The draft white paper from the Remote Sensing Band on the FIA/NLCD partnership cites the need for a detailed description of how the decision tree method operates. The draft white paper identified the need to supply this information to the Statistics Band for final evaluation. The following information should be useful.

Each member of the FIA Management Team should have already received one hardcopy report on preliminary results from the FIA/EDC pilot tests that evaluated FIA plots for automated mapping of forest cover types.

Chengquan, Huang; Yang, Limin; Homer, Collin; Coan, Michael; Rykhus, Russell; Zhu, Zhiliang; Lister, Andrew; Hoppus, Michael; Tymcio, Ronald; Cooke, William; McRoberts, Ronald; Wendt, Daniel; and Weyermann, Dale. 2002 Synergistic use of FIA plot data and Landsat 7 ETM+ images for large area forest mapping. 3rd Annual FIA Symposium, Oct 17-19 2001, Traverse City, MI.

This report concentrates on the results, but does have a brief presentation on the processing procedures.

The hardcopy might not have arrived yet via USPO in some locations.The 13-Meg MSWord document can be opened at

ftp://edcftp.cr.usgs.gov/pub/edcuser/huang/outgoing/report2fia.doc

This file will be automatically purged by the USGS system from their anonymous ftp site on 01/03/02. If you need to down-load the report after that date, please correspond with Chengquan Huang at

The "machine-learning" or "expert-system" algorithm used by EDC is the Classification and Regression Tree technique. The results from this technique are expressed as sets of if-then rules that are easier to understand by users than other statistical methods (e.g., maximum likelihood classification, logistic regression and neural nets). The Classification and Regression Tree technique is a mature procedure in applied statistics, with many examples in expert-systems, marketing, finance, investments, auditing, pharmaceuticals, epidemiology, and genetics (see ). The seminal reference is:

Breiman, L., J.H. Friedman, R.A.Olshen, and C.J. Stone. 1984. Classification and Regression Trees, Wadsworth International Group, Belmont CA 358 p.

The following reference has a short description on application of Classification and Regression Trees in NLCD-1992. This paper also goes into a lot of other detail on NLCD-1992 that remains relevant to NLCD-2000

Vogelmann, James E.; Howard, Stephen M.; Yang, Limin; Larson, Charles R.; Wylie, Bruce K.; and Van Driel. 2001. Completion of the 1990s National Land Cover Data Set for the conterminous United States from Landsat Thematic Mapper Data and ancillary data sources. Photogrammetric Engineering and Remote Sensing. 67(6). 650-662

The following references concentrate on Classification and Regression Tree applications in remote sensing:

Hansen, M., Dubayah, R., and Defries, R. 1996. Classification trees: an alternative to traditional land cover classifiers. International Journal of Remote Sensing 17:1075-1081.

Friedl, M.A. and C.E. Brodley. 1997. Decision tree classification of land cover from remotely sensed data. Remote Sensing of Environment 61:399-409.

Lawrence, Rick L. and Andrea Wright. 2001. Rule-based classification systems using Classification and Regression Tree (CART) analysis. Photogrammetric Engineering and Remote Sensing. 67(10):1137-1142.

The Classification and Regression Tree algorithm has been implemented in several software systems:

CART by StatWorks ( ).

S-Plus ( )

C5 ( ). This is the software used by EDC. It has the advantage of being designed to

Analyze very large databases, such as MRLC-2000 Landsat/ancillary data, with appropriate FIA plot data

Provide C source code so that classifiers can be embedded into EDC image-database systems.