betaFeedback

WorksPhysics and Astronomy

A Five-Dimensional Geometric Formulation of Geophysical Fluid Dynamics

Abstract

We apply a geometric formulation of Newtonian dynamics to inviscid compressible geophysical flow, with gravity and the effects of rotating or moving coordinates incorporated in the metric tensor. This construction relies on the introduction of an auxiliary fifth coordinate, which allows mass and momentum transport, internal energy, and pressure work to be combined into a single mass–momentum tensor. The covariant divergence of this tensor leads to the continuity equation for mass density, the three momentum equations, and the thermodynamic equation, with the system closed by an equation of state. A time-independent metric also yields conservation of total energy. We recover familiar forms of the equations and illustrate the construction in rotating spherical and terrain-following coordinates. The thin-shell and spherical-geopotential approximations, with constant effective gravity, simplify the metric while retaining vertical gravity and the traditional Coriolis terms. Emphasis is placed on the connection between the geometric structure of the formulation and the equations commonly used in atmospheric dynamics.

Introduction

Writing down the equations governing atmospheric motion is one thing; writing them down in a coordinate system fit for computation is quite another. The atmosphere is a compressible fluid, rotating with the planet and stratified by gravity. Operational numerical weather prediction commonly employs spherical coordinates rotating with the Earth, often combined with a terrain-following vertical coordinate. These modelling choices introduce many coordinate- and frame-dependent terms. Model developers frequently need to change coordinate systems or simplify the equations while preserving consistency with Newtonian mechanics. Deriving or modifying the individual terms separately is error-prone and can compromise conservation properties. Here, we present a geometric formulation that represents these terms, including Coriolis and centrifugal accelerations, through the metric tensor and its associated connection. This formulation provides a set of rules for transforming the equations and treating the geometric terms consistently.

These transformation rules are provided by tensor analysis [5]. Tensor equations are covariant: their mathematical form is preserved under a change of coordinates or reference frame. The tensor transformation rules ensure that these equations represent the same physical statements in different coordinate systems. Once the atmospheric equations are written in tensor form, the metric and inertial terms associated with a change of coordinates or reference frame can be derived systematically. Several four-dimensional tensor formulations of the fluid equations have been developed along these lines [1, 5, 16, 19]. In particular, Charron et al. [5] showed how the governing equations used in atmospheric models can be obtained from a compact four-dimensional tensor formulation.

In the four-dimensional formulation of Charron et al. [5], the coordinate and inertial terms are treated systematically, but gravity remains an explicit external force. We extend this formulation by incorporating the gravitational force into the geometry, while retaining the ability to use rotating and terrain-following coordinates alongside their approximations. A geometric description of Newtonian gravity for point particles was introduced by Élie Cartan, who represented the trajectories of freely falling particles as geodesics of a four-dimensional connection [3, 4]. This formalism incorporates Newtonian gravity into the geometry, but the gravitational connection is not determined by a single non-degenerate four-dimensional metric tensor. Introducing a fifth dimension allows such a metric description, as shown by Eisenhart [8]. In this construction, Newtonian trajectories in a gravitational potential are realized as the projection of geodesics in the extended five-dimensional geometry to the four-dimensional Newtonian space-time. The auxiliary coordinate, denoted by x4x^{4}, introduces an additional degree of freedom allowing gravity to be incorporated into the five-dimensional metric structure. It does not correspond to a direction in physical space or time. Translation invariance in this coordinate introduces a symmetry associated with the conservation of mass. A modern differential-geometric treatment of this idea was later developed by Duval, Burdet, Künzle, and Perrin [6]; Duval, Gibbons, and Horváthy [7].

The Eisenhart–Duval treatment of gravity can be applied to fluid dynamics by introducing a symmetric mass–momentum tensor TABT^{AB} that represents mass and momentum transport as well as internal energy and pressure work [2]. On the extended five-dimensional manifold, the conservation laws can be expressed compactly through its covariant divergence,

∇ATAB=0.(1)\nabla_{A}T^{AB}=0. \tag*{(1)}

This expression gives the continuity equation, the three momentum equations, and the thermodynamic equation. Together with an equation of state, these equations form a closed system of partial differential equations.

The aim of this paper is to present an explicit five-dimensional formulation of compressible atmospheric dynamics in rotating and moving coordinates, building on the Eisenhart–Duval construction and the null-fluid formulation [2, 6, 7, 8]. The main contribution is the derivation of the metric, geometric source terms, and balance laws for the coordinate systems and approximations used in atmospheric modelling. We construct the rotating spherical metric and its transformation to a generalized, possibly time-dependent vertical coordinate, and derive the corresponding continuity, momentum, internal-energy, and total-energy balances. We also show how a thin-shell metric, combined with a spherical effective potential and constant effective gravity, recovers the nonhydrostatic shallow-atmosphere equations with traditional Coriolis terms. These results extend the atmospheric tensor formulation of Charron et al. [5] by incorporating gravity into the metric and treating its consequences for the governing equations and their approximations.

The paper is organized as follows. Sections 2–4 introduce the geometric formulation and the kinematic relations. Sections 5 and 6 present the fluid equations and their alternative forms. Sections 7 and 8 discuss applications to rotating spherical and generalized vertical coordinates. Section 9 introduces the thin-shell and spherical-geopotential approximations, with constant effective gravity, and derives the resulting equations with traditional Coriolis terms. The integral form of the balance laws is given in Appendix § 14, and a list of symbols is provided in Appendix § 15.

Geometric Preliminaries

We introduce the tensor calculus and geometric identities needed to formulate the fluid equations. Familiarity with multivariable calculus is assumed but no prior knowledge of differential geometry is required. For a more comprehensive mathematical treatment, the reader is referred to Isham [11].

Manifolds, coordinate systems, and tensors

An nn-dimensional smooth manifold M\mathcal{M} is a topological space that locally resembles Euclidean space Rn\mathbb{R}^n, meaning that every point has a neighbourhood that maps bijectively and continuously onto an open subset of Rn\mathbb{R}^n with a continuous inverse. Such a mapping is called a chart. A collection of compatible charts covering the manifold is called an atlas. Each chart assigns coordinates to every point in its domain. Some charts in the atlas overlap on their common domains, and on each overlap, the two sets of coordinates are related by smooth invertible functions called transition maps.

Throughout this paper, we adopt the following index conventions: uppercase Latin indices A,B,C,…A, B, C, \ldots take values in {0,1,2,3,4}\{0,1,2,3,4\}; lowercase Greek indices μ,ν,…\mu,\nu,\ldots take values in {0,1,2,3}\{0,1,2,3\}; and lowercase Latin indices i,j,k,…i,j,k,\ldots take values in {1,2,3}\{1,2,3\}. An index appearing once as a superscript and once as a subscript is summed over the range appropriate to that index.

In the five-dimensional setting used in this study, a general transformation from a coordinate system S\mathcal{S}, with coordinates x0,x1,x2,x3,x4x^0,x^1,x^2,x^3,x^4, to S~\widetilde{\mathcal{S}}, with coordinates x~0,x~1,x~2,x~3,x~4\widetilde{x}^0,\widetilde{x}^1,\widetilde{x}^2,\widetilde{x}^3,\widetilde{x}^4, takes the form

x~A=x~A(x0,x1,x2,x3,x4),(2)\widetilde{x}^{A}=\widetilde{x}^{A}(x^0,x^1,x^2,x^3,x^4), \tag*{(2)}

The transformation is required to be smooth and invertible on the overlap of the coordinate charts. Further restrictions can be imposed depending on the physical properties of the system. In § 3, we restrict these transformations to preserve the absolute time of Newtonian mechanics and leave the auxiliary coordinate unchanged.

Physical quantities are described by fields that assign a scalar, vector, or tensor to each point of the manifold. Their components transform according to definite rules under a change of coordinates, so that the physical quantities they represent remain unchanged.

A scalar is a quantity whose numerical value at a point is the same in all coordinate systems. Pressure, temperature, and density are examples of scalars. A vector at a point PP is defined as a tangent to a curve passing through PP. If we denote its components in the coordinate-induced basis as VAV^{A}, they transform contravariantly between two coordinate systems:

V~B=∂x~B∂xAVA.(3)\widetilde{V}^{B} = \frac{\partial\widetilde{x}^{B}}{\partial x^{A}} V^{A}. \tag*{(3)}

A complementary object is the covector (or covariant vector), of which the gradient of a scalar field is an example. For a scalar field ff, the components WA=∂f/∂xAW_{A} = \partial f/\partial x^{A} transform according to

