HomeMy WebLinkAboutGIS pilot assessment
N.C. Dept. of Transportation
7¤³« £ 0±¤£¨¢³¨®
-®£¤« !²²¤²²¬¤³
For Kinston Bypass TIP # R-2553 Corridor 36, Lenoir Co.
Weatherford, Morgan D
3-5-2016
Summary
The Merger Team selected Corridor 36 to assess the wetland prediction models for the Kinston
Bypass Pilot Project. The corridor is approximately 500 feet wide and 28 miles in length for a total of
1,718 acres. Traditional field delineations were performed along the entire corridor using Regional
Supplement to the Corps of Engineers Wetland Delineation Manual: Atlantic and Gulf Coastal Plain
Region. A total of 166.12 acres of wetlands were delineated constituting 9.7% of the corridor. The
model correctly predicted 126 acres or 76% of these wetlands. The model also predicted 86% percent of
the non-wetlands correctly. Overall, the models correctly predicted 86% of the total corridor correctly.
The model also over-predicted 219 acres of wetland. The predictions can be broken down into five
areas: riparian and non-riparian Rolling Coastal Plain (RCP), riparian and non-riparian Carolina Flatwoods
(CF) and the Southeastern Floodplains and Low Terraces (SFLT) which is almost exclusively riparian. Of
these areas the RCP non-riparian model had the highest overall performance while the RCP riparian
model had the lowest overall performance. The greatest influence on the poor performance of the RCP
riparian model was the over-prediction of wetlands. The Carolina Flatwoods non-riparian model had the
lowest accuracy in predicting actual wetlands. A review of the data shows that a major contributor to
this was likely incorrectly mapped non-hydric soils.
i
Introduction
The NCDOT ICI/Onsite Mitigation Group developed wetland prediction models for the purpose of
estimating wetland impacts for the Kinston Bypass (R-2553) Detailed Study Alternatives (DSA) as part of
a GIS Pilot Project. Ultimately, the goal of the pilot project was to reduce fieldwork and costs and
reduce project delivery times by relying on GIS resources more during the project planning phase. In
particular, critical layers such as wetland locations were updated with the intention of selecting the
Least Environmentally Damaging Practicable Alternative (LEDPA) using GIS technology and using minimal
fieldwork. However, prior to selecting the LEDPA, the Kinston Bypass project planning was halted due to
de-prioritization according to the Strategic Mobility Formula. However, NCDOT and the Merger Team
sought to continue with the GIS Pilot Project to obtain tangible results and provide feedback for future
modeling efforts. As a result, a DSA was chosen by the Merger Team to perform a full field-delineation
to compare actual wetland locations to predicted wetland locations in a statistical analysis. Further GIS
analysis was performed to provide more detailed characterizations of the predictions and errors
followed by a Discussion Section providing future work and potential adjustments to the model if
necessary.
Methodology
The Merger Team chose Corridor 36 as the test corridor due to the variation of ecoregions and wetland
types located in the footprint (Figure 1). The corridor is 500 ft. wide and approximately 28 miles in
length. The wetlands were delineated using the Regional Supplement to the Corps of Engineers Wetland
Delineation Manual: Atlantic and Gulf Coastal Plain Region and their functional ratings were
determined according to the NC Wetland Assessment Method. The boundaries were mapped with a
GPS and converted to an ArcGIS shapefile with assigned attributes including individual wetland name,
NC WAM wetland type and NCWAM functional ratings. In all, 76 individual wetland polygons were
delineated in the corridor for a total of 166.12 acres. Note, some polygons represent the same
contiguous wetlands that were broken down into more than one NC WAM wetland type or assessment
area. Table 1 shows the approximate acreage by each NCWAM wetland type.
