Validity of the Landsat surface reflectance archive for aquatic science: Implications for cloud-based analysis

This repository contains the dataset and codes used to generate and evaluate accuracy of Landsat Collection 2 Level 2 surface reflectance (ρ_s) products for aquatic applications. More information could be obtained in the published study: Maciel et al. 2023

For reproductibility purposes, R-Code is provided to calculate Landsat Level-2 products accuracy based on in-situ (ρ_s) simulated to Landsat-5/TM, Landsat-7/ETM+, Landsat-8/OLI and Landsat-9/OLI-2.

The simulation (i.e., application of the Relative Spectral Response) for each sensor was performed using the R package bandSimulation available here.

The compilation of this dataset is based on three different data sources.

1. GLObal Reflectance community dataset for Imaging and optical sensing of Aquatic environments (GLORIA) (Lehmann et al., 2023)
2. Augmented GLORIA (Pahlevan et al. 2022)
3. Dataset from Instrumentation Laboratory for Aquatic Ecosystems (LabISA-INPE, Brazil) (Maciel et al., 2021)

Figures and statistical analysis

For reproduction of this study, users should first clone de GitHub repository and execute the R scripts. A .CSV file (Matchups.csv) is provided with the in-situ simulated reflectance and Landsat extracted data. A detailed description of Metadata is available in the file Metadata-LO-Letters-data.docx.

The script "01_Generate_correlations_heatmap.R" run the statistical analysis and figures plot for the Figure 2 of the main manuscript and for the Figure S2 of the Supplementary Material. It also calculates all the statistical metrics.

Microsoft Planetary Computer data extraction

The extraction of ρ_s was performed using the Microsoft Planetary Computer (MPC). The MPC hosts a STAC catalog with a copy of the USGS Collection 2 Level 2 products. The rSTAC package was used to provide the connection between STAC and R and for the retrieval of the data. Extraction was performed using the MPC platform, but the code used is available in the Scripts folder Scripts/Extract_landsat_5.R.

We strongly reccomend the readers to use the MPC platform to extract the data. The example is provided for Landsat-5/TM - but users could change the sensor in the script easilly.

Time-series plots

The time-series was obtained using the Google Earth Engine platform and the script used to generate the time-series is available here. After the generation of the time-series for the specific location, user should save the graph with the reflectance values and also the graph with the pixel count number. It allows to filter for dates with a small number of pixels in our area of interest. The time-series analysis R code is available to plot the time-series.

More information about the time-series is provided in the Supplementary Material of Maciel et al. (2023) publication.