Optimising impact of radar reflectivity and lidar backscatter obs on forecasts

To study and optimize the impact of cloud radar and lidar observations on numerical weather prediction (NWP) forecasts, Four-Dimensional variational (4D-Var) assimilation experiments have been performed using CloudSat cloud radar reflectivity and CALIPSO lidar backscatter observations. The work is an extension of WP-2000 within the PEARL Cloud project, which optimised the impact of assimilating cloud radar and lidar observations on NWP analyses. In addition to the three-month assimilation period considered in WP-2000, this work-package increases the number of assimilation cycles considered to up to 11 months, by combining four independent assimilation experiments. The extended period also allows the seasonality of the observations and model to be considered and the appropriateness of a bias correction scheme trained on one-months data to be tested.
Results indicate that assimilating CloudSat radar reflectivity and CALIPSO lidar backscatter has a positive impact on the globally-averaged forecast skill of large-scale variables such as temperature, wind and humidity at both short- and medium-range timescales. The impact is significant for the forecast of uppertroposphere temperature at short-range forecasts, and for other variables, such as vector wind, at forecast lead-times between 4-6 days. The positive impact is also verified against independent observations of top-of-atmosphere radiation fluxes, where significantly positive reductions in the fit to CERES daily average short-wave and long-wave radiation are found at forecast lead times between 1 and 5 days. Further positive impacts are possible when the sub-optimal use of lidar backscatter in one of the experiments is fixed and the reasons for regional degradations in forecast skill, such as in mid-level temperature and wind forecasts in the North Eastern Pacific stratocumulus to cumulus transition zone, are investigated.