Follow Us
|
|
Colors/magnitudes: explanations and caveats
Introduction
MAST is now adding new ground-based survey colors and magnitudes for the Kepler Field of View: so far this includes UK-IRT, KIS (Part 1), and Sloan (SDSS/DR9). The KIC u-magnitudes have been removed from MAST's Target Search form (but not the KIC form) because of their poor quality.An executive summary of results for optical wavelengths is that the UBV survey should be preferred over the KIS survey. However, a broad secondary scatter distribution when some photometric target groups are include makes our recommendations for using SDSS g, r, and i magnitudes somehwhat nuanced (see UPDATES below). Nonetheless, the SDSS magnitudes magnitudes do bring in a number of new Kepler targets (thanks to the SDSS U mangnitudes, particularly high temperature targets) and furthermore clarify some problems with the KIS and KIC magnitudes, particularly for bright magnitudes; see the graphs below. Users should combine KIC and KIS colors and magnitudes only with care, as they are based on different zeropoint magnitude systems (AB and Vega/Johnson; see table below). Graphical comparisons of KIC vs. UBV and KIC vs. SDSS magnitudes and colors are given separately in this tarball.
The reference for the AB magnitude system is
Oke, J.B. and Gunn, J. 1983, ApJ, 266, 713O, and for the Vega system
Johnson & Morgan (ApJ, 117, 486, 1953J). The table below lists
the magnitude system for each survey in the Kepler Colors Table.
We give a link for the rules connecting survey acronyms with the filter magnitude names. You should also know that because of the zeropoint differences for the magnitude systems MAST will minimize the combinations of optical magnitude systems in the following way: colors formed from one filter from one KIC/KIS/UBV survey and a second filter from another. Therefore, because there will be no mixed colors, e.g., no (g_KIS - r_KIC). The only gr colors represented will be understood to be either gr (KIC) or gr_KIS.
Also, we won't comment on the Halpha equivalent widths in the KIS survey, which are parameterized by the rHα KIS index. A useful reference on the calibration of this index is Drew et al. 2005, MNRAS, 362, 753D.
Magnitude Surveys
SURVEY NAME | AREAL COVERAGE | MAG SYSTEM | SEARCH RADIUS (") |
|---|---|---|---|
KIC (griJ) |
~full |
AB |
1-3 (varied) |
GALEX (NUV, some FUV) |
~1/2 |
AB |
2.5 |
2MASS (J,H,K) |
(selected objs.) |
Vega |
1 |
UK_IRT (J-mags) |
full |
Vega |
1 |
INT/KIS (U,g,r,i, Hα) |
~1/2 | Vega |
1 |
UBV (Everett et al.) |
full |
Vega |
1.5 |
SDSS/DR9 (native Sloan) |
~1/10 |
AB |
1 |
In all cases filter wavelength centroids and transmissions should be shifted by -5Å from the tabular values because optical rays converge as they pass through the filter.
The Kepler Isaac Newton Telescope Survey (KIS) and
Everett (UBV) surveys are described in Greiss et al. 2012, AJ, 144, 24G and
Everett, Howell, & Kinemuchi 2012, PASP, 124, 316E.
We have revised the September release of this page with "UPDATES" marked
"Fall, 2012" to reflect our analysis of SDSS General Release (SDSS/DR9)
data for the Sloan u, g, r, i, and z magnitudes.
In late 2012 or early 2102 we intend to include part 2 of KIS, and PanStarrs
as these catalogs become available.
It is important to add that although our matching radius for SDSS
targets is only 1", it is clear that many matches, e.g. to the KIC,
should be made out to 2". Additions to proposers' target list of SDSS
objects that are non-KIC objects should be checked carefully against nearby
KIC targets for this reason. As with the GALEX matchings we decided
maintain a smaller "gold standard" radius to avoid false matches.
Filter transmissions
Filter transmission curves may be accessed via the links below for the
KIS Sloan and UBV filters to a table (first three
tabular entries):
INT/ING/KIS filters (UgriHα) link here.
Select the Halpha, U (RGO) or g, r, i (Sloan-Gunn) filters in various ascii
and image formats.
Similarly:
SDSS Project (ugri) link here.
UBV link here. (Select the first three rows in the table.)
Graphs (magnitude scattershots)
The graphs below plot only a subset (typically 5-10%, except for plots
with titles saying "all") of the
available data; we plot a subset so as not to overwhelm the plot
with excessive outliers that can serve to "bleed" across the figures
and confuse the reader.
