A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L Variational data assimilation and forecast error statistics “All models are wrong …” (George Box) “All models are wrong and all observations are inaccurate” (a data assimilator) Ross Bannister, 11th July 2011 University of Reading, r.n.bannister@reading.ac.uk A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L Distinction between ‘errors’ and ‘error statistics’ When people say ‘errors’ they sometimes (but not always) mean ‘error statistics’ Error: The difference between some estimated/measured quantity and its true value. E.g. εest = xest – xtrue or εy = y – ytrue Errors are unknown and unknowable quantities Error Statistics: Some useful measure of the possible values that ε could have. E.g. a PDF PDF(ε) σ= √<ε2> E.g. second moment of ε, <ε2> (called a variance), or <ε2>1/2 = σ (standard deviation). If only the variance is known, then the PDF is approximated as a Gaussian. P(ε) ~ exp – ε2/2<ε2> ε ε Error statistics are knowable, although often difficult to determine – even in the Gaussian case. Here, error statistics = second moment (ie assume PDFs are Gaussian and unbiased, <ε> = 0). A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L Plan A. Erroneous quantities G. Flavours of variational data assimilation B. Error statistics are important H. Control variable transforms C. ‘Observation’ and ‘state’ vectors I. The ‘BLUE’ formula D. ‘Inner’ and ‘outer’ products J. A single observation example E. Error covariances K. Forecast error covariance statistics F. Bayes’ theorem L. Hybrid data assimilation A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L A. Erroneous quantities in data assimilation Information available about the system before observations are considered. Data that are being fitted to • Observations (real-world data) • Prior data for the system’s state • Prior data for any unknown parameters } y H approx (x true ) model y Imperfections in the assimilation system • Model error within the da system • Representivity Data that have been fitted • Data assimilation-fitted data (analysis, ie posteriori error statistics) y H exact (x true ) instrument (x true x resolved x unresolved ) true true y H exact (x resolved x unresolved ) instrument true true unresolved y H exact (x resolved ) Hx instrument true true H exact (x resolved ) representivity instrument true y representivity instrument A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L B.1 Error statistics are important 1. Error statistics give a measure of confidence in data (eg here obs error stats) No assim Assim with large obs errors Assim with small obs errors A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L B.2 How are error statistics important? 2. Error statistics of prior data imply relationships between variables time x2 Background forecast (no assim) Analysis forecast (consistent with prior and ob errors) x1 x1 and x2 cannot be varied independently by the assimilation here because of the shape of the prior joint PDF. • Known relationships between variables are often exploited to gain knowledge of the complicated nature of the prior error statistics (e.g. changes in pressure are associated with changes in wind in the mid-latitude atmosphere (geostrophic balance)). A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L C. ‘Observation’ and ‘state’ vectors y = x = extra parameters The observation vector – comprising each observation made. There are p observations. The structure of the state vector (for the example of meteorological fields u, v, θ, p, q are 3-D fields; λ, φ and ℓ are longitude, latitude and vertical level). There are n elements in total. A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L D. ‘Inner’ and ‘outer’ products The inner product (‘scalar’ product) between two vectors gives a scalar (a1 a2 an ) b1 b1 aTb a b a1b1 a2b2 anbn b n a, b n , a T b 1 The outer product between two vectors gives a matrix a1 (b1 b2 bn ) a1b1 a a b abT 2 2 1 a a b m m 1 a1b2 a 2 b2 