Davis Vantage Pro weather station setup and calibration
This page will log my efforts at setting up and calibrating my Vantage Pro
weather station. All in all, this is taking me a lot more time that I
expected, but most of that comes from pursuing various interesting
The Vantage Pro console displays one set of data, but it shares a
different set via the communications protocol. In particular, most
derived data isn't available through the protocol, e.g. dewpoint and
wind chill. Oddly, sunrise & sunset are, even though you don't
even need a weather station to compute that!
Directly measured data
This was the first sensor I checked. Normally I just stick it in an
ice bath (sometimes a double bath), but that's impractical with the the
electronics of the ISS (Integrated Sensor Suite). So I set it on an inside
table and put several other thermometers around it. They all read the
same within 0.1 or 0.2 degrees Fahrenheit. I expected as much, but it's
worth checking this basic data, especially when it's so easy.
I have the passive ventilated radiation shield. On a couple days
with mixed sun and clouds, the change between sunny and shaded conditions
seemed to be only a degree or so, good enough for me. Given the accuracy
of the temperature sensor, active ventilation is worthwhile.
This was a high priority, as I wanted to get some data before the weather
turned cold. Instead of pouring water into the funnel, I had the idea of
weighing some ice and letting it melt in the funnel. This would provide
a flow closer to a real rain event. This appears to be a really good
way to check calibration, but you really want temperatures over 50F so
that the ice melts during the day, and to come up with a reasonable
measurement, you need a fair amount of ice. For a resolution of one
tip per hundred, you need about one hundred tips, or enough ice to equal
one inch of rain. That's 463 grams, just over a pound
You probably want to do testing when the dewpoint is
close to freezing to minimize condensation and evaporation.
I did some limited tests, I'm not sure where the data is, but the error
appears to be a few percent. Serious testing will have to wait until spring.
I should be done with this, as I have a Thommen altimeter (a wonderful
device) and my wife has a GPS receiver with WAAS correction. However,
a substantial distraction is coming from determining my altitude. I've
used a value from a DeLorme's Topo USA, which made me think that my
GPS receiver, without WAAS, was reading too high. I'm beginning to think
that the map is wrong and the GPS right. There is a benchmark near home,
I need to make some checks against that. I made one, but I need to do
several. Juggling the equipment and the dog becomes a challenge, especially
with winter gloves.
This is easy, as my property is well aligned with true North/South. As
long as the weathervane turns readily, it all should work. An easy way to
find true north is to find the time of the sun's meridian passage (averaging
the sun rise and sunset times works fine). At that time, the shadow of
a vertical pole is on a north/south line.
Wind speed has to be one of worst-measured variables. The NWS standard is
to have the anemometer at 10 meters (33 feet). How about 3 meters? With
trees and buildings around? It'll do. There's little reason to check the
accuracy, but a long time ago, probably around the late 1960s, Scientific
Aerican's The Amateur Scientist had an article about people who
calibrated the deflection of a pingpong ball on a string with a wind tunnel.
I'm not sure how that compares to other ways to check it's probably quite good.
I'm not sure how this is measured, and I don't know what sort of quirks it
has. I remember a "toy" humidity gauge that worked on measuring the
length of hair changing, but apparently there's a humidity-sensitive
capacitor that sounds a lot more likely a device to use.
I suspect this is off by several percent because I've never seen
a humidity reading above 97% or so, even in rain or fog. On the other
hand, we've had snow cover since December 2nd, and ice attracts water
vapor better than most ofther surfaces. This needs further study, and
I've just added the math to compute the frost point, see below.
This is derived data, but the weather console provides it anyway. Sigh.
It seems quite accurate as it is within a minute of both the "official"
set at the US Naval Observatory and with a program I wrote long ago.
It's a tad annoying that the weather station protocol doesn't provide some
basic data, leaving it up to the recipient to compute the values. At the
top of the list of things I'd expect to be given is dew point, a much more
useful quantity than relative humidity for most uses.
Poking around on the web, I've found several equations to take temperature
and relative humidity and derive the dew point by first calculating what
the water vapor pressure is. I'm currently using one from Buck. One reason
is that it's in a format that is easy to rearrange to convert vapor pressure
This winter has been very cold, currently the dew point is -17F. There is
remarkable little moisture in the air and small changes have a big impact on
the dew point. So do small errors. This will be a better variable to monitor
in the spring.
Water vapor condenses on ice more readily than water. Given that
we've had snow cover since December 2nd, this is one reason why the
weather console reports relative humidities less than 100%. Buck also
has an equation for the frost point, and I've added that to my reports
and graphs to see if the temperature meet or crosses the frost point.
I finally had a night with good radiational cooling and frost.
It looks like the relative humidity sensor is just fine. The figure
on the right shows the temperature changes on a clear and calm night
here in the Contoocook River valley. This is also a very good example why
you never hear references to the dew point as being the likely low for the
night. While the vapor to frost
phase transition releases some 15% more latent
heat than the vapor to water transition, there is so little water vapor in
saturated cold air that the released heat can't prop up the temperature.
In 2001, a
new windchill formula was released. The old one was based
on measurements in Antarctica to see how long it took for containers
of water to freeze. The new formula is much better and looks at heat
loss from a model of a human head. It also is designed to make a wind
chill value at 3 mph be close to the air temperature. Unfortunately,
chill is defined to not exist at wind speeds below 3 mph. Worse,
the equation behaves rather oddly and increases very quickly at lower
wind speeds. I suspect it might make sense in a weightless environment
where the lack of convection means warm air pools around the body. Still,
sunny and zero is fairly pleasant when the wind is calm. I'd like to have
a number attached to that.
Contact Ric Werme or
return to his home page.
Last updated 2004 February 2 (Groundhog Day).