LimnoSat-US: A Remote Sensing Dataset for U.S. Lakes from 1984-2020

LimnoSat-US is an analysis-ready remote sensing database that includes reflectance values spanning 36 years for 56,792 lakes across > 328,000 Landsat scenes. The database comes pre-processed with cross-sensor standardization and the effects of clouds, cloud shadows, snow, ice, and macrophytes removed. In total, it contains over 22 million individual lake observations with an average of 393 +/- 233 (mean +/- standard deviation) observations per lake over the 36 year period.  The data and code contained within this repository are as follows:

HydroLakes_DP.shp:  A shapefile containing the deepest points for all U.S. lakes within HydroLakes.  For more information on the deepest point see https://doi.org/10.5281/zenodo.4136754  and Shen et al (2015).

LakeExport.py: Python code to extract reflectance values for U.S. lakes from Google Earth Engine.

GEE_pull_functions.py: Functions called within LakeExport.py

01_LakeExtractor.Rmd: An R Markdown file that takes the raw data from LakeExport.py and processes it for the final database.

SceneMetadata.csv: A file containing additional information such as scene cloud cover and sun angle for all Landsat scenes within the database.  Can be joined to the final database using LandsatID.

srCorrected_us_hydrolakes_dp_20200628: The final LimnoSat-US database containing all cloud free observations of U.S. lakes from 1984-2020.  Missing values for bands not shared between sensors (Aerosol and TIR2) are denoted by -99.  dWL is the dominant wavelength calculated following Wang et al. (2015).  pCount_dswe1 represents the number of high confidence water pixels within 120 meters of the deepest point.  pCount_dswe3 represents the number of vegetated water pixels within 120 meters and can be used as a flag for potential reflectance noise.  All reflectance values represent the median value of high confidence water pixels within 120 meters. The final database is provided in both as a .csv and .feather formats.  It can be linked to SceneMetadata.cvs using LandsatID.  All reflectance values are derived from USGS T1-SR Landsat scenes.