NC WAM TypeAcreageRiparian/Non-riparian
Basin Wetland0.48 Non
Hardwood Flat0.01 Non
Pine Flat41.77Non
Pocosin26.75Non
Seep0.04 Non
Bottomland Hardwood Forest 19.37Rip
Headwater Forest 9.44 Rip
Non-tidal Freshwater Marsh6.40 Rip
Riverine Swamp Forest61.85Rip
Total166.12 Rip-97.06ac. Non-69.05 ac.
Table 1: Total acres of delineated wetlands in Corridor 36 by NC WAM wetland type.
1
Jurisdictional streams were also field delineated and mapped with polygon shapefiles. Non-
jurisdictional ditches that had no ordinary high water mark were mapped as well and placed into a
different shapefile.
The final wetland model results were developed from and delivered as a raster, or grid, where each cell
measured 20 ft by 20 ft. To match this format, the polygon for Corridor 36 was converted to a raster in
ArcGIS with 20 ft cell size. This corridor raster was then converted to a point file by using the centroid of
each grid cell to represent each 20 ft cell (Figure 2). A total of 187,117 points represented the entire
1,718 acre corridor and were the basis of the analysis. By combining these points with the delineations,
predicted wetlands and other locational information, such as riparian vs non-riparian areas, a thorough
assessment of the model could be conducted using frequency analyses in statistical software and spatial
analyses in ArcGIS.
The point shapefile was first overlaid with theactual field-delineated wetland polygon shapefile. Each
point was assigned an attribute indicating whether it was located in an actual wetland or in a non-
wetland area (Figure 3). NCDOT wanted to record if the actual wetland was a riparian or non-riparian
wetland type. Each point was assigned another attribute to indicate this based on the NCWAM wetland
type captured for each wetland. Next, each point was assigned an attribute to indicate whether it was
located on a predicted wetland (Figure 4). The riparian area shapefile NCDOT digitized for the entire
study area was overlaid to determine whether each point was located in a riparian or non-riparian area
as determined by NCDOT (Figure 5). A total of 313 acres of the corridor (18.2%) was mapped as a
riparian area. Again, an attribute was assigned to indicate this in the attribute table. Each point was
then assigned an attribute for the appropriate ecoregion in which it was located. Lastly, the corridor
was divided into new location and existing location where the corridor follows existing US 70 to the east
and west of Kinston (Figure 6). An attribute was assigned for each point to indicate if it was located on
the new location portion or the existing location portion.
The attribute table was then saved as an Excel spreadsheet. A SAS program was written to read the
spreadsheet and analyze the results. A total of seven frequency procedures were written to create
frequencies, or contingency tables, that represent the total number of the cells and percentages that
were accurately predicted as compared to the delineated wetlands in the corridor for several conditions.
These purpose of these procedures are as follows and the resulting contingency tables are provided in
Appendix A.
Frequency Procedure 1: Compared the total number of cells designated “1” or “0”, or wetland and non-
wetland respectively, for the actual delineations to the number of cells designated as “1” or “0” for the
predicted wetlands for the entire corridor. This procedure yielded one contingency table.
Frequency Procedure 2: Compared the number of cells designated as wetland and non-wetland for the
actual delineations to predicted wetlands (same as above) but the corridor was split into the three
ecoregions in which the corridor was located and reported as such. This procedure yielded three
contingency tables, one for each ecoregion.
2
Frequency Procedure 3: Similar to Procedure 2 but the ecoregions were further divided into riparian or
non-riparian. This procedure yielded six contingency tables, a riparian and non-riparian table for each of
the three ecoregions.
Frequency Procedure 4: Compared the number of cells designated as wetland and non-wetland for
actual delineations to predicted wetlands for the riparian and non-riparian areas over the entire
corridor. This procedure yielded two contingency tables, one for riparian areas and one for non-riparian
areas.
Frequency Procedure 5: Compared the number of wetland and non-wetland cells for actual delineations
to predicted wetlands for the new location portion of the corridor and along the existing US-70 location.
This procedure yielded two contingency tables, one for the new location and one for the existing
location.
