Estimation of Leaf Area Index Using Lidar, High Resolution Multispectral Imagery, and Advanced Image Segmentation.

Plant Leaf Area Index (LAI) is a determinant factor of evapotranspiration in forested watersheds. Remote measurement of LAI is a difficult task because of the problem associated with accurate plant speciation. Advanced image segmentation techniques could provide very accurate forest tree speciation using Light Detection and Ranging (LiDAR) and high resolution multispectral imagery. The objectives of the study were to: 1) use advanced image segmentation for plant speciation to develop LAI raster layer using vegetation indices derived from LiDAR and multispectral imagery, and 2) validation of estimated LAI using field measurements at two study areas. The study areas are: 1) a managed pine forest plantation (Carteret, NC) and 2) a blueberry orchard in the middle of an unmanaged pine forest (Woodbine, GA). LiDAR data for both sites were pre-processed with Quick Terrain Modeler to develop a highly accurate normalized digital surface model nDSM. The nDSM and the 1m resolution 4-band National Agricultural Imagery Program (NAIP) imagery were required for Object Based Image Analysis (OBIA) in order to segment and classify images. The image segmentation technique employed in this study utilizes image textural information to better classify plant species. Using the nDSM, plant species (trees, bushes, and grass) were separated first based on height. Different trees and bushes were distinguished using the textural and band ratio (NIR/R, NDVI, SAVI, TNDVI) analysis features. Geometry (area, border length, skeleton features, compactness, and density) of segmented features was taken into consideration to distinguish the plants based on their canopy shape, size, etc. Rule sets were developed to complete segmentation and the subsequent classification. LAI raster was developed from the SAVI raster created using the classified image. Finally the LAI from the raster image were validated with the LAI data collected from the field using LICOR instrument for the assessment of prediction accuracy of the proposed method.