Increase contrast using standard techniques used in remote sensing to increase the difference between different land cover types. This applies to all satellite images in this project. The Analysis section will contain any detailed results from these procedures. Principal component analysis will be carried out and results compared to other enhancements. Through iterative process, results of classification, together with error analysis will determine the optimum band selection for best classification. This process is similar to optimization, or the steps of knowledge discovery in networked environment of databases (Maceachren et al, 1999). The general steps are: 1) data selection (satellite images), pre-processing (enhancement), transformation and information extraction (classification), interpretation and evaluation (error analysis).

Classification is a process in which all the pixels in an image that have similar spectral signatures are identified (Lillesand and Kiefer, 1994). The largest two advantages of ER Mapper over other software in SIS lab is that the data volume produced with each procedure is very small (store algorithms only), and very fast image processing (ER Mapper, 1995). Generally use supervised classification when have some knowledge of the image and can specify regions explicitly. Yamagata (1997) described this process as follows: Each image pixel is allocated exclusively to one of a small number of known categories, producing an image containing thematic information. The resulting thematic map can be used to estimate the area of each category, if the numbers of boundary pixels or mixed pixels are small. This applies to both IKONOS and Landsat TM images, but the Landsat image gets the benefit of training areas defined by prior IKONOS classification. Attempt to define training areas in the Scotty Creek basin where both images overlap.

After each classification output, compare results to land cover of known locations, estimate errors qualitatively and quantitatively if possible. Use aerial photographs, low altitude oblique photos of ground sites, existing maps, and cross-compare the Landsat classification with IKONOS classification if possible. The IKONOS images are limited to Scotty Creek basin (see Picture 25) where most of the ground truth data comes from.

In IDRISI, use overlay function (image algebra), context specific functions (grouping), and any other methods (Clark Labs, 1999) to separate wetlands obtained from Landsat image classification into connected and disconnected wetlands. The connectivity refers to the surface hydrology network. Wetland classification using IDRISI has been demonstrated by Ahvenniemi (1998) in Finland and Nemliher (2000) in Estonia (eastern Europe).

Simulation of the draining process of wetlands in Scotty Creek basin will be conducted using spatial analysis operations in IDRISI. Distance from any wetland cell to the connected drainage system, (streams and lakes) computes with a COST surface.