Frequency Procedure 6: Compared the number of wetland and non-wetland cells for actual delineations
to predicted wetlands for each NC WAM type. However, because non-wetlands do not have an NC
WAM wetland type, there is no meaningful interpretation for the non-wetland cells. In other words, this
is a count of the number of cells in each delineated NC WAM wetland type and a count of those cells
that were predicted correctly as a wetland. Essentially this is an assessment of positional accuracy for
each wetland. This procedure yielded 10 contingency tables, one for each of the nine wetland types
plus one for the non-wetlands (which is automatically generated but has no interpretation in this case).
Frequency Procedure 7: Showed the number of cells for each delineated riparian and non-riparian
wetland and the proportion located in a riparian or non-riparian area according to NCDOT’s riparian area
map. Essentially, this compares how well the delineators and the predictions matched up on riparian
determinations of wetlands. This procedure yielded 1 contingency table.
Other analyses were performed in ArcGIS to further assess the performance of the model. The cell
points, wetland polygons and other layers were used in a variety of select-by-location and select-by-
attribute procedures to better understand the nature of the predictions and errors, particularly to
qualify the over-predictions. Results from NCDOT’s ditch model were used in the analysis to provide an
initial determination of the potential interactions of estimated lateral effect and predicted wetlands.
Results
Procedure 1
As stated above, the contingency table for the first procedure as produced by SAS along with the
directions on how to read it are provided in Appendix A. Below is a summary of the information from
that table that divides all of the points into four prediction categories; correctly predicted wetland,
correctly predicted non-wetlands, wetlands predicted as non-wetland and non-wetlands predicted as
wetlands.
Of the 187,117 total cells in the corridor, 145,192 cells (78.6%) were both actual non-wetlands and
predicted non-wetlands, or correctly predicted non-wetlands. Note, this is not interpreted as 78% of
3
non-wetland cells were correctly predicted. 13,703 cells (7.3%) of the total were both actual wetlands
and predicted wetlands, or correctly predicted wetlands. Adding these two numbers shows that
158,895 cells or 85% of the cells in the corridor were predicted correctly. 4,372 cells (2.3%) of the total
cells were actual wetlands but predicted as non-wetland while 23,850 cells (12.8%) of the total were
actual non-wetlands predicted as wetlands.
The Prediction Category for All Points in the Corridor
13%
2%
7%
78%
Correctly Predicted Non-WetlandCorrectly Predicted Wetland
Wetlands Predicted as Non-wetlandsNon-wetlands Predicted as Wetlands
Chart 1. The percentages of the correct and incorrect predictions for all points in the corridor.
The contingency table also provides two common measures of model performance, specificity and
sensitivity and the corresponding Type I and Type II error rates. Specificity in this case is the proportion
of actual non-wetland cells predicted as non-wetlands which is 85.9%. This results in a Type I error
(false positive rate or over-prediction) of 14.1% which is equal to 1-Specificity. Sensitivity is the
proportion of actual wetland cells predicted as wetlands which is 75.8%. This results in a Type II error
(false negative rate) of 24.2%, equal to 1-Sensitivity. The contingency tables also tell us that of the
predicted non-wetland cells, 97.1% were actually non-wetland while 2.9% were actually wetland and of
the predicted wetland cells, 63.5% were actually non-wetland while 36.5% were wetland.
In terms of acreage, 126 acres of the 166 acres of the actual wetlands were predicted correctly, 77 acres
of the riparian type wetlands and 49 acres of the non-riparian type. The model also over-predicted 219
acres of wetlands in the corridor, 155 non-riparian acres and 64 riparian acres. From the past modeling
efforts and discussions with the agencies, it was not unexpected that over-predictions would be present.
It is important to understand the nature of these over-predictions which is discussed in more detail later
in the Results.
4
Procedure 2
The contingency tables for each ecoregion are provided in Appendix A however a summary table for all
three ecoregions is provided below (Table 2). Carolina Flatwoods (CF) model had the highest error rates
and lowered the overall model performance. This was not unexpected due to the difficulties in creating
a satisfactory model for this ecoregion. The models for the Rolling Coastal Plain (RCP) and Southeastern
Floodplains and Low Terraces (SFLT) had significantly lower error rates.
