{
  "id": "drag-force",
  "title": "Drag Force",
  "lang": "en",
  "source_lang": "en",
  "summary": "Drag force is the component of fluid force opposing relative motion, arising from surface stresses and the transfer of momentum through a fluid.",
  "category": "science",
  "subcategory": "physics",
  "type": "concept",
  "aliases": [
    "Fluid drag",
    "Fluid resistance",
    "Aerodynamic drag",
    "Hydrodynamic drag"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/drag-force/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/drag-force/",
      "json": "https://aiwiki.page/zh/drag-force.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Drag force is the component of the force exerted by a fluid on an object that acts opposite a specified direction of relative motion. It occurs when objects move through gases or liquids, when fluids flow around stationary objects, and when different parts of a fluid move relative to one another. Air resistance on a vehicle and water resistance on a swimming animal are familiar examples. Drag governs settling particles, aircraft performance, ship propulsion, wind loads on structures, and the gradual orbital decay of satellites in the upper atmosphere.\n\nDrag results from the distribution of pressure and viscous stresses over an object's surface. Its magnitude depends on the object's size, shape, orientation, and surface condition, as well as the fluid's properties and the character of the flow. Although often represented by a simple formula involving speed and a drag coefficient, drag encompasses several distinct mechanisms, including skin friction, flow separation, lift-induced effects, and wave generation. Predicting it requires connecting surface-scale processes with the motion and momentum of the surrounding fluid."
    },
    {
      "heading": "Physical definition and momentum transfer",
      "text": "In a uniform incoming flow, drag is the component of the total fluid force parallel to that flow; lift is the component perpendicular to it. For an object moving through otherwise stationary fluid, the drag force normally opposes its velocity relative to the fluid. These descriptions are equivalent after changing reference frames. More complicated flows, such as shear flows or rapidly accelerating flows, require an explicitly stated reference direction: the total hydrodynamic force need not point opposite the object's instantaneous velocity.\n\nAt the continuum scale, the force on an impermeable body is obtained by integrating the fluid stress over its wetted surface. If \\(\\mathbf n\\) points outward from the body into the fluid, the force exerted by the fluid is\n\n\\[\\mathbf F=\\int_S\\left(-p\\mathbf n+\\boldsymbol{\\tau}\\cdot\\mathbf n\\right)\\,dS,\\]\n\nwhere \\(p\\) is pressure and \\(\\boldsymbol{\\tau}\\) is the viscous stress tensor. Projecting this expression onto the chosen drag direction gives the drag. The pressure contribution acts normal to each surface element; the viscous contribution can have both tangential and normal components. For incompressible Newtonian flow, viscous stresses are proportional to local velocity gradients, with dynamic viscosity \\(\\mu\\) as the proportionality coefficient.\n\nThe same force can be understood through conservation of momentum. A body changes the momentum carried by the surrounding fluid, and the corresponding reaction acts on the body. For a stationary body in steady flow, a downstream wake commonly contains less streamwise momentum than the undisturbed incoming stream. With appropriate pressure and lateral-flux corrections, this momentum deficit provides an alternative way to calculate drag. A wake survey is therefore not merely a visualization technique: it can supply a quantitative force measurement.\n\nDrag is closely associated with energy transfer, but force and dissipation are not identical concepts. If a body moves at speed \\(U\\) through stationary fluid and experiences resistance of magnitude \\(D\\), the mechanical power required to maintain that motion is\n\n\\[P=DU.\\]\n\nThis power can initially enter organized fluid motion, turbulence, surface waves, or sound before ultimately being converted largely into heat. For a fixed obstacle in a flowing stream, the obstacle itself performs no mechanical work because it does not move; the energy sustaining the flow instead comes from a pressure difference, a pump, gravity, or another external source.\n\nThe pressure and viscous contributions are useful bookkeeping categories, but they are dynamically coupled. Viscosity can alter a thin region near a surface, change where flow separates, and thereby produce a large change in pressure drag over the entire body."
