Skip to content
gena edited this page Sep 9, 2019 · 30 revisions

Welcome to the eo-bathymetry wiki!

Links

Deliverables

For EMODnet:

  • Coastline as a vector for LAT, MLW, MSL, MHW, HAT(?)
  • Intertidal bathymetry

Roadmap

  • GD: export scene boundary (envelope polygon) + time stamp into GeoJSON (2017)

  • GH: extract model results for these time stamps and export back into GEE as (cell_geom, time, value) FC. Alternative is to extract water levels from http://volkov.oce.orst.edu/tides/tpxo8_atlas.html

  • GD: for every tile, detect a set of very accurate isolines based on cloud-free images

  • GD: use CDF-clip (-25%) combined with Otsu thresholding, maybe multi-class

  • GD, GH: assign water level values for isolines by overlapping them with model results - WL

  • GD, GH: interpolate/extrapolate water depth isolines into a surface - WL, use Krigging

  • GD, GH: perform linear regression between WL ~ P, where P - water occurrence

  • GD, GH: transform water occurrence into bathymetry using regression coefficients from the previous step

  • GD, GH: estimate LAT, MLW, MSL, MHW, HAT(?) from bathymetry image

  • for LAT, HAT - extrapolate bathymetry (Krigging)

  • extra - use optical-based depth estimates for extrapolation, only if visible bands correlate with bathymetry.

Supplementary:

  • extract time monitoring stations times series (in-situ)

these could be found on the web via EU dataportal, Dutch RWS or opendata: wetwetwet.nl:

  • given a polygon (maybe satellite image bounds or a processing box), find intersecting model cells and extract water levels for these cells at given times (extracted from image start_time)

(time, model_domain_id, cell_id, value)

  • convert all model grids to shapefiles, to import to EE

  • select boxes to perform processing

  • implement prototype algorithm to generate (partial) water masks from satellite date

  • horizontal bias removal

  • See also FAST project results:

Risks

  • Computation takes several hours for a single tile, ~20-30%
  • Too complex algorithm, too many components

Compute statistics for TOA values

  • Multi-class Otsu
  • Histogram smoothing (KDE)

Visualisation

Bathymetry from optical images

Effect of water on light tranmission:

Notes from the meeting (28-11-2017)

  • export multiple isolines (polygons), at least for some tests
  • northern parts - check multiple sensors for coastline only (S1, MODIS, etc., look only on summer)
  • add search algorithm for lowest water level, using 30 years of data
  • generate confidence images (on coarser scale)
  • where to stop inland?
  • compare with legal baselines
  • method to compare fuzzy coastlines