Update on seasonal forecast system based on SLAV
Update on seasonal forecast system based on SL-AV model at Hydrometcentre of Russia. Decadal prediction plans. M. A. Tolstykh (2, 1) and D. B. Kiktev (1), R. B. Zaripov (1), E. N. Kruglova (1), I. A. Kulikova (1) Hydrometcentre of Russia (1); Institute of Numerical Matematics Russian Academy of Sciences (2) 2015
Plan n n Seasonal forecast system at HMCR Plans for running decadal forecasts at INM RAS
Seasonal version of SL-AV model n n Semi-implicit semi-Lagrangian vorticity-divergence dynamical core of own development (Tolstykh 2010), mostly ALADIN/LACE parameterizations. Current version: Resolution 1. 4 x 1. 125 degrees lon/lat, 28 levels. Described in (Tolstykh et al, Izvestia RAN, Ser. Ph. A&O, 2010) with updates in (Tolstykh et al Izvestia RAN, 2014) n Verification of seasonal forecasts (also for MGO (St. Petersbourg) forecasts) http: //neacc. meteoinfo. ru/verificationforecast
Running forecasts n n n d 0 -63: 1. 1 x 1. 4 L 28 once a week, 20 member ensemble initialized on 00 Z every Wednesday forced py persisted SST anomalies (mean for 2 weeks) from NCEP (Reynolds SST OI v 2). Perturbation from a breeding cycle. Reforecast suite with 10 members spanning 30 years (1981 -2010) run in real-time. m 0 -4: Forecast suite is the same as d 0 -63, but forecast lead time is 4 months. Runs on the last Wednesday of a month. Re-forecast suite with 10 members spanning 30 years (1981 -2010) 00 Z and 12 Z 26 -30 of each month (24 -28 for February). Input to APCC, WMO LC LRF multi-model ensemble. Forecasts are also available at North Eurasia Climate Centre site (neacc. meteoinfo. ru)
Scores and forecasts at NEACC site Skill scores: ROC_A - ROC Score Above Normal ROC_N - ROC Score Near Normal ROC_B - ROC Score Below Normal RO - sign consistency coefficient ACC - anomaly correlation coefficient RMSE - root mean square error Some plots from WMO LC are also available
Activities since 16 th session n n n NEACC site (forecasts, diagnostics – scores, indices, …) Implementing CHFP and S 2 S output standards. New snow albedo parameterization in SL-AV, cloudiness retuning. Moving from NCEP 2 to ERA Interim data for hindcasts. Too fast transition to new technology -> serious bug in operational seasonal forecasts in January Corrected by the end of February
Comparison of hindcasts errors for 2011 and 2014 versions. RMSE averaged over 4 seasons and 28 years (1983 -2010), for H 500(dam) , MSLP (h. Pa), T 850, Т 2 м, H 200 (dam) with respect to reanalysis. S 20 - (90 - 20 S), N 20 – (20 -90 N), TR – (20 S-20 N).
New version of the model n n More accurate SW and LW radiation parameterization(CLIRAD SW + RRTM LW). ALARO-0 microphysics Increased horizontal resolution (0. 72°x 0. 9° lat-lon) Promising preliminary results (especially in dynamics), poor precip so far, tuning is underway
Plans for decadal forecasts at INM RAS n n n n INMCM 5 with improvements - higher model top, better stratosphere resolution, - new cloudiness and condensation parameterizations, - aerosols description. Two resolutions considered: `Optimistic’: Atm. 2˚x 1. 5˚L 73, Oc. 0. 5˚x 0. 25˚L 40 (twice higher ocean horizontal resolution than in CMIP 5) Atm. 2˚x 1. 5˚L 21, Oc. 1˚x 0. 5˚L 40
Thank you for attention!
- Slides: 10