Monitoring of Riverine Aquatic Vegetation Using Satellite PlanetScope Imagery: Feasibility, Limitations and Prospects

Spectral interference induced by the water and the spatial resolution of many satellite images (≥ 10 m) limit the efficiency of remote sensing for monitoring riverine aquatic plant stands. In this study, the potential of using PlanetScope satellite images (3 m in spatial resolution, ~daily acquisition) for monitoring seasonal and interannual aquatic vegetation surface area was evaluated. Airborne images (≤ 0.2 m) acquired on four dates on three aquatic plant stands were used to create, through visual interpretation, reference maps indicating whether the pixels of each PlanetScope image acquired at ±8 days correspond to aquatic vegetation or nonvegetated aquatic areas. For each PlanetScope image, the green normalized difference vegetation index (GNDVI) was calculated and centred on the mean (GNDVI centred ) to distinguish aquatic vegetation from nonvegetated aquatic areas while minimizing variations in their spectral signature over time. Reference maps from the date when aquatic vegetation was the least developed were used to calculate the GNDVI centred classification threshold. To reduce classification errors from radiometric inconsistencies, the frequency at which pixels of PlanetScope images acquired at ±8 days from airborne images were classified as aquatic vegetation was also calculated. Aquatic vegetation was then empirically defined as pixels with a frequency ≥ 85%. Although the classification of PlanetScope images indicates that low abundances of aquatic vegetation cannot be detected, our results show that large changes in stand surface area can be monitored using a multidate classification threshold, thus providing new opportunities for the monitoring of riverine aquatic vegetation on large scales.