The NLCD 2001 is created by partitioning the U.S. into mapping zones. A total of 66 mapping zones were delineated within the conterminous U.S. based on ecoregion and geographical characteristics, edge matching features and the size requirement of Landsat mosaics. Mapping Zone 49 encompasses whole or portions of several states, including the states of Illinois, Wisconsin, Michigan, Indiana, Iowa, Missouri, and Kentucky. Questions about the NLCD mapping Zone 49 can be directed to the NLCD 2001 land cover mapping team at the USGS/EROS, Sioux Falls, SD (605) 594-6151 or mrlc@usgs.gov.
Conceptually, the descriptive tree is a classification tree generated by using the final minimum-map- unit land cover product (1 acre) as training data, and Landsat and other ancillary data as predictors. The goal of the descriptive tree is to summarize the effects of boosted trees (10 sequential classification trees) into a single condensed decision tree that can be used as a diagnostic tool for the classification process. This descriptive tree can be used to assess the relative importance of each of the input data sets on each land cover class. Such information may also be useful to customize the minimum-mapping-unit classification to meet a user's specific needs through raster modeling. Descriptive trees usually capture 60 to 80% of the information from the original land cover data.
The leaf or terminal nodes of the descriptive tree are assigned to sequential numbers (called node numbers) and mapped across the entire mapping zone on a pixel-by-pixel basis. These node numbers can then be matched with the various conditional statements associated with each respective terminal node. This spatial layer appears similar to a cluster map, but is the result of a supervised classification - not an unsupervised clustering. This node map can potentially be used as input by users to customize NLCD land cover, by linking the spatial extent of an individual node with the rules of the conditional statement.
The Land Cover spatial classification confidence data layer is provided to users to help determine the per-pixel spatial confidence of the NLCD 2001 land cover prediction from the descriptive tree. The C5 algorithm produces an estimate (a value between 0% and 100%) that indicates the confidence of rule predictions at each node based on the training data. This spatial confidence map should be considered as only one indicator of relative reliability of the land cover classification, rather than a precise estimate. Users should be aware that this estimate is made based on only training data, and is derived from a generalized descriptive decision tree that reproduces the final land cover data. However, this layer provides valuable insight for a user to determine the risk or confidence they choose to place in each pixel of land cover.
A logic statement from a descriptive tree classification describes each classification rule for each classified pixel. An example of the logic statement follows:
IF tasseled-cap wetness > 140 and imperviousness = 0 and canopy density < 4, then classify as Water
This logic file can be used in combination with the spatial node map to identify classification logic and allow modifications of the classification based on user's knowledge and/or additional data sets.
Additional information may be found at <http://www.mrlc.gov/mrlc2k_nlcd.asp>.
To conduct the land cover classification using DT, a large quantity of training data is required. For mapping Zone 49, training data were collected from several combined sources including ancillary land cover maps such as USGS Multi-resolution Land Cover (MRLC) 1992 maps for Illinois, Wisconsin, Michigan, Indiana, Iowa, Missouri, and Kentucky, USGS Digital Orthophoto Quarter Quadrangles (DOQQs), USGS GAP datasets for Illinois, Michigan, and Indiana, USDA - National Agriculture Statistics Services (NASS) Cropland Data Layers for Illinois and Indiana, and the Illinois Department of Natural Resources, Illinois Natural History Survey, Illinois State Geological Survey, Illinois Department of Agriculture, United State Department of Agriculture National Agricultural Statistics Services 1:100,000 Scale Land Cover of Illinois 1999-2000, Raster Digital Data, Version 2.0, September 2003. The Land cover classes from ancillary land cover map datasets were cross-walked to NLCD 2001 equivalent classification codes prior to use. A unique source image for random sampling of training points was created from multiple datasets (two or more) to take advantage of previous land cover mapping efforts. ERDAS Imagine models were designed to intersect map images to determine the spatial extent of areas where two or more existing maps agreed on the land cover classification. The resulting land cover image was split into separate images for each class. A convolve model was used on each single-class image to create separation between classes reducing the probability that training points over transitional pixels would be selected. Convolved single-class images were then recombined to create the final image for training point sampling. Training points were generated using the ERDAS Imagine Classifier module Accuracy Assessment tool. Classes were randomly sampled on an individual basis in a proportion roughly approximating the percentage of pixels of the sample class in the training image. The ERDAS NLCD Tool module and the utility Convert Pixels to Ascii were used interchangeably to generate independent variable values for use in the .data file. Once an initial classification was completed, a number of subsequent iterations were necessary to improve the classification result. A series of PERL scripts specifically written for this project were used to make adjustments to the See 5 .data file as required to generate an acceptable map.
Note that the training data were used to map all land cover classes except for four classes in urban and sub-urban areas (developed open space, low intensity developed, medium intensity developed, high intensity developed). All urban and suburban land cover classes were mapped and quality assessed separately through a sub-pixel quantification of impervious surfaces using a regression tree modeling method.
Following the development of the best classification through decision tree modeling, additional steps were required to complete the final land cover product. The four classes in urban and suburban areas were determined from the percent imperviousness mapping product (described in the next section). The threshold for the four classes is: (1) developed open space (imperviousness < 20%), (2) low-intensity developed (imperviousness from 20 - 49%), (3) medium intensity developed (imperviousness from 50 -79%), and (4) high-intensity developed (imperviousness > 79%). In Zone 49, tertiary (local) roads were edited in additional steps to reduce the anomalous occurrence of urban pixels due to these roads in rural areas. In the first step, all local roads were removed from the NLCD thresholded impervious layer, creating a modified urban layer. A mask was developed to include urban areas and agricultural areas where systematically spaced local roads were present. This mask was used to extract local roads from the original ancillary layer. The selected local (extracted) roads to be included in the final land cover product were recombined with the modified NLCD urban layer to create the final urban landcover. Other classes of forest and non-forest were combined with the urban classes to complete the land cover product. Finally visual inspection of the classification was made with areas/pixels that were wrongly classified delineated first as an "area of interest" (AOI), subsequently then limited manual editing was done to eliminate the classification error within the AOI.
