5:15 PM - 6:30 PM
[AHW22-P18] Analysis of coastal seagrass bed distribution using UAV and near-infrared camera data
Keywords:Coastal sea grass, UAV, Remote sensing
In this study, we analyzed a combination of visible and near-infrared image data obtained using an unmanned aerial vehicle (UAV) with the aim of understanding the distribution of seagrass beds in coastal shallow waters. Algal beds that thrive on dried-up beaches (hereafter referred to as dried-up beds ; DUB) were identified using NDVI, one of the commonly used vegetation indices. In addition, algal beds thriving in seawater (hereafter referred to as submerged beds ; SMB) were identified using the newly developed index NSWI (Normalized Seaweed Index). These results were compared to the results of a ground survey conducted at the same time.
The distribution of NDVI calculated from the data during the lowest tide in July 2020 was generally consistent with the distribution of DUBs. NSWI was able to identify most of the DUBs under conditions where the tidal depth was less than 40 cm.
However, under conditions where the tidal depth exceeded 40 cm, the SMB could not be successfully identified by NSWI using the obtained raw data because of the large attenuation of the reflection from the seafloor as it travels through the long underwater optical path. For this reason, the attenuation of light at each wavelength was corrected using the Lambert-Veil law, and the NSWI was recalculated using the corrected data for improved analysis. The results show that the improved NSWI can identify SMBs under the conditions of tidal depths between 40 and 170 cm.
Using the data from eight flights in July 2020 in the specific area analyzed in this study, we calculated the combined area of DUB and SMB, and found that the area was almost constant regardless of the water depth under the condition of a tidal depth of 140 cm or less. The average of eight data runs showed an area value of 3104 m2 , with a variance of 462 m2. This suggests that the area of seagrass beds in the shallow water can be estimated with a random error of about ±14.9% using the method in this study.
The distribution of NDVI calculated from the data during the lowest tide in July 2020 was generally consistent with the distribution of DUBs. NSWI was able to identify most of the DUBs under conditions where the tidal depth was less than 40 cm.
However, under conditions where the tidal depth exceeded 40 cm, the SMB could not be successfully identified by NSWI using the obtained raw data because of the large attenuation of the reflection from the seafloor as it travels through the long underwater optical path. For this reason, the attenuation of light at each wavelength was corrected using the Lambert-Veil law, and the NSWI was recalculated using the corrected data for improved analysis. The results show that the improved NSWI can identify SMBs under the conditions of tidal depths between 40 and 170 cm.
Using the data from eight flights in July 2020 in the specific area analyzed in this study, we calculated the combined area of DUB and SMB, and found that the area was almost constant regardless of the water depth under the condition of a tidal depth of 140 cm or less. The average of eight data runs showed an area value of 3104 m2 , with a variance of 462 m2. This suggests that the area of seagrass beds in the shallow water can be estimated with a random error of about ±14.9% using the method in this study.