CoastSat Versions Save

Contributions from @2320sharon and @DanieTheron to improve the download updates and the cloud mask.

Main changes:

• changed Sentinel-2 to collection to S2_HARMONIZED (otherwise there is a shift after 2022/01/01)
• avoid downloading imagery that is already in folder
• function to create an animation with the mapped shorelines

Improved Sentinel-2 cloud masking using the s2cloudless collection on Google Earth Engine https://developers.google.com/earth-engine/datasets/catalog/COPERNICUS_S2_CLOUD_PROBABILITY. This should make Sentinel-2 imagery much more usable for shoreline mapping.

Some minor updates including:

• new way to convert coordinate systems (with pyproj)
• added IMAGE_QUALITY flag in metadata
• @2320sharon added a commit to save images as .jpg files instead of matplotlib plots

Update to the workflow to download and preprocess Sentinel-2 imagery. Makes sure the bands that come at different resolutions (10m, 20m and 60m) are reprojected onto the same pixel grid with gdal_warp.

CoastSat 2.0 MAJOR update:

• New download and preprocess functions for Landsat images: this improvement provides better alignment between panchromatic and multispectral bands and fixes an occasional issue of a shift between L5 and L7/L8 in the resulting shoreline time-series. Note that the Sentinel-2 download and preprocessing remains exactly the same, but should be updated too eventually (I did not have time). Changes to SDS_download.py and SDS_preprocess.py were made.
• New quality-control functions: to help with a fully automated shoreline extraction, including:
  • improved automated shoreline mapping when the manual quality-control parameters (check_detection and adjust_detection) are turned off. Changes to SDS_shoreline.py were made.
  • enhanced functionalities to compute the intersections (dealing with loops, multiple intersections etc), new function SDS_transects.compute_intersections_QC().
• New post-processing functions: includes a despiking algorithm SDS_transects.reject_outliers() for the shoreline time-series and functions to compute seasonal and monthly averages, SDS_transects.seasonal_averages() and SDS_transects.monthly_averages().
• Easier way to install the environment suggested by @YonniYe in #278.

Updates from previous release:

• Compatibility with Landsat 9 imagery
• Users can choose whether to use Landsat Collection 1 or the newly released Collection 2 (note that Google Earth Engine are still uploading Collection 2 to their servers). Images after 2022/01/01 are only available in Collection 2.
• More bug fixes from @2320sharon on merging Sentinel-2 images that overlap in the region of interest
• New installation instructions (without the environment.yml file)

minor bugfix as Inf values were crashing the classifier

Bugfix when merging overlapping S2 images in SDS_download.py

New functionality that allows users to plot the histogram of MNDWI pixel intensities and adjust the threshold that is used for shoreline detection as show below:

this functionality can be activated by setting adjust_detection to True (default value set to False).