Improving ZTF Calibration

Zeropoint uncertainties variations by field.
Data Flags for ZTF DR1.

Flags for ZTF DR1 photometry

Field Selection

Goal:
Select a small sample of ZTF data that is representative of the overall dataset. Use this to determine the images and photometry that should be flagged within ZTF DR1.

Data requested for five separate sets of fields to sample crowding, airmass, and distance for equatorial pole:
1, 10 fields equally distributed along l with |b| < 10 deg,
2, 10 fields along l with 10 < |b| < 20,
3, 20 fields along l with 20 > |b|,
4, 10 fields along RA with dec < -15 (not included in those above),
5, 10 fields along RA with dec > 50 (not included in those above).
Received data for only 42 test fields.


The distribution of fields used for ZTF DR1 data flagging.

Analysis:
Issues with the data are most clearly seen with the values of ZP and ZP_rms and the number of calibrator star matches. These parameter selections are mainly based on the quality discriminators presented in the ZSDS (section 13.4).


Variation in ZP_rms over time for g (blue points) and r-bands (red). The two lines give the ZP_rms flag thresholds for each photometry band.


Variation in frame ZP over time for g (blue points) and r-bands (red). The two upper green lines give two thresholds used to discriminate possible bad and likely bad data (the final ZP thresholds vary by quadrant).

Observations containing likely bad data are clearly seen to have uncertain values of ZP from the calibration. Likewise, the bad observations have smaller ZP values that are likely caused by a combination of clouds and lunar illumination.


The distribution of frame ZP and ZP_rms values. Left: ZTF r-band data. Right: ZTF g-band data. The flagged data is given by blue points and the presumed good data by the red points.

Well calibrated (and presumably good) data is located in a distinct region of ZP vs ZP_rms space. However, some of this data may still be usable, so multiple cuts were ZP devised.


The distribution of ZP with airmass. Left: The ZTF r-band data. Right: The ZTF g-band data. The solid lines show the ZP cuts used, and the trend. The dashed lines show the airmass trend. The cyan lines show the values from Ofek for PTF.

As expected there is a clear trend in ZP with airmass. For r-band this is ~0.15 mags/airmass and for g this is ~0.2 mags/airmass. These values vary slightly from the values found for PTF of g=0.23, r=0.112 by Ofek. Hayes and Lathem (1975) gave values for Palomar of g at 4800A=0.206, r at 6250A=0.115, z at 7000A=0.07. In comparison, for PS1, Tonry et al. (2012) gives g=0.22, r=0.13, i=0.09, z=0.05. Further work is need if these values are to be used for something other than quality cuts.

Results:
For ZTF r-band, data is flagged with any of:

zp > 26.65 - 0.15*Airmass
zp_rms > 0.05
numps1calmatches < 120
zp < zp1_threshold[quadrant] - 0.15*Airmass
or
zp < zp2_threshold[quadrant] - 0.15*Airmass

For g-band:
zp > 26.7 - 0.2*Airmass
zprms > 0.06
numps1calmatches < 80
zp < zp1_thres[quadrant] - 0.2*Airmass
or
zp < zp2_thres[quadrant](col 3) - 0.2*Airmass


ZP uncertainties in varying field selections

During the analysis of ZTF fields for DR1 it was clear that there were significant differences in the distributions of ZP uncertainties between fields. These variations roughly follow the density of sources in those fields.


Zero point rms vs the number of PS1 star calibrators in a quadrant for DR1 fields.
Selections are: black (|b| < 10 deg), 2= Red (10 < |b| < 20), 3= Green (20 > |b|), Blue (field 539, l=44.3, b=3.6 deg).


G and r-band zero points vs number of the number of ZTF catalog sources within a quadrant. The colour selections are as above

Conclusions:
1) The RMS for ZP values in high galactic latitude fields are generally large due to the lack of calibrator stars within a quadrant. As the number of stars increases at slightly lower Galactic latitudes the additional calibrator stars reduce the scatter.
2) At very low Galactic latitudes, crowding effects mean larger numbers of stars are required to reduce the ZP scatter. When there are more than ~90,000 stars in a quadrant crowding/blending simply drives the ZP scatter up.
3)The g-band data appears much more affected by crowding than the r-band data since the RMS stays high even with large numbers of calibrator stars. However, there may be another cause.