    },
    {
      "heading": "Drag coefficients and similarity parameters",
      "text": "For many engineering flows, drag is written as\n\n\\[D=\\frac12\\rho U^2 C_D A,\\]\n\nwhere \\(\\rho\\) is fluid density, \\(U\\) is a specified relative speed, \\(A\\) is a reference area, and \\(C_D\\) is the drag coefficient. The quantity \\(\\tfrac12\\rho U^2\\) is the dynamic pressure. This expression defines \\(C_D\\); it does not establish that drag always varies as the square of speed. If the coefficient changes with speed, the resulting force follows a different dependence.\n\nReference-area conventions matter. Automobiles and spheres commonly use frontal projected area, aircraft usually use wing planform area, and ship-resistance coefficients often use wetted surface area. Consequently, two numerical drag coefficients cannot be meaningfully compared without knowing their reference areas and operating conditions. For many road-vehicle comparisons, the product \\(C_DA\\), called drag area, is more directly informative than \\(C_D\\) alone.\n\nThe central similarity parameter for incompressible flow around a rigid body is the Reynolds number,\n\n\\[\\mathrm{Re}=\\frac{\\rho UL}{\\mu}=\\frac{UL}{\\nu},\\]\n\nwhere \\(L\\) is a characteristic length and \\(\\nu=\\mu/\\rho\\) is kinematic viscosity. Reynolds number measures the relative importance of inertial and viscous effects. At very low values, drag on a sphere is proportional to speed. At higher values, wake formation, boundary-layer transition, and separation can change both the coefficient and the force law. A characteristic length must always accompany a quoted Reynolds number: sphere diameter, wing chord, and ship length describe different physical scales.\n\nOther dimensionless parameters become important when additional mechanisms enter. The Mach number, \\(M=U/a\\), compares speed with the local speed of sound and controls compressibility effects. The Froude number, \\(\\mathrm{Fr}=U/\\sqrt{gL}\\), characterizes gravity-related free-surface effects, particularly ship waves. Relative surface roughness, \\(k_s/L\\), can alter wall friction and transition. For small particles in rarefied gases, the Knudsen number, the ratio of molecular mean free path to body size, determines whether continuum and no-slip assumptions remain suitable.\n\nThese parameters explain both the usefulness and the limits of scale models. A geometrically similar model can reproduce full-scale behavior only when the governing dimensionless conditions are adequately matched. In an ordinary water tank, matching a ship's Froude number generally does not also match its Reynolds number. Resistance prediction must therefore separate effects that scale differently.\n\nAn illustrative calculation shows the practical scale of aerodynamic resistance. With \\(\\rho=1.2\\ \\mathrm{kg\\,m^{-3}}\\), \\(U=30\\ \\mathrm{m\\,s^{-1}}\\), and \\(C_DA=0.60\\ \\mathrm{m^2}\\), drag is \\(324\\ \\mathrm N\\), requiring about \\(9.7\\ \\mathrm{kW}\\) of aerodynamic power. If the drag coefficient remains unchanged, doubling speed multiplies drag by four and the required power by eight."
    },
    {
      "heading": "Creeping flow, particle settling, and terminal speed",
      "text": "When Reynolds number is much smaller than unity, viscous stresses dominate fluid inertia. This regime is called creeping flow or Stokes flow. For a rigid sphere of radius \\(a\\), translating steadily through an unbounded incompressible Newtonian fluid with a no-slip surface, the drag is\n\n\\[D=6\\pi\\mu aU.\\]\n\nGeorge Gabriel Stokes derived this result in 1851. In terms of sphere diameter \\(d=2a\\), the equivalent drag coefficient is\n\n\\[C_D=\\frac{24}{\\mathrm{Re}}, \\qquad \\mathrm{Re}=\\frac{\\rho Ud}{\\mu}.\\]\n\nThe inverse dependence of \\(C_D\\) on Reynolds number makes this fully consistent with the general drag-coefficient definition: substituting it yields drag proportional to \\(U\\), not \\(U^2\\).\n\nStokes flow illustrates why pressure drag cannot simply be equated with inertial resistance. For the ideal Stokes-flow sphere, one-third of the total drag comes from pressure and two-thirds from viscous shear. Both contributions scale with viscosity and speed. Pressure remains necessary to enforce incompressibility and redirect the flow even when fluid inertia is neglected. The equations are linear under these assumptions, permitting superposition and underpinning many calculations of particle interactions.\n\nA settling sphere reaches terminal velocity when downward weight is balanced by upward buoyancy and drag. In the Stokes regime, its terminal speed is\n\n\\[U_t=\\frac{(\\rho_p-\\rho_f)gd^2}{18\\mu},\\]\n\nwhere \\(\\rho_p\\) and \\(\\rho_f\\) are particle and fluid densities. This formula explains the strong size dependence of slow sedimentation: otherwise identical particles with twice the diameter have four times the terminal speed, provided both remain within the low-Reynolds-number regime. It is used in sedimentation analysis, particle characterization, and descriptions of small droplets or suspended solids.\n\nThe assumptions impose important limits. Nearby walls increase resistance, so a sphere settling in a narrow tube does not behave like one in an unbounded fluid. Concentrated suspensions exhibit collective interactions and hindered settling. Nonspherical particles require orientation-dependent resistance coefficients. Clean droplets and bubbles can have mobile interfaces and internal circulation, producing drag different from that of a rigid sphere; surface-active contaminants can suppress this mobility.\n\nFor sufficiently small airborne particles, gas molecules no longer maintain the ordinary no-slip boundary condition. Slip corrections then reduce drag relative to the uncorrected Stokes prediction. At still smaller scales or lower gas densities, a molecular description is needed.\n\nParticle motion can also fluctuate because of Brownian motion. For a dilute spherical particle in thermal equilibrium with an unbounded fluid, the Stokes–Einstein relation connects the translational diffusion coefficient \\(\\mathcal D\\) to the same viscous resistance:\n\n\\[\\mathcal D=\\frac{k_{\\mathrm B}T}{6\\pi\\mu a}.\\]\n\nThus the coefficient describing resistance to imposed motion also determines the scale of thermally driven diffusion under the applicable continuum assumptions."
    },
    {
      "heading": "Boundary layers, separation, and bluff-body resistance",
      "text": "At high Reynolds number, much of the fluid outside the immediate vicinity of a body can behave approximately as if viscosity were negligible. Nevertheless, the no-slip condition requires fluid velocity at an ordinary solid surface to match the surface velocity. The transition from that velocity to the outer-flow velocity occurs within a boundary layer, where strong velocity gradients make viscous effects important despite the large overall Reynolds number.\n\nSkin-friction drag is the streamwise contribution of wall shear stress. For an incompressible laminar boundary layer over a smooth flat plate aligned with a uniform stream, with zero pressure gradient, the Blasius solution gives the mean skin-friction coefficient\n\n\\[\\overline C_f=\\frac{1.328}{\\sqrt{\\mathrm{Re}_L}},\\]\n\nwhere the reference area is the wetted area of one side of the plate and \\(\\mathrm{Re}_L\\) uses plate length. This result is not a general formula for every smooth body: pressure gradients, leading-edge geometry, roughness, and transition modify the boundary layer.\n\nA turbulent boundary layer transports momentum toward the wall more effectively than a laminar one and generally produces greater wall shear under comparable conditions. At the same time, that stronger transport can help near-wall fluid resist an adverse pressure gradient—a pressure increase in the direction of flow. When fluid near the wall loses sufficient downstream momentum, the boundary layer can separate from the surface. Separation creates a region of recirculating flow and often a broad wake with reduced base pressure.\n\nThis produces the fundamental distinction between streamlined and bluff bodies. A streamlined body is shaped to keep flow attached over much of its surface, reducing pressure drag even if it has substantial wetted area. A bluff body, such as a circular cylinder in crossflow, usually experiences extensive separation and substantial pressure drag. Making a body shorter or reducing its surface area does not necessarily reduce total resistance if the resulting shape creates a larger separated wake.\n\nThe drag crisis of a sphere is a particularly striking example. A smooth sphere commonly has a drag coefficient around \\(0.4\\)–\\(0.5\\) across part of the subcritical regime. At Reynolds numbers of a few hundred thousand, boundary-layer transition can delay separation and sharply reduce drag. The exact transition depends on surface roughness and incoming turbulence. Golf-ball dimples exploit this mechanism: although they increase local surface friction, they can reduce total resistance by changing separation and wake pressure.\n\nConsequently, surface smoothness, laminar flow, and low drag are not interchangeable. Smooth surfaces are valuable where skin friction dominates, but carefully introduced roughness can reduce pressure-dominated drag in particular regimes. The appropriate interpretation depends on the complete flow field, not on surface friction considered in isolation."