a m b2 a1bn a 2 bn a m bn a m , b n , abT mn When ε is a vector of errors, <εεT> is a (symmetric) error covariance matrix A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L E. Forms of (Gaussian) error covariances The one-variable case P(x) 2 P( ) exp 2 2 2 2 1 P( x) 1 2 2 exp ( x x )2 σ= √<ε2> x xtrue x x (unbiased) 2 2 <x> x The many variable case P ( x) 1 1 exp (x x ) T S 1 (x x ) 2 2 n S ε x x true x x (unbiased) 12 S 2 3 1 2 22 P(x) 0 x1 S is a genericsymbol- specificsymbols will be used later x2 A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L F. Bayes’ theorem and the variational cost function Bayes theorem links the following • PDF of the observations (given the truth) • PDF of the prior information (the background state) • PDF of the state (given the observations – this is the objective of data assimilation) P(x | y ) 1 A P ( y | x) P ( x) 1 1 1A exp (y h(x))T R 1 (y h(x)) exp (x x b ) T Pf-1 (x x b ) 2 2 1 1 1A exp (y h(x))T R 1 (y h(x)) (x x b ) T Pf-1 (x x b ) 2 2 1 1 ln P(x | y ) ln A (y h(x))T R 1 (y h(x)) (x x b ) T Pf-1 (x x b ) 2 2 maximizeP(x | y ) minimize ln P(x | y ) minimizeJ (x) J ( x) 1 1 (x x b ) T Pf-1 (x x b ) (y h(x))T R 1 (y h(x)) 2 2 Lorenc A.C., Analysis methods for numerical weather prediction, Q.J.R.Meteor.Soc. 112 pp.1177-1194 (1986) A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L G.1 Flavours of variational data assimilation 1. Weak constraint 4D-VAR General form J ( x) 1 (x(0) x b (0))T Pf-1 (x(0) x b (0)) 2 1 T (y (t ) h t [x(t )])T R t1 (y (t ) h t [x(t )]) 2 t 0 1 T T (x(t ) m t t 1[x(t 1)])T (Q 1 )t ,t ' (x(t ' ) m t 't '1[x(t '1)]) 2 t 1 t '1 x ( 0) x ( 1 ) x x(T ) Q : model error covariance matrix Simplified form : assume model errors are uncorrelated in time (white noise) (Q 1 ) t ,t ' tt 'Q t1 last termbecomes 1 T (x(t ) m t t 1[x(t 1)])T Q 1 (x(t ) m t t 1[x(t 1)]) 2 t 1 A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L G.2 Flavours of variational data assimilation 2. Strong constraint 4D-VAR Assume perfect model: Q0 x(t ) mt t 1[x(t 1)] need to determinex(0) only x x(0) J ( x) 1 (x x b ) T Pf-1 (x x b ) 2 1 T (y (t ) h t [m t 0 (x)])T R t1 (y (t ) h t [m t 0 (x)]) 2 t 0 The ‘strong constraint’ approximation is valid if model error is negligible x x(0) A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L G.3 Flavours of variational data assimilation Linearisation of m 2. Incremental 4D-VAR (quasi-linear) Linearisation of h remember: x x(0) Let x x ref x x(t ) m t 0 (x) m t 0 (x ref x) m t 0 (x ref ) M t 0x M t 0 x(t ) x x ref Let x(t ) x ref (t ) x(t ) y model (t ) h t [x(t )] h t [x ref (t ) x(t )] h t [x ref (t )] H tx(t ) y model (t ) Ht x(t ) x ref (t ) Apply to strong constraint cost function Let x b x ref x b Let y (t ) y (t ) h t [m t 0 (x ref )] 1 1 J (x) (x x b ) T Pf-1 () J (x) (x x b ) T Pf-1 () 2 2 1 T 1 T ref T 1 T 1 ( y ( t ) h [ m ( x )] H M x ) R ( ) ( y ( t ) H M x ) R () t t 0 t t 0 t t t 0 t 2 t 0 2 t 0 The incremental form is exactly linear and is valid if non-linearity is ‘weak’ A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L H. Control variable transforms 1 J (x) (x x b ) T B -1 () 2 where x x x ref , x b x b x ref , 1 T (y (t ) H t M t 0x) T R t1 (), 2 t 0 y (t ) y (t ) h t [m t 0 (x ref )] herex is the' controlvariable' Let x U , x b U b 1 J ( ) ( b ) T U T B -1U () 2 here is the' cont rolvariable' 1 T (y (t ) H t M t 0 U ) T R t1 () 2 t 0 Design U such thatUTB-1U I 1 1 T T J ( ) ( b ) () (y (t ) H t M t 0 U ) T R t1 () 2 2 t 0 Advantages of the new control variable: • No B -matrix to worry about • Better conditioning of the problem In practice we design U and B follows (the implied B = U U T) The design of U is called ‘background error covariance modelling’ B nn A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L I. The cost function and the ‘BLUE’ formula 1 1 (x x b ) T Pf-1 (x x b ) (y h(x))T R 1 (y h(x)) 2 2 1 1 (x x b ) T Pf-1 (x x b ) (y h(x b ) H (x x b ))T R 1 (y h(x b ) H (x x b )) 2 2 x J Pf-1 (x x b ) H T