This is a work-in-progress adaptation of Digital Earth Australia (DEA) and Digital Earth Africa Coastlines products for the Pacific nations. Please refer to the following for a description of the algorithms on which we based this work:
Bishop-Taylor, R., Nanson, R., Sagar, S., Lymburner, L. (2021). Mapping Australia's dynamic coastline at mean sea level using three decades of Landsat imagery. Remote Sensing of Environment, 267, 112734. Available: https://doi.org/10.1016/j.rse.2021.112734
Bishop-Taylor, R., Sagar, S., Lymburner, L., Alam, I., Sixsmith, J. (2019). Sub-pixel waterline extraction: characterising accuracy and sensitivity to indices and spectra. Remote Sensing, 11 (24):2984. Available: https://doi.org/10.3390/rs11242984
We have adapted the algorithms where necessary to account for differences in the study area and the available input datasets, as well as for performance. A full writeup is in process.
Coastline creation consists of five primary steps:
- Calculating tides for each location and available landsat image. The code for
this is in
dep_coastlines/calculate_tides.py
. (See the notes in that file on tide model data location.) - Filtering landsat images based on the tide values and creating annual
mosaics. This is done in
dep_coastlines/tide_corrected_mosaics.py
. - Post-processing the annual mosaics to remove clouds, and mask the analysis
zone to the coastal zone. In
dep_coastlines/clean_rasters.py
. Note that this step requires a model file. - Delineating the coastline using the cleaned annual mosaic (a second step in the script in #3).
- Calculating rates of change over time and merging data into final products
(see
dep_coastlines/continental.py
).
Processing is done over an grid of arbitrary size, as defined in
dep_coastlines/grid.py
using Landsat Collection-2 Level-2 (tiers 1 and 2) data
loaded from the Microsoft "Planetary Computer". The tides dataset is TPXO9.
Steps 2, 3, and 4 were accomplished using a deployment of
argo. The workflows for these are saved in
.argo/
. Steps 1 and 5 were completed locally using a modest
workstation-level computer. The dockerfile for step 1 is Dockerfile.tides
. The
remaining step use Dockerfile
.