    },
    {
      "heading": "Lift-induced drag, compressibility, and wave generation",
      "text": "Some drag mechanisms arise from flow structures extending far beyond a body's immediate surface. A finite aircraft wing generates lift by changing the momentum of the surrounding air. Because the wing has finite span, its trailing vortex system produces downwash. This changes the direction of the local relative airflow and gives the aerodynamic force a rearward component, conventionally called induced drag.\n\nFor an approximately planar wing in a commonly used finite-wing model,\n\n\\[C_{D_i}=\\frac{C_L^2}{\\pi e\\,\\mathrm{AR}},\\]\n\nwhere \\(C_L\\) is lift coefficient, \\(\\mathrm{AR}=b^2/S\\) is aspect ratio, \\(b\\) is span, \\(S\\) is wing area, and \\(e\\) represents departure from the ideal span loading. This expression is most useful within the assumptions of the underlying lifting-line approximation. Larger span generally reduces induced drag for a specified lift and dynamic pressure, but structural mass, flexibility, and operating constraints limit practical span.\n\nFor steady level flight at fixed weight, induced drag typically decreases as speed rises, whereas much of the non-lift-related resistance increases with speed. A simplified drag polar,\n\n\\[C_D=C_{D0}+kC_L^2,\\]\n\ncaptures this competition over a restricted operating range. It helps explain why aircraft have distinct speeds for minimum drag and minimum power, rather than becoming progressively more efficient as speed decreases. At very low speed, large lift coefficients and eventual flow separation invalidate the simple model.\n\nCompressibility introduces additional mechanisms. Near transonic conditions, portions of the flow around a body can become supersonic even while the free stream remains subsonic. Shock waves may then form, changing pressure distributions and sometimes causing boundary-layer separation. Wave drag can rise steeply through this regime. Swept wings, thinner sections, and area distribution shaped to moderate streamwise changes in cross section are established ways of managing these effects.\n\nWave drag also occurs in liquids, but with a different restoring mechanism. A ship moving near a free surface generates gravity waves that carry energy away. Its wave-making resistance depends strongly on Froude number and hull geometry. Interference among waves generated by different hull regions can create pronounced changes in resistance with speed. This is why geometric similarity and Reynolds-number matching alone do not fully characterize ship resistance.\n\nThese drag categories are not always strictly additive in a unique physical sense. A shock can change separation, lift can alter boundary-layer development, and a free surface can modify the pressure distribution responsible for other resistance components. Engineering decompositions are useful models, but the measured total force emerges from one coupled flow."
    },
    {
      "heading": "Unsteady motion, vortex shedding, and complex fluids",
      "text": "The usual steady drag formula becomes incomplete when the body or surrounding fluid accelerates rapidly. Some force is associated with accelerating nearby fluid, an effect described by added mass. For a sphere in an unbounded incompressible inviscid fluid, the added mass is one-half the mass of the displaced fluid. This inertial force can exist even in an idealized flow with no steady viscous drag, demonstrating why every fluid-force component should not automatically be labeled resistance.\n\nIn viscous unsteady flow, the force can also depend on the history of relative motion. For a small sphere under low-Reynolds-number assumptions, a history term arises because vorticity generated at the surface diffuses into the surrounding fluid over time. The current force therefore depends not only on present speed and acceleration but also on earlier changes in motion. Particle models for oscillatory flows, rapidly changing suspensions, and sediment transport may need these contributions in addition to steady drag.\n\nSeparated wakes can generate their own time dependence even when the incoming flow and body position are steady. Alternating vortex shedding behind a circular cylinder begins near Reynolds number 47 under idealized low-disturbance conditions. At higher Reynolds numbers, a vortex street can produce periodic transverse forces and fluctuating drag. The characteristic frequency is commonly expressed by the Strouhal number,\n\n\\[\\mathrm{St}=\\frac{fL}{U}.\\]\n\nFor a circular cylinder, values near \\(0.2\\) occur over a substantial range of practical conditions, although the value is not universal.\n\nWhen vortex shedding couples with structural motion, vortex-induced vibration can occur. The shedding frequency may synchronize with the moving structure over a range of speeds, a phenomenon called lock-in. Offshore risers, cables, bridge components, and heat-exchanger tubes provide important engineering examples. Mean drag alone is insufficient for their analysis because fluctuating forces and fatigue loading may govern performance.\n\nFluid constitutive behavior introduces another source of complexity. In a Newtonian fluid, shear stress is proportional to shear rate with a viscosity independent of that rate at fixed thermodynamic conditions. Polymer solutions, slurries, pastes, and many biological fluids can instead be shear-thinning, shear-thickening, viscoelastic, or yield-stress materials. An object moving through them may experience resistance that cannot be represented by inserting a single constant viscosity into a Newtonian correlation.\n\nFlexible bodies and living organisms also alter their shape in response to loading. Leaves can align with a current, aquatic vegetation can bend, and swimming animals can actively control surface motion and wake structure. Their drag can therefore vary with speed partly because their geometry changes. In these cases, resistance is a coupled fluid–structure problem rather than a property of a fixed shape."
