Video Map-generating and visualizing video map tiles from EO data

Gennadii Donchyts, Fedor Baart, Giorgos Dimopoulos, P.J.M. van Oosterom, Christine Rogers, Cindy van de Vries, Martijn Meijers

Research output: Contribution to conferenceAbstractScientific

Abstract

The rapid increase in volume, resolution, and frequency of EO datasets poses the need for new scalable, cloud-based, tools for interactive exploration, visualization, and analysis. The size of data increases in several dimensions concurrently: temporal resolution go from the order of several days to several hours (CubeSats, reanalysis), spatial resolution increases from the order of hundreds of meters to order of meters (newer satellites, downscaled indicators), the spatial coverage increases from selected areas of interest to global and the timespan increases from years to decades (satellites, reanalysis).

Many datasets (optical satellite images, reanalysis of climate and coastal and ocean currents) provide a new look at the world, in details never seen before. We want to make these datasets visible, show the results, preferably without reducing dimensions and giving up all the newly achieved efforts of gathering the most detailed, most rich datasets ever created. This does leave us with a technical challenge. How can we create interactive animations of PB of geospatial data?

Here we show a new toolset and tiling scheme that helps to scale up interactive visualizations of the largest global datasets in a form of georeferenced video maps.

Interactive visualisation of sea level and currents (based on the datasets from MEASUReS, HYCOM, GTSM) provides an example of a visualisation covering several decades of sea-level rise measurements and model reanalysis. We present a way to efficiently encode data in video and reuse it in the client-side, to visualize or to perform GPU-enabled analysis. An interactive visualization of Earth’s changes captured by Sentinel and Landsat satellites and turned into video map tiles provides an example of how to disseminate large volume datasets while still being able to run state of the art algorithms on the client-side to detect land-use changes, using libraries such as TensorFlow.js or directly using computational shaders. The tools build on Google Earth Engine as a source of multi-temporal EO and reanalysis data. We show how the tools are integrated into mapping libraries (Mapbox GL, Leaflet). The storage format is based on a standardized layout of tiles (TMS). The client-side and client synchronization of video tiles are based on the emerging efforts of the w3c webtiming group.
Original languageEnglish
Number of pages1
Publication statusPublished - 2019
EventAGU Fall Meeting 2019 - San Francisco, United States
Duration: 9 Dec 201913 Dec 2019
https://www.agu.org/fall-meeting

Conference

ConferenceAGU Fall Meeting 2019
Country/TerritoryUnited States
CitySan Francisco
Period9/12/1913/12/19
Internet address

Bibliographical note

Session on Data and Rising Technologies - SWIRL

Fingerprint

Dive into the research topics of 'Video Map-generating and visualizing video map tiles from EO data'. Together they form a unique fingerprint.

Cite this