COS 171-10 - Using LiDAR and object-based image analysis to map wetlands in Mt. Rainier National Park

Thursday, August 9, 2012: 4:40 PM
F149, Oregon Convention Center
Meghan A. Halabisky, School of Environmental and Forest Sciences, University of Washington, Seattle, WA and L. Monika Moskal, School of Environmental and Forest Sciences, University of Washington
Background/Question/Methods

Wetlands have been identified by wildlife, vegetation, and aquatic groups as sensitive and critical habitat. However, management of wetlands has been hampered historically by a lack of information on their spatial distribution. Manual photo interpretation using aerial photographs is the primary method for mapping wetlands. However, in high relief and forested areas where terrain and forest canopy can obstruct the aerial view it is difficult to accurately identify wetlands using aerial imagery alone. LiDAR is an active remote sensing technique that emits laser light toward the ground and measures the time of the return back to the sensor creating a three dimensional dataset. LiDAR is uniquely capable of detecting wetlands in steep terrain and below forest canopy as it can penetrate forest canopy and eliminates shadowing effects. The goal of this research was to combine multiple types of data:  LiDAR, aerial imagery, & thematic data to improve the detection of wetlands in Paradise Meadows within Mt. Rainier National Park through the use of object based image analysis (OBIA). An additional goal was to create a method that could be replicated using freely available data, and therefore could be used without additional costly data acquisition. We used the National Wetland Inventory to assess the accuracy of our wetland classification.

Results/Conclusions

The overall accuracy of our wetland classification was 74.68% with a kappa statistic of 61.57%. The producer’s accuracy was 92.00% for open water wetlands and 97.30% for emergent wetlands. The largest source of error came from errors of commission; wetlands that we classified that the NWI did not detect. Within our study area, the NWI identified 202.29 acres, while our classification identified 303.28 acres of wetland habitat. Through examination of the Mt. Rainier amphibian dataset and aerial imagery this error of commission is likely the result of additional wetlands that had not been previously identified by the NWI due to effects of shadows from forest canopy and steep terrain. This research demonstrates that the combination of LiDAR and OBIA can improve the detection of wetlands in forested and high relief areas. The additional benefit of this method is that it can be used to detect wetlands over a broad landscape through batch processing techniques at a relatively low cost.