    },
    {
      "heading": "Experimental measurement and numerical prediction",
      "text": "Direct drag measurement usually employs a force balance connected to the test object. Wind tunnels, towing tanks, and water channels provide controlled incoming flows, while internal or external balances resolve the relevant force components. Successful measurement requires more than a sensitive sensor: support interference, tunnel-wall effects, blockage, flow nonuniformity, vibration, and balance cross-talk can all change the result or its interpretation.\n\nCalibration establishes the relationship between the instrument response and applied force, but it does not remove every experimental bias. A model support can disturb the flow even when its own force is subtracted. A confined test section can accelerate flow around a model or change its wake. Surface roughness and transition location must also represent the intended conditions. Results should therefore include operating conditions, coefficient definitions, correction methods, and measurement uncertainty.\n\nIndirect methods infer drag from the flow or from an object's motion. Wake measurements use momentum balance, sometimes supplemented by pressure measurements. Surface pressure taps can be integrated to estimate pressure forces, while wall-shear techniques address skin friction. Free-flight or coast-down experiments infer resistance from measured deceleration after accounting for other forces. For ships, towing measurements provide total resistance at specified speeds, but extrapolation to full scale requires attention to Reynolds-number and free-surface effects.\n\nComputational fluid dynamics predicts drag by solving approximations to the governing flow equations and integrating pressure and viscous stresses over the computed surface. Reynolds-averaged methods model the effects of turbulence on the mean flow and are widely used in engineering design. Large-eddy simulation resolves larger turbulent motions while modeling smaller ones. Direct numerical simulation resolves all dynamically relevant turbulent scales but is prohibitively expensive for many full-scale, high-Reynolds-number applications.\n\nDrag is a demanding numerical quantity because relatively small errors in wall shear, separation position, or base pressure can materially change the integrated force. Accurate geometry is not sufficient by itself. Mesh resolution, near-wall treatment, turbulence and transition models, domain size, boundary conditions, and time averaging all affect predictions. A visually plausible wake does not establish that the force is accurate.\n\nVerification and validation address different questions. Verification examines whether the mathematical model is solved correctly, including discretization and convergence errors. Validation compares predictions with physical observations for the intended regime. Agreement between two simulations using similar assumptions is not equivalent to independent experimental validation. For design decisions involving small drag differences, uncertainty in the difference may be as important as uncertainty in either absolute result."