W~B=∂xA∂x~BWA.(4)\widetilde{W}_{B} = \frac{\partial x^{A}}{\partial\widetilde{x}^{B}} W_{A}. \tag*{(4)}

Higher-rank tensors are defined by requiring their components to transform as products of the Jacobian matrix ∂x~A/∂xB\partial\widetilde{x}^{A}/\partial x^{B} (for each upper index) and its inverse ∂xA/∂x~B\partial x^{A}/\partial\widetilde{x}^{B} (for each lower index). For example, a rank-two tensor with one contravariant and one covariant index transforms as

Q~ABA=∂x~A∂xC∂xD∂x~BQCDC.\widetilde{Q}^{A}_{\phantom{A}B} = \frac{\partial\widetilde{x}^{A}}{\partial x^{C}} \frac{\partial x^{D}}{\partial\widetilde{x}^{B}} Q^{C}_{\phantom{C}D}.

Each upper index contributes a factor of the Jacobian, whereas each lower index contributes a factor of the inverse Jacobian. A quantity is a tensor if and only if its components transform according to this rule.2

Metric tensor

A central geometric object in what follows is the metric tensor gABg_{AB}. It can be understood through the line element, which assigns a squared interval to an infinitesimal coordinate displacement:

dℓ2=gAB dxA dxB.d\ell^{2} = g_{AB}\,dx^{A}\,dx^{B}.

In Euclidean space expressed in Cartesian coordinates, the spatial line element is the squared distance dℓspace2=dx2+dy2+dz2d\ell_{\mathrm{space}}^{2} = dx^{2} + dy^{2} + dz^{2}. In spherical coordinates with longitude λ\lambda, latitude φ\varphi, and radius rr, it becomes

dℓspace2=r2cos⁡2φ dλ2+r2 dφ2+dr2.d\ell_{\mathrm{space}}^{2} = r^{2}\cos^{2}\varphi\,d\lambda^{2} + r^{2}\,d\varphi^{2} + dr^{2}.

The metric coefficients convert coordinate increments into physical distances.

The metric defines a smoothly varying inner product on the tangent space of the manifold, assigning a scalar gABVAWBg_{AB}V^{A}W^{B} to any pair of tangent vectors VAV^{A} and WAW^{A} defined at the same point. This provides a correspondence between vectors and covectors: contracting a vector with the metric lowers an index, VA=gABVBV_{A} = g_{AB}V^{B}, whereas contraction with the inverse metric gABg^{AB}, defined by gACgCB=δBAg^{AC}g_{CB} = \delta^{A}_{B}, raises indices according to VA=gABVBV^{A} = g^{AB}V_{B}.

In the Eisenhart–Duval construction, the metric is a pseudo-Riemannian metric, i.e. it is symmetric and non-degenerate but need not be positive definite. Its signature specifies the numbers of positive and negative directions in a tangent-space basis that diagonalizes the metric. Unlike the Euclidean dot product, however, a pseudo-Riemannian metric need not assign a positive squared norm to every non-zero vector.

The metric can be used to define a volume associated with a coordinate cell. Whereas the product of coordinate increments gives the size of a cell in coordinate space, the metric supplies the factor that converts this coordinate volume into geometric volume. In five dimensions, the resulting volume element is

dV5=∣det⁡g∣ d5x.dV_{5} = \sqrt{\lvert\det g\rvert}\,d^{5}x.

where det⁡g\det g is the determinant of the covariant metric tensor in the chosen coordinates and d5x\mathrm{d}^{5}x is the coordinate volume measure. The absolute value accounts for the metric’s indefinite signature. This volume element is invariant under a coordinate transformation: although the coordinate cell and the metric factor change, their product represents the same geometric volume.

Affine connection

An affine connection provides a rule for comparing vectors at neighbouring points of the manifold and for differentiating tensor fields in a geometrically meaningful way.

