elliesch / landsat_sst_algorithm

Adding Bayesian regression approach

Landsat_SST_algorithm

NLSSTpipeline produces atmospheric corrected sea surface temperatures (SST) from Landsat thermal infrared imagery. The atmospheric correction is produced using an NLSST algorithm. We calculate the NLSST algorithm coefficients using ECMWF ERA5 atmospheric column water vapors at 37 atmospheric levels input into the MODTRAN radiative transfer model to estimate the relationship between Top of Atmosphere (TOA) thermal brightness temperatures and sea surface temperatures. We use coincident MODIS water vapor measurements to correct each Landsat scene using the NLSST algorithm. A cross-calibration is included for the SSTs using MODIS. To create the cross-calibration we created uncalibrated Landsat SSTs near the Cosgrove and Dotson ice shelves and used them to matchup with MODIS and produce the calbration in the LandsatCalibration notebook.

LandsatCalibration calculates the bias and trend for calibration between MODIS and Landsat SSTs. It creates matchups between MODIS and Landsat for region near Cosgrove and Dotson ice shelves, West Antarctica and uses an Orthoganal Distance Regresson (ODR) to produce a calibration equation for Landsat, which is then used to calibrate all SST pixels used for analyses. The current calibration (20230627) is significant to the 90% confidence level (p=0.08).

ERADownload is the code required to download all ERA-5 atmospheric profiles and sea surface temperature reanalysis data for the training of the MODTRAN atmospheric correction model. MODTRAN4_prep prepares the ERA-5 data for intake for the MODTRAN software run by the MODTRAN C script.

Data includes matchup data (MODISvLandsat) that has already been created in LandsatCalibration. Uncalibrated SSTs produced by NLSSTpipeline for input into LandsatCalibration were too big to include so the user must run NLSSTpipeline first to be able to run LandsatCalibration. Data required to generate the atmospheric correction (modtran_atmprofiles) and the output generated by the notebook (TCWV) are found in Data/AtmCorrection.

This algorithm was constructed to be used in the virtual cloud and was built using the CryoCloud cryosphere community JupyterHub (https://doi.org/10.5281/zenodo.7576601).

This work is covered under an MIT license.

Project contact info:
Tasha Snow, PhD
Colorado School of Mines
[email protected]