Without County Overlay:
With County Overlay:
___________________________________________________________________________________ Procedures
The following procedures were adapted from Viazanko, Andrea, "Predictive Model of Illegal Dumpsites in Westmoreland and York Counties, Pennsylvania" (2017). Theses and Dissertations (All). 1561. The sample size for the Illegal Dumpsite dataset is 46 .
Roads:
Road shapefiles were obtained from TDOT.
Euclidean Distance was used upon the roads polylines.
Extract by Mask was used to cut the extent down to the Upper Cumberland Region.
Road/Euclidean Distance values were extracted to UC Illegal Dumpsites point layer.
The following bar chart was created from this layer;
This chart yielded the following value ranges, ranked from highest to lowest occurrence among dumpsites:
0-419.00 with 43 occurrences
2939.3-3359.3 with 3 occurrences
These value ranges were reclassed as follows:
0-419.00 = 1
2939.3-3359.3 = 2
Raster surface was then set aside until final step.
Landcover:
NCLD 2016 Landcover data was used.
Extract by Mask was used to cut the study area down to the Upper Cumberland Region.
The NCLD dataset was then extracted to the UC Illegal Dumpsites Point Layer.
Landuse was then summarized by count and the following bar chart produced;
It was determined, using this chart, that the following land uses occurred most often;
Developed, Open Space with 32 occurrences.
Deciduous Forest with 5 occurrences.
Developed, Low Intensity and Hay/Pasture with 3 occurrences per category.
Developed, Medium Intensity and Mixed Forest with 1 occurrence per category
The NCLD raster surface was then reclassified using the following values (Method 1);
Developed, Open Space = 1
All other categories = 2
The NCLD raster surface was also reclassified using a different set of values, as follows (Method 2);
Developed, Open Space and Deciduous Forest = 1
Developed, Low Intensity and Hay/Pasture = 2
All other categories = 3
These raster surfaces were then set aside until the final steps.
Population:
2017 Block Group polygon data from the Census Burea (obtained with the help of Chuck Sutherland) was used.
Calculate Geometry was used to calculate area, expressed as square kilometers, for each block group.
Calculate field was then used to calculate population density per square kilometer for each block group. Formula was as follows;
Population (B01001m1)/Area = Population Density per Square Kilometer
Polygon to Raster was then used upon the blockgroup.
Values from this raster were then extracted to UC Illegal Dumpsites layer. The following bar chart was created using this data;
The most common value ranges from this chart are as follows;
0.87-5.7 with 36 occurrences
5.8-10.6 with 5 occurrences
20.4-25.4 with 2 occurrences
10.7-15.5 with 1 occurrence
35.2-40.1 with 1 occurrence
The raster surface was then reclassified as such;
Pop Density/sqr kilom
0-5.7 = 1
All other values (5.8-40.1) = 2
Raster surface was then set aside until final steps.
Slope:
TNDEM (obtained from Chuck Sutherland)
Extract by mask was used to cut the study area down to the Upper Cumberland Region.
Slope was ran on the resulting file using percent rise rather than degree.
Slope values were then extracted to UC Illegal Dumpsites Point Layer. The following bar chart was created from this data;
The chart showed the following ranges were typically found within dumpsites, ranked most to least;
0.74-5.4 with 13 occurrences
5.5-10.1 with 11 occurrences
10.2-14.8 with 6 occurrences
14.9-19.6 with 5 occurrences
19.7-24.3 with 3 occurrences
24.4-29.0 with 2 occurrences
The raster surface was then reclassified with the following values;
0-10.1 = 1
10.2-19.6 and 29.1-33.7 = 2
19.7-29.0 and 33.8-38.4 = 3
Raster surface was then set aside until final steps
Final Steps
Weighted Sum was used to combine the above raster files using the following weights;
Roads = 0.47
Population = 0.23
Slope = 0.15
Land Use = 0.15
The productive surface was then symbolized based upon a stretch type of equalized histogram.
Two models were produced from this process. Only the first, which used the Land Use reclassifcation found in Method 1, is shown as there is no discernabke difference between models made with Method 1 Land Use data and Method 2 Land Use data. _______________________________________________________________________________________
I conclude that this model needs further refinement due to the following limitations; A sample size of 46 dumpsites may be too small to accurately create reclassify variables for use within a predictive surface.
Reclassification scheme may need to be tweaked once further statistical analysis of each variable is conducted.