Ecoregion% of Corridor Specificity %Sensitivity %Overall %
CF 38.179.5 59.577.9
RCP 18.589.8 83.489.3
SFLT 43.493.4 87.892.6
Table 2. Model assessment for each ecoregion.
Procedure 3
The riparian and non-riparian areas were separated and reported for each ecoregion and the summary
is provided below (Table 3). The CF was relatively consistent in overall performance for both the
riparian and non-riparian areas while having the overall lowest sensitivity. The RCP Non-riparian and the
SFLT models were similar in overall performance as well. The RCP Riparian model had the highest
overall error rate, mostly due to over-predictions. The RCP Non-riparian model had the best
performance of any model. In fact, every non-riparian model performed better overall than its riparian
counterpart. Though it should be noted no non-riparian wetlands were delineated in the SFLT ecoregion
because the ecoregion is almost exclusively in a riparian area.
Ecoregion Riparian/ Specificity %Sensitivity % Overall %
Non-riparian
CF Non80.253.378.5
CF Rip75.770.674.8
RCP Non95.190.094.8
RCP Rip60.475.963.9
SFLT Non94.0N/A94.0
SFLT Rip87.288.087.7
Table 3. Model assessment for riparian/non-riparian areas in each ecoregion.
Procedure 4
The riparian and non-riparian models were then assessed along the entire corridor without dividing into
ecoregions. As indicated above, the non-riparian models predicted less actual wetlands correctly but did
correctly predicted a higher percentage of overall cells than the riparian models (Table 4). The main
reason for this is because the riparian models over-predicted the number of wetland cells.
Riparian/Non-riparianSpecificity %Sensitivity % Overall %
Non 88.470.0 87.4
Rip 70.780.5 73.6
Table 4. Model assessment strictly for riparian/non-riparian areas.
5
Procedure 5
The corridor was divided into three parts: two parts represented the portion of Corridor 36 that was
located along existing US 70 to the east and west of Kinston and the third represented the new location
portion. These two conditions were then used to analyze the model performance along an existing road
and associated disturbed land as compared to more forested and agricultural land (Table 5). The overall
performance was fairly consistent though the new location had significantly higher sensitivity. This
means wetlands immediately adjacent to US 70 were more problematic to predict.
Existing/New LocationSpecificity %Sensitivity % Overall %
Existing 89.855.8 88.8
New 84.477.4 83.5
Table 5. Model assessment for Corridor 36 on existing location versus new location.
Procedure 6
This procedure analyzed how many wetland cells the model predicted for every NC WAM wetland type
(Table 6). This is essentially a measure of locational accuracy of the model because this does not take
into account predicted wetland cells that are adjacent or near the actual wetland. Because non-
wetlands do not have a NC WAM type, specificity is not applicable and does not account for over-
predictions. Also note, the first contingency table for Procedure 6 in the Appendix includes all of the
actual non-wetland cells. This table is automatically generated and only provides redundant information
from Table 1. According to these numbers, pocosins were the most challenging to predict. NCDOT will
look at this particular wetland type further. One possible aspect of pocosins that may help explain why
they are difficult to predict is related to their vegetation. Pocosins are typically dominated by dense,
waxy evergreen vegetation. Some literature has suggested these particular areas are often difficult to
obtain sufficient true bare earth LiDAR points which can lead to high error rates in the elevation data.
NC WAM Wetland TypePredicted Correctly % Total # of Actual Acres Not Predicted
Wetland Cells
Riverine Swamp Forest 90.4%67415.9
Pine Flat 87.5%45305.2
Pocosin 43.5%291115.0
Bottomland Hardwood 60.7%21077.6
Forest
Headwater Forest 74.5%10372.4
Non-Tidal Freshwater 56.4%6932.8
Marsh
Basin Wetland 45.0%49 0.2
Seep 100%50.0
Hardwood Flat 0%20.02
Table 6. Number of correctly predicted wetland cells for each NC WAM wetland type.
