7.11 IR Photometry Errors

The estimated errors for the inverse sensitivity (zeropoints) of the WFC3 IR channel computed from observations of the white dwarf standards plus a G-type star are ~1% (statistical) and 1-2% absolute (systematic, uncertainties in the models used as absolute standards, WFC3 ISR 2020-10). The signal to noise ratio (SNR), per observation, for the standard stars is such that the formal Poisson error is ~0.2% for the wide band filters.  However, Poisson noise is not the limiting factor at SNR > 100 due to poorer than expected repeatability (WFC3 ISR 2020-10). Data taken using different observation configurations (apertures, sample sequences and numbers, staring versus scan mode) can show differences in photometry. For example, 1 sigma relative photometric repeatability values have been measured at +/- 1.5% (WFC3 ISR 2019-07, staring mode) and 0.65%  (WFC3 ISR 2021-05, scanning mode). 

A new calibration for the WFC3/IR inverse sensitivities (zeropoints) was released in 2020 (WFC3 ISR 2020-10). The zeropoint changes (0.5-2%, depending on filter) were due to updated HST CALSPEC models and an increase in the Vega reference flux (2020 AJ 160,21) as well as new IR flatfields (Section 7.8). At the time of writing, there is no time-dependence in the IR inverse sensitivities measurements. For optimum results as well as consistency with WFC3/UVIS and ACS, the new 2020 WFC3/IR zeropoints should be used; data retrieved from MAST after Oct 2020 will have the most current photometric calibration incorporated into the science image header keywords.

The IR photometric repeatability is occasionally worse than expected based on the nominal calibration/Poisson error.  Some scatter in measurements within a visit can be due to self-persistence of bright sources releasing charge into pixels used in subsequent exposures (WFC3 ISR 2019-07, WFC3 ISR 2016-11).  Self-persistence causes sources to appear brighter by around 2% on average when many saturating stars have been exposed onto the array (individual stars may be even brighter). Sufficient dithering (~10 pixel steps) can mitigate persistence effects, bringing average repeatability for many stars down to 0.5% (WFC3 ISR 2019-07). On the other hand, high-precision photometry from spatial scans suggests that larger dithers may cause observations of the same source to have a higher level of scatter (WFC3 ISR 2021-05). Some of the dependence of repeatability on the extent of dithers may be due to limitations in the flat field accuracy although visit-to-visit comparisons of scan photometry of the same sources at the same positions also show offsets (Figure 7.14), implying other causes as yet not understood. As evident from Figure 7.14, the overall photometric repeatability, a combined effect of the intra-visit scatter and the inter-visit offsets, typically surpasses the Poisson error in the high SNR regime.

Currently, evidence for time-dependent sensitivity in the IR channel is mixed. Monitoring of staring mode images of individual standard stars (2009-2021) do not show a discernible time-dependence although data are noisy with short-term 1-sigma repeatability of +/- 1.5% (WFC3 ISR 2019-07). However, studies using other data have shown low levels of sensitivity declines: 

G102, G141 standard star data 2009-2020:  0.17% /yr+/- 0.015% and 0.085%/yr +/- 0.014%, respectively  (AJ 157, 229)

G102, G141 standard star data 2009-2021:  0.162%/yr +/- 0.013% and 0.076%/yr +/- 0.012%, respectively (Bohlin, R., 2021 priv.comm.)

F160W stellar cluster PSF-fitting photometry 2009-2020: ~0.2%/yr +/- 0.04%  although some evidence of drop in throughput in 2012 (WFC3 ISR 2020-05)

F140W scanned data 2015-2021: 0.024%/yr +/- 0.008% (WFC3 ISR 2021-05)

F098M scanned data 2020-2021:  0.044%/yr +/- 0.070% (WFC3 ISR 2021-05)

F110W images of M4 cluster data 2012-2017: 0.21%/yr (unpublished; decline drops to ~0.1%/yr if only first-in-visit images are used but uncertainties are large) 

Supplemental calibration observations are being obtained to further probe for sensitivity losses.

For users who need precise photometry, a complete error analysis would include Poisson noise of the science target as well as the additional sources of photometric error of flat fields, spatial repeatability (~1% peak to peak), temporal repeatability (2% peak to peak), readnoise, uncertainty in the encircled energy (approximately 0.5% at the infinite aperture), uncertainty in the gain (Section 7.2), and processing errors (sky subtraction, dark subtraction, correction for persistence).