The sounder instrumentation on the GOES satellites multispectrally determines precipitable water (Look here [or here] for details; a graphic is available here). Here are some things to keep in mind when viewing the GOES data on these pages. The points below that are important will be repeated when appropriate in the discussion of the GPS/GOES differences.

- The sounder instrument uses 9 footprints to compute precipitable water,
and at least 4 of those cannot be sensed as cloudy by the GOES instrument.
At the GOES subsatellite point (at the [0N, 75W] for GOES-8
and at [0N, 135W] for GOES-10).
the footprints are 8 kilometers (roughly circular)
and are spaced 10 km apart. The subsatellite point is not where we are
acquiring data, however, and as you move poleward and laterally
away from the subsatellite point, the footprint becomes more and more
elliptical. For example, at Madison, the footprint it 11 km in the E-W and
16 km in the N-S; Over the CART site in Oklahoma, it is 12 km in the E-W
and 15 km in the N-S. Precipitable water with the sounder is done with
a 3x3 box (9 footprints), so that means the precipitable water is a point
representative of a 33x49 km box in WI and a 36x45 km box over the CART
site. The precipitable water from the GOES sounder is valid for an area
larger than the precipitable water from the GPS data (GPS precipitable water
is a measurement for a cylindrical volume with a circular diameter of
about 30 km); one can
envisage scenarios where this would introduce errors into the analyses
done here (For example, if there is a very tight gradient).
- The sounder data used here is from within a 1 degree lat/lon box
centered on the GPS receiver. In other words, I look for the data points
within about 50-100 km of the receiver and that's what I use to compare.
The lat/lon of that point is determined to be the centroid of the (up to)
nine GOES footprints from which data are used to compute the precipitable
water. If the footprints are represented by the letters below:
A B C D E F G H I J K L M N O P Q R

and the sky is cloudy at A, G, and M, the lat/lon will obviously be a little different than if C, I, and O are cloudy. The precipitable water values are computed using 9 points (ABCGHIMNO), then the algorithm steps over 3 columns and uses the 9 points to the right of row CIO (DEFJKLPQR). Each GOES footprint is used only once to compute precipitable water. -
The algorithm uses 9 points to reduce noise --
*i.e.*, for smoothing. CPU demands are behind using each point only once; perhaps in the future after using (ABC,GHI,and MNO) to compute the value centered near H the algorithm will use (BCD, HIJ, and NOP to compute a point near I. For now, after the value near H is computed, the value near K is computed using DEF, JKL, and PQR. - I asked myself while doing this -- how much natural variability will occur over the one-degree box centered on the GPS receiver site and how will this affect the comparisons between sounder precipitable water and GPS precipitable water? To address, this, I have plotted data at nearby sites (North Liberty, IA vs. Milwaukee WI; Lamont, OK and Haskell, OK; Haviland KS and Vici OK) to see how the two compare. The data are also presented as a scatterplot which can be seen here.

Click here for study results.

Click here for more information on GPS
precipitable water and things to keep in mind about it.