6
Procedure 7
As part of the modeling process, NCDOT determined that separate models were needed to predict
wetlands in riparian and non-riparian wetlands. To address this, riparian area boundaries for the entire
study area were mapped and the appropriate models were only applied to those areas. This procedure
was designed to report the proportion of actual wetland cells that were determined in the field to be
riparian and were located within in the NCDOT determined riparian area. The same proportions were
also found for the actual non-riparian wetland cells. Both types were highly consistent as 94.8% of the
actual riparian cells were inside the NCDOT determined riparian area and 95.2% of the non-riparian
wetlands matched correctly. In total, 95% of the actual wetlands were identified correctly as riparian or
non-riparian using the NCDOT mapped riparian area.
GIS Analysis
In addition to the frequency tables, the predictions were assessed on a wetland polygon basis, rather
than a cell by cell basis. The initial assessment was a simple selection by location in ArcGIS where the
delineated wetland polygons that intersected at least one predicted grid cell point were selected. Of the
76 wetland polygons, 19 of the polygons did not intersect at least one predicted wetland grid cell point.
Upon further inspection, some of the polygons were in fact located on the predicted wetlands but by
chance did not intersect a point that represented the grid cells. An example of this situation is provided
in Figure 7. After these situations were accounted for, only 12 wetlands (16%) of the total number of
polygons were completely missed by the model. These 12 wetlands totaled to 2.44 acres or 1.5% of the
total delineated wetland acreage.
As stated before, over-predictions were responsible for a large percentage of the model error and were
not completely unexpected. During model development for Carthage Bypass, NCDOT had discussions
with the USACE regarding model expectations. On that particular project, it became apparent that a
more conservative model was preferable and that over-predicting wetlands would be much more
favorable to under-predicting. This frame of though was carried over into the Kinston Bypass model.
To characterize the over-predictions, NCDOT examined them further. Almost 58 acres of the over-
predictions (26%) were attributed to three actual individual wetlands. The next task was to examine if
the other over-predictions were merely “exaggerating“ an actual wetland or trying to “fabricate” a
wetland that did not exist. This is an important though difficult distinction to make because we are only
given a slice of information through the landscape in the form of a corridor and only a limited amount of
surrounding field data is available. Nevertheless, NCDOT looked at the number of these over-
predictions within various distances of an actual wetland in the corridor (Graph 2). Approximately 52%
of the over-predictions were within 300 feet of an actual wetland. This does provide evidence that a
large number of over-predictions had a spatial relationship with an existing wetland and the over-
predictions aren’t merely random in nature.
7
60
50
40
Percentage of
30
Total Over-
Predicted Cells
20
10
0
50100150200250300
Distance (ft) From An Actual Wetland Boundary
Graph 2. The percentage of over-predictions and various distances from an actual wetland boundary.
Another source of error, particularly as it relates to over-predictions is the effect of ditches on wetlands.
The wetland models do not explicitly take into account the lateral effect of ditches. After some
discussions and field visits with the USACE, the NCDOT lateral effect model was developed. The results
of this model are polygons which estimate the lateral effect of the ditches throughout the study area.
Though it is yet to be determined how to implement this model, NCDOT wanted to examine how the
wetland results could be potentially affected by the implementation. Approximately 367 acres of the
entire corridor was located in a polygon estimating the lateral effect of a ditch. Of this, just over 38
acres of over-predictions were within the predicted lateral effect. By applying the lateral effect model,
these 38 acres could be predicted as non-wetland instead of wetland, depending on how it is
implemented. Only 4 acres of predicted wetlands that were actual wetlands could potentially be
removed and would have a marginal effect on the model results and error rates.
