This repo is currently frozen so that the approach can be evaluated, but there were a few things to capture about the Viewing* and Hit* notebooks in the event this code is revisited:
[ ] Time selectors are a bit awkward; should have an integrated time/date selector.
[ ] Add a time-scrubber widget, letting the user watch vessels and satellites move while scrubbing. For this to be practical, will need sample positions on a regular time base (rasterized time) so that selecting a time is an indexing operation rather than a search.
[ ] There's a lot of work that could be done using the timing of available pings to predict
[ ] It would be nice to have a visualization that shows the possible locations that a vessel could have been, given a finite set of pings and a maximum velocity (either the actual maximum for that vessel type, or a default of 20 knots). Each ping pins the location at a specific time, then the longer the time to the next ping, the larger the uncertainty in location, becoming lower again as the time approaches the next ping. This envelope would be interesting to visualize, to show the quality of data we have about a particular vessel's location.
[ ] A lot more could be done with drilldown and linked selection, to provide metadata about a selected vessel or satellite.
[ ] Individual boat markers (e.g. thin triangles) should be pointing in the direction of the vessel's reported heading.
[ ] Plot TLE on Robinson or other less-distorted projection?
[ ] Could datashade all precomputed tracks; not sure what that will show, particularly given the distortions caused by projections.
[ ] Lots of work could be done to optimize the user experience and add progress reporting/spinners, to make users appreciate that apps are working even when a step is slow.
This repo is currently frozen so that the approach can be evaluated, but there were a few things to capture about the Viewing* and Hit* notebooks in the event this code is revisited: