SSWA.TV

Weather Station Quality Control

General Pass Rates
Must pass QC Levels 1 and 2.

Perfect Pass Rates
Must pass QC Levels 1, 2, and 3.

Quality Control Description

Quality control (QC) is performed by NOAA MADIS.

General vs Perfect Pass Rates
A "general pass" is our term for a measurement that has passed through at least Levels 1 and 2 of the QC system, but possibly failed at Level 3. This is the pass rate we recommend using when discussing our station's performance. QC Levels are explained in greater detail below, but essentially Level 3 compares our measurement to an estimated expected value. The expected value is only updated hourly, but has a generous range, which means it works great during most weather. However, because of this hourly schedule it cannot catch fast changes (rainfall causing a quick temperature drop, front causing a rapid pressure change, thunderstorm creating a burst of high winds, etc). This is why we often ignore this QC Level, especially when fast changing measurements are common. A "perfect pass" is our term for a measurement that passes every single level of the QC system, including Level 3. We only use it during times of steady weather, which gives us a good idea of our stations biases and overall performance.

QC Levels
MADIS has three levels of QC that it goes through in ascending numerical order. A failed measurement does not continue to the next level. Not all levels can be applied to every measurement.
Here is a list of which levels (corresponding to the sections below) are used for each measurement we report:
TEMP .... 1 2a 2b 2c 3
RELH .... 1 2a 2b 2c 3
SEAP .... 1    2b 2c 3
WSPD .... 1    2b 2c 3
WGST .... 1    2b     
WDIR .... 1       2c 3
PR24 .... 1           
PR01 .... 1           
A measurement that only goes through and passes Level 1 is marked as C (coarse). A measurement that only goes through and passes Levels 1 and 2 is marked as S (screened). A measurement that goes through and passes Levels 1, 2, and 3 is marked as V (verified).
Level 1 - Validity
This first level restricts measurements to a strict range. For example, temperature readings must be within -60°F and 130°F. Any reported values outside of this range would be marked as X (rejected/erroneous).
Here is a full list of ranges for each variable our station runs through QC:
TEMP .... -60 -  130 °F 
RELH ...... 0 -  100 %  
SEAP .... 568 - 1100 mb 
WSPD ...... 0 -  287 mph
WGST ...... 0 -  287 mph
WDIR ...... 0 -  360 °  
PR24 ...... 0 -   44 in 
PR01 ...... 0 -   44 in 
Level 2a - Internal Consistency
This level contains a few different checks, but none of them apply to our station. Dewpoint is checked to see if it is below the air temperature, station pressure is checked for rate of change and to see if it aligns with the reported sea level pressure. The network we report to does not have the ability to report dewpoint or station pressure, these are derived by the MADIS system and so these checks are skipped for our station. In the event we ever expanded the SSWA station network, we would become an independent network and these checks would be run on actual values instead of calculated values.
Level 2b - Temporal Consistency
Level 2b is similar to Level 1 except instead of limiting the current measurements to a range, this level limits the rate of change of measurements to a range. Simply, the current value is compared to the value from one hour ago and if the difference is more than the acceptable limit, the observation will be marked as Q (questioned).
Here is a full list of ranges for each variable our station runs through QC:
TEMP .... 35  °F/hour
SEAP .... 15  mb/hour
WSPD .... 23 mph/hour
Relative humidity is not on this list because it is not checked. Instead, dewpoint is used with a max rate of 35 °F/hour. It was not included because technically this is a value that is calculated from our masurements on their end, not a value we report directly, and not a value we show on this page. The RELH field in the tables above is actually the QC result for dewpoint.
Level 2c - Statistical Spatial Consistency
This level checks to see if any measurement has failed QC more than 75% of the time in the past week. If a measurement meets that condition, it will be automatically marked as failed until the weekly failure rate is below 25%. This sublevel is only applied to measurements that also go through Level 3.
Level 3 - Spatial Consistency
The final level checks a measurement against a calculated expected value. If a measurement deviates greatly from the expected value it fails this level and is marked as Q (questioned). In many cases, most of the weight of the expected value at a station would be found by looking at other nearby stations. However, with our station, this is not possible because the closest reliable station in the network is about 80 miles away. In these cases, most of the weight of the expected value comes from a weather analysis grid which is used as a sort of background to fill in gaps. The biggest problem with this approach for remote stations is that the analysis grid is only updated hourly. This means, for example, if the analysis grid says the temperature should be 79 °F at 10 PM and 73 °F at 11 PM, the QC system must assume the temperature should be 79 °F from 10:00 PM to 10:59 PM, then 73 °F from 11:00 PM to 11:59 PM. Of course, there is also an acceptable range, so just because a station reports 76 °F at 10:30 PM does not mean it will fail this level since it would fall into the acceptable range. This approach works great during steady weather, but problems start to show during periods of fast change, such as rainfall causing the temperature to drop quickly, a front moving through and causing the pressure to change rapidly, or a thunderstorm creating a short period of high winds. When any of these or similar events happen, it will cause a station's reported measurements to fail this level, even though the values are accurate. The MADIS system is smart though, so it puts less weight onto this QC level. Any measurements that fail here are still allowed into the system, as long as they aren't impossibly far off.

DB:22.30ms
TL:22.58ms
1793247641