Download presentation
Presentation is loading. Please wait.
Published byColin Cain Modified over 8 years ago
1
MERRA reanalysis budget lessons about GCM errors in the W. Atlantic / IAS region Brian Mapes Univ. of Miami
2
MERRA reanalysis MERRA = Modern Era (from 1979) Reanalysis for Research and Applications Uses GEOS-5 GCM (formerly NSIPP) ½ x 2/3 deg resolution 3D state estimation process Budgets balance exactly albeit with “analysis tendencies” built by Incremental Analysis Updating (IAU)
3
some analyzed values Z at some point time (6h increments) ΔZ/Δt = Ż model + Ż ana ΔZ/Δt = (Ż dyn + Ż phys ) + Ż ana MERRA’s variables Z [T,u,v,q v ] satisfy: Oops -- missed by E Z ! Adjust Ż ana by -2E Z /Δt and try again from t-Δt/2 E Z oops Cost: 1.5x free model run Ż ana (t) saved as a data set EZEZ
4
Learning from analysis tendencies (ΔZ/Δt) obs = (Ż dyn + Ż phys ) + Ż ana If state is kept accurate (LS flow & gradients), then (ΔZ/Δt) obs and advective terms Ż dyn will be accurate and thus Ż ana ≅ -(error in Ż phys ) ✔ ✔
5
Vertically integrated qv tendencies: mst = moist physics (convection + LS condens.) ana = analysis tendency (~ –phys. error) Summer climatologies:
7
Analysis is moistening this area, presumably because physics is an excessive sink (raining too much)
8
June, July similar: analysis steadily adding qv (physics raining it out too much?)
9
Precipitation bias
12
rainy bias in all reanalyses
14
Vert. int. of analysis tendency of Very similar pattern in Jun, Jul (A map of radiosonde T biases?)
15
T tendencies make more sense vertically resolved T is a structural variable (mass field in geostrophic balance) Also have analysis tendencies of u,v Are the analysis increments of T and wind dynamically related?
16
Dynamical side: U analysis tendencies overlaid with {u,v} vectors 1000mb 850 700 500
17
U analysis tendencies overlaid with {u,v} vectors 1000mb 850 700 500 60-90W Zonal mean 60-90W Zonal mean 60-90W Zonal mean 60-90W Zonal mean
18
Analysis tends. on mass and wind fields: fighting the model’s excessive baroclinic vortex Cyclonic analysis torque: Anticyclonic analysis torque Lower thickness wanted for thermal wind balance Higher thickness wanted
19
Analysis tends. on mass and wind fields: fighting model’s excessive baroclinic vortex Cyclonic analysis torque: Anticyclonic analysis torque
20
Way too much: analysis opposes mean deep convective heating (with this model’s peculiar characteristic profile) open contours: model physics heating
21
GEOS-5 model’s peculiar heating profile in tropical deep convection Strange cooling spike at 700: misplaced melting (should be 550 mb) dT/dt_mst in TOGA COARE 15 day time section at a grid point
22
largest value on globe: >3 m/s per day u deceleration by analysis tendency (obs fighting model) largest value on globe: >3 m/s per day u deceleration by analysis tendency (obs fighting model)
23
warm SST season (later than summer)
24
Conclusions Analysis tendencies ~ -model systematic errors – Obs chronically nudging model a certain way MERRA W. Atl. has too much deep convection – excessive precip relative to GPCP/CMAP – excessive qv sink (fought by pos. qv anal. tendency) – excessive baroclinic vorticity couplet 600mb/200mb – includes largest u200 analysis tendency on the planet – seasonality: late summer (warm SST season) – T’ errors in thermal wind balanced sense w/ above – Just excessive strength: opposes mean moist Q – profile is distinctive, erroneous: melting at 700 mb instead of 550 Familiar: shallow v. deep convection problem
25
Thank you http://www.ncdc.noaa.gov/gibbs/image/GOE-12/WV/2006-07-14-15
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.