Discrepancy between dust and synchrotron intensity models with observations at 100µm and 408MHz

[erussier]

I checked models d9, d10 and d12 at 100µm (one of IRAS frequencies), these are smoothed to the IRIS resolution (4.3arcmin):

This is the IRIS map:

Also, I checked models s4, s5 and s7 at 408MHz (Haslam frequency), these are smoothed to the Haslam map resolution (56arcmin):

And the Haslam map at 408MHz:

Here is the link to the notebook I used to generate these figures: https://gist.github.com/erussier/cf19c05437bf05c1397e148a3b988253

[giuspugl]

thanks a lot for looking into this. below my two cents on this:

1. small scales in the amplitudes do not show strong departure from the templates shown, e.g. d9 and s4
2. Clear departures are due to small scales coadded into the spectral parameters, see d10, s5, s7.
3. Differences between s5 and s7 are solely due to the curvature of the synch spectral index.

[delabrou]

It is clear that s5 and s9 are completely off at 408 MHz, and the reason for this is that the 23GHz template was scaled for 408MHz with a prescription that is incompatible with what is then used to scale the 23GHz map back to 408 MHz. It is a limitation of the models, which should be properly documented (for now) and ultimately fixed (ASAP). Dust is less dramatic, but still unsatisfactory, both in terms of the (too Gaussian, too stationary) small scales, and clear incompatibility between the 100 micron simulated maps and the IRAS observations. For future models we probably should constrain the scaling laws so they do not generate very inconsistent foreground maps anywhere in the potential frequency range of use (400MHz to 3000 GHz at least, I would advocate). For now we should give a "range of acceptability" (hopefully 10-1000GHz, but 20-500 GHz instead if we find discrepancies too large outside of that range).