The purpose of this page is to really emphasize how important it is to do all you can to make your weather data as accurate as possible. As amateurs, we are not beholding to any standards other than the ones we set for ourselves. Our credibility for providing accurate weather data to other people depends on how accurate our instruments are and how well we mount them and care for them. This page describes all the nasties that can affect your good data so you know the limitations and strengths of the hardware and software and what you can do to keep things as accurate as possible.
We checked out four weather sensors in a calibrated wind tunnel to see how well they performed. That data is the basis of our wind speed calibration values. Above 10MPH, the sensor units agreed with the wind tunnel reading within +/-2%. Variances between units were less than 5%. Below 10MPH the unit starts to read low, falling off to about 10% low at 5MPH and 30% low at 3.5MPH.
The basic 1Wire weather station hardware reads one of 16 equally-spaced directions for each five second sample. This is lower resolution than most users want. Short-term wind direction is highly variable due to atmospheric turbulence and when you take a bunch of samples and average them together, the variation is averaged out and you are left with a representative value of wind direction.
The weather station temperature sensor is supposed to be accurate to within +/0.5 degrees C over the 0-70 degree C range. Below 0 degrees C, it is nonlinear, but for temperatures in the lower United States, errors will be less than 2 degrees. Because the temperature sensor is inside the outdoor plastic unit in a sealed environment, it can become hotter than ambient when the sun is shining directly on the housing. This can result in significant temperature error. We ran a one-week test using the standard outdoor units, a homebrew radiation shield, and a commercial radiation shield and compared them. Our worst case error was +16 degrees F on a calm sunny day. On our web page is a set of plans and the directions to construct an inexpensive radiation shield which will reduce this error by a factor of four in light winds and more in higher winds. Dallas is considering an external temperature sensor that could be placed in a protected environment. If that becomes available, we will announce it and add software support for it as soon as practical.
One of the greatest things about this weather station is that all the data coming down the cable is already digital. Cable length does not impact reading accuracy at all.
Since we know the errors introduced at low wind speeds, we can compensate for it in firmware. It is on our ďto doĒ list. Watch the release notes to know when we add this feature.
In our case, there are three kinds of noise in the wind direction data. The first is variation in wind direction (turbulence). The second is sampling error due to the limited number of data points available. With averaging we can still claim accuracy within six degrees for the averaged data. See the FAQ page for lots of discussion about this. The third source of error we canít control is an offset error when you first calibrate the wind direction sensors. That is a single sample, and if you do point perfectly accurate, we could still be off up to 11.25 degrees. That one hurts, but there isnít much we can do about it. Fortunately, this is the measurement most tolerant of errors.
There is little the firmware can do to compensate for the solar heating of the outdoor unit. The best approach here is to use a radiation shield to keep the solar radiation off of the sensor and to dissipate the heat due to the direct solar radiation. The home-made radiation shield on our web page will reduce these errors by a factor of about four, so it is a good option for anyone who doesnít mind building something.
All data coming down the cable is verified with one of two types of CRCs, so that erroneous data will be eliminated. There is quite a bit of code not only checking the data, but making sure bad or missing data does not mess up the averaging routines. One of the (soon to be) weather display screens will show cumulative communication error counts. Hopefully these will always read 0, but occasional errors may be possible. If you see lots of errors, then you have some kind of cabling problem that needs addressing.
Your part in making sure accurate wind data get out is to make sure the sensor unit is in a good place to get a good sample of the weather. Below is a good description of what is needed by Ray McKnight WB3ABN RMcKnight@PACNORWEST.USCG.mil
STANDARDS FOR WEATHER SENSOR INSTALLATION
To aid those Amateurs who install home/amateur weather stations, I have extracted the various recommendations regarding the proper methods for installing the various sensors. These recommendations are contained in FMC-S4-1994, "Federal Standards for Siting Meteorological Sensors at Airports", and meet both FAA and NWT guidelines.
In following these standards you will ensure your weather station is providing the most accurate readings possible. Obviously, you will need to adapt these guidelines to your particular installation and may not be able to obtain the exact results listed. These standards, at minimum, should provide guidance in the mechanics involved in obtaining accurate sensor readings.
I will address the sensors most commonly found in a home unit:
surface wind speed and direction
ambient air temperature
dew point temperature
(this does not address manufacturer specific installation issues)
Atmospheric pressure: Should usually be installed in a weatherproof building or enclosure (most home units include this sensor in the main display). Avoid areas that are affected by pressure changes caused by compression due to closing doors in small rooms, avoid jarring, vibration and rapid temperature changes. Avoid direct sunlight, drafts from open windows, and air currents from heating or cooling systems. Air flowing over a sensor can cause artificially low readings. It should be at least 3 feet above ground level, but less than 100 feet above average terrain. If installed in a closed room, venting to the outside may be necessary.
Wind speed and direction: Should be oriented to TRUE NORTH, not magnetic north! BE AWARE that in some areas true north may vary by 15 degrees or more from a raw compass reading. Call the nearest Coast Guard unit for your local magnetic variation. The site should be relatively level but small gradual slopes are OK. This sensor should mounted 30-33 feet above average terrain, EXCEPT it should be at least 15 ft above any obstruction within 500 feet. If installed on a tower, it should be at the top, but a side bracket can be used and should be at least 3 feet away from the tower side. Towers should be of an open design to permit free air flow. This will be the hardest sensor to install at a home location to meet the intended guidelines. Don't cut down your neighbors trees to reduce obstructions! Achieve the best compromise possible. You should also check this sensor at least monthly to ensure free movement of the wind cups and vane, as well as proper orientation to true north.
Temperature and dew point: Should be at 5 feet (+/- 1ft) above ground level, or 2 feet above average maximum snow depth. It must be adequately ventilated but needs protection from direct sunlight, earth radiated heat, etc. Also it should not be influenced by artificial conditions like concrete or blacktop, heat radiating from buildings, heating/cooling exhausts, etc. Avoid placing above house roofs or near windows, doors or roof eaves. Vegetation near a sensor should be clipped to 10 inch height or less.
Lightning detector: Install according to manufacturers directions. Metal obstructions should be no closer than 2 times their height.
Precipitation accumulation: Must be level (measure the upper rim or orifice). Surrounding terrain should be as flat as possible. It should be as close as possible to ground level, and free from surround obstructions (no closer than twice their height). Avoid hard surfaces which may allow splashing into sensor. To avoid losses due to wind, install an "alter-type" wind shield. In areas with rain or sub-freezing temperatures, it must be heated to measure snow or freezing rain. In home units, this is usually the least accurate sensor. Check periodically to ensure sensor is free of debris.
Consult the manufacturer for guidelines in extending cables. You should also keep RF sources as far away as possible from sensor cables and the processor unit. Cable shielding may be necessary. You will also need to set your corrected sea level pressure. Call the nearest airport and ask the control tower for CORRECTED SEA LEVEL PRESSURE, not airfield local altimeter pressure.
Hope this helps you get the most accurate readings possible. NWS DOES use APRS reported weather, so we should strive to be as accurate as we can!