Conclusions
This assessment provides the performance of NCDOT’s initial attempt at modeling wetlands on this
scale. Overall, the models correctly predicted 86% of the total corridor correctly and correctly predicted
76% of the wetlands. The Pilot Project goals of producing wetland maps for the study area of known
accuracy and consistency were achieved.
8
Figures
9
q
Figure 1: Corridor 36 of Kinston Bypass R-2553 DSAs
01.536
Corridor 36
Miles
R-2553 500ft DSAs
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
q
Figure 2: Raster of Corridor 36 and the Points at Centroid of Each Raster Cell
0100200400
Corridor 36 Raster Points at Cell Centroid
Feet
Corridor 36 Raster
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
q
Figure 3: Raster Cell Points Assigned "1" or "0" for Actual Wetlands
Corridor 36 Points
True_Wetla Attribute
0
1
0100200400
Delineated Wetlands
Feet
Corridor 36 Raster
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
q
Figure 4: Raster Cell Points Assigned "1" or "0" for Predicted Wetlands
Corridor 36 Points
Cor_ModelW Attribute
0
1
0100200400
Predicted Wetland
Feet
Corridor 36 Raster
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
q
Figure 5: Raster Cell Points Assigned "1" or "0" for NCDOT Mapped Riparian Area
Corridor 36 Points
RipZone Attribute
0
1
0125250500
NCDOT Mapped Riparian Area
Feet
Corridor 36 Raster
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
q
Figure 6: New Location and Existing Location Sections of Corridor 36
Corridor on Existing Location
0124
Corridor on New Location
Miles
Streets
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
q
Figure 7: Example of Wetland That Did Not Intersect A Point
Corridor 36 Points
Delineated Wetlands
0102040
Predicted Wetlands
Feet
Corridor 36 Raster
NC OneMap, NC Center for Geographic Information and Analysis, NC 911 Board
Appendix A
10
Overall Accuracy
The FREQ Procedure
05:44 Friday, March 04, 2016 1
Procedure 1
Table of True Wetland by Modeled Wetland
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Modeled
Modeled
Percent
Non-wetland Wetland
Row Pct
Col Pct
01 Total
Actual Non-Wetland 0 145192 23850 169042
77.59 12.75 90.34
85.89 14.11
97.08 63.51
Actual Wetland 1 4372 13703 18075
2.34 7.32 9.66
24.19 75.81
2.92 36.49
Total 149564 37553 187117
79.93 20.07 100.00
The first column represents the possible values of the actual wetlands, 0 (non-wetland) and 1 (wetland).
The top row represents the possible values of modeled wetlands, 0 (non-wetland) and 1 (wetland). The
four inner boxes highlighted in red each have four numbers that correlate to the appropriate values of
each. The first number in each box gives the number of cells with that combination of values out of the
total number of cells. The number immediately below this is the percentage. For example, 145,192
cells (78.6%) of the 187,117 total cells in the corridor were both actual non-wetlands and predicted
non-wetlands. Note, this is not interpreted as 78% of non-wetland cells were correctly predicted.
13,703 cells (7.3%) of the total were both actual wetlands and predicted wetlands. Adding these two
numbers shows that 158,895 cells or 85.9% of the cells were predicted correctly. 4,372 cells (2.3%) of
the total cells were actual wetlands but not predicted as such while 23,850 cells (12.8%) of the total were
non-wetlands predicted as wetlands. The third number gives a more detailed interpretation of the
numbers by providing two common measures of model performance, specificity and sensitivity and the
corresponding Type I and Type II error rates. The third numbers are read across the rows in both
columns and show how each actual wetland and non-wetland were modeled. Specificity in this case is
the proportion of non-wetland cells predicted as non-wetlands which is 85.9%. This results in a Type I
error (false positive rate) of 14.1% which is equal to 1-Specificity. Sensitivity is the proportion of
wetland cells predicted as wetlands which is 75.8%. This results in a Type II error (false negative rate)
of 24.2%, equal to 1-Sensitivity. The fourth numbers are read in both rows down both columns. This
number shows the proportion of each modeled wetland and non-wetland cell that was an actual wetland or
non-wetland. For example, the first column shows that of the modeled non-wetland cells, 97.1% were
actually non-wetland while 2.9% were wetland. The second column shows that of the modeled wetland
cells, 63.5% were actually non-wetland while 36.5% were non-wetland.