Plot 1) u (ukic) vs. Ukis :
The ukic vs. Ukis-magnitude scatter plot shows 3-4
parallel sequences each offset from the 1:1 relation with KIS magnitudes but
distributed over a range of 1½ magnitudes. For this reason the authors
of the KIC catalog have asked MAST to withdraw their u-magnitudes (only
2092 objects) from the Target Search pages. They may still be accessed
via the MAST/KIC web page, but we doubt that they are useful.
UPDATE (Fall, 2012): A scattershot plot of KIC objects with the latest Sloan survey release, SDSS/DR9, shows once again that serious problems exist with the U magnitudes between the two surveys. The linear regression between u_kic and u_sdss is described by a large intercept of 6.09 magnitudes (at mu_kic = 0) and a slope of 0.62. The ranges for KIC and SDSS magnitudes are, respectively, 12--19, and 14--18. The SDSS Project has documented a red leak for its U filter, whereas no information exists on this point from the KIC Project.
Plot 2) g(kic) vs. gkis :
We find a offset of of -0.03 magnitudes. However, when one takes
account of the zeropoint differences between the KIS (Vega)
and KIC(AB) systems, as given by
Gonzáles-Solares et al. (2011, MNRAS, 416, 927) as:
Here AB referred explicitly to observations by the SDSS/DR9 in other regions
of the sky. However, in principle it could apply to other AB-based systems,
such as the KIC. If we adopt an average color from our sample
This plot and the r and i scatter plots disclose two other points that are
more bothersome. At faint magnitudes (> 16-17, depending on the filter)
a faint, nonparallel secondary sequence develops. We will show evidence
below that the secondary magnitude sequence is a problem with the KIC data.
At bright magnitudes (mkis < 12) the KIS magnitudes are too
bright. Greiss et al. were aware of this problem for <12 KIS objects and
ascribed it to partial saturation of images. Note, however, the KIS "class"
value for saturation is only used for a minority of objects in the range
10-12th magnitude.
UPDATE (Fall, 2012):
A comparison of gkic with gsdss magnitudes in
"Plot 2b" (below) shows an
intercept at gkic = 14 of 14.0 mags, i.e., no offset. Because
of the asymmetrically distributed scatter we estimate the error on the
intercept to be ±0.05 mags. In addition, notice that a "flange feature"
in the upper right of the diagram mimics the feature in Plot 2. This tells
us that many of the KIC magnitudes in the gkic 18-21 arise
from systematic errors in the KIC survey exist (the KIC does not provide
estimated magnitude errors). Similar features
are found in the comparisons for the KIC r and i (but not z) magnitude
figures, Plots 3 and 4. On the other hand, a second flange, the scatter above
the regression line in the magnitude range 12-14, is not present in Plot 2b
as it is in Plots 2-4, indicating just as clearly that the fault seems
to lie with the SDSS magnitudes; the errors for these particular objects
are much larger than the errors the SDSS pipeline indicates.
Plot 2b) g(kic) vs. gsdss :
UPDATE (Fall, 2012):
The mean regression line for rkic with rsdss in the
magnitude range 11-18 has a near unitary (0.988) slope. Because of the
asymmetrically distributed scatter an intercept at rKIC = 14 of
-0.04 mags. in the regression relation is probably statistically insignificant.
Our remarks for Plots 2 and 2b concerning two "flange" features also apply here.
UPDATE (Fall, 2012):
The mean regression line for ikic with isdss
magnitudes has a slope of 0.963 between magnitudes 11 and 16.
Because of the asymmetrically distributed scatter
an intercept at ikic = 14 of 14.02 mags. in the regression
relation is probably statistically insignificant. The remarks for Plots 2
and 2b concerning two "flange" features apply here.
UPDATE (Fall, 2012):
The corresponding (g-i)kic vs. (g-i)sdss plot exhibits
a derived regression slope of 1.027 and intercept of -0.016 mags. for a
(g-i)kic color of 0 (an intercept of 0 at a mean color of
0.6 mags.). However, there is a large secondary scatter
around this relation, indicating that we have not yet isolated the
population of moderately bright (< 16th magnitude) in our comparison.
UPDATE (Fall, 2012):
The corresponding (r-i)kic vs. (r-i)sdss plot exhibits
grave enough problems to state that a mean regression line is worthless.
The scatter plots shows two dense concentrations of points, with the
secondary one being only much slightly redder in (r-i)kic and
much redder in (r-i)sdss. Users should probably refrain altogether
from using (r-i)sdss for color selections of targets. Users are
referred to the plot named ri_risdss.png in the tarball.
Apart from these features, other investigators and we have found
(g-r) color dependences in the differences between KIS and SDSS magnitudes
(this seems to be true for at least r and i filter magnitudes as well as g).