Ordinary partial derivatives of tensor components do not in general yield tensors because the coordinate basis vectors themselves vary from point to point. The affine connection defines the corresponding covariant derivative, which accounts for this variation. For a vector VAV^{A}, the covariant derivative in the BB-direction is

∇BVA=∂VA∂xB+ΓBCAVC,(5)\nabla_{B}V^{A} = \frac{\partial V^{A}}{\partial x^{B}} + \Gamma^{A}_{BC}V^{C}, \tag*{(5)}

where ∇B\nabla_{B} denotes covariant differentiation, whereas ∂VA/∂xB\partial V^{A}/\partial x^{B} is the ordinary partial derivative. Similarly, the covariant derivative of a rank-two contravariant tensor TABT^{AB} in the CC-direction is

∇CTAB=∂TAB∂xC+ΓCDATDB+ΓCDBTAD.(6)\nabla_{C}T^{AB} = \frac{\partial T^{AB}}{\partial x^{C}} + \Gamma^{A}_{CD}T^{DB} + \Gamma^{B}_{CD}T^{AD}. \tag*{(6)}

Note that neither the partial derivatives of the components nor the Christoffel symbols in (5) and (6) transform tensorially; only the complete sum of terms in each expression obeys the tensor transformation law.

The metric tensor gABg_{AB} determines a unique torsion-free, metric-compatible affine connection, called the Levi–Civita connection. Metric compatibility means that the covariant derivative of the metric vanishes:

∇CgAB=0.\nabla_{C}g_{AB} = 0.

Geometrically, this means that parallel transport preserves the scalar product of vectors. Algebraically, raising or lowering indices before or after covariant differentiation yields the same result. The coefficients of the Levi–Civita connection in a coordinate basis are the Christoffel symbols of the second kind:

ΓBCA=12gAD(∂gDB∂xC+∂gDC∂xB−∂gBC∂xD).(7)\Gamma^{A}_{BC} = \frac{1}{2}g^{AD}\left(\frac{\partial g_{DB}}{\partial x^{C}} + \frac{\partial g_{DC}}{\partial x^{B}} - \frac{\partial g_{BC}}{\partial x^{D}}\right). \tag*{(7)}

The torsion-free condition appears as symmetry in the two lower indices, ΓBCA=ΓCBA\Gamma^{A}_{BC} = \Gamma^{A}_{CB}.

The covariant divergence of a vector is obtained by contracting the indices of the covariant derivative (5). Using the identity

1∣det⁡g∣∂∣det⁡g∣∂xA=ΓACC,\frac{1}{\sqrt{\lvert\det g\rvert}}\frac{\partial\sqrt{\lvert\det g\rvert}}{\partial x^{A}} = \Gamma^{C}_{AC},

this expression becomes

∇AVA=1∣det⁡g∣∂∂xA(∣det⁡g∣ VA).\nabla_{A}V^{A} = \frac{1}{\sqrt{\lvert\det g\rvert}}\frac{\partial}{\partial x^{A}}\left(\sqrt{\lvert\det g\rvert}\,V^{A}\right).

Similarly, contracting the derivative index with a tensor index in (6) defines the covariant divergence

∇ATAB=1∣det⁡g∣∂∂xA(∣det⁡g∣ TAB)+ΓACBTAC,(8)\nabla_{A}T^{AB} = \frac{1}{\sqrt{\lvert\det g\rvert}}\frac{\partial}{\partial x^{A}}\left(\sqrt{\lvert\det g\rvert}\,T^{AB}\right) + \Gamma^{B}_{AC}T^{AC}, \tag*{(8)}

which is used below to express local balance laws in terms of mass and momentum transport.

Five-Dimensional Newtonian Geometry

In Newtonian mechanics, a particle subject to no force moves with constant velocity along a straight line in an inertial frame. Gravity changes this motion through a force derived from a potential. The Eisenhart–Duval construction incorporates this potential into a five-dimensional metric, so that a particle moving under gravity alone follows a geodesic of the extended geometry. This gives a geometric description of gravitational motion while retaining Newtonian mechanics. Time remains absolute and spatial distances at fixed time retain their Euclidean meaning.

