodology.

Table 5.5  Forest Clearings by Area & Class

It should be noted that, although filtering was applied to remove any cleared forest areas of less than 1 acre (5 raster pixels) to coincide with the minimum mapping unit during the post-classification processing phase of each land use and land cover map, 290 cleared forest vector-polygons (converted raster pixels) were split or considered separate polygons from their original raster pixel clusters due to the differing nature of the data formats.

In the raster data realm (coded grid cells), pixels clusters can be considered contiguous on the basis of either 4 or 8-way connectivity, with 4-way connectivity being characteristic of right, left, top and bottom connections and 8-way connectivity being characteristic of 4-way connectivity plus additional 4 corner diagonal connections. Conversely, in the vector realm (coded points, lines, and polygons), individual polygons that are connected to other polygons (previously considered contiguous pixel clusters diagonally in the raster data realm) are considered as separate entities even though they share a common vertex or node. As a result, 290 polygons with an area of less than an acre were segmented and considered separate from their original contiguous cleared forest areas.

Since the total area of cleared forest in Carroll County would remain the same (sum of the areas of all cleared forest polygons) by assuming that all 290 polygons (with an area of less than one acre) had not been separated from their original contiguous areas, the remaining 1098 larger cleared forest polygons (with an area greater than one acre) are considered the correct and accurate total. Essentially, it is assumed that the 290 cleared forest polygons (with an area of less than one acre) were never separated from the larger 1,098 cleared forest polygons (with an area greater 1 acre). As a result, the number of cleared forest polygons in each of the 6 remaining area classes from Table 5.5 would remain the same. The largest and smallest cleared forest polygons (with an area greater than one acre) were also assumed to remain the same.

Although these are important assumptions, they are not critical to the overall assessment of the cleared forest areas since all spatial analysis has been compiled considering the total area of cleared forest, or the sum of the areas of all cleared forest polygons. With that in mind, a second table was constructed to display these results. (Table 5.6) Table 5.6 displays a more true description of the actual forest to pasture conversion with respect to the valid number of cleared areas, 1098. However, there are a number of important assumptions made in this table, as previously discussed. It shouldbe noted that by recalculating Table 5.6, the average area of forest clearing increased from 5.73 acres (Table 5.5) to 7.24 acres (Table 5.6), an increase of 1.51 acres. This increase in the average size of forest clearing is considered to be accurate since the total area of cleared forest did not change, only the number of cleared forest polygons, which is also considered to be accurate at 1,098.

Table 5.6  Reclassed Forest Clearings by Area & Class

*Assumes that the number of forest clearings in each area class remained the same.

** Assumes that the minimum and maximum cleared forest areas remained the same.

When considering the number of forest clearings per area class in Table 5.6, the resulting trend is not overly surprising. One might assume there to be a greater number of smaller area forest clearings, 940 forest clearings of 1 – 10 acres in area class one. This is the dominant area class and comprises 86 percent (85.6) of the total number of forest clearings. The number of forest clearings declines in each class as the area of each class increases with the exception of the last area class, forest clearings greater than 50 acres. The number of forest clearings in area class two, 10 – 20 acres, comprises roughly 7 percent (6.7) of the total number of forest clearings, 74 cleared forest areas. In area class three, 20 - 30 acres, the number of forest clearings comprises roughly 3 percent (3.2) of the total number of forest clearings, 35 forest clearings. In area class four, 30 – 40 acres, the number of forest clearings comprises roughly 2 percent (1.6) of the total number of forest clearings, 17 forest clearings. In area class five, 40 – 50 acres, the number of forest clearings only comprises about one-half of 1 percent (0.6) of the total number of forest clearings, 7 forest clearings. And in area class six, > 50 acres, the number of forest clearings slightly increases and comprises slightly more than 2 percent (2.2) of the total number of cleared forest areas, 24 cleared forest areas.

One note of interest is the number of forest clearings in area class six, > 50 acres. Statistically, this is the fourth largest area class, out of the six classes, with respect to the number of forest clearings. The following table (Table 5.7) describes the 24 forest clearings in both acres and square miles. There were fifteen forest clearings with an area of 50 – 100 acres. There were eight forest clearings with an area of 100 – 200 acres. And there was one forest clearing greater than 200 acres. The largest forest clearing was 232.82 acres or 0.36 square miles. These twenty-four cleared forest areas comprise about 30 percent (30.2%) of the total amount of forest that was cleared from 1984 to 1999, out of 12.43 square miles.

Table 5.7  Area Class Six, Cleared Forest Areas > 50 Acres

* Assumes that all of the largest 24 cleared forest areas did not change as a result of considering 1,098 as the total number of cleared forest polygons, similar to Table 5.6.

An additional, yet important assumption is also made considering the amount of cleared forest areas or polygons between 1 - 10 acres. It is possible that some of the smaller cleared forest areas of less than a couple of acres were perhaps, not actually clearcut and converted to pastures. This potential misclassification is a result of inherent limitations of the spatial resolution of the Landsat Thematic Mapper satellite data, 30 by 30 meter pixels (900 square meters or 0.22 acres). In addition, spectral resolution is also characteristic of inherent limitations by assuming that every land class incorporated in this study has a constant and consistent spectral signature. Natural processes affecting the landscape such as fire, wind and storm damage, and natural vegetative succession could also account for small amounts of forest loss over the fifteen years considered in this study.