The completed single pixel product was then generalized to a 1 acre (approximately 5 ETM+ 30 m pixel patch) minimum mapping unit product using a "smart eliminate" algorithm. This aggregation program subsumes pixels from the single pixel level to a 5-pixel patch using a queens algorithm at doubling intervals. The algorithm consults a weighting matrix to guide merging of cover types by similarity, resulting in a product that preserves land cover logic as much as possible. Agricultural classes (class codes 81 and 82) were generalized to approximately 22 acres (100 ETM+ 30 m pixel patch) using the multiple MMU option in the NLCD Smart Eliminate Tool.
Acquisition dates of Landsat ETM+ (TM) scenes used for land cover classification in Zone 49 are as follows:
SPRING-
Index 7 for Path 22/Row 31 on 03/25/00 = Scene_ID 7022031000008550
Index 8 for Path 22/Row 32 on 04/29/01 = Scene_ID 7022032000111950
Index 8 for Path 22/Row 33 on 04/29/01 = Scene_ID 7022033000111950
Index 9 for Path 22/Row 34 on 04/13/01 = Scene_ID 7022034000110350
Index 4 for Path 23/Row 30 on 04/25/00 = Scene_ID 5023030000011610
Index 5 for Path 23/Row 31 on 05/03/00 = Scene_ID 7023031000012450
Index 4 for Path 23/Row 32 on 04/25/00 = Scene_ID 5023032000011610
Index 4 for Path 23/Row 33 on 04/25/00 = Scene_ID 5023033000011610
Index 6 for Path 23/Row 34 on 04/26/03 = Scene_ID 7023034000311650
Index 1 for Path 24/Row 30 on 04/24/00 = Scene_ID 7024030000011550
Index 2 for Path 24/Row 31 on 04/27/04 = Scene_ID 5024031000411810
Index 3 for Path 24/Row 32 on 04/30/99 = Scene_ID 5024032009912010
Index 3 for Path 24/Row 33 on 04/30/99 = Scene_ID 5024033009912010
Index10 for Path 25/Row 32 on 02/27/00 = Scene_ID 7025032000005850
LEAF ON (Summer)-
Index 6 for Path 22/Row 31 on 07/02/01 = Scene_ID 7022031000118350
Index 5 for Path 22/Row 32 on 07/16/03 = Scene_ID 5022032000319710
Index 4 for Path 22/Row 33 on 08/01/03 = Scene_ID 5022033000321310
Index 7 for Path 22/Row 34 on 06/16/01 = Scene_ID 7022034000116750
Index 8 for Path 23/Row 30 on 07/09/01 = Scene_ID 7023030000119050
Index 9 for Path 23/Row 31 on 08/24/03 = Scene_ID 5023031000323610
Index 9 for Path 23/Row 32 on 08/24/03 = Scene_ID 5023032000323610
Index 9 for Path 23/Row 33 on 08/24/03 = Scene_ID 5023033000323610
Index10 for Path 23/Row 34 on 07/04/02 = Scene_ID 5023034000218510
Index 3 for Path 24/Row 30 on 06/19/00 = Scene_ID 5024030000017110
Index 1 for Path 24/Row 31 on 07/14/03 = Scene_ID 5024031000319510
Index 1 for Path 24/Row 32 on 07/14/03 = Scene_ID 5024032000319510
Index 2 for Path 24/Row 33 on 07/27/02 = Scene_ID 5024033000220810
Index11 for Path 25/Row 32 on 07/31/01 = Scene_ID 5025032000121210
Index12 for Path 22/Row 32 on 06/16/01 = Scene_ID 7022032000116750
Index12 for Path 22/Row 33 on 06/16/01 = Scene_ID 7022033000116750
LEAF-OFF (Fall)-
Index 1 for Path 22/Row 31 on 10/19/00 = Scene_ID 7022031000029350
Index 2 for Path 22/Row 32 on 11/07/01 = Scene_ID 7022032000131150
Index 2 for Path 22/Row 33 on 11/07/01 = Scene_ID 7022033000131150
Index 2 for Path 22/Row 34 on 11/07/01 = Scene_ID 7022034000131150
Index 6 for Path 23/Row 30 on 10/24/99 = Scene_ID 7023030009929750
Index 6 for Path 23/Row 31 on 10/24/99 = Scene_ID 7023031009929750
Index 6 for Path 23/Row 32 on 10/24/99 = Scene_ID 7023032009929750
Index 7 for Path 23/Row 33 on 10/10/00 = Scene_ID 7023033000028450
Index 8 for Path 23/Row 34 on 10/16/02 = Scene_ID 7023034000228950
Index 3 for Path 24/Row 30 on 11/16/99 = Scene_ID 7024030009932050
Index 4 for Path 24/Row 31 on 11/05/01 = Scene_ID 7024031000130950
Index 5 for Path 24/Row 32 on 10/20/01 = Scene_ID 7024032000129350
Index 4 for Path 24/Row 33 on 11/05/01 = Scene_ID 7024033000130950
Index 9 for Path 25/Row 32 on 11/12/01 = Scene_ID 7025032000131650
Landsat data and ancillary data used for the land cover prediction -
Data Type of DEM composed of 1 band of Continuous Variable Type.
Data Type of Slope composed of 1 band of Continuous Variable Type.
Data Type of Aspect composed of 1 band of Categorical Variable Type.
Data type of Position Index composed of 1 band of Continuous Variable Type.