Code for predicting turbidity dynamics in small reservoirs in central Kenya using Sentinel-2 1C data and machine learning.
# reservoir-turbidity-prediction
Workflow and code for predicting turbidity dynamics in small reservoirs in central Kenya using Sentinel-2 1C data and machine learning as done in Steinbach, S., Bartels, A., Rienow, A., Thiong'o Kuria, B., Zwart S. & Nelson, A. Predicting turbidity dynamics in small reservoirs in central Kenya using remote sensing and machine learning. Int. J. Appl. Earth Obs. Geoinf.
Stefanie Steinbach
This repository contains the workflow and main code scripts used for the paper titled "Predicting turbidity dynamics in small reservoirs in central Kenya using remote sensing and machine learning." The provided scripts use R for ancillary data processing, Python and specific packages installed in a Conda environment for turbidity modelling, and R for machine learning-based predictor selection. Credit for Python support: Torben Dedring
# Input imagery and pre-processing
Sentinel-2 1C imagery is downloaded from the Copernicus Dataspace and all bands resampled to 10 m spatial resolution and a spatial subset created in ESA SNAP. The
# Prepare ancillary data
C2RCC can use default configuration for the retrieval of Inherent Optical Properties (IOPs) or select them from ancillary data. Here, we used ancillary data; NCEP-NCAR for temperature (converted to °C) and surface level pressure, CHIRPS for rainfall (mm), TOMS for ozone (DU), and the SRTM DEM for elevation (masl), all available in the Google Earth Engine Data Catalog. For each Sentinel-2 image file, a CSV file is created with the respective information from the ancillary data, salinity set to 0.0001 (minimum value which is assumed for freshwater), and the other parameters set as desired. The CSV file is named according to the respective Sentinel-2 image file, e.g., Subset_S2A_MSIL1C_20230108T074311_N0509_R092_T37MBV_20230108T092516_resampled.csv. Parameter retrieval and naming can be automated in R using C2RCC_Ancillary_data_table_creation.R.
| Parameter | Value |
|----- …