Frosting of radiometers

[BaptisteVandecrux]

In the winter or at night during the rest of the year, the radiometer can get frosted and the dlr and ulr values may not be reliable.

This is visible at EGP and CEN2 through the prolonged periods of low difference between dlr and ulr.:

Such a low difference is unlikely because, in the winter, atmospheric inversion prevails and the surface should be colder than the atmosphere (ulr < dlr).

We need to find a semi-automated way to spot those periods and, possibly, a standard test using another sensors (humidity at saturation? direct measurement of inversion from temperature measurements?) that would confirm the frosting of the radiometer.

[BaptisteVandecrux]

After going through the data, I can see that the difference between dlr and ulr can also be quite small during overcast periods. Here is an example at CEN2:

Where the first week of August, 21-26 August and 6-11 Sept are overcast periods, with enough shortwave radiation to make frosting unlikely.

So we cannot go for a blunt test only on the difference between dlr and ulr.

I can see two other observations that could help discriminating frosting periods:

1. At stations where two levels of temperature measurements (t_u and t_l) are available (like CEN2), we could look whether the absence of inversion is also visible from t_u – t_l
2. At stations where there is only one level of measurement, we could use humidity measurements to look for frost-prone periods. The issue being that overcast periods also come with higher humidity values.
3. Maybe the temporal variability of dlr and ulr could help discriminating between overcast and frosted periods.
   I also thought that, in summer, overcast conditions (that may cause low dlr-ulr) can be spotted from the shortwave radiation. But how to spot overcast period in wintertime?

[mvantiggelen]

Hi Baptiste,
This is something that is also very common in the Antarctic AWS data. A robust way of flagging rimed radiometer data could be flagging when there is both high relative humidity (e.g. RH > 80%) and low wind speeds (e.g. U<3 ms-1 or even <1 ms-1). With 2-level data that should be even more robust. A different way is to compare the observed surface temperature from LWup, and the Ts from a SEB model. During rimed periods the two are often very different.
Best,
Maurice