Our investigation of the differences between g magnitude distributions
in any pair of the three Sloan filter surveys - KIC, KIS, and SDSS itself -
confirms a color dependence among all three. Thus, it appears that any of
the surveys suffers a color dependence relative to any of the others.
The worst case again seems to be for the SDSS system, which shows both
a larger slope with color and an asymmetric distribution about the mean.
The Kepler team's Working Group on Stellar Properties is looking further
into these issues.
Plot 3) r vs. rkis :
Likewise, the offset in the r magnitude scatter plot of -0.15 altogether
vanishes when an average color
Plot 4) i vs. ikis :
Again the seemingly large offset of -0.48 magnitudes is largely due
to the zeropoint difference between the KIS (Vega) and KIC
(AB) magnitude systems, namely:
Given a mean value <(r - i)> = 0.35 for our population, the predicted
offset is -0.44, which differs by -0.03 mags. from the value given in
our plot.
Plot 5) zkic vs. zsdss :
Plot 5 shows that the faint magnitude "flange" feature in the earlier
plots, and attributed to KIS magnitude problems is not present for
the z magnitudes. On the other hand, the upper left flange,
attributed to KIC problems above, is still present in this filter and
The departure of the slope in this diagram from 1.0 is probably negligible.
The offset at zKIC = 14.0 is zero.
Plot 6) Ukis vs. UUBV (where U represents
UUBV):
This figure shows an excellent relation between the U magnitudes in the KIS
and UBV/Kitt Peak system. The latter was adopted for the Everett et al. study.
There is a shift of about +0.08 (i.e., in the sense,
KIS is too faint). This is a sizeable discreprancy. The authors report
(Steeghs, priv. comm.) that they believe such differences are caused by
incomplete evaluations of different throughput and transmission curves
(including red leaks) in the two systems. These are particularly difficult
to evaluate for the ultraviolet filters because of the lack of extensive
standard star data and of course the dependence of the U-filter results
on observing conditions. It is still not clear which survey, UBV, SDSS,
or KIS provides the best U (or u) magnitudes.
UPDATE (Fall, 2012):
A comparison of Usdss with UUBV magnitude regression
exhibits a slope of 1.00. However the intercept is -0.84 magnitudes.
We have no previous published U vs. U calibration to refer to, but we infer
this large offset is mainly a consequence of the different zeropoints for the
AB and Vega magnitude systems, respectively (differences in red red leaks
between the filters could also be a factor). Likewise, we have no other
valid U vs. U plots to cite concerning the the dropoff in UUBV
at the faint and especially bright magnitude ends of our Usdss
vs. UUBV plot (see tarball file).
Plot 7) gkis vs. B (where B represents BUBV):
This scatter plot shows an excellent general correlation of the
magnitudes in the blue-green filters of the UBV and KIS systems. For
example, there is no secondary sequence at faint magnitudes. This shows
that the spurious feature is inherent in the KIC g, r, and i magnitudes.
In addition, the slope is 1.0,
which was also found in a study by Jester et al. (2005, AJ, 130, 873) for
AB-based g magnitudes observed in the SDSS survey. In fact Jester et al.'s
mean relation is:
where the g and r magnitudes are meant for SDSS but could as well correspond
to the KIS survey. Taking a mean value of <(g-r)kis> = 0.67,
Jester's relation predicts an intercept of 0.42 magnitude, which
agrees very well with our value of 0.41.
The turn-up of the sequence at bright filters appears is
the reflex of the turn-down noted above for the gkic vs.
gkis plot.
Plot 8) rkis vs. V (where V represents VUBV):
This plot shows another satisfying agreement, on the whole, in both random
and systematic errors. Transposing Jester et al.'s relation, one finds:
where all quantities are in the Vega system. Assuming <(B-V)> = 0.56 for our
population, we find a close correspondence with our intercept of 0.33
to the predicted offset of 0.35 mags. A turn-up is again visible
at the bright magnitude end of the relation, casting doubt on the reliability
of KIS magnitudes at the bright end, an issue that Greiss et al. noted.
Simple Color Plots
Plot a) (g-r)kic vs. (g-r)kis :
Because of the different magnitude systems adopted by the KIC and KIS, this plot
shows that the intercept in the (g-r)kic vs. (g-r)kis
plot is nonzero. Somewhat surprisingly, the slope is 0.83, very different
from 1.0, in the sense that the KIS color shows a smaller range than the
KIC color does. In an earlier preprint of their
paper, Greiss et al. showed a dependence of the magnitude difference,
(gsdss-gkic) with color (gkic -
rsdss. This is another way of showing the difference in slope
depicted above, and it occurs in the same sense. In this case, the "odd
man out" is the (g - r)kis color, which suggests
in this case that the nonunitary slope in color comes from the KIS and
not the KIC. MAST will investigate this further when the SDSS/DR9 data
for the Kepler field are ingested.
