L-CONNECT (river-Lake CONNECTivity)

L-CONNECT framework combines machine learning algorithms and spectral similarity features to predict water surface connectivity between floodplain lakes and their main river. The assumption is that the spectral similarity between river and lake waters is a good proxy for hydrological connectivity. This methodology applies when the main river drives the changes in the optical water properties (i.e., flux of high sediment load waters) of its flooding lakes. Here, hydrological connectivity means the dynamic flow interactions between large rivers during the flooding season and their floodplain lakes, which occurs by channelized and diffuse overbank flows, and affects surface water connectivity.

Requirements and Usage:

L-CONNECT framework is coded in Python 3.8 and it requires Python packages to run: numpy, pandas, geopandas, rasterio, gdal.

To run the L-CONNECT, it is necessary a suitable use of the environment lconnect by command line:

        conda env create -f environment.yml
        conda activate lconnect
        cd into the lconnect directory
        python api.py

Input Parameters:

The input parameters must be manualy filled in the file parameters.txt. L-CONNECT framework requires different input data: path_IMG, path_CLOUD, path_gridPoint, path_ROI, path_OUTPUT, sensor.

• path_IMG: directory with images;
• path_CLOUD: directory with cloud masks - udm2 (only SuperDove data);
• path_gridPoint: path with a file of points (in shapefile) from reference waterbody (e.g., river);
• path_ROI: path with a file of polygons (in shapefile) from target waterbody (e.g., lakes);
• path_OUTPUT: output directory;
• sensor: applied sensor (sd or msi).

Output Parameters:

Two files .csv are available in path_OUTPUT: reference_spectra.csv and OutputLakesParameters.csv.

• reference_spectra: reference spectra obtained from reference waterbody (monthly average);
• OutputLakesParameters: target waterbody parameters obtained after classification. In this file, the following information is available: date, id_lake, spectral similarity features, lakes spectra, hydrological connectivity or Conn.

The hydrological connectivity is represented by values 0-not-connected and 1-connected.

Notes:

• The images .TIFF (in path_IMG) must be stacked;

• The msi images must be identified with YYYYMMDD;

• The msi images must be in surface reflectance (from 0 to 1);

• The sd images must be in surface reflectance (from 0 to 10,000);

• Spectral bands used for SuperDove: B441, B490, B531, B565, B610, B665, B705, B865;

• Spectral bands used for MSI/Sentinel-2: B490, B560, B665, B705, B740, B783, B842.

Reference:

Rejane S. Paulino, Vitor S. Martins, Evlyn M.L.M. Novo, Daniel A. Maciel, Deborah L. Correia-Lima, Claudio C.F. Barbosa, Marie P. Bonnet, Antje Uhde, A framework based on spectral similarity to estimate hydrological connectivity in Juruá River floodplain lakes using 3-m PlanetScope data. Journal of Hydrology, v. 625, 130156, 2023. (https://doi.org/10.1016/j.jhydrol.2023.130156)