Ground based microwave radiometer – cloud detection, sky-clearing and O-B statistics

Ground based microwave radiometers (MWR) are currently in the focus of meteorological agencies which intend to deploy MWR in network setups. The centralized processing of MWR data products within ACTRIS and the imminent integration of MWR into the EUMETNET E-PROFILE network are two prominent examples for this development.

Assimilation experiments with clear-sky MWR brightness temperatures (T_B) at DWD show a positive impact on the numerical weather prediction. However, cloudy-sky conditions involve large random differences between model and observation and therefore, the most frequent reason for rejecting data from data assimilation is the suspected presence of clouds. A balanced detection of clouds and artificial sky-clearing are possible strategies to mitigate this effect.

We present the resulting effect of a neural network (NN) based liquid water cloud detection on observation minus background (O-B, ICON-D2) statistics and compare it to established cloud detection schemes. The NN relies solely on the observed T_B and is trained with T_B computed with a line-by-line radiative transfer model (Rosenkranz 2022, non-scattering) from ERA5 reanalysis data.

We will also present O-B statistics with artificially cleared T_B observations. The spectral signature of the liquid water is subtracted from the observed T_B spectrum using a NN trained on ERA5. The cloud detection and sky clearing allow a thorough discussion of 2-years O-B statistics, which provides some insights on the model performance and monitoring instrument errors. The presented progress on sky clearing of T_B also provide a possible way forward to profit even more from assimilation of MWR T_B in numerical weather prediction models.