Now open for application
Closed for application
CU41.CU-Alpha.12

Improving small-scale ocean predictions with probabilistic forecasting

  • Reference person
    Andrea
    Storto
    andrea.storto@cnr.it)
  • Host University/Institute
    Consiglio Nazionale delle Ricerche
  • Internship
    NO
  • Research Keywords
    Ocean predictability
    Deep learning
    Probabilistic forecasting
  • Reference ERCs
    PE10_8 Oceanography (physical, chemical, biological, geological)
    PE10_21 Earth system modelling and interactions
    PE6_12 Scientific computing, simulation and modelling tools
  • Reference SDGs
    GOAL 7: Affordable and Clean Energy
    GOAL 13: Climate Action
    GOAL 14: Life Below Water
  • Studente
  • Supervisor
  • Co-Supervisor

Description

Ensemble forecasting is increasingly attracting the attentionof the ocean forecasting community, as recent studies pointed out the abilityof probabilistic systems to extend the forecast horizon of ocean meso- andsubmeso- scale processes. In this Ph.D. project, the grant recipient willdevelop state-of-the-art stochastic physics schemes for ocean generalcirculation models to enable probabilistic forecasting and assess their impacton the predictability of small-scale oceanic features in realistic case studies.The project will build upon stochastic schemes previously implemented for theNEMO general circulation model, complemented with new methods eventually basedon deep learning to adaptively infer the ocean model's small-scale uncertaintyand subgrid variability.

Suggested skills:

Background in geophysics, Earth Sciences, environmentalmodelling or applied mathematics is preferable, with experience in numericalcodes.

Research team and environment

The group on Ocean and Climatevariability at CNR ISMAR, Rome, focusses on understanding the changes in theoceans over a broad range of scales, using mostly numerical tools such asgeneral circulation models and reanalyses. It includes about 10 staff members(including reserachers, postdocs and students), and can benefit from localclusters plus agreements with HPC centers to run the simulations.