The Level 1 dataset has an overall classification accuracy of about 91.2%. For each Level 1 class, 120 pixels were selected and assessed for accuracy. The results were compiled in an error matrix, as shown below. Then each column in the error matrix was multiplied by the proportion of that cover type on the landscape, and compiled to yield an overall accuracy estimate for the entire image. Detailed reference information was not available to complete an assessment of level 2 accuracy. 

                              Reference Data
Classified     10-Agri 20-For 30-Urb 40-Water 50-Wet/Shrub/Grass  TOTAL
10-Agri            115      0      4        0            1          120
20-Forest            2    113      0        0            5          120
30-Urban             5      5    103        0            7          120
40-Water             0      2      0      117            1          120
50-Wet/Shrub/Grass   0      4      0        0          116          120
  TOTAL            122    124    107      117          130
  
                  User's Accuracy  Producer's Accuracy
10-Agri              96.36%              94.26%
20-Forest            97.55%              91.13%
30-Urban             28.31%              96.26%
40-Water             72.48%             100.00%
50-Wet/Shrub/Grass   93.65%              89.23%

Impervious Surface Area Accuracy Assessment:

To measure the accuracy of the Landsat-derived impervious surface estimates, an independent sample of assessment sites were selected for each of the Landsat images. The sites were randomly selected and compared with high resolution aerial imagery to determine the amount of impervious surface within each accuracy assessment sites. Accuracy was determined with a regression of actual vs. predicted plots results are listed below.

path 27 row 29 (TCMA area)

Residual standard error: 10.31 on 133 degrees of freedom
Multiple R-squared: 0.9377,     Adjusted R-squared: 0.9372 
F-statistic:  2001 on 1 and 133 DF,  p-value: < 2.2e-16 


path 27 row 28 (North of the TCMA)

Residual standard error: 14.9 on 117 degrees of freedom
Multiple R-squared: 0.8362,     Adjusted R-squared: 0.8348 
F-statistic: 597.2 on 1 and 117 DF,  p-value: < 2.2e-16 


path 26 row 29 (Wisconsin, east of the TCMA)

Residual standard error: 11.5 on 107 degrees of freedom
Multiple R-squared: 0.8912,     Adjusted R-squared: 0.8901 
F-statistic:   876 on 1 and 107 DF,  p-value: < 2.2e-16

RMS error is less than 1/4 - pixel or 7.5 meters

RMS error is less than 1/4 - pixel or 7.5 meters

Raw Landsat images were georeferenced to less than ¼ pixel RMS error. Early- and late-summer images were stacked to create a single image with 14 bands of spectral information. A guided clustering classification approach was applied. Training areas were delineated for Level 1 classes (agriculture, forest, urban, water, wetland/shrubland/grassland). One class at a time, training area pixels were clustered into subclasses based on their spectral characteristics. Spectral signature separability was evaluated for the subclasses (within, then among level 1 classes), and signatures were merged or deleted as necessary. Then a maximum likelihood classifier was used to classify the entire image. Subclasses were subsequently combined back into level 1 classes. Level 2 classes for forest, urban, and wetland/shrubland/grassland were derived by assessing subclass clusters. For agricultural pixels, level 2 classes were drawn directly from the 2007 NASS Cropland Data Layer classification (resampled to 30 m resolution), because adequate reference data was not available. The level 2 'Extraction' class was derived from a 2005 Metropolitan Council shapefile, and is only available for Anoka, Carver, Dakota, Hennepin, Ramsey, Scott, and Washington Counties. The urban/developed class was reclassified to percent impervious surface area.

Impervious Surface Area:
Landsat Thematic Mapper data were used to classify and map impervious surface area for the Urban class. Impervious area is mapped as a continuous variable from 0 to 100 percent for each 30-meter pixel in raw form. For distribution impervious has been reduced to 6 classes.

Rectified Landsat imagery was transformed to tasseled cap values using ERDAS Imagine default processes. Only band two (greeness) was used for the impervious transformation.

Model calibration sites were randomly selected and high resolution aerial imagery was used to determine the amount of impervious surface within each calibration site. Fifty or more sites, distributed across the image area, were collected for each image. Imperious surface area was determined from the FSA-NAIP imagery. Percent impervious surface area was calculated for each site by hand digitizing impervious surface within the site.

Modeling Impervious Surface Area:

The measurements of impervious surface area from the calibration sites were used to develop a least squares regression model relating percent impervious to the spectral-radiometric response of the Landsat TM data for each of the images. The Landsat parameter used to estimate percent impervious was tasseled cap greenness which is sensitive to the amount of green vegetation and therefore is inversely related to the amount of impervious.  The regression model used was different for each Landsat image.

0 Unclassified
11  Row crops
12  Small grains
13  Pasture / hay
14  Other crops
20 Forest
21 Deciduous forest
22 Coniferous forest
23 Mixed forest		
30 Urban
32 Turf grass
33 Extraction (gravel pits, etc) [7-county TCMA only]
40 Water (open water)
50 Wetland / Grassland / Shrubland
51 Woody wetland
52 Emergent herbaceous wetland
53 Shrubland
54 Grassland / herbaceous
61 Impervious intensity >0% - 10%.
62 Impervious intensity >10% - 25%.
63 Impervious intensity >25% - 40%.
64 Impervious intensity >40% - 60%.
65 Impervious intensity >60% - 80%.
66 Impervious intensity >80% - 100%.

LEVEL 1 CLASSES
0  Unclassified
10  Agriculture (combine 11-19)
20  Forest (combine 21-29)
30  Urban (combine 31-39 and 60-69)
40 Water
50  Wetland/shrubland/grassland (combine 51-59)

Level 2 Classes
11 Agriculture Land cover Row crops.
12 Agriculture Land cover Small grains.
13 Agriculture Land cover Pasture / hay.
14 Agriculture Land cover Other crops.
21 Forest Land cover Deciduous forest
22 Forest Land cover Coniferous forest
23 Forest Land cover Mixed forest
(31) Urban Land Cover Developed land (impervious 0-100%)
32 Urban Land Cover Turf grass
33 Urban Land Cover Extraction (gravel pits, etc) [7-county TCMA only]
40 Open water
51 Woody wetland Land Cover
52 Emergent herbaceous wetland Land Cover
53 Shrubland Land Cover
54 Grassland / herbaceous Land Cover
61 Acres of Urban Impervious intensity greater than 0 and less than or equal to 10.
62 Acres of Urban Impervious intensity greater than 10 and less than or equal to 25.
63 Acres of Urban Impervious intensity greater than 25 and less than or equal to 40.
64 Acres of Urban Impervious intensity greater than 40 and less than or equal to 60.
65 Acres of Urban Impervious intensity greater than 60 and less than or equal to 80.
66 Acres of Urban Impervious intensity greater than 80 and less than or equal to 100.

This data may be used for educational and non-commercial purposes, provided proper attribution is given. Secondary distribution of the data is permitted, but not supported by the University of Minnesota. By accepting the data, the user agrees not to transmit this data or provide access to it or any part of it to another party unless the user includes with the data a copy of this disclaimer.

see website or contact info http://land.umn.edu