A personal highlight: we include iceberg melt patterns, a topic I’ve worked on for years and care deeply about.

Icebergs don’t just melt, they drift over vast distances, cool their surroundings, and shape regional ocean properties.

Capturing this spatial and vertical complexity matters. (3/7)

Today @jens-d-mueller.bsky.social gave a great talk on the ocean carbon sink during the record-warm year 2023 & an outlook for 24. #EGU25

This relates to our joint work @awi.de @ecmwf.int that is presented Thursday morning, room 0.14. See you there:

meetingorganizer.copernicus.org/EGU25/EGU25-...

and (iii) better propagation and symmetry characteristics of the Madden–Julian Oscillation.
(7/n)

(ii) impacts of kilometre-scale urban areas on the diurnal temperature cycle in #cities
(which may relate to the coming @ipcc.bsky.social
Special Report on Climate Change and Cities);
(6/n)

We also provide first examples of significant advances in the realism and thus opportunities of these km-scale simulations, such as (i) a clear imprint of resolved #Arctic sea ice leads on atmospheric temperature; (5/n)

… and (iv) eddy-resolving features in large parts of the mid- and high-latitude oceans (finer than 5 km grid spacing) to resolve mesoscale #eddies and sea ice leads.

Compare the map of eddy variability (std of sea surface height) in the model and #AVISO:
(4/n)

(iii) improved intense #precipitation characteristics;
(3/n)

Main improvements are (i) better conservation properties of the coupled model system (water and energy budgets), which also benefit @ecmwf.bsky.social operational 9 km model; (ii) a realistic top-of-the-atmosphere (TOA) radiation balance throughout the year; (2/n)

How much warming? That is where simple modelling can help. Turns out that the temperature response to the observed reduction in planetary albedo since December 2020 (estimated from a simple energy-balance #model, doi.org/10.1175/JCLI...) may explain the warming ‘gap’ of about 0.2K. (7/8)

To put simply, considering the globe as a whole, high clouds and cloud-free scenes result in warming of the Earth's atmosphere, as less energy escapes into space than arrives from the sun. For low clouds, it is the opposite: their decline leads to warming. (6/8)

One trend appears to have mainly affected the planetary #albedo: the decline in low-altitude #clouds in the northern mid-latitudes and the tropics. The Atlantic Ocean particularly stands out - the region where the most unusual temperature records were observed in 2023. (5/8)

However, the AI-based models tend to #drift towards known conditions from the training data set, as seen in this "hairy plot" of global-mean 2m temperature. Individual hairs are daily forecasts from 12:00 UTC. The future state is from IFS-FESOM in the H2020 project nextGEMS nextgems-h2020.eu
(4/7)

3 AI-based models for weather forecasting ( #Panguweather, @deep-mind.bsky.social's #GraphCast
@stephanhoyer.com, @ecmwf.bsky.social's AIFS) produce skillful forecasts across different climate states, e.g. in proxy #preindustrial, present-day & future 2.9K warmer #climates.

RMSE for 2m temp: (2/7)