05:44 Friday, March 04, 2016 2
Accuracy by Ecoregion
The FREQ Procedure
Procedure 2
ECO=63h(Carolina Flatwoods)
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
593581527474632
73.0118.7991.80
79.5320.47
95.6579.38
1
270039676667
3.324.888.20
40.5059.50
4.3520.62
Total
620581924181299
76.3323.67100.00
05:44 Friday, March 04, 2016 3
Accuracy by Ecoregion
The FREQ Procedure
Procedure 2
ECO=65m(Rolling Coastal Plain)
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
58224662664850
81.759.3091.06
89.7810.22
98.2255.49
1
105653146370
1.487.468.94
16.5883.42
1.7844.51
Total
592801194071220
83.2416.76100.00
05:44 Friday, March 04, 2016 4
Accuracy by Ecoregion
The FREQ Procedure
Procedure 2
ECO=65p(Southeastern Floodplains and Low Terraces)
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
27610195029560
79.805.6485.44
93.406.60
97.8230.60
1
61644225038
1.7812.7814.56
12.2387.77
2.1869.40
Total
28226637234598
81.5818.42100.00
05:44 Friday, March 04, 2016 5
Accuracy by Ecoregion & Riparian?Non-riparian
The FREQ Procedure
Procedure 3
ECO=63h RipZone=0
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
509251257263497
75.1618.5593.71
80.2019.80
96.2484.72
1
199122684259
2.943.356.29
46.7553.25
3.7615.28
Total
529161484067756
78.1021.90100.00
05:44 Friday, March 04, 2016 6
Accuracy by Ecoregion & Riparian?Non-riparian
The FREQ Procedure
Procedure 3
ECO=63h RipZone=1
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
8433270211135
62.2719.9582.22
75.7324.27
92.2461.40
1
70916992408
5.2412.5517.78
29.4470.56
7.7638.60
Total
9142440113543
67.5032.50100.00
05:44 Friday, March 04, 2016 7
Accuracy by Ecoregion & Riparian?Non-riparian
The FREQ Procedure
Procedure 3
ECO=65m RipZone=0
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
52242270354945
89.514.6394.14
95.084.92
99.3546.76
1
34430783422
0.595.275.86
10.0589.95
0.6553.24
Total
52586578158367
90.109.90100.00
05:44 Friday, March 04, 2016 8
Accuracy by Ecoregion & Riparian?Non-riparian
The FREQ Procedure
Procedure 3
ECO=65m RipZone=1
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
598239239905
46.5430.5277.06
60.3939.61
89.3663.70
1
71222362948
5.5417.4022.94
24.1575.85
10.6436.30
Total
6694615912853
52.0847.92100.00
05:44 Friday, March 04, 2016 9
Accuracy by Ecoregion & Riparian?Non-riparian
The FREQ Procedure
Procedure 3
ECO=65p RipZone=0
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
25223159926822
94.005.9699.96
94.045.96
99.96100.00
1
11011
0.040.000.04
100.000.00
0.040.00
Total
25234159926833
94.045.96100.00
05:44 Friday, March 04, 2016 10
Accuracy by Ecoregion & Riparian?Non-riparian
The FREQ Procedure
Procedure 3
ECO=65p RipZone=1
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
23873512738
30.744.5235.26
87.1812.82
79.787.35
1
60544225027
7.7956.9564.74
12.0487.96
20.2292.65
Total
299247737765
38.5361.47100.00
05:44 Friday, March 04, 2016 11
Accuracy by Riparian/Non-riparian
The FREQ Procedure
Procedure 4
RipZone=0
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
12839016874145264
83.9411.0394.97
88.3811.62
98.2175.94
1
234653467692
1.533.505.03
30.5069.50
1.7924.06
Total
13073622220152956
85.4714.53100.00
05:44 Friday, March 04, 2016 12
Accuracy by Riparian/Non-riparian
The FREQ Procedure
Procedure 4
RipZone=1
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
16802697623778
49.1820.4269.61
70.6629.34
89.2445.50
1
2026835710383
5.9324.4630.39
19.5180.49
10.7654.50
Total
