Deliverable No. 4.2: Evaluation of the contribution of the Arctic observing system to forecast skill from short/medium-range-to-seasonal time scales

APPLICATE – GA 727862 Deliverable 4.2 APPLICATE WP4 (‘Support for Arctic observing system design’) demonstrated that numerical models can be successfully used to assess the importance of current Arctic observations for predictive skill in the Arctic and beyond and for guiding the design of future Arctic observing system. This deliverable summarizes part of this effort made in Task 4.2. This task aimed (i) to assess the importance of current Arctic observing systems for predictive skill from hours to months ahead, (b) to identify which enhancements to the observing systems themselves would contribute to improved predictive skill in the Arctic and beyond and (c) to guide the design of future observing systems in Arctic region. These objectives were accomplished by making use of data denial experiments, in which the quality of the initial conditions of weather predictions is artificially degraded by removing the contribution of a certain observation type, and of historical simulations with various coupled global circulation models. Coordinated idealized ‘data denial’-type experiments performed with two coupled climate models HadGEM3-GC3 and EC-EARTH, were used to quantify the impact of accurate sea ice thickness initialization on the Arctic sea-ice and atmospheric circulation forecast skill up to a year ahead. These idealized ‘data denial’ experiments consisted in initializing ensemble simulations using degraded sea ice thickness information and comparing the loss of skill with respect to control simulations assumed as the ‘truth’. This provided a theoretical understanding on the improvement of sea ice and atmospheric predictions linked to the sea ice thickness assimilation in coupled prediction systems. It was shown that an accurate sea ice initialization on the 1st of January improves the skill in all performed experiments. In particular, there is a high skill gain for the sea ice volume and extent in January-February and a re-emergence of skill gain from September onwards. The 2m air temperature shows a prediction skill gain in January with some improvements in September. The fact that the two models show a good agreement in terms of the regions where they show either a skill gain or loss increases the confidence in these findings. Observing System Experiments performed for the first time for Arctic atmospheric observations with a global state-of-the-art numerical weather prediction system (the ECMWF Integrated Forecasting System) were used to demonstrate the value of these observations for short and medium-range weather forecasts. It was demonstrated that all observing systems contribute to forecast skill in the Arctic and mid-latitudes and have complementary positive impacts on forecast skill. Conventional observations play the most important role during winter, emphasizing the need to maintain and further develop conventional observational networks, which are sparse and costly to maintain in polar regions. Satellite microwave observations play the most important role during summer, while for winter the use of these observations is found to be suboptimal due to issues with their assimilation, in particular over snow and sea ice. This is one of the success stories in APPLICATE highlighting that the quality of weather predictions could be further enhanced by increasing the uptake of these observations in numerical weather prediction systems. For this to be achieved investments are needed in all key components of numerical weather prediction systems, i.e. coupled modelling, ensemble and data assimilation techniques and use of observations. Similarly, to the idealized data denial experiments, the fact that the results obtained at ECMWF are corroborated by coordinated observing system experiments performed at other weather centres in the framework of YOPP, increases the confidence in these results and in the ways forward presented herein and in D4.4. Finally, Ponsoni et al. (2020) used coupled historical simulations performed in the framework of CMIP with several coupled climate models to test the hypothesis that an ideal sampling strategy, characterized by only a few optimal sampling locations of key parameters, can provide in situ data for statistically reproducing and predicting the interannual sea ice volume (SIV) variability. Indeed, it was shown that four well-placed locations are sufficient for reconstructing about 70% of the SIV anomaly variance. These stations are placed at the transition Chukchi Sea– central Arctic–Beaufort Sea (79.5◦N, 158.0◦W), near the North Pole (88.5◦N, 40.0◦E), at the transition central Arctic–Laptev Sea (81.5◦N, 107.0◦E), and offshore the Canadian Archipelago (82.5◦N, 109.0◦W), in this respective order. These results provide recommendations for the ongoing and upcoming observational initiatives, in terms of an Arctic optimal observing design, for studying and predicting the pan-Arctic SIV and its interannual variability.