Day 4

Detailed paper information

Back to list

Paper title Supervised Machine Learning for Optical Water Types Classification in Lakes
  1. Ana Belen Ruescas Universitat de València Speaker
  2. Katalin Blix University of Tromsø
Form of presentation Poster
  • A7. Hydrology and Water Cycle
    • A7.06 EO for monitoring water quality and ecological status in inland waters
Abstract text The objective of this work is to develop new classifications of optical types of water, using remote sensing reflectance (Rrs) measured by satellite as basis. The Rrs of several lakes with different water types are selected and labelled by an expert, identifying each optical water type (OWT) manually. The area of study are different reservoirs and lakes on the eastern Iberian Peninsula, and the Rrs are extracted from Sentinel 2-MSI atmospherically corrected imagery.

The OWT classifiers used here are framed in what is known as supervised classifiers, since they use the previous information given by the user to determine the classes to be detected. In order to classify these data we need atmospherically corrected images, for which the Case 2 C2RCC (Case 2 Regional Coast Color) algorithm developed by Doerffer et al. (2016) and available at SNAP has been applied. The collection of the refectance samples for training and testing has also been carried out using SNAP GUI.

Jupyter Notebooks are in place for the training, testing, application and validation of the models. The classifications generated can help to better understand the seasonal and spatial variations of the studied water masses, being a basic support in the monitoring programs of lakes and reservoirs. It is possible to use the OWT classification as final products to analyse changes in water types related to the different water dynamics of the lakes, or they can considered an intermediate products that could help in the subsequent selection of the water quality data extraction algorithm (for example, chlorophyll concentrations or total suspended matter) generated and adapted to specific types of water (Eleveld et al., 2017, Stelzer et al. 2020).

Results on the classifications tested, and validation of those results, will be analysed and considerations taken about the transfer learning to other lakes in Europe.