R 1 (y h(x b ) H (x x b )) J ( x) x J 0 at min of cost function(x x a ) Pf-1 (x a x b ) H T R 1 (y h(x b ) H (x a x b )) 0 (Pf-1 H T R 1 H )(x a x b ) H T R 1 (y h(x b )) 0 x a x b (Pf-1 H T R 1 H ) 1 H T R 1 (y h(x b )) (the' information form') Using theSherman - Morris- Woodburyformula(Pf-1 H T R 1 H )Pf H T H T R 1 (R HPf H T ) we get x a x b Pf H T (R HPf H T ) 1 (y h(x b )) (the' covarianceform') A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L J. A single observation example Analysis increment of the assimilation of a direct observation of one variable. x a x b Pf H T (R HPf H T ) 1 (y h(x b )) 0 1 Pf 0 0 ( 0 1 0 0) 2 Pf y 1 0 1 0 ( y x b (r )) 0 0 1 Pf [ y2 Pf (r , r )]1 ( y x b (r )) 0 0 y x b (r ) x a (r ' ) x b (r ' ) Pf (r ' , r ) 2 y Pf (r , r ) Obs of atmospheric pressure → A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L K.1 Forecast error covariance statistics • In data assimilation prior information often comes from a forecast. • Forecast error covariance statistics (Pf) specify how the forecast might be in error εf = xf – xtrue, Pf = <εf εfT>. • How could Pf be estimated for use in data assimilation? • • • • Analysis of innovations (*). Differences of varying length forecasts (‘NMC method’). Monte-Carlo method (*). Forecast time lags. • Problems with the above methods. 11 1 2 1 3 2 3 Pf or B 2 1 2 2 3 3 3 1 3 2 • A climatological average forecast error covariance matrix is called ‘B’. A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L K.2 Analysis of innovations We don’t know the truth, but we do have observations of the truth with known error statistics. Definition of observation error : y = ytrue + εy = h(xtrue) + εy Definition of forecast error : xtrue = xf – εf Eliminate xtrue : y = h(xf – εf) + εy ≈ h(xf ) - Hεf + εy ‘Innovation’ : y - h(xf ) ≈ εy - Hεf LHS (known), RHS(unknown) Take pairs of in-situ obs whose errors are uncorrelated (for variable v1, posn r and v2, r+Δr) y(v1,r) - xf (v1,r) ≈ εy(v1,r) - εf(v1,r) y(v2,r +Δr) - xf (v2,r +Δr) ≈ εy(v2,r +Δr) - εf(v2,r +Δr) Covariances <[y(v1,r) - xf (v1,r)] [y(v2,r +Δr) - xf (v2,r +Δr)]> = <[εy(v1,r) - εf(v1,r)] [εy(v2,r +Δr) - εf(v2,r +Δr)]> = <εy(v1,r) εy(v2,r +Δr)> - <εy(v1,r) εf(v2,r +Δr)> - <εf(v1,r ) εy(v2,r +Δr)> + <εf(v1,r) εf(v2,r +Δr)> ↑ ↑ ↑ ↑ Obs error covariance Zero (obs and forecast errors Forecast error covariance between (v1, r) and (v2, r+Δr) uncorrelated) between (v1, r) and (v2, r+Δr) zero unless v1=v2 and Δr=0 (one particular matrix element of Pf or B) <> average over available observations and sample population of forecasts A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L K.3 Monte-Carlo method (ensembles) • N members of an ensemble of analyses. • Leads to N members of an ensemble of forecasts. • The ensemble must capture the errors contributing to forecast errors. • Initial condition errors (forecast/observation/assimilation errors from previous data assimilation). Ensembles • Model formulation errors (finite resolution, unknown x parameters, …). • Unknown forcing. • Can be used to estimate the forecast error covariance matrix, e.g. • Pf ≈ < (x-<x>) (x-<x>) T > = 1/(N-1) ∑i=1,N (xi - <x>) (xi - <x>)T • Problem: for some applications N << n. • n elements of the state vector (in Meteorology can be 107). • N ensemble members (typically 102). • Consequence – when Pf acts on a vector, the result is forced to lie in the sub-space spanned by the N ensemble members. 1 N 1 N T Pf x i i x i ( iT x) N 1 i 1 N 1 i 1 t A B1 B2 C D E F G1 G2 G3 H I J K1 K2 K3 L L. Hybrid data assimilation Variational data assimilation (with B-matrix) •B is full-rank •B is static Ensemble data assimilation •Sample Pf is flow dependent •Sample Pf is low rank Suppose x V U V V , x Vb U V bV , x E U E E , x Eb U E bE x x V x E J ( ) 1 ( V bV ) T () 2 1 ( E bE ) T () 2 1 T (y (t ) H t M t 0 [U V V U E E ])T R t1 () 2 t 0 V E , V n , E N , n N U V U(as in standardincremental variational data assimilation) U E x1 x x N x Comments and Questions ! ?