Discussion on the flux initialization in multiband data

[mkuemmel]

It turns out that the flux initialization has quite an influence on the photometric result when running SE++ on mulit-band data. As an example I am using a simulated dataset for Euclid which consists of VIS and H band data. The zeropoints are quite different:
MAG_ZEROPOINT = {'VIS':24.48941, 'H': 29.8512}

When initializing the flux with:

for band, group in mesgroup:
flux_total[i] = get_flux_parameter()

The comparison with the true universe in the H band is:

get_flux_parameter() allows a range of 10e-3, 10+3. Due to the high zeropoint offset the upper range is reached, which leads to the totally off measurements in the upper right corner and in the range [20-22] mag especially for red sources.

Enhancing the range with:

for band, group in mesgroup:
flux_total[i] = FreeParameter(1.0, Range((1.0e-07,1.0e+05), RangeType.EXPONENTIAL))

leads to:

Taking into account the zeropoint offset with:

BAND_SCALE = {'VIS':1.0, 'H': 140}
for band, group in mesgroup:
flux_total[i] = get_flux_parameter(scale=BAND_SCALE[band])

results in:

Enhancing the allowed range around the scaled flux with:

BAND_SCALE = {'VIS':1.0, 'H': 140}
for band, group in mesgroup:
flux_total[i] = FreeParameter(lambda o: o.get_iso_flux() * BAND_SCALE[band], Range(lambda v,o: (v * 1E-9, v * 1E9), RangeType.EXPONENTIAL))

results in:

[marcschefer]

Good tests, thanks.

I'm not really that surprised, I remember from early tests that results were very sensitive to initial values and how the parameters were set up. I didn't expect the last one to be so bad when using 1e-9 to 1e9 range, though.

[mkuemmel]

True, also FreeParameter(1.0, Range((1.0e-07,1.0e+05), RangeType.EXPONENTIAL)) show large offsets at bright mags that I do not really understand.