Email sent to ecowitt
Lucy
Please forward my information and suggestions below to your engineering group
Regards
John
In the reviews tab for the WH41 on Ecowitt.com there is a link to a comparison between the Ecowitt WH41 and three expensive (US$ 20,000 +) air quality gauges at the Rubidoux air quality monitoring station in California. Ecowitt replied to this comparison:
The sensor is sensitive to liquid droplets - rain/fog/sprinkling. When the Dew Point is close to the outdoor temperature (T - D < = 2C), the PM2.5 reading will be very high (which is not the real condition).
That might be the reason of "overestimatedPM2.5 mass concentration measured by" other sensors.
I looked at data from a weather station near the air quality monitoring station on Wunderground (Rubidoux KCARIVER89 Elev 259 m, 33.99 °N, 117.43 °W). These data show that for up to 8 hours a day for many days during the three month test period, PM2.5 will have been over estimated because the air temperature was within 2C of the dew point. The data for these periods should have been excluded from the comparison.
I suggest that Ecowitt contact the authors of the comparison to obtain the raw data, exclude the data for the periods when the air temperature was within 2C of the dew point, and then post the results on Ecowitt.com.
The correlations between the WH41 and the FEM Grimm reference instrument (linear regressions) had R-squared coefficients of 0.39 (5 minute mean), 0.50 (1 hour mean) and 0.70 (24 hour mean). The longer periods damp out the effects of the erroneous data due to high dew points. Excluding the data for periods when the air temperature was within 2C of the dew point would have a better correlation (higher R squared).
For comparison the linear regressions for one hour mean values between the reference instruments gave R-squared values 0.64 to 0.82. I expect that after the erroneous data have been excluded from the analyses, R squared for the WH41 to FEM Grimm comparison will be close to, if not within the range of, the R-squared values for the comparisons between the reference gauges. I.e. the WH41 data is as good as the reference gauge data if the data for periods when the air temperature was within 2C of the dew point are excluded.
The linear regressions all have the form y = mx + c. For example, the relationship between the 24 hour means recorded by the FEM Grimm and WH41 unit 5378 was:
FEM Grimm PM2.5 = 0.4244*(WH41 PM2.5) – 0.8208.
An offset of +0.8208 and a gain of 0.4244 would calibrate the WH41 to the FEM Grimm so that the 24 hour means matched. On my GW1000 with 1.6.3 firmware, the only calibration option for the PM2.5 measurement is an offset. It would be useful to have a gain setting as well for users that have access to data from nearby air quality monitoring stations or other reliable sources of calibration data.
I suggest that Ecowitt consider modifying the software (or firmware) in their consoles so that only PM2.5 data for periods when the air temperature is greater than 2C above the dew point is used.