Leverhulme Trust Research Project / DTC project joint meeting

7 March 2025

MUMIP LT/DTC meeting Friday 7 March, 1:30pm UK time

Minutes

 

Slides

Edward Groot

 

Present

Hannah Christensen (Oxford)
Julia Simonson (CU Boulder)
Kathryn Newman (NOAA)
Jeff Beck (NOAA)
Romain Roehrig (Meteo France)
Hugo Lambert (Exeter)
Keith Williams (Met Office)
Edward Groot (Oxford)

 

Apologies

Xia Sun (NOAA), Wahiba Lfrah (Meteo France)

 

Introduction

Julia Simonson was welcomed to the group.

 

Kathryn Newman

 

  • Working on two runs, RAP and GFS physics using forcing with UFS
  • RAP completed, currently post processing
  • GFS is 20 days in
  • Will deliver data to NCAR machine

 
RR – what analysis will you do?

KN - Tendency PDFs, conditional PDFs. Funding ends in May

EG - plans to perform intercomparison with other data

 

Hugo Lambert and Keith Williams

 

  • UM data on Jasmine
  • Naming convention is following native variable names
  • Hugo happy to put files on Casper (US machine)
  • Met Office moving computer in May

KW – have original column of test data with tendencies output

Action: KW/HL to transfer this single column data to JASMIN

 

Romain Roehrig

 

Recap: ARPEGE dataset is complete and on JASMIN

 
Showed bias plots of humidity, temperature, and wind speed

  • Humidity shows complex structure in biases
  • Temperature is warmer in ARPEGE in the boundary layer and colder in free troposphere
  • Wind speed is generally higher

 
Work to assess biases as a function of atmospheric conditions:

  • Initially considering five days of data and a quarter of the domain
  • K-means clustering on profiles of relative humidity and buoyancy plume (related to atmospheric stability - see Beucler et al 2024). Profiles from ICON and ARPEGE after three hours 
  • Detect five regimes in relative humidity and buoyancy plumes showing different levels of stability and instability
  • Regime occurrence in ICON and ARPEGE has already large differences three hours into runs
  • Considered composites of humidity for each regime. Stable regime is very consistent between ARPEGE and ICON, but unstable regime is very different 
  • Looked at conditional precipitation rates

 
Future work:

  • Refined clustering
  • Additional drivers (SST, large scale forcing)
  • Apply analysis to other MUMIP models
  • Can we calibrate ARPEGE to mimic ICON? 

 
Plus

  • More simulations with alternative surface forcing
  • New version of ARPEGE
  • New set of forcing data using AROME-Global DYAMOND2 global simulation

 
HC - considered diurnal cycle? Correlations between different regimes in space and time?

HL - could cluster on outcomes - i.e. if you know you are interested in rain rates cluster on this variable, and see which inputs lead to those outcomes instead of specifying extra clustering variables

EG - complements Oxford analysis - will discuss with Wahiba in more detail

JB - what is the truth? Can we compare to observational precip estimates?

RR - difficult as ICON is free running, not simulating particular weather events

 

Edward Groot

 

Assessment of dynamical tendencies from different models

  • High correlation between dynamic tendencies of temperature for GFS, RAP, IFS. Space ARPEGE slightly different – perhaps due to cloud initialisation
  • q tendencies similar across the same three models
  • U, V similar for IFS and ARPEGE while GFS and RAP look more different - smoother. 
  • This picture consistent across many columns

 
RR - happy to re-run ARPEGE for a few days with different initialisation to test if this is the source of the difference. How are other models initialised?

EG - IFS has no way of initialising ql and qi - they develop instantaneously

RR - if you don’t have these fields specified at initialisation, energetically inconsistent with ICON

RR - surprised that the models are so similar! Dynamical tendency will include impact of vertical advection, which will differ due to different vertical profiles

 
Alternative surface forcing:

  • New dataset available with IFS forced using SST+ surface fluxes
  • High correlations in tendencies between this data and original SST forcing.

 
RR - could be good to estimate ICON ‘dynamical tendencies’ by combining vertical advection with horizontal advection

 
Outlook

  • Papers (one or perhaps 2) being written up on CAPE work.
  • EGU talk planned

 

Hannah Christensen

 

Overview paper

  • Would like to write up overview paper to be submitted by October
  • Including protocol and overview of possible analysis

 
Action: HC to write skeleton and distribute before next meeting, for feedback

 

AOB

 

HL - CFMIP workshop in Exeter, 7-10 July

HC - Global hack-a-palooza on validation of k-scale simulations. UK node in Oxford, May 12th-16th