Quantification of tidal effects on Landsat imagery

This project aims on calculating the impact of tidal range on shoreline detection results when working with Landsat imagery at 30 meters spatial resolution. A fully automatic approach using the Google Earth Engine Python API has been chosen to guarantee transferability to other study areas. For exploring coastline dynamics it is import to know about the effects of tidal variantions on the detection results and to eventually correct for them, especially when working with time series. The presented tool only requires the area of interest (which can be selected interactively), an hourly sea level data csv, an osm reference shoreline and the sea level station coordinates and user paths as input. Sea level data can be obtained from the UHSLC Legacy Data Portal.

The area around Qui Nhong in Central Vietnam has been chosen as study area. Central Vietnam belongs to a mesotidal environment.

Method

1. Identification of low and high tide dates at Landsat acquisition time within the selected AOI

   • Cleaning of sea level data/ extraction of valid years
   • Calculation of Landsat overpass time for each year
   • Filtering of sea level data by Landsat acquisition hour
   • Identification of low and high tide peaks (+/- 1 day)
   • creation of low and high tide date lists for each available year
   • Output: low and high tide date lists

   

2. Selection of nearly cloud-free images at low and high tides

   • Filtering Landsat SR archive by 3-day date ranges
   • Calculation of cloud cover within AOI
   • Filtering of selected images by cloud cover (< 20%)
   • Cloud-masking on selected images (using sr_cloud_qa band, for LS8 pixel_qa band)
   • Creation of daily mosaics
   • Removal of mosaics covering < 2/3 of the AOI
   • Creation of low and high tide image pairs (each high tide image receives the closest available low tide image or vice versa, depending on the length of the list)
   • Output: list of image pairs

3. MNDWI calculation and binarization (on GEE)

   • Calculation of the Modified Difference Normalized Water Index (MNDWI): (Green-SWIR1/Green+SWIR2)
   • Binarization using Otsu's threshold
   • Download rasters from Google Drive to local machine
   • Output: shorelines geojson for each image pair

4. Shoreline extraction (on local machine)

   • Contour extraction using the skimage.measure.find_contours Python function
   • Clipping of contours to buffered OSM shoreline
   • Output: shorelines geojson for each image pair

   

5. Calculation of shoreline displacement between low and high tide

   • Creation of transects perpendicular to OSM shoreline (100 m spacing) with landwards transect origin
   • Calculation of shoreline/ transect intersections
   • Measuring the distance of each intersection to the transect origin (the difference between the distance of each high tide intersection and the distance each low tide intersection equals the landwards shoreline displacement from high to low tide)
   • Output: transects geojson with shoreline displacement information

   

6. Statistics

   • Calculation of median shoreline displacement and standard deviation for each image pair
   • Calculation of overall median shoreline displacement and standard deviation for all image pairs
   • Output: bar plot with error bars