You are here:
A COMPARISON OF INTER-ANALYST DIFFERENCES IN THE CLASSIFICATION OF A LANDSAT TEM+ SCENE IN SOUTH-CENTRAL VIRGINIA
Citation:
IIAMES, J. S. AND R. CONGALTON. A COMPARISON OF INTER-ANALYST DIFFERENCES IN THE CLASSIFICATION OF A LANDSAT TEM+ SCENE IN SOUTH-CENTRAL VIRGINIA. Presented at ASPRS 2006 Annual Conference, Reno, NV, May 01 - 05, 2006.
Impact/Purpose:
Our research objectives are to: (a) develop new methods using satellite remote sensor data for the rapid characterization of LC condition and change at regional to national scales; (b) evaluate the utility of the new NASA-EOS MODIS (Moderate Resolution Imaging Spectrometer) leaf area index (LAI) measurements for regional scale application with landscape process models (e.g., biogenic emissions and atmospheric deposition); (c) provide remote sensor derived measurement data to advance the development of the next generation of distributed landscape process-based models to provide a predictive modeling capability for important ecosystem processes (e.g., nutrients, sedimentation, pathogens, etc.); and (d) integrate in situ monitoring measurement networks with UAV and satellite based remote sensor data to provide a continuous environmental monitoring capability.
Description:
This study examined inter-analyst classification variability based on training site signature selection only for six classifications from a 10 km2 Landsat ETM+ image centered over a highly heterogeneous area in south-central Virginia. Six analysts classified the image at the 30 m ETM+ resolution varying only location and number of training sites. These classifications were then degraded to coarser resolutions, assigning the dominant land cover to the new cell resolution. Analyst-to-analyst differences were noted at the varying scales as well as overall accuracy assessment results compared to a land cover map digitized from an August 3, 2002, Ikonos panchromatic image. Results indicated that highest accuracies for all six analysts occurred at the 450 m scale resolution (i.e. 20.25ha), corresponding to a 364m2 (13.25 ha) average patch size for all classes. Spectral separability for training site data was analyzed for each of the six classifications. These tests included a Euclidean Distance, Transformed Divergence, and Jeffries-Matusita Distance evaluation. All spectral separability tests pointed to areas of class confusion within each interpretation, but prediction of an analyst accuracy ranking based on separability amongst all six interpretations was not achieved. This study was initiated to examine land cover variability between analysts as it applies to the process of creating leaf area index (LAI) surface maps used in the validation of medium resolution LAI products.