A Deep Learning model for reconstructing Northern Hemisphere atmospheric blocking
Atmospheric blocking events are persistent, high-impact weather patterns in the middle latitudes, that divert the jet stream and storm tracks for days to weeks.
Heat waves, cold spells, droughts, and flooding are often linked to persistent blocks.
Future changes in atmospheric blocking due to greenhouse-gas forcing are highly uncertain due to climate model biases, the complex nature of the phenomenon, and its strong natural variability.
Paleoclimate proxy records extend observational records to help reduce these uncertainties, but have low temporal resolution.
How can we extract meaningful synoptic-scale signals from low-resolution paleoclimate records?
PaleoBlockNet is a Deep Learning (DL) model that infers summertime (JJA) frequencies of blocked days from seasonal surface temperature fields.
If you are interested in exploring PaleoBlockNet's reconstructions, an interactive web interface can be accessed at https://www2.hawaii.edu/~ckaramp/paleoblocknet/
Karamperidou, C., Extracting Paleoweather from Paleoclimate: A Deep Learning Reconstruction of Northern Hemisphere Summertime Atmospheric Blocking over the Last Millennium, Nature Communications Earth & Environment, in revision (2024)
Christina Karamperidou - ckaramp@hawaii.edu
Research Group: https://www2.hawaii.edu/~ckaramp