Sentinel-2 Water Quality Models and Google Colab Workflow for the Yucatán Peninsula
Description
This dataset contains the calibration and validation datasets, together with the Google Colab workflow, used to develop and apply Sentinel-2-based water quality models for inland aquatic systems. The workflow integrates the Modified Atmospheric Correction for Inland Waters (MAIN) and multiple linear regression (MAIN-MLR) approaches to retrieve five key water quality parameters: chlorophyll-a (Chl-a), total suspended solids (TSS), turbidity (NTU), colored dissolved organic matter (CDOM), and Secchi disk depth (SDD). The dataset includes five Excel files corresponding to each water quality parameter. Each workbook contains two separate datasets: (1) calibration data used for model development and regression analysis, and (2) independent validation data used to evaluate model performance. The accompanying Google Colab notebook provides the complete remote sensing workflow, including Sentinel-2 image preprocessing, cloud and shadow masking, atmospheric correction, water masking, calculation of remote sensing reflectance (Rrs), and generation of water quality products. These data and scripts are provided to facilitate reproducibility, evaluation, and further application of Sentinel-2 remote sensing approaches for monitoring inland waters.
Files
Steps to reproduce
The uploaded Excel workbooks contain the calibration and validation datasets used to develop and evaluate the Sentinel-2-based water quality retrieval models. Separate workbooks are provided for Total Suspended Solids (TSS), Chlorophyll-a (Chl-a), Turbidity (NTU), Colored Dissolved Organic Matter (CDOM), and Secchi Disk Depth (SDD). The calibration datasets include in situ measurements, Sentinel-2 spectral bands (B1-B8A), natural log-transformed (ln) variables, and spectral band ratios used as predictor variables in the multiple linear regression (MLR) models. The validation datasets contain independent field observations, model predictions, and error calculations used to assess model performance. The datasets were generated through field sampling and Sentinel-2 remote sensing conducted between February 2024 and January 2026 across 37 continental aquatic systems in the southeastern Yucatán Peninsula, Mexico. Field observations were paired with cloud-free Harmonized Sentinel-2 Level-1C imagery acquired within ±7 days. Images were identified in Google Earth Engine (Gorelick et al., 2017) and processed using a Google Colaboratory Python workflow. The workflow applies cloud and cloud-shadow masking, water masking, and the Modified Atmospheric Correction for Inland Waters (MAIN) to generate remote sensing reflectance (Rrs) following the methodology of Page et al. (2019). Mean reflectance values for Sentinel-2 bands B1-B8A were extracted using 50 × 50 m buffers (30 × 30 m for smaller water bodies) centered on each sampling location. The extracted reflectance values were combined with field measurements of SDD, Chl-a, TSS, NTU, and CDOM. Predictor variables included Sentinel-2 spectral bands and spectral band ratios. Water quality variables were natural log-transformed, and multiple linear regression (MLR) models were developed in JMP Pro using a 70% calibration and 30% validation approach. The accompanying Google Colaboratory Python script reproduces the satellite image processing workflow and applies the calibrated retrieval equations to generate spatial estimates of each water quality parameter. Reference: Page, B.P., Olmanson, L.G., Mishra, D.R., 2019. A harmonized image processing workflow using Sentinel-2/MSI and Landsat-8/OLI for mapping water clarity in optically variable lake systems. Remote Sens. Environ. 231, 111284. https://doi.org/10.1016/j.rse.2019.111284
Institutions
- Instituto Tecnológico de ChetumalQuintana Roo, Chetumal