Data assimilation and forecasting the weather (!)
Eugenia Kalnay
and many friends
University of Maryland

Content
 Forecasting the weather - we are really getting betterÉ
 Why: Better obs? Better models? Better data assimilation?
 Intro to data assim: a toy example, we measure radiance and we want an accurate temperature
 Comparison of the toy and the real equations
 An example from JMA comparing 4D-Var and LETKF (a type of Ensemble Kalman Filter)

Typical 6-hour analysis cycle
The observing system a few years agoÉ
Typical distribution of observations in +/- 3hours
Typical distribution of observations in +/- 3hours
Model grid points (uniformly distributed) and observations (randomly distributed). For the grid point i only observations within a radius of influence may be considered
Some statisticsÉ
Some comparisonsÉ
Slide 10
Comparisons of Northern and Southern Hemispheres
Satellite radiances are essential in the SH
More and more satellite radiancesÉ
Intro. to data assimilation: a toy example
 Assume we have an object, a stone in space
 We want to estimate its temperature T (oK) accurately by measuring the radiance y (W/m2)
that it emits. We have an observation model:
 We also have a forecast model for the temperature
 We will derive the data assim eqs (KF and Var) for this toy system (easy to understand!)
 Will compare the toy and the real huge vector/matrix equations: they are the same!

Toy temperature data assimilation, measure radiance
Toy temperature data assimilation, measure radiance
Toy temperature data assimilation, measure radiance
Toy temperature data assimilation, variational approach
Toy temperature data assimilation, variational approach
Typical 6-hour analysis cycle
Toy temperature analysis cycle (Kalman Filter)
Toy temperature analysis cycle (Kalman Filter)
Summary of toy system equations (for a scalar)
Summary of toy system equations (cont.)
Summary of toy system equations (cont.)
Equations for toy and real huge systems
Interpretation of the NWP system of equations
Interpretation of the NWP system of equations
Summary of NWP equations (cont.)
Comparison of 4-D Var and LETKF at JMA
T. Miyoshi and Y. Sato
Comparison of 4-D Var and LETKF at JMA
T. Miyoshi and Y. Sato
Comparison of 4-D Var and LETKF at JMA
T. Miyoshi and Y. Sato
Comparison of 4-D Var and LETKF at JMA
18th typhoon in 2004, IC 12Z 8 August 2004
T. Miyoshi and Y. Sato
Comparison of 4-D Var and LETKF at JMA
RMS error statistics for all typhoons in August 2004
T. Miyoshi and Y. Sato
Summary
Data assimilation methods have contributed much to the improvements in NWP.
A toy example is easy to understand, and the equations are the same for a realistic system
Kalman Filter (too costly) and 4D-Var (complicated) solve the same problem (if model is linear and we use long assimilation windows)
Ensemble Kalman Filter is feasible and simple
It is starting to catch up with operational 4D-Var
Important problems: estimate and correct model errors & obs. errors, optimal obs. types and locations, tuning additive/multiplicative inflation, parameters estimation,É
Tellus: 4D-Var or EnKF? In press
Workshop in Buenos Aires NovÕ08