Assessment of classification methods and elaboration of the potential of coarse resolution satellite imagery for forest cover mapping at the Continental region
Malkoc, E.
Environmental spatio-temporal heterogeneity characterizes major parts of the Continental biogeographical region's vegetation, which extends across transition zones between humid and dry climates. Continental forest areas in particular are dynamic and changing through time and space with respect to composition and stand structure due to the Continental region's main characteristics. Satellites provide a general view of the whole Earth that is unavailable to any other forest measurement method. To measure forests globally, satellite imagery is a practical necessity, especially on a global and/or regional scale. Satellite remote sensing is a unique data source for wide regional monitoring as a result of its information content and its frequent coverage. This study evaluates the current and future technical capacity of Proba-V satellite imagery to measure and monitor global forests. MODIS, SPOT-Vegetation and simulated Proba-V imagery are going to be used for assessing the classification methods and elaboration of the potential of coarse resolution satellite imagery for forest cover mapping at the Continental region. Land-cover classifications were performed on each satellite data using-pixel based classification techniques, more specifically, the Maximum Likelihood (ML), Support Vector Machines (SVM) and Artificial Neural Networks classifiers. The stratified random sampling approach was used in order to derive a reference map from Landsat 5 (TM). All the results were evaluated using the reference map derived from Landsat 5 (TM) and ground truth data from Google Earth with high spatial resolution imagery.
显示更多 [+] 显示较少 [-]