Participants: Margarita Choulga , Gianpaolo Balsamo 

Shared with: Christel Prudhomme , Christoph Rüdiger , Vincent Fortin (ECCC)


Tasks discussed:

  1. Target period 2018-2022; target timing 24-hour as starting point, and 6-hour as the target.
  2. Confluence GloPA page. Daniel Varela Santoalla to tell if possible to create a confluence page for the GloPA (Global Precipitation Analysis) pilot project. 
  3. Margarita started to look at the GRIDPP (GitHub - metno/gridpp: Software to post-process gridded weather forecasts) algorithm in its current implementation (found already a LSM Orography improvements). 
  4. The Leave-Out-Procedure needs to re-run for new observations or new algos. There are 3 parameters to calibrate, the horizontal decorrelation, the vertical decorrelation, and the ratio of sigma_o/sigma_b (telling how much to believe the model and how much to believe in the obs). - 10x10 degree per month leaving 20% of stations out to calibrate/validate
  5. Collecting key Precipitation datasets (PARDIG-LR and NRT, IMERG, CaPA, EFAS-TP, MSWEP2, MSWEP3) for needed hours and period.
    1. https://developers.google.com/earth-engine/datasets/catalog/projects_gcp-public-data-weathernext_assets_PARDIG_1_LR
    2. https://developers.google.com/earth-engine/datasets/catalog/projects_gcp-public-data-weathernext_assets_PARDIG_1_NRT
    3. https://developers.google.com/earth-engine/datasets/catalog/JAXA_GPM_L3_GSMaP_v8_operational
    4. https://developers.google.com/earth-engine/datasets/catalog/NOAA_CPC_Precipitation
  6. Margarita started to look into spatial parallelism to boost the speed and tailor the analysis to the observation density (e.g. 10x10 deg).
  7. CaPA use ground-based in situ plus satellite data product (IMERG version to be confirmed by Vincent). - share initial ideas with Vincent, ask his GitHub user for collaboration and code.
  8. Titanlib QC is currently switched off. Margarita to look into SMHI report. RMK_118  SMHI Gridded Climatology RMK_118 SMHI Gridded Climatology.pdf 
  9. Make the calibration part of the production? This is motivated by 2 things. The background errors are varying month by month and year by year. The observation density and the satellites are also not steady in time.
  • No labels