Deliverable No. 5.3 Report on individual impacts of improved process-representation, treatment of snow, ensemble generation and increased resolution on the weather and climate prediction performance

This document provides an overview of individual impacts of improved description of Arctic processes, model resolution and ensemble generation in weather and climate predictions based on work in APPLICATE work package 5.3. The report provides recommendations on which of the developments to be included in the enhanced (stream 2) predictions performed for Deliverable 5.4, whose purpose is to assess the added-value of APPLICATE on weather and climate predictions.

This document shows that for short- and medium-range predictions enhancement in prediction skill can be achieved by introducing a sea ice model in the Meteo France weather prediction systems (improve near-surface temperature) and a multi-layer snow scheme in the European Center for Medium Weather Forecasting (ECMWF) forecast systems (improve near-surface temperature and snow depth). These changes are recommended to be included in their respective systems. The superiority of surface assimilation compared to dynamical downscaling of global forecasts in regional forecast systems further underlines the importance of the surface processes.

Regional high-resolution systems will benefit from both a further increase in resolution and introduction of Ensemble Prediction Systems (EPSs) to account for the uncertainty in the predictions. Both these enhancements are recommended, but require a substantial increase in operational computer power and should therefore be considered depending on the use of the prediction systems.

Improving the oceanic and atmospheric resolution in seasonal predictions give rather inconclusive results. Both positive and negative impacts are found in different systems. More studies are needed before recommendations on operational use of increased computer power can be given.

In climate prediction systems finer oceanic resolution can improve the representation of the Arctic ocean with realistic atmospheric forcing. However, large-scale atmospheric circulation biases in the fully coupled system deny such an improvement.

No substantial improvements and small sensitivity in seasonal prediction skill is found by introducing a more realistic description of sea ice melt ponds. This may be related to the existing tuning of the sea ice albedo which is reduced to account for the missing melt ponds. It is therefore not recommended to include a more realistic melt pond description without any further tuning of the sea ice models.

The description of the sea ice thickness distribution in seasonal prediction systems has also been studied. A sensitivity is found on the number of sea ice categories, but 

no evident benefit from including additional categories beyond the default configuration (5 levels) which it is recommended to keep.

Stochastic perturbations to represent the uncertainty in a seasonal prediction system show a neutral impact in the ocean model, but a deterioration in the atmosphere. It is therefore not recommended to use these perturbations scheme as more work, e.g. on tuning, is needed.