Day 4

Detailed paper information

Back to list

  1. Kiana Zolfaghari University of Waterloo, Canada Speaker
  2. Nima Pahlevan NASA Goddard Space Flight Center
  3. Caren Binding Environment and Climate Change Canada, Burlington, ON, Canada
  4. Daniela Gurlin Wisconsin Department of Natural Resources
  5. Stefan Simis Plymouth Marine Laboratory
  6. Antonio Ruiz-Verdú Image Processing Laboratory (IPL) - University of Valencia (Spain)
  7. Lin Li Department of Earth Sciences, Indiana University-Purdue University, Indianapolis, IN, USA
  8. Christopher J. Crawford Earth Resources Observation and Science (EROS) Center, U.S. Geological Survey, Sioux Falls, SD, USA
  9. Andrea VanderWoude Great Lakes Environmental Research Laboratory, NOAA, Ann Arbor, MI, USA
  10. Reagan Errera Great Lakes Environmental Research Laboratory, NOAA, Ann Arbor, MI, USA
  11. Arthur Zastepa Environment and Climate Change Canada, Burlington, ON, Canada
  12. Claude R. Duguay University of Waterloo
Form of presentation Poster
  • A7. Hydrology and Water Cycle
    • A7.06 EO for monitoring water quality and ecological status in inland waters
Abstract text Cyanobacterial Harmful Algal Blooms are an increasing threat to coastal and inland waters. These blooms can be detected using optical radiometers due to the presence of phycocyanin (PC) pigments. However, the spectral resolution of best- available multispectral sensors limits their ability to diagnostically detect PC in the presence of other photosynthetic pigments. To assess the role of spectral resolution in the determination of PC, a large (N=905) database of co-located in situ radiometric spectra and PC collected from a number of inland waters are employed. We first examine the performance of select widely used Machine Learning (ML) models against that of benchmark algorithms for hyperspectral remote sensing reflectance (Rrs ) spectra resampled to the spectral configuration of the Hyperspectral Imager for the Coastal Ocean (HICO) with a full-width at half-maximum of < 6 nm. The ML algorithms tested include Partial Least Squares (PLS), Support Vector Regression (SVR), eXtreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP). Results show that the MLP neural network applied to HICO spectral configurations (median errors < 65%) outperforms other scenarios. This model is subsequently applied to Rrs spectra resampled to the band configuration of existing hyper- (PRecursore IperSpettrale della Missione Applicativa; PRISMA) and multi-spectral (OLCI, MSI, OLI) satellite instruments and of the one proposed for the next Landsat sensor. The performance assessment was conducted for a range of optical water types separately and combined. These results confirm that when developing algorithms applicable to all optical water conditions, the performance of MLP models applied to hyperspectral data surpasses that of those applied to multispectral datasets (with median errors between ~ 73% and 126%). Also, when cyanobacteria are not dominant (PC:Chla is smaller than 1), MLP applied to hyperspectral data outperforms other scenarios. The MLP model applied to OLCI performs best when cyanobacteria are dominant (PC:Chla is equal or greater than 1). Therefore, this study quantifies the MLP performance loss when datasets with lower spectral resolutions are used for PC mapping. Knowing the extent of performance loss, researchers can either employ hyperspectral data at the cost of computational complexity or utilize datasets with reduced spectral capability in the absence of hyperspectral data.