Passive microwave observations from satellite radiometers are widely used to make remote-sensing measurements of the Earth’s atmosphere and surface characteristics. Current missions operate in the spectral range of 1 – 200 GHz but this will be extended in the future to 229 GHz for the EPS-SG MWS instrument and to frequencies over 600 GHz for the ICI mission. Total column water vapour, cloud liquid water path, ocean surface wind speed and direction, sea-surface temperature and salinity, and profiles of humidity and temperature, are all derived from microwave observations. The top-of-atmosphere (TOA) spectral signals in this spectral range can, depending on the state of the atmosphere, comprise a significant component due to emission and reflection from the ocean surface. This is particularly true of microwave imagers (where data quality assessment and operational use at NWP centres rely on radiative transfer modelling including surface terms) and the surface-sensitive channels of microwave temperature and humidity sounders (e.g. AMSU-A channel 5 and window channels).  It is therefore critical that uncertainties in the ocean surface microwave radiative transfer are accurately calculated. This requirement spans applications ranging from the assimilation of Level-1 products (for example) in reanalysis efforts, to the generation of Level-2 (and higher) products at all levels of maturity, ranging from near-real-time operational products to climate data records.

Several emissivity models have been developed over the last two decades to support the assimilation of microwave-imager data at operational NWP centres and to support applications based on retrievals of the ECVs listed above from satellite-based microwave imager observations. These models account for several processes influencing the emissivity of the ocean surface, including: polarised reflection of the ocean’s (dielectric) surface derived from the Fresnel equations, large scale roughness due to wind-driven waves, small scale roughness due to capillary waves, and the radiative effect of foam at progressively higher wind speeds. An ocean surface emissivity model, which is widely used in the remote sensing and operational NWP community, is the Fast Ocean Emissivity Model (FASTEM), which forms part of the RTTOV fast radiative transfer model. Following the initial formulation by English and Hewison (1998), FASTEM has been developed over the last 20 years, with many recent developments guided and informed by an analysis of biases observed between satellite observations and simulations based on NWP models (Bormann et al (2011); Bormann et al (2012); Meunier at al (2014); and Kazumori et al (2015)). The current version of FASTEM (version 6) includes the dielectric constant model and wind speed terms developed by Liu et al (2011), the foam parameterisations of Stogryn (1972) and O’Monahan and Muircheartaigh (1986), and the wind-direction dependence terms developed by Kazumori et al (2015).

A number of studies have been carried out to estimate the uncertainties of ocean emissivity models (e.g. Guillou et al 1996; Guillou et al; 1998, Greenwald et al; 1999). However, most studies which estimated uncertainties were carried out before the latest versions of FASTEM, which include considerable updates made by Liu et al (2011) and Kazumori et al (2015), and also tended to focus on one aspect of the model or one frequency. Therefore, despite a number of studies being carried out to validate the FASTEM model, it still lacks traceable estimates of the uncertainties associated with the computed emissivities in the 1-200 GHz range.  This gap has been identified as an important deficiency in using NWP-based simulations for the validation of new satellite missions.

FASTEM is an approximate (fast) parameterisation of an underlying reference model (English et al., 2017). Such a reference model has three main components: (i) the dielectric model predicting the polarised reflection and refraction for a flat water surface (Lawrence et al. 2017); (ii) the roughness model which represents the ocean roughness due to large scale swell and wind-induced waves; and (iii) the foam model which commonly parameterises the ocean foam coverage as a function of wind speed and assigns a representative emissivity to the foam fraction. For a true reference model, each of these components should be associated with traceable uncertainties.