Assessment of the Suitability of PlanetScope SuperDove Imagery for Aquatic Remote Sensing at Lake Junin in Central Peru

Kurzfassung

This thesis evaluates the suitability of multispectral PlanetScope SuperDove satellite-data for physics-based bathymetry and determination of chlorophyll-a concentrations for a high-altitude lake in Peru. A field campaign was conducted in June 2023 to obtain in-situ data for regional adaptation of the processing methodology and for validation purposes. The collection and analysis of in-situ measurements is presented, accompanied by an analysis of their uncertainties. A validation dataset for bathymetry was obtained using two different methods for in-situ water depth measurements. From this, a historical data set from a lake bathymetry survey in 2006 was adapted, and an up-to-date validation raster was derived (Mean average error was 0.39 m between raster pixels and in-situ measurements). Chlorophyll-a concentration was determined from water samples by two different laboratories and from inverse modelling of in-situ water spectra in the software WASI. A good agreement was found for water depth above 2 m (mean average error: 0.8 µg/l). Three SuperDove images from consecutive days during the field campaign were investigated regarding image noise, sensitivity to atmospheric correction parametrization and consistency in consecutive overpasses. Results are summarized for two areas of interest in optically deep and shallow water. Here it was shown that the band specific reflectance between images were inconsistent, especially so in optically deep water with up to 45% difference in the derived remote sensing reflectance. Additionally, initialization in the software ACOLITE for atmospheric correction showed high sensitivity to the extent of the provided image scene, which impacted estimated parameters for the atmospheric correction. Here, errors over 10% were observed for all spectral bands. Reflectance spectra from three SuperDove images were compared with in-situ measurements from 21 field stations. Excluding the near-infrared band, the coefficient of determination for a linear regression model was 0.73, 0.91 and 0.91 for the three satellite images. Finally, outcomes from processing SuperDove imagery in WASI fell below the desired threshold for a deep water and shallow water model as water spectra derived from SuperDove imagery were distorted. As a result of the study, SuperDove proved to be unsuitable for determining water depth and chlorophyll-a in Lake Junin.