MDW hydrogen layer: known issues
This layer is a first pass, and it's rough in places. Bright nebulae hold up well. On the faint sky, most of what you see is error, and most of the visible seams come from how I joined the fields.
What this layer is
I took the star-removed images from Data Release 1 of the MDW Sky Survey, 2,110 fields of the northern sky in the red light of glowing hydrogen, and reprocessed them onto one HEALPix map. HEALPix divides the sky into cells of equal area.
It isn't an MDW product. The values changed in processing, and I anchored the sky level to the WHAM survey. MDW's team hasn't reviewed it. The tiles are for viewing, not for measurement. Please use MDW's own release for science.
The viewer shows it on a gentle stretch that keeps faint-sky error dim. That changes how it looks, not what's wrong with it.
How far to trust it
I compared this layer with the narrowband NSNS hydrogen survey on 404,673 shared cells, each about 14 arcminutes wide. The scatter between them stays near 2.5 to 3 rayleighs (R, a unit of this kind of light) at every brightness, so faint sky sits inside it and bright nebulae rise far above it.
| Brightness in NSNS | Share of cells | Correlation with NSNS | Scatter compared with signal |
|---|---|---|---|
| Under 3 R | 70% | 0.20 | About 2.7 times the signal |
| 3 to 10 R | 21% | 0.29 | About 0.9 times |
| 10 to 30 R | 6% | 0.35 | About 0.35 times |
| 30 to 100 R | 2% | 0.57 | About 0.17 times |
| Over 100 R, such as the North America Nebula | 0.6% | 0.988 | About 0.08 times |
This can't say which survey holds which part of the scatter, so some of it may belong to NSNS. The two scale at about 77 R per MJy/sr, close to the nominal 86.7. That leans on NSNS's calibration, so it isn't a direct measurement. The factor I used to anchor to WHAM is unmeasured.
Problems from my processing
| Issue | What happens |
|---|---|
| Field grid and dark bands | MDW's fields each keep their own sky level, and I joined them with one plane per field, anchored to WHAM at 1 degree. That doesn't close the steps. On field 0312 the step passes 0.34 MJy/sr on a level of 2.56. I haven't measured how many fields do this. |
| Weak sky fit | The fit used only fully measured cells, which left some fields with too little to work with. |
| Missing fields | Fields 1797, 2109, 2111, 4119 and 4120 are left out. Parts of their footprints show as holes: from about 13% of the footprint on 2109 and 2111 to about 67% on 1797. Neighboring fields cover the rest. |
| Satellite trails | My combine has no outlier rejection, so a trail in one field carries into its overlaps (seen on field 2112). |
| Tiny flux loss | A one-sided loss per field that grows toward the pole, up to 7.4 parts in 10 million. It's too small to see. |
| Limited checking | I checked results against the source images on three fields only: 0001, 0312 and 2112. Field 0001 was clean apart from a faint ripple I can't explain. |
Things I saw in the source files
These passed through into the layer. I may be misreading some of them, and I'd welcome a correction.
| Issue | What I saw |
|---|---|
| Star filling that diverged | In 952 fields, 586,345 cells touch filled-in star holes whose value lies outside that field's measured range. I had no guard against this in this build, and filling that went wrong but stayed in range isn't counted. |
| Fields with no fill mask | 36 fields in this layer are older products with no fill mask (37 in the release), so their filled areas count fully. Some cells reach −187 MJy/sr on a level near 1. Fields: 0288, 0309, 0316, 0317, 0364, 0365, 0409, 0534, 0538, 0539, 0584, 0617, 0640, 0644, 0646, 0648, 0649, 0733, 0750, 0756, 0961, 0962, 1123, 1135, 1160, 1223, 1252, 1286, 1394, 1515, 1591, 1603, 1717, 1799, 1866, 2066. |
| Exact-zero pixels | Some stacks hold pixels of exactly 0.0. I confirmed them on the three fields I checked. The scan hints at many more, but I haven't measured how many. |
| Field 0312's distortion | The distortion terms in its header look wrong. I checked only this field and haven't proven it. |
What I can't explain yet
- Some black blobs and bright squares across the map.
- M 82 looks wrong. I don't know why yet.
What happens next
I don't plan to rebuild this on Data Release 1. I'll rebuild on Data Release 2 once it's out, with the fixes I've made since.
Terms: I found a CC BY 4.0 license on the release's catalog record and MDW's note that all commercial rights in the data are reserved, but no license for the images themselves. This project is free and non-commercial, and I'm asking the team what they'd prefer. If you're on the MDW team and want this layer changed, credited differently or taken down, write to me at hello@skycruncher.io and I'll do it right away.
Back to the viewer, where the full credit is under Licenses and credits.