Jemrest




Home \| Instructions to Authors \| SASE \| Editorial Board \| Articles \| Subscriptions \| Conference

Abstract

Torbick, N, P. Lawrence and Kevin Czajkowsk. 2006. Application and assessment of an expert system for wetlands identification in northwest Ohio, USA. JEMREST 2:26-37

DOI: 10.4029/2006jemrest2no14

The loss of wetlands has gained considerable attention over the past dew decades,. Regulations and practices now promote the conservation and monitoring of these fragile ecosystems. Complexities of wetlands regulations have made detailed and accurate remotely-sensed classifications difficult. With increases in ancillary spatial data, methods to exploit multispectral remotely sensed imagery for advancing wetlands mapping can now be developed. The objective of this paper was to create an expert system to identify four primary wetland types of interest: forest, prairie, riparian, and coastal. Extensive wetlands fieldwork integrated with geospatial agricultural and soils information were utilized to restrict Enhanced Thematic Mapper Plus spectral reflectance to create an updated wetlands inventory. The expert system demonstrated abilities to improve wetlands classification with overall accuracies between 84% and 96%. Challenges were presented by in-class variability, spectral overlap, and the complexity of jurisdictional regulations. A final tradoff between ommissiion error and misclassification was reached. As new imagery becomes available this expert system can be applied to update wetlands maps on a regular basis.

Email info@jemrest.org for full text