Deblending metric

[jmeyers314]

The distribution of the best seeing will affect our ability to deblend objects. To start to probe this, @pmelchior and I looked at the spatial distribution of the best seeing. For example, here's a figure showing the 10th percentile best seeing in the i-band over the wide-fast-deep footprint.

It appears that our ability to identify blends may vary considerably over the survey footprint.

[drphilmarshall]

Very interesting. Great stuff! @pmelchior https://github.com/pmelchior would
you be interesting in adding a few words as well as this figure to the WL
science case in the white paper? I think it belongs there, since in the end
both the rotational dithering and the best seeing stacks will affect the WL
figure of merit, whose proxy is likely to be some sort of shear systematic
statistic I guess. If you are up for this, please do start a pull request
and let me know when you think you might be ready for it to be merged.
This could be one for version 2 of the paper, for example.

[pmelchior]

@drphilmarshall: I can add an explanation to why the seeing variation matters. It's not just the effective depth, which changes the statistics budget, but the shear calibrations depend strongly on seeing. The more homogenous the observed seeing is, the easier it will be to apply accurate calibrations (from deeper reference fields or simulations) to the observations.

What's the timescale for version 2?