Allow user customization of the spacetime chunk for which new predictions are made

[Kodiologist]

Maybe there's a slick way to do this with dynamic branching. There would be a target that collects spacetime chunks, however they're defined, and then the prediction code would use one dynamic branch per chunk. I'm not sure under what circumstances targets will delete the cache file for a dynamic branch that no longer exists.

[allanjust]

examples of desired spacetime chunks might be all grid cells containing AQS monitoring sites (so -- generated from a set of points) or grid cells covering a single (multi) polygon (e.g., greater NYC)

[allanjust]

sounds like tar_prune can remove targets that are no longer part of the pipeline but I don't think it is run by default

[Kodiologist]

Since prediction is fast (taking Coco about 17 minutes for the whole contiguous US for Aqua in January 2019), it may be effective in the short term to implement this by making predictions for the whole study area and then returning only the ones requested.

[Kodiologist]

This is mostly implemented now. What I ended up doing is writing a fairly generic function for making new predictions that can be used to define a target or called outside of a target. What's missing is some kind of user interface, presumably through the configuration file. Currently, there is none.