    },
    {
      "heading": "Engineering, environmental, and biological roles",
      "text": "In transport, drag determines a major part of the power required to sustain speed. Road vehicles are shaped to manage separation at the rear, underbody flow, wheel wakes, cooling airflow, and crosswind behavior. Reducing frontal area or drag coefficient can lower aerodynamic power demand, but changes must coexist with requirements for passenger space, stability, visibility, and thermal management. Aerodynamic drag is distinct from tire rolling resistance and drivetrain losses, which follow different mechanisms.\n\nAircraft design balances drag against lift, structural requirements, payload, and operating conditions. A long slender wing may reduce induced drag but increase structural demands; a thicker wing can accommodate structure and fuel but alter compressibility and profile resistance. Marine vehicles face a different balance among skin friction, viscous pressure resistance, wave-making resistance, and appendage effects. Surface fouling on a hull changes roughness and can increase the power needed to maintain a given speed.\n\nDrag can also be deliberately increased. Parachutes convert the motion of a descending payload into a broad, momentum-deficient wake, reducing descent speed. Air brakes and drag chutes dissipate vehicle kinetic energy through the surrounding flow. Drag-based wind machines extract useful mechanical power from fluid motion, although many high-performance turbines rely primarily on lift. In all these cases, resistance is part of the intended operation rather than simply a loss to be minimized.\n\nIn environmental flows, drag mediates momentum exchange between moving air or water and rough surfaces, plants, buildings, and sediment grains. Vegetation can slow floodwater while also changing local turbulence and sediment deposition. Particle settling influences water clarity and the transport of suspended material in a river or the ocean. Whether a grain remains suspended depends not only on its settling resistance but also on turbulent mixing and upward fluid motions.\n\nBiological locomotion spans very different drag regimes. A bacterium operates at low Reynolds number, where viscous forces dominate and reciprocal shape changes cannot generally produce net propulsion under the assumptions of Stokes flow. This constraint is known as the scallop theorem; rotating flagella and nonreciprocal beating patterns avoid its restriction. Large fish and marine mammals operate at much higher Reynolds numbers, where inertia, boundary layers, body shape, and wake organization become central.\n\nSatellites provide an application in a very different environment. Even the tenuous upper atmosphere exerts drag on spacecraft in low Earth orbit. The resulting loss of orbital energy generally lowers the orbit. Atmospheric density varies with altitude and solar activity, while gas–surface interaction and rarefaction complicate the drag coefficient. Deployable drag sails exploit this resistance to accelerate orbital decay rather than relying on it being negligible."
    },
    {
      "heading": "Historical development and continuing research",
      "text": "Early resistance theories attempted to connect fluid forces with the impact of particles on moving bodies. Isaac Newton treated resisting media in the 1687 Principia and developed models that captured aspects of speed and shape dependence under specific assumptions. Such approaches were important precedents, but they could not represent the full pressure field, viscous stresses, and separated wakes of real continuum flows.\n\nA central conceptual difficulty emerged in Jean le Rond d'Alembert's 1752 analysis: steady, incompressible, inviscid, irrotational flow around a body can predict zero drag despite the resistance observed in experiments. This result, d'Alembert's paradox, is not resolved by simply assuming that all drag is surface rubbing. Real viscosity changes the flow structure and can create a separated wake with substantial pressure drag, even when viscosity is small outside thin surface layers.\n\nNineteenth-century work established complementary foundations. Stokes obtained the low-Reynolds-number sphere result in 1851. William Froude's nineteenth-century ship-model experiments helped establish practical resistance testing and free-surface similarity. Osborne Reynolds's 1883 experiments on flow through pipes clarified the importance of the balance between inertia and viscosity and the distinction between laminar and turbulent behavior. These advances made it possible to organize apparently disparate resistance measurements by governing flow regimes.\n\nLudwig Prandtl's boundary-layer theory, introduced in 1904, provided the decisive connection between nearly inviscid outer flow and viscous surface regions at high Reynolds number. Heinrich Blasius's 1908 flat-plate solution supplied a quantitative laminar example. Frederick Lanchester and Prandtl subsequently developed vortex-based descriptions of finite-wing behavior that clarified induced drag. These theories did not replace experimentation; they supplied structures within which test results could be interpreted and generalized.\n\nCurrent drag research frequently concerns the interaction of mechanisms rather than a single missing force law. Transition prediction remains difficult when surface roughness, free-stream turbulence, pressure gradients, and acoustic disturbances interact. Separated turbulent flows remain sensitive to modeling assumptions. Flexible surfaces, multiphase mixtures, rarefied gases, and strongly non-Newtonian fluids add further scales and constitutive uncertainties.\n\nDrag-reduction methods under investigation or specialized use include riblets, polymer additives, air lubrication, suction, compliant coatings, and active flow control. Each targets a particular mechanism. Riblets can modify near-wall turbulent motion; suitable dilute polymers can reduce turbulent friction in some internal flows; air layers can reduce water contact with a surface; suction can delay transition or separation. Their net benefit depends on durability, contamination, operating range, and the energy or equipment required to maintain the effect. A reduction measured under one set of conditions is therefore not a universal material or geometric property."
    }
  ],
  "links": [
    "force",
    "pressure",
    "viscosity",
    "momentum",
    "heat",
    "density",
    "reynolds-number",
    "buoyancy",
    "brownian-motion",
    "boundary-layer",
    "calibration",
    "measurement-uncertainty",
    "turbulence",
    "river",
    "ocean",
    "atmosphere-of-earth",
    "isaac-newton"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}