Introduction The analytical model A numerical application Summary Mean mass transport induced by internal Kelvin waves, with application to the circulation in the Van Mijen fjord in Svalbard. Eivind Støylen eivind.stoylen@geo.uio.no & Jan Erik Weber j.e.weber@geo.uio.no Department of Geosciences, MetOs section University of Oslo 14.October 2008 Introduction The analytical model Outline 1 Introduction Wave drift Internal Kelvin waves 2 The analytical model The geometry Linear theory Nonlinear theory 3 A numerical application The Van Mijen fjord 4 Summary Summary Further research A numerical application Summary Introduction The analytical model A numerical application Summary Wave drift Stokes (1847): A mean Stokes drift Ūs arise from propagating wave motion Movie: Wikipedia Longuet-Higgins (1953): The Lagrangian drift Ūl is the sum of Ūs and a viscositydependent Eulerian mean Ūe . Introduction The analytical model A numerical application Summary Internal Kelvin waves Propagating on a pycnocline along a boundary, due to Coriolis Amplitude decay exponentially outwards to the internal Rossby radius a of order 1-10 km Commonly observed in large stratified lakes/fjords Introduction The analytical model A numerical application Summary Previous work Wind-induced internal Kelvin wave in a large stratified lake: Csanady (1972): Derives a linear large-amplitude internal wave, the first to suggest significant mass transport near shore Wunsch (1973): From Csanady, derives Stokes drift in the linear wave. Eulerian mean is neglected Ou and Bennett (1979): Derives the mean second order drift, retaining the viscous Eulerian mean current. Comparison with observations in Lake Kinneret, Israel Gómez-giraldo et al. (2006): Field data and numerical model in Lake Kinneret. Verifies an internal Kelvin wave propagating around the basin Introduction The analytical model A numerical application Two-layer model Hydrostatic pressure distribution H2 much larger than H1 u2 , v2 negligible Ice lid; η negligible No-slip at z = H1 Variable under-ice pressure Ps Summary Introduction The analytical model A numerical application Summary Tidally driven internal Kelvin wave Monochromatic, constant frequency ω Coast at y = 0 Anticipate v1 = 0 everywhere Guess ξ˜ = Ae−αx−y/a eiψ Phase function: ψ = kx + ly − ωt a: internal Rossby radius α: horizontal damping coefficient Introduction The analytical model A numerical application Summary Linear theory Separate u1 in barotropic and boundary-layer part u1 = ũ1 + û1 . From first order Navier-Stokes and continuity: ξ˜ = Ae−αx−y/a cos ψ c1 A −αx−y/a ũ1 = − e cos ψ − sin ψ H1 2 −(1−i)(H1 −z)/δ û1 = <{−ũ1 e } Boundary-layer thickness δ = Small parameter ≡ δ/(2H1 ) α is of order p 2ν/ω H1 Introduction The analytical model A numerical application Summary Nonlinear theory Integrate; Z U= Second order in wave steepness: H1 u1 dz ξ Oscillatory and mean part; U = Ũ+ Ū, Ũ ∼ e (x) − τi )/ρ1 (y) − τi )/ρ1 −f V̄ − c21 ξ¯x = HS1 + (τs f Ū − c21 ξ¯y = HS2 + (τs −iωt Average over a wave period 2π/ω (x) (y) Ūx + V̄y = 0 HS1, HS2: Nonlinear wave forcing terms Introduction The analytical model A numerical application Summary Nonlinear theory Integrate; Z U= Second order in wave steepness: H1 u1 dz ξ Oscillatory and mean part; U = Ũ+ Ū, Ũ ∼ e (x) − τi )/ρ1 (y) − τi )/ρ1 −f V̄ − c21 ξ¯x = HS1 + (τs f Ū − c21 ξ¯y = HS2 + (τs −iωt Average over a wave period 2π/ω (x) (y) Ūx + V̄y = 0 HS1, HS2: Nonlinear wave forcing terms Introduction The analytical model A numerical