Lower Bua River In-situ water quality, GEE Sentinel 2 reflectance values and Fish CPUE data
Description
The data has 4 categories which are In-situ water quality data obtained during the monthly monitoring surveys, fish catch per unit effort data obtained each month on landing sites, point type shapefile for the sampling stations and 2 reflectance values datasets obtained from sentinel 2 level 2A data which is already atmospherically corrected by Sen2Cor. The reflectance values were in two categories- one timeseries data from the year 2018 to 2022 which produced 557 observations processed at less than 3% cloudy cover percentage; and the other training data reflectance values from 5 stations processed at less than 1 to 3 % cloud cover percentage depending on the image availability with preference given to images which has the least cloud cover percentage and were colocated with the In-situ water quality surveys at +/- one day of sampling. The procedure for obtaining the reflectance values and tiff images for the study area was done in google earth engine after importing the shapefile as an assert with the attached link , the same with obtaining the time series reflectance values. Timeseries reflectance data, in-situ data and fish CPUE data were then further processed in python notebook using machine learning methods. The attached .py file contains the detailed python procedure
Files
Steps to reproduce
Steps for reproducing the datasets 1Extracting the reflectance values from sentinel 2 level 2A data from 5 sampling stations which were subsequently used for training various Machine Learning methods, selected the best performing method which was later used in predicting water quality values after factoring in the timeseries reflectance values in the below step. But this step procedure can be obtained can be checked using this link: https://code.earthengine.google.com/7a5c8b9c406f4527ac03b0af0cd11b34 2. Extracting the timeseries reflectance values used to derive timeseries water quality data (used the procedure provided in the following link in Google Earth Engine Sentinel-2 surface reflectance data (COPERNICUS/S2_SR_HARMONIZED) were filtered from 2018 to 2022 with less than 3% cloud cover, cloud-masked using the QA60 band, and spectral reflectance values were extracted at five predefined sampling points along the Bua River at 10-metre resolution before being exported as a CSV time series to Google Drive and can be found using the following link : https://code.earthengine.google.com/3bec8cee504811a0ee790352769d77ef) 3. Machine learning methods performed use python notebook: Sentinel-2 surface reflectance time series extracted from Google Earth Engine (2018–2022, less than 3% cloud cover) across five sampling stations along the Bua River were used to train and validate multiple machine learning models — including Random Forest, Gradient Boosting, XGBoost, and Ridge regression — via Leave-One-Out Cross-Validation to predict seven water quality parameters (Secchi depth, chlorophyll-a, temperature, pH, electrical conductivity, salinity, and turbidity), after which the best-performing models were applied to the full reflectance time series to generate monthly water quality predictions that were validated against in situ observations and used in a time-lag correlation analysis with fish catch per unit effort (CPUE) data for nine species, with all outputs exported as CSV files and diagnostic plots saved to a designated project directory.