icebreaker5221 Posted May 21, 2011 Share Posted May 21, 2011 It looks like the operational ECMWF recieved a nice upgrade this week. I know a little bit about data assimilation, and it sounds like they're upgrading the 4d-var background error covarience matrices and making them flow-dependent. They have also reduced AMSU-A observation errors, so I'm assuming AMSU data has proven sufficiently reliable that they can weigh the data more heavily. This combined with the aforementioned improvement should be particularly useful in allowing ECMWF to better initialize low-data regions. There is also an upgrade to the physics (cloud scheme). Per ECMWF verification, the upgrades result in statistically-significant improvements to model verification in both the mid-latitudes and the tropics at all forecast times. Full release: http://www.ecmwf.int...t/ecmwfnews/276 A new version of the ECMWF forecasting and analysis system, Cycle 37r2, was implemented on 18 May 2011. The new cycle includes both meteorological and technical changes. The main meteorological changes included in this cycle are the use of flow-dependent background errors in the data assimilation, reduced AMSU-A observation errors and improvements to cloud scheme. The changes show clear benefits in terms of objective upper-air scores in the medium range. Temperature and winds are improved throughout the troposphere in the extra-tropics. The improvements to the cloud scheme increase the humidity in the upper troposphere, providing a better fit to observations and improving humidity scores in the extra-tropics in the early forecast range. The impact of Cycle 37r2 is shown in the figure below. Impact of Cycle 37r2. The figure shows the normalised difference in anomaly correlation of 500 hPa geopotential height (bold blue lines) between Cycle 37r2 and the previous operational cycle (36r4) together with the 95% confidence interval (light blue bars) for the Northern and Southern hemispheres and Europe. The plot for the tropics shows normalized difference in root-mean-square error of 850 hPa vector wind. The positive impact of Cycle 37r2 is important in all areas and at almost all forecast ranges. Link to comment Share on other sites More sharing options...
baroclinic_instability Posted May 22, 2011 Share Posted May 22, 2011 It looks like the operational ECMWF recieved a nice upgrade this week. I know a little bit about data assimilation, and it sounds like they're upgrading the 4d-var background error covarience matrices and making them flow-dependent. They have also reduced AMSU-A observation errors, so I'm assuming AMSU data has proven sufficiently reliable that they can weigh the data more heavily. This combined with the aforementioned improvement should be particularly useful in allowing ECMWF to better initialize low-data regions. There is also an upgrade to the physics (cloud scheme). Per ECMWF verification, the upgrades result in statistically-significant improvements to model verification in both the mid-latitudes and the tropics at all forecast times. Full release: http://www.ecmwf.int...t/ecmwfnews/276 A new version of the ECMWF forecasting and analysis system, Cycle 37r2, was implemented on 18 May 2011. The new cycle includes both meteorological and technical changes. The main meteorological changes included in this cycle are the use of flow-dependent background errors in the data assimilation, reduced AMSU-A observation errors and improvements to cloud scheme. The changes show clear benefits in terms of objective upper-air scores in the medium range. Temperature and winds are improved throughout the troposphere in the extra-tropics. The improvements to the cloud scheme increase the humidity in the upper troposphere, providing a better fit to observations and improving humidity scores in the extra-tropics in the early forecast range. The impact of Cycle 37r2 is shown in the figure below. Impact of Cycle 37r2. The figure shows the normalised difference in anomaly correlation of 500 hPa geopotential height (bold blue lines) between Cycle 37r2 and the previous operational cycle (36r4) together with the 95% confidence interval (light blue bars) for the Northern and Southern hemispheres and Europe. The plot for the tropics shows normalized difference in root-mean-square error of 850 hPa vector wind. The positive impact of Cycle 37r2 is important in all areas and at almost all forecast ranges. I saw this earlier on their website too. Good stuff here--a nice improvement to the DA system. Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.