application Summary Solve mean second-order equations From Longuet-Higgins (1953): Ū = Ūl = Ūs + Ūe Z Ūs = H1 u1 dt · u1x (V̄s = 0) Friction as drag on Eulerian current: (x) τs = −cd |Ūe |Ūe /H12 Eulerian equations −f V̄e − c21 ξ¯x = − ∂ 3 cd c1 Ūs − 2 Ūe2 ∂x 2 H1 f Ūe − c21 ξ¯y = 0 Ūex + V̄ey = − ∂ Ūs ∂x Introduction The analytical model A numerical application Summary Solve mean second-order equations From Longuet-Higgins (1953): Ū = Ūl = Ūs + Ūe Z Ūs = H1 u1 dt · u1x (V̄s = 0) Friction as drag on Eulerian current: (x) τs = −cd |Ūe |Ūe /H12 Eulerian equations −f V̄e − c21 ξ¯x = − ∂ 3 cd c1 Ūs − 2 Ūe2 ∂x 2 H1 f Ūe − c21 ξ¯y = 0 Ūex + V̄ey = − ∂ Ūs ∂x Introduction The analytical model A numerical application Solutions c1 A2 −2(αx+y/a) e , V̄s = 0 2H r 1 αH1 Ūe = c1 Ae−αx−y/a cd Ūs = At x = 0: ũ1 = 0.18 m/s ūl = 0.096 m/s v̄l = O(10−3 ) m/s r A αH1 −2y/a −2αx −y/a −αx V̄e = αac1 A (1 − e )e + (1 − e )e 2H1 cd r A αH1 ξ¯ = Ae−2αx + A(1 − e−y/a )e−αx 4H1 cd Summary Introduction The analytical model A numerical application The Van Mijen fjord Figure: Bergh (2004) Summary Introduction The analytical model A numerical application Summary Preliminary model result The internal Kelvin Wave is indeed present Propagating along the coast, damped by friction Large amplitudes at the sounds Figure: Interface amplitudes (Støylen 2008) Parameter T ∆ρ/ρ ∆x η0 H1 Value 119 hrs 0.001 200 m 52.2 cm 30 m Introduction The analytical model A numerical application Summary Drift pattern in the Van Mijen fjord? Eastward flow at southern coast Friction ⇒ flow to interior Return flow at northern wall Mass conservation Introduction The analytical model A numerical application Summary Internal Kelvin wave under ice in a two-layer system Integrated between material boundaries to retain Lagrangian properties Alongshore: Eulerian flux comparable to Stokes flux Normal to coast: Non-zero Eulerian flux due to friction Model results suggest a tidally driven internal Kelvin wave in the Van Mijen fjord Summary Introduction The analytical model A numerical application Further research Tidally driven model runs Barotropic response, interaction between the sounds Include wind stress in model Set-up of internal waves? Upwelling scenarios Include non-linear terms in the numerical model Further developing the analytical model Cooperation with Unis, Svalbard. Field measurements Summary Introduction The analytical model A numerical application Summary Thank you Mean mass transport induced by internal Kelvin waves, with application to the circulation in the Van Mijen fjord in Svalbard. Eivind Støylen eivind.stoylen@geo.uio.no & Jan Erik Weber j.e.weber@geo.uio.no Department of Geosciences, MetOs section University of Oslo 14.October 2008 Additional references Gómez-Giraldo, A., Imberger, J., Antenucci, J.P., 2006: Spatial structure of the dominant basin-scale internal waves in Lake Kinneret. Limnol. Oceanogr., 51, 229-246 Appendix: Parameters and values Choice of values for plot: A=10 m Parameters: c1 = p g0 H1 T=12.4 hrs f = 1.42e-4 s−1 k = ω/c1 H1 =30m a = c1 /f ∆ρ/ρ = 0.001 α = k/2 g = 9.81 m/s2 l = f /(2c1 ) ν = 3e-4 m2 /s cd = 10−3 Appendix: Some assumptions and definitions Navier-Stokes (N-S) in lower layer with large H2 and negligible η: 0 = −g0 ξ˜x − Psx /ρ1 , g0 ≡ g(ρ2 − ρ1 )/ρ1 Surface pressure gradient balanced by interface gradient. Upper layer N-S: u1t = g0 ξ˜x + ν∇2 u1 fu1 = g0 ξ˜y Geostrophy in y-direction. Assume frequency dependency: ˜ = −iω{ũ1 , û1 , ξ} ˜ ∂/∂t{ũ1 , û1 , ξ}