A few examples Shallow water equation derivation and solutions. Goal: Develop the mathematical foundation of tropical waves

Transcription

1 A few examples Shallow water equation derivation and solutions Goal: Develop the mathematical foundation of tropical waves

2 Previously: MCS Hovmoller Propagating w/ wave velocity From Chen et al (1996)

3 Precipitable Water Anomalies

4 Kelvin Wave

5 Rossby Wave

6 Dry adiabatic primitive equations Recall from earlier lectures: Vector momentum equation in rotating coordinates d v! dt = "2 #! $ v! " 1 % &p + g! + F! r Continuity: " #! v = 0 To derive mathematical expressions of the various types of large-scale tropical waves, we invoke a simplified form of these equations the shallow water equations [SWE]. The approach here follows Matsuno 1966.

7 SWE: Principal Assumptions The fluid can be approximated by two layers, both of which are horizontally and vertically homogeneous at rest. In the upper layer, both pressure and density are horizontally invariant; in the lower layer, only density is constant. The fluid is hydrostatic.

8 From the hydrostatic assumption: SWE Derivation (I)! p!z =!"g Taking the horizontal gradient of the hydrostatic equation in either layer gives: "! p %! H $ ' =! H #!z & ((!g) = 0 =! (!z! H p) with the last equality following from the horizontal invariance of density in either layer. The last equality implies that the lower level horizontal pressure gradient is also invariant with height. It is convenient to replace pressure by the depth of fluid, h, in the lower layer. That is: h h h!! p! (" H p)dz = " H!!z!z dz = #" H! "gdz = #"g" H h 0 0 0

9 SWE Derivation (II) Consider the set-up as illustrated schematically on the right. Here, we take h(x,y,t) = H + η(x,y,t). We need to estimate horizontal variations in pressure in the lower layer. h H ρ u ρ l p 1 2 δx p δz For points 1 and 2 we have [considering, w/o loss of generality, x only]: p l1 = p +! p 1 = p + " u g!z Thus: p l2 = p +! p 2 = p + " l g!z #( p +! lim l g!z) " (p +! u g!z) & % ( = g! l "! u!x!0$!x ' ( ) lim The horizontal momentum equations in the lower layer become:!u!t + v! H! " H u # fv = # 1 " p l! l "x = # g $ "# "x!v!t + v! H! " H v + fu = # 1 " p l! l "y = # g $ "# "y!x!0 (!h /!x)!x!x = g(! l "! u )!h!x Here, gʹ is reduced gravity: g " = g # $ # l u # l

10 SWE Derivation (III) From the continuity equation:!w!z =!" #! H v H Integrating from the surface up to h, and noting that the horizontal velocity is independent of height [since the pressure gradient is height independent] yields: h h!w " dz = w(h)! w(0) =! " # H $ v H dz =! h# H $ v H!z 0 The vertical velocity vanishes at the surface. On the other hand, w(h) is the rate at which the interfacial height changes, i.e., # w(h) =!!t + v! & % H! " H (h $ ' Thus, using h(x,y,t) = H + η(x,y,t), gives:!"!t =!! v H " # H!! (H +!)# H "! v H 0

11 Linearizing the SWE We want to solve the SWE in perturbation form, i.e., the wave solutions are calculated with respect to a background state (U, V, and Ν): # u& # U + u )& % ( v % ( = % ( V + v ) % ( $ "' $ * + ")' For a motionless background state (U=V=Ν=0) and ignoring terms in quadratic in perturbation quantities gives: " u # "t = f v # $ "%# "x " v # "t = $ f u # $ "%# Here, φʹ is: #" = g " h " "y!"!!t = " g! H# H $ v! H! A further assumption, appropriate close to the equator, is the equatorial betaplane approximation, i.e., we can write f in Cartesian coordinates as: f " #y; # = 2$ R e