It is extremely difficult to precisely document the total loss of forest in Carroll County without accurate ground truth data to coincide with the September 18, 1984 Landsat Thematic Mapper image. One possible solution would be to obtain and compare historic aerial photographs of the entire county with the satellite derived change map. However, this would offer more of a visual analysis as any potential quantitative assessment would be difficult to derive as a result of differing formats, analog verse digital, and spatial and spectral characteristics. In addition, this is assuming that historic aerial photographs for the entire county exist within reasonable temporal constraints.

The Carroll County office of the Natural Resource and Conservation Service (NRCS) located in Berryville, Arkansas has an aerial photograph data archive containing coverage of the entire county for the years of 1979, 1982, 1990, and 1994. These aerial photographs were acquired by the Arkansas Highway and Transportation Department (AHTD). The specific dates of acquisition were unknown at the time of inquiry, however, most aerial photographs are obtained during the leaf-off or fall season to allow for greater visual assessment. Aerial photographs from 1982 could offer a potential means for change detection accuracy assessment, however, due to the numerous differences between the data types, the 2 year temporal difference, and project time constraints this prospective issue was not investigated further.

Nonetheless, utilizing Landsat Thematic Mapper satellite imagery has been documented as an important source of information for change detection in numerous studies and arenas (Chapter Three, Literature Review) and is considered a viable resource for land use and land cover data. It is important, however, to recognize that there are some inherent limitations involving multispectral classification of the landscape.

Areas of forest conversion were also analyzed for their spatial distribution throughout Carroll County. Although only slightly less than two percent (1.94%) of the entire county was classified as experiencing forest loss from satellite imagery, it is important to determine where areas of land conversion are taking place in order to assess potential effects to the surrounding landscape.

The spatial distribution of cleared forest areas in Carroll County were analyzed by watershed; specifically, utilizing a 12 digit hydrologic boundary of the upper White River. The "12 digit" refers to a classification scheme used by the U.S. Geological Survey that ranks a hydrologic unit area by topographic and hydrologic characteristics and size. A 12 digit hydrologic watershed boundary classifies sub-watersheds ranging in size from 10,000 to 40,000 acres. All of Carroll County resides within the upper White River watershed. By performing a simple overlay operation of the county boundary over the watershed boundary, the specific 12 digit sub-watersheds within Carroll County can be further analyzed. By performing this overlay operation, it was determined that 39 different sub-watersheds are contained within the county.

An additional overlay operation was performing analyzing the amount of areas of cleared forest within each sub-watershed. The area of cleared forest within each sub-watershed was compared to the area of each sub-watershed to determine the percentage of each sub-watershed that had experienced land conversion. This was calculated in square miles and is presented in Table 5.8. In addition, Map Foldout 5.4 portrays the areas of cleared forest within each 12 digit hydrologic watershed boundary in Carroll County.

Based upon this analysis, it was determined that forest conversion had occurred in 37 out of the 39 sub-watersheds present in Carroll County. However, it was also determined that this did not present a sound, quantifiable spatial distribution. The sub-watersheds are not the same size in area, and portions of some sub-watersheds are not contained within the county boundary. For example, the sub-watershed with the highest percentage of cleared forest was also the second smallest sub-watershed, Boatright Creek located in southern Carroll County with an area of 0.36 square miles (231.26 acres) and 0.03 square miles (17.40 acres) of cleared forest or 8.33%. The majority of this sub-watershed resides in Madison County, to the southwest.

Map Foldout 5.4

Table 5.8  Percentage of Sub-Watersheds Cleared of Forest in Carroll County, 1984-1999

* The names of the sub-watersheds were derived from USGS 7.5 minute, 1:24,000 topographic quadrangle maps. Similar names were utilized in some instances with regard to the larger rivers and streams in Carroll County such as specifying upper, middle, and lower sub-watersheds. Some sub-watershed names may not be correct or be outdated due to the fact that some sub-watershed names were not present on the quadrangle maps and alternative names were derived from other physical features in close proximity.

The sub-watersheds with the second and third highest percentages of cleared forest were also characteristic of small areas. For example, Marble Creek, also located in south western Carroll County, is characterized by an area of 2.02 square miles (1,290.26 acres) and 0.14 square miles (86.99 acres) of cleared forest or 6.93%, and Old North Clifty Creek, located in western Carroll County, is characterized by an area of 0.18 square miles (114.25 acres) and 0.01 square miles (8.70 acres) of cleared forest or 5.56%. The majority of Marble Creek also resides in Madison County and good portion of Old North Clifty Creek is now inundated by Beaver Lake.

In contrast, the largest sub-watershed in Carroll County is the middle Osage Creek located in southeastern Carroll County with 39.12 square miles (25,036.18 acres). This sub-watershed experienced 1.10 square miles (704.59 acres) of cleared forest, or 2.81%. The largest amount of cleared forest area within a single sub-watershed occurred in lower Yocum Creek located in northeastern Carroll County, 37.79 square miles (24,188.17 acres), with 1.45 square miles (930.76 acres) of cleared forest or 3.84%.

The average percentage of the sub-watersheds cleared of forest was 2.20%. However, this is a misleading statistic due to the unbalanced size of each sub-watershed. It is nevertheless surprising that forest clearing occurred in 37 out of 39 (94.87%) sub-watersheds in Carroll County. The only two sub-watersheds which did not experience forest to pasture conversion were Woods Hollow Creek, located predominantly in Madison County, and Terrapin Creek which is located predominantly in Boone County, to the east.