188281533334161
55.1244.88100.00
05:44 Friday, March 04, 2016 13
Accuracy by New/Existing Location
The FREQ Procedure
Procedure 5
New_Locati=0
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
42704487747581
87.279.9797.24
89.7510.25
98.6286.61
1
5987541352
1.221.542.76
44.2355.77
1.3813.39
Total
43302563148933
88.4911.51100.00
05:44 Friday, March 04, 2016 14
Accuracy by New/Existing Location
The FREQ Procedure
Procedure 5
New_Locati=1
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
10248818973121461
74.1713.7387.90
84.3815.62
96.4559.44
1
37741294916723
2.739.3712.10
22.5777.43
3.5540.56
Total
10626231922138184
76.9023.10100.00
05:44 Friday, March 04, 2016 15
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=' '
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
0
14519223850169042
85.8914.11100.00
85.8914.11
100.00100.00
Total
14519223850169042
85.8914.11100.00
05:44 Friday, March 04, 2016 16
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Basin Wetland
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
272249
55.1044.90100.00
55.1044.90
100.00100.00
Total
272249
55.1044.90100.00
05:44 Friday, March 04, 2016 17
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Bottomland Hardwood Forest
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
82812792107
39.3060.70100.00
39.3060.70
100.00100.00
Total
82812792107
39.3060.70100.00
05:44 Friday, March 04, 2016 18
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Hardwood Flat
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct0Total
1
22
100.00100.00
100.00
100.00
Total
22
100.00100.00
05:44 Friday, March 04, 2016 19
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Headwater Forest
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
2657721037
25.5574.45100.00
25.5574.45
100.00100.00
Total
2657721037
25.5574.45100.00
05:44 Friday, March 04, 2016 20
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Non-tidal Freshwater Marsh
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
302391693
43.5856.42100.00
43.5856.42
100.00100.00
Total
302391693
43.5856.42100.00
05:44 Friday, March 04, 2016 21
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Pine Flat
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
65938714530
14.5585.45100.00
14.5585.45
100.00100.00
Total
65938714530
14.5585.45100.00
05:44 Friday, March 04, 2016 22
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Pocosin
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
164412672911
56.4843.52100.00
56.4843.52
100.00100.00
Total
164412672911
56.4843.52100.00
05:44 Friday, March 04, 2016 23
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Riverine Swamp Forest
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct01Total
1
64560966741
9.5790.43100.00
9.5790.43
100.00100.00
Total
64560966741
9.5790.43100.00
05:44 Friday, March 04, 2016 24
Accuracy by NCWAM Type
The FREQ Procedure
Procedure 6
NCWAM_type=Seep
TableofTrue_WetlabyCor_ModelW
True_Wetla(True_Wetla)Cor_ModelW(Cor_ModelW)
Frequency
Percent
RowPct
ColPct1Total
1
55
100.00100.00
100.00
100.00
Total
55
100.00100.00
05:44 Friday, March 04, 2016 25
Accuracy of Riparian Area by WAM Type for Actual Wetlands
The FREQ Procedure
Procedure 7
TableofTrue_RipbyRipZone
True_Rip(True_Rip)RipZone(RipZone)
Frequency
Percent
RowPct
01Total
ColPct
0
71383597497
39.491.9941.48
95.214.79
92.803.46
1
5541002410578
3.0755.4658.52
5.2494.76
7.2096.54
Total
76921038318075
42.5657.44100.00