Estimating reed loss caused by Locusta migratoria manilensis using UAV-based hyperspectral data.

Locusta migratoria manilensis has caused major damage to vegetation and crops. Quantitative evaluation studies of vegetation loss estimation from locust damage have seldom been found in traditional satellite-remote-sensing-based research due to insufficient temporal-spatial resolution available from most current satellite-based observations. We used remote sensing data acquired from an unmanned aerial vehicle (UAV) over a simulated Locusta migratoria manilensis damage experiment on a reed (Phragmites australis) canopy in Kenli District, China during July 2017. The experiment was conducted on 72 reed plots, and included three damage duration treatments with each treatment including six locust density levels. To establish the appropriate loss estimation models after locust damage, a hyperspectral imager was mounted on a UAV to collect reed canopy spectra. Loss components of six vegetation indices (RVI, NDVI, SAVI, MSAVI, GNDVI, and IPVI) and two "red edge" parameters (D_r and SD_r) were used for constructing the loss estimation models. Results showed that: (1) Among the six selected vegetation indices, loss components of NDVI, MSAVI, and GNDVI were more sensitive to the variation of dry weight loss of reed green leaves and produced smaller estimation errors during the model test process, with RMSEs ranging from 8.8 to 9.1 g/m;. (2) Corresponding model test results based on loss components of the two selected red edge parameters yielded RMSEs of 27.5 g/m² and 26.1 g/m² for D_r and SD_r respectively, suggesting an inferior performance of red edge parameters compared with vegetation indices for reed loss estimation. These results demonstrate the great potential of UAV-based loss estimation models for evaluating and quantifying degree of locust damage in an efficient and quantitative manner. The methodology has promise for being transferred to satellite remote sensing data in the future for better monitoring of locust damage of larger geographical areas.