In New York State, the golden-winged warbler (GWW), a state species of special concern, has recently been found to nest in swamp forest habitat in Sterling Forest State Park in the Hudson Highlands ecoregion. These swamp forest breeding territories are often embedded in a mosaic of wetland cover types. An accurate map of wetlands in the Hudson Highlands would be a useful input to a GWW habitat suitability model and could help conservation managers better allocate limited resources towards GWW monitoring and habitat management. This study tests the effectiveness of single and dual-season dual-polarization L-band ALOS PALSAR data and National Wetlands Inventory ground reference data for classifying wetlands at high resolution in a mountainous landscape using a maximum likelihood decision rule. PALSAR imagery was chosen over Landsat for its ability to detect surface conditions beneath a vegetation canopy and for its higher spatial resolution, necessary for capturing land cover characteristics at a scale that would be meaningful for characterizing GWW territories. Dual-season dual-polarization imagery produced the highest overall accuracy (72 %, kappa = 0.5669) and the highest producer’s accuracy for forested wetlands (~58 %). Errors in classifying forested wetlands were dominated by misclassification to the Urban class, most likely due to the similarity of SAR double bounce trunk-ground and building-ground interactions. Emergent and scrub-shrub wetlands proved difficult to segregate and were classified with maximum accuracies of 49 and 14 %, respectively. These results are comparable to those of similar studies and could be improved with more current ground reference data. In addition, the classification captured the mosaic of wetland cover types present in GWW study sites and territories in Sterling Forest State Park indicating PALSAR imagery could be used for GWW habitat analyses.
Read full abstract