Assimilating observations sensitive to cloud and precipitation

TitleAssimilating observations sensitive to cloud and precipitation
Publication TypeMiscellaneous
Year of Publication2017
AuthorsGeer, AJ, Ahlgrimm, M, Bechtold, P, Bonavita, M, Bormann, N, English, S, Fielding, M, Forbes, R, Hogan, R, Hólm, E, Janiskova, M, Lonitz, K, Lopez, P, Matricardi, M, Sandu, I, Weston, P
Secondary TitleTechnical Memorandum

Satellite radiances were originally used only in clear-sky conditions, but now they can be assimilated in cloudy and precipitating areas with the ‘all-sky’ approach. This has significantly increased the influence of microwave humidity-related observations, which now contribute around 20% of observation impact (measured using adjoint techniques) and improve medium-range scores by around 3%. Much of the benefit comes through 4D-Var tracing, which infers dynamical initial conditions from observed humidity, cloud and precipitation. Initial conditions can be further improved by extending all-sky assimilation to temperature-sounding microwave and infrared radiances, though there are still many challenges, and observation operators need more development. Research will be undertaken to assimilate new types of data, such as satellite-based cloud and precipitation radar, and satellite radiances at frequencies not previously used (e.g. sub-mm or solar). The use of ground-based precipitation radar, rain gauges and lightning imagers is also being developed. All these observations bring new information, not just on the location and mixing ratio of hydrometeors, but also on subgrid variability (such as the cloud overlap), particle shape and size distribution, and even on particle orientation.

To support these new observations will require an increasingly accurate representation of moist physical processes. The forecast model will start to represent more details of the microphysics and sub-grid distribution of cloud and precipitation, and crucially, these new variables and the assumptions behind them will start to be constrained directly by observational data, using it in the most optimal way, e.g. in its original form as a radiance or a backscatter. By helping to improve the modelling of cloud and precipitation, cloud and precipitation-sensitive observations will give benefit at all forecast ranges. A further important step will be to archive more of the variables required by cloud and precipitation observation operators (for example, hourly precipitation accumulations, or the mixing ratios of convective precipitation) to support forecast validation and verification against observations. The data assimilation system must also continue to provide the right tools, including a move towards cloud control variables. Also the tangent-linear and adjoint representations of the moist physical processes must evolve in tandem with the forecast model itself.