UPDATE (Fall, 2012):
Although (g-r) color plots have been particularly worrisome for the KIC
and KIS surveys, the correlated scatter properties are much worse for the
SDSS (g-r) color versus the other surveys, as indicated by the discussion
on the "flange" features in the SDSS magnitude plots noted above.
However, when these features are filtered, the mean regression relations
are not bad. For example, the intercept and slopes in the (g-r)kic
vs. (g-r)sdss plots are -0.03 mags and 1.02, respectively. For
a mean KIC (g-r) color of 0.6, this corresponds to an offset of -0.014 mags.,
which is negligible.
If the flange features in the upper regions of the g, r magnitude plots are
not filtered, broad and heavy banded features appear in the (g-r)
plots. These distributions are broad and asymmetric around the mean relation
and are a signal that faint stars or other groups of stars with
poor SDSS errors have not been properly eliminated from the sample.
(This is certainly still true in our SDSS color plots with i filter data.)
Plot b) (g-i)kic vs. (g-i)kis :
This plot shows a slope that is close to but not quite equal to the
value of 1.0 we would expect based on the individual magnitude plots,
#1 and #3, given above. The nonzero intercept follows from those plots too.
Plot c) (r-i)kic vs. (r-i)kis :
This plot shows a slope that is closer to unitary than the previous
one. The nonzero intercept was discussed in the individual
magnitude plots, #1 and #3 above.
Plot d) (U-g)kis vs. (U-B) (where U and B refer to UBV):
Here we plot the UV-blue color in the KIS and Johnson systems. Although
they exhibit a decidedly nonunitary slope, 0.80, this is to be expected
as this is close to the predicted slope of 0.78 by Jester et al. The offset
of -0.24 is quite different from the one found by Jester et al. because they
considered the (u-g) color based on SDSS's AB system; (U-g)kis
and (U-B) are both based on the Vega system.
UPDATE (Fall, 2012):
The (U-g)SDSS vs. (U-B) plot, limited to stars brighter
than gSDSS < 18.0 mag, and with error cuts in uSDSS
and gSDSS < 0.1 mags., exhibits a slope of 0.87 and an intercept
at color (U-g)SDSS = 0 of -1.08 magnitudes gives a good mean
regression relation but values of the slope and intercept which differ
substantially from the empirical relation from Jester et al. from the
d-b) (U-g)sdss vs. (U-B) (where U and B refer to UBV):
UPDATE (Fall, 2012):
The corresponding (U-g)sdss vs. (U-B) plot exhibits an intercept
of -1.08 mags. and a slope of 0.87.
The Jester et al. 2005 predicts values of -0.88 mags. and 0.78, respectively.
Perhaps these differences can be attributed to the apparently
poor Usdss magnitudes discussed in the UPDATE for Plot 1. As for
Plot d, the causes for these different coefficients are not yet known, but
there are a few reasonable suspects. First, the Everett et al. (2012) note
that their calibrations for the U (and also V filters) are tied to the
KIC Teff scale, which according to Pinsonneault is in error.
These errors are particularly severe for hot stars, for which the contribution
in the U-bands are substantial. A second suspect is that the filter
tranmission passbands are especially different for the U filter in the
SDSS and KIS (RGO U filter) surveys.
Plot e) (g-r)kis vs. (B-V) :
This plot shows almost a perfect regression relation an intercept of 0.02
mags. and a slope of 0.97. We notice that the exact solution depends
somewhat on the amount of scatter admitted in the initial solution
(here ±0.2 mags.). For example, mall departures from coefficients
of 1.0 and 0.0 , respectively, can be induced by including or excluding
various amounts of the asymmetric scatter around the mean relation.
Although it is yet to be confirmed from SDSS DR9 data, we
suspect the asymmetric scatter at the red end of the blue is associated
with the unexpected (g-r)kis dependence in relation to the
(g-r)kic noted for Plot a).
UPDATE (Fall, 2012):
The mean regression for the (g-r)sdss vs (B_V) color diagram
gives intercept and slope values of 0.20 mags and 0.86, respectively.
The departures from a 1:1 are due to the different filter bandpasses and
also the different magnitude systems used (Vega va. AB). The Jester et al
(2005) relation suggests values of 0.22 mags. and 0.98. The scatter in
the plot in the tarfile is suppressed because by necessary filtering of
the objects comprising the "flange features."
 |
|
|
|
|
|
|
|
|
|