A second calibration-study for the Delta_T UVB sensor


February 1997

Francis Massen & Nico Harpes

This report supersedes the paper " A first calibration-study for the Delta_T UV_A and UV_B sensors"

1. Preliminary comments


The two UV sensors from Delta_T (UV_A , channel #7 and UV_B, channel #8) which operate since the start in 1994 of Meteo_LCD are relative low-cost sensors, whose accuracy and even calibration was doubtful from beginning on. When we added the much more precise and sophisticated UVB and UVA Biometers from Solar Light Co., we found the following:

2. Calibration of the Delta_T UVB sensor


First checks showed that a consistent good relationship between the readings of the UVB Biometer which serves as the reference and the low-cost Delta_T UVB sensor can not be found over the whole range of measured effective UVB irradiances. The overall relationship seems to be a parabolic one. For best results, we tried to fit the Delta_T UVB readings to the assumed correct readings of the UVB Biometer by using two different time periods of the year:
15 Feb. 1996 to 31 Mar. 1996 (1106 readings)



Overall climatological parameters
mean minimum maximum
temperature 2.8 C -10.8 C 17.4 C
solar irradiance 108 W/m2 0 W/m2 782 W/m2
UVB 152 mMED/h 0 mMED/h 1640 mMED/h


The computed statistical correlation between the readings of both sensors is r=0.98.

The following figure shows the Delta_T UVB readings (in W/m2) on the x-axis, and the corresponding readings (in mMED/h) of the UVB_Biometer on the y-axis.




A good fit can be given by the parabola (forced through the origin, as the very small zero offsets may be neglected):

mMED/h = 165.36*(Delta_T_UVB_reading)^2 + 190.89*(Delta_T_UVB_reading)


The goodness of the fit is R=0.99.
The standard error is:
std. error = SQRT((Sum_of(y_i -y_fit)^2)/1104) = 44 mMED/h

where y_i is the correct irradiance in mMED/h and y_fit the number obtained by applying the parabole given above to the readings (in W/m2) of the Delta_T UVB sensor.

If the effective UVB irradiance in W/m2 eff. should be found out of the Delta_T UVB readings, the parabola to apply is:

W/m2 eff. = 0.0114757*(Delta_T_UVB_reading)^2 + 0.0132474*(Delta_T_UVB_reading)


01 Jun.1996 to 24 Jul.1996 (2592 readings)



Overall climatological parameters
mean minimum maximum
temperature 16.5 C 7.4 C 32.6 C
solar irradiance 277 W/m2 0 W/m2 1229 W/m2
UVB 491 mMED/h 0 mMED/h 2790 mMED/h


The computed statistical correlation between the readings of both sensors is r=0.98.
The following figure shows the Delta_T UVB readings (in W/m2) on the x-axis, and the corresponding readings (in mMED/h) of the UVB_Biometer on the y-axis.




A reasonable good fit can be given by the parabola:

mMED/h = 41.59*(Delta_T_UVB_reading)^2 + 350.17*(Delta_T_UVB_reading)


The figure shows that the spreading of the values around the parabola is much greater than for the precedent case. As a consequence the standard error of this fit is considerably higher:

std. error = SQRT((Sum_of(y_i -y_fit)^2)/2590) = 106 mMED/h

where y_i is the correct irradiance in mMED/h and y_fit the number obtained by applying the parabole given above to the readings (in W/m2) of the Delta_T UVB sensor.

If the effective UVB irradiance in W/m2 eff. should be found out of the Delta_T UVB readings, the parabola to apply is:

W/m2 eff. = 0.00289*(Delta_T_UVB_reading)^2 + 0.0243*(Delta_T_UVB_reading)


3. Conclusion


This analysis shows that the readings of the Delta_T UVB sensor can be used to compute a reasonable good approximation of the correct effective UVB irradiance; the accuracy of the result is poorer than that obtained by the readings of the Solar Light UVB_Biometer mod. 501. Deviations between both instruments are smaller during the colder periods with less solar irradiation. For best results, a separate parabolic fit should be used for the winter-spring-autumn and summer periods.
Whenever available, the readings of the Solar Light UVB_Biometer should be used as a measure of the correct effective UV irradiance; the fitted readings of the Delta_T UVB sensor can help in the case where data from the UVB_Biometer are not available.
Back to Home-Page of Meteo_LCD