An optimal estimation algorithm for the retrieval of fog and low cloud thermodynamic and micro-physical properties

Abstract

Abstract. A new generation of cloud radars, with the ability to make observations close to the surface, presents the possibility of observing fog properties with better insight than was previously possible. The use of these instruments as part of an operational observation network could improve the prediction of fog events, something which is still a problem for even high-resolution numerical weather prediction models. However, the retrieval of liquid water content (LWC) profiles from radar reflectivity alone is an under-determined problem, something which ground-based microwave radiometer observations can help to constrain. In fact, microwave radiometers are not only sensitive to temperature and humidity profiles but are also known to be instruments of reference for the liquid water path. By providing the thermodynamic state of the atmosphere, to which the formation and evolution of fog events are highly sensitive, in addition to accurate liquid water path, which can be used to constrain the LWC retrieval from the cloud radar alone, combining microwave radiometers with cloud radars seems a natural next step to better understand and forecast fog events. To that end, a newly developed one-dimensional variational (1D-Var) algorithm designed for the retrieval of temperature, specific humidity and liquid water content profiles with both cloud radar and microwave radiometer (MWR) observations is presented in this study. The algorithm was developed to evaluate the capability of cloud radar and MWR to provide accurate LWC profiles in addition to temperature and humidity in view of assimilating the retrieved profiles into a 3D- and 4D-Var operational assimilation system. The algorithm is firstly tested on a synthetic dataset, which allows the evaluation of the developed algorithm in idealised conditions. This dataset was constructed by perturbing a high-resolution forecast dataset of fog and low-cloud cases by its expected errors. The algorithm is then tested with real data from the recent field campaign SOFOG-3D, carried out with the use of LWC measurements made from a tethered balloon platform. As expected, results from the synthetic dataset study were found to contain lower errors than those found from the retrievals on the dataset of real observations. It was found that LWC can be retrieved in idealised conditions with an uncertainty of less than 0.04 g m−3. With real data, as expected, retrievals with a good correlation (0.7) to in situ measurements were found but with a higher uncertainty than the synthetic dataset of around 0.06 g m−3 (41 %). This was reduced to 0.05 g m−3 (35 %) when an accurate droplet number concentration could be prescribed to the algorithm. A sensitivity study was conducted to discuss the impact of different settings used in the 1D-Var algorithm and the forward operator. Additionally, retrievals of LWC from a real fog event observed during the SOFOG-3D field campaign were found to significantly improve the operational background profiles of the AROME (Application of Research to Operations at MEsoscale) model, showing encouraging results for future improvement of the AROME model initial state during fog conditions.

Role of thermodynamic and turbulence processes on the fog life cycle during SOFOG3D experiment

“…For the absolute humidity, the maximum dry bias of the MWR is around 1.4 g m −3 in the lowest troposphere up to 1.7 km and becomes wet above (0.3 g m −3 ). Martinet et al (2022) showed that the LWP accuracy has been validated in clear-sky conditions and has shown errors between 1 and 14 g m −2 . These error range are in the scope of those defined in the literature (Crewell and Löhnert, 2003;Marke et al, 2016).…”

“…every 10 min with a zenith pointing each 1 s. Using neural networks, brightness temperatures measured by the MWR at all elevation angles (the lower elevations angles added to measurements at zenith) are inverted to temperature and humidity variables. More details on this method can be found in Martinet et al (2022). Comparing temperature and humidity profiles retrieved by the MWR with radiosonde data, Martinet et al (2022) found that air temperature has cold biases below 0.5 K in absolute value below 2 km but increases up to 1.5 K above 4 km altitude.…”

“…More details on this method can be found in Martinet et al (2022). Comparing temperature and humidity profiles retrieved by the MWR with radiosonde data, Martinet et al (2022) found that air temperature has cold biases below 0.5 K in absolute value below 2 km but increases up to 1.5 K above 4 km altitude. The low biases in the lowest atmosphere allow a good estimation of the lowest temperature inversion under focus in this study.…”

“…The dissipation processes are more difficult to study than the fog formation processes due to the complexity of fog's scale. At the state of the art, based on case studies, numerical weather prediction models (Philip et al, 2016;Bell et al, 2022) and high-resolution models (Price et al, 2018;Ducongé et al, 2020;Fathalli et al, 2022) including LESs (Bergot et al, 2015;Mazoyer et al, 2017) have the ability to simulate fog formation in several complex areas. However, they have difficulties in simulating the processes driving fog evolution over land in real time (Steeneveld et al, 2015;Price et al, 2015;Román-Cascón et al, 2016;Waersted et al, 2019;Pithani et al, 2020;Boutle et al, 2022).…”

Vertical Profiles of Liquid Water Content in fog layers during the SOFOG3D experiment

“…The observation strategy combined vertical profiles derived from remote sensing instruments (microwave radiometer (MWR), Doppler cloud radar and Doppler lidars) and balloon-borne in-situ measurements of fog microphysics and thermodynamics. Bell et al (2022) and Vishwakarma et al (2023) combined cloud radar reflectivity with temperature and humidity profiles and LWP retrieved from MWR, to better estimate the vertical profile of LWC in the fog layer. They demonstrated that LWC retrieval is highly sensitive to the prescribed droplet concentration, and that agreement with in situ data is highly dependent on cloud-fog heterogeneity.…”