The Eisenhart–Duval metric

We use coordinates xA=(x0,x1,x2,x3,x4)x^{A} = (x^{0}, x^{1}, x^{2}, x^{3}, x^{4}), where x0=tx^{0} = t and x4x^{4} denotes the auxiliary coordinate. We restrict coordinate changes to synchronous transformations, which preserve absolute time and leave the auxiliary coordinate unchanged:

References

  1. [1]Avis, L. M. (1976). A space-time tensor formulation for continuum mechanics in general curvilinear, moving, and deforming coordinate systems (Tech. Rep. No. NASA-TR-R-462). Hampton, VA: NASA Langley Research Center.
  2. [2]Banerjee, N., Dutta, S., & Jain, A. (2016). Null fluids: A new viewpoint of Galilean fluids. Physical Review D, 93(10), 105020.DOI
  3. [3]Cartan, É. (1923). Sur les variétés à connexion affine et la théorie de la relativité généralisée (première partie). In Annales scientifiques de l’école normale supérieure (Vol. 40, pp. 325–412).
  4. [4]Cartan, É. (1924). Sur les variétés à connexion affine, et la théorie de la relativité généralisée (première partie)(suite). In Annales scientifiques de l’école normale supérieure (Vol. 41, pp. 1–25).
  5. [5]Charron, M., Zadra, A., & Girard, C. (2014). Four-dimensional tensor equations for a classical fluid in an external gravitational field. Q. J. R. Meteorol. Soc., 140, 908–916. Retrieved from https://doi.org/10.1002/qj.2185
  6. [6]Duval, C., Burdet, G., Künzle, H. P., & Perrin, M. (1985). Bargmann structures and Newton–Cartan theory. Phys. Rev. D, 31, 1841–1853.pubmed.ncbi.nlm.nih.gov/9955910
  7. [7]Duval, C., Gibbons, G. W., & Horváthy, P. (1991). Celestial mechanics, conformal structures and gravitational waves. Phys. Rev. D, 43, 3907–3922.pubmed.ncbi.nlm.nih.gov/10013357
  8. [8]Eisenhart, L. P. (1928). Dynamical trajectories and geodesics. Ann. Math., 30, 591–606.DOI
  9. [9]Gal-Chen, T., & Somerville, R. C. J. (1975). On the use of a coordinate transformation for the solution of the Navier–Stokes equations. Journal of Computational Physics, 17(2), 209–228. doi: 10.1016/0021-9991(75)90037-6
  10. [10]Hirsch, C. (2007). Numerical computation of internal and external flows: The fundamentals of computational fluid dynamics. Elsevier.
  11. [11]Isham, C. J. (1989). Modern differential geometry for physicists (Vol. 32). World Scientific.
  12. [12]Klemp, J. B. (2011). A terrain-following coordinate with smoothed coordinate surfaces. Monthly Weather Review, 139(7), 2163–2169. doi: 10.1175/MWR-D-10-05046.1DOI
  13. [13]Lamb, H. (1945). Hydrodynamics. New York: Dover Publications. Retrieved from https://archive.org/details/hydrodynamics001amb/page/14/mode/2up (Reissue of the 6th edition (1932))
  14. [14]Laprise, R. (1992). The Euler equations of motion with hydrostatic pressure as an independent variable. Monthly Weather Review, 120(1), 197–207. doi: 10.1175/1520-0493(1992)120<0197:TEEOMW>2.0.CO;2doi.org/10.1175/1520-0493(1992)120<0197:teeomw>2.0.co;2
  15. [15]Phillips, N. A. (1966). The equations of motion for a shallow rotating atmosphere and the ‘traditional approximation’. J. Atmos. Sci., 23, 626–628.doi.org/10.1175/1520-0469(1966)023<0626:teomfa>2.0.co;2
  16. [16]Pudykiewicz, J. A. (2017). Metric properties of the terrain following coordinates on ellipsoid with mountains (Tech. Rep.). Environnement et Changement climatique Canada: Recherche en Prévision Numérique Atmosphérique. (Available at https://www.researchgate.net/publication/316669106)DOI
  17. [17]Ringler, T. D. (2011). Momentum, vorticity and transport: Considerations in the design of a finite-volume dynamical core. In Numerical techniques for global atmospheric models (pp. 143–183). Springer.DOI
  18. [18]Schär, C., Leuenberger, D., Fuhrer, O., Lüthi, D., & Girard, C. (2002). A new terrain-following vertical coordinate formulation for atmospheric prediction models. Monthly Weather Review, 130(10), 2459–2480.
  19. [19]Vinokur, M. (1974). A new formulation of the conservation equations of fluid dynamics (Tech. Rep. No. NASA-TM-X-62415). Moffett Field, CA: NASA Ames Research Center.

Paper details

vlnc:W_mm2wq26b
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Claude, Codex

View statement
Looking for related papers…
Contents