Closed jjkohler closed 3 years ago
I can fix by adding another special-case naming the layer in WebGLVideoTile.ts, but the better approach would be to fix the code to take advantage of the new timeline metadata to make a more principled adaptation between the timeline, its rate of playback, and its range compared to the TimeMachineLayer.
I'd much prefer the second approach, but it's somewhat involved and will require a lot of testing. Could we rationalize delaying this until after release? Are there important regressions in behavior for stories and layers that are in use?
@rsargent Here's an example of this issue, paired with a publicly available layer. During the first weeks of 2020 data, the landsat layer actively cycles through years worth of data, before freezing at some point.
@rsargent another example of mismatch between landsat and an overlaying data layer: https://staging.earthtime.org/explore#v=36.14414,-115.17283,10.201,latLng&t=1.55&ps=50&l=blsat,uf&bt=20000101&et=20151231&startDwell=0&endDwell=0
2000 in urban fragility is actually showing 1984 (and 2015 in UF is 1999 in landsat) -- since it is just showing what is at those same frames Golden behavior for context: https://golden.earthtime.org/explore#v=36.14414,-115.17283,10.201,latLng&t=1.55&ps=50&l=blsat,uf&bt=20000101&et=20151231&startDwell=0&endDwell=0
Another example of the data layer gratuitously advancing landsat, for years that landsat imagery is not actually present.
Landsat Advancing only through the 1930s and then stopping https://staging.earthtime.org/explore#v=34.27548,-109.4899,5.912,latLng&t=6.75&ps=50&l=blsat,gm_co2&bt=19000101&et=20111231&startDwell=0&endDwell=0
Should be fixed with e00ff3ef7783d07bd93f08f64187fffba7af24ee. @jjkohler, would you be willing to check?
Fixed regression with 8980e7de8f1ca9669a6e757ee036c968934bdc1c
@jjkohler please check again?
It appears that this fix now addresses all the known cases.
When land sat is turned on with data for a time frame beyond the last year of satellite imagery (e.g. 2020), it continues to advance, using time increments arbitrarily to advance through different years of data. This can be seen at the above link.