12 Solving the linearized SWE (I) Consider separable solutions of the form: $ u "' * u ˆ (y)- & ) v " & ) =, ˆ /, v (y) / %#"( +, ˆ #(y)./ e i(kx01t ) Then: "i#ˆ u = $yˆ v " ik ˆ % "i#ˆ v = "$yˆ u " % ˆ & %y "i# ˆ ' $ = " g % H ikˆ u + &ˆ v ) ( &y *, + Solving the first equation above for u ˆ and substituting into the remaining two, and then eliminating ˆ ", yields a single equation in v ˆ : " 2 v ˆ "y + -' # 2 ) 2 g $ H % k 2 % k ( # & * +, % & 2 y 2 /. g $ H 0 2 v ˆ = 0 1

13 Solving the linearized SWE (II) " 2 v ˆ "y + -' # 2 ) 2 g $ H % k 2 % k ( # & * +, % & 2 y 2 /. g $ H 0 2 v ˆ = 0 1 We require the above equation to have solutions vanishing as y. This requirement follows from the equatorial beta-plane approximation which is not valid at high latitudes, so solutions must be equatorially trapped. For decaying solutions, the following meridional dispersion relationship must hold: " g H # & $ 2 ( g " H % k 2 % k ' $ # ) + = 2l +1; l = 0,1,2,... l is the meridional * wavenumber The general form of the solutions is: ˆ v (") = v 0 H n (")e #" 2 / 2 ; " = & ( ' $ % g H ) + * 1/ 2 y H n (ξ) is the n th Hermite polynomial. The first three Hermite polynomials are: H 0 =1; H 1 =2ξ; H 2 =4ξ 2-2

14 Gravity/Rossby waves Since the meridional dispersion relationship is cubic in ω, it has 3 solutions. These are interpreted as eastward and westward propagating equatorially trapped inertiagravity waves and westward propagating Rossby waves. The case for l=0, i.e., v is a Gaussian centered on the equator, must be treated separately from l > 0. In particular, the dispersion relationship can be written as: Inertia-gravity Rossby-gravity ω & ( ' " g # H $ % " $ k )& + ( *' " g # H + k ) + = 0 * Rossby westward eastward k The root coming from the second term in parentheses is not permitted, as this term is required not to vanish by the derivation. The remaining two roots correspond to an eastward propagating gravity wave and a westward propagating mode that asymptotes to Rossby wave-like behavior in the short wavenumber limit.

15 Equatorial Kelvin wave (I) The equatorial Kelvin mode has no meridional velocity perturbations, so "i#ˆ u = "ik ˆ $ 0 = "#yˆ u " $ ˆ % $y "i# ˆ $ = " g % H(ikˆ u ) Eliminating ˆ " between the first and third equations gives: (" 2 # g $ Hk 2 ) u ˆ = 0 % " 2 k = c 2 = g $ H 2 On the other hand, eliminating ˆ " between the first and second equations gives: 0 = "#yˆ u " c $ˆ u $y Integrating this equation gives: u ˆ = u 0 e ("#y 2 / 2c) Note that for this solution to decay away from the equator, c>0, i.e., the Kelvin wave mode propagates eastward.

16 Equatorial Kelvin wave (II) The structure of the equatorial Kelvin mode looks like: The e-folding decay width of the Kelvin wave structure is given by: For c = 30 ms -1, Y K = 1600 km. # Y K = 2c & % ( $ " ' 1/ 2 Contours show lower tropospheric pressure with positive (negative) anomalies denoted by solid (dashed) lines. The contour interval is onefourth the maximum amplitude of the anomaly, and the zero contour is not shown. Anomalies of convergence (divergence) that are greater than two-thirds the maximum amplitude are shaded dark (light) gray. From Majda & Stechmann In the Kelvin wave, the meridional force balance is exact geostrophic balance between the meridional pressure gradient and zonal velocity; the equatorial Kelvin mode owes its existence to the change in sign of the Coriolis force at the equator.