Evaluation of hyperspectral microwave observations for global NWP: Representation of expected uncertainties

In preparation for assessing the impact of a future hyperspectral microwave (hyperMW) sounding instrument on global Numerical Weather Prediction (NWP), it is essential to adequately define the expected instrument uncertainties. Accurate uncertainty estimates are particularly important for obtaining realistic results from the Ensemble of Data Assimilations (EDA) framework, which will be used for the impact evaluation. Two key aspects are addressed here: the application of initial perturbations to simulated brightness temperatures (BTs) and the specification of an accompanying observation error model. Simulated BTs are perturbed using random and correlated components of instrument noise, based on pre-launch estimates for the Hyperspectral Microwave Sounder (HyMS). Noise-equivalent delta temperature (NEDT) values of 0.53 K and 0.54 K are used for the temperature and humidity-sounding bands respectively, following super-obbing of nine fields of view. As the correlations reduce rapidly with channel separation, only nearest and next nearest neighbour noise correlations are used. Background departures reveal noisier behaviour than for current broadband sensors, as expected, particularly in temperature-sounding channels. The hyperMW observations are assimilated within an all-sky framework where the accompanying error model accounts for increased observation error in the presence of cloud. Cloud signals are estimated using indicators derived from single low-noise broadband super-channels centred at 52.80 and 190.32 GHz for the temperature and humidity bands, respectively. These indicators capture larger representation errors well, as demonstrated by their relationship with the standard deviation of background departures. However, for higher-peaking temperature-sounding channels, cloud-related representation errors are small relative to instrument noise, making constant errors more appropriate than the situation-dependent errors used for heritage sensors. Strong inter-channel error correlations are present, particularly in the humidity band, with many off-diagonal values exceeding 0.5 in clear-sky and 0.9 in cloudy conditions. Recent technical advances at ECMWF now allow such correlations to be explicitly included in the all-sky assimilation scheme. Further EDA sensitivity experiments will refine the error settings and determine the final assimilation configuration.