Evan Hockridge

Thomas Gonya

Todd Horn

The Impacts of Altitude on High-Resolution Multispectral Remote Sensing for Hardwood Forest Species Delineation

Hardwood forest inventory is a time consuming and expensive process, especially for small forest owners (Cite). To expedite the process, satellite remote sensing techniques have been used with limited success due to low spatial resolution; unless expense high-resolution satellite data is purchased (Immitzer et al., 2012). Recent studies have shown that unmanned aerial systems (UAS) can be a means to collect affordable high-resolution forestry data that is as accurate as traditional ground surveys in the Central Hardwood Forest region of the United States (Hockridge, 2018). However, to fully match the quality of ground surveys, UAS remote sensing methods will need to successfully delineate species of trees in the imagery.

Our project seeks to identify tree species through UAS multispectral imagery utilizing the RedEdge Altum sensor integrated on the C-Astral Bramor ppX airframe. The sensor’s access to the LWIR and NIR bands and aircraft’s low altitude flight capability should allow for species specific spectral signatures to be identifiable. The aircraft will be flown at forested sites local to the West Lafayette, IN region at approximately 300 feet. It is unknown how much flight altitude’s effect on spatial resolution will impact the accuracy of the species delineation, so the Altum sensor will also be mounted on a fixed wing manned airplane that will fly surveying transects at altitudes of 2,000 and 3,000 feet so that data can be compared. The airborne datasets will be processed into multi-band orthomosaics and then analyzed to determine indices that can be used to identify tree species. From there, maps of tree species will be generated for both manned and unmanned aircraft data, and the accuracy associated with different spatial resolutions will be assessed when compared with ground surveys of tree species. From this, an ideal altitude for tree identification can be determined.

References

Immitzer, M., Atzberger, C., & Koukal, T. (2012). Tree Species Classification with Random Forest Using Very High Spatial Resolution 8-Band WorldView-2 Satellite Data. Remote Sensing, 4(9), 2661-2693. doi:10.3390/rs4092661