Analytical Branch-Wise Drag Model

• The Analytical Branch-Wise Drag Model is a family of formulations that decomposes drag dynamics into explicit branches defined by sign conditions, transient effects, and hierarchical or regime-based criteria.
• Its methodology spans diverse applications—including coronal mass ejections, landslide motion, porous media flows, and turbulent boundary layers—yielding quantitative predictions and performance metrics.
• The approach emphasizes regime separation and component partitioning for clear physical interpretation while relying on simplifying assumptions that may limit its applicability in highly complex systems.

Searching arXiv for recent and relevant papers on analytical drag models and branch-wise formulations. I’ll look up arXiv entries related to analytical drag models, ensemble drag-based models, and branch-wise decompositions. “Analytical Branch-Wise Drag Model” (Editor’s term) denotes an analytical drag formulation in which the governing force law, reduced dynamics, or closure is decomposed into explicit branches associated with distinct sign conditions, thresholds, geometric classes, or scale-separated regimes. Across the literature considered here, branch structure appears in drag-based coronal mass ejection propagation, unsteady wave drag, sparse-data drag-reduction response modeling, landslide drag, rough-surface drag decomposition, canopy drag partition, filtered multiphase drag, solute drag, dusty-disc instability, atmospheric geostrophic drag, and flexible-blade reconfiguration (Dumbović et al., 2018, Gierczak et al., 2020, Fernex et al., 2019, Pudasaini, 2022, Jain et al., 2022, Buono et al., 2024, Chen et al., 2018, Humadi et al., 2014, Longarini et al., 2022, Narasimhan et al., 2023, Wei et al., 6 Jan 2025). This suggests that the term does not denote a single canonical equation; rather, it identifies a family of analytical constructions in which drag or drag-governed dynamics is written as a finite set of interpretable branches instead of a single undifferentiated empirical law.

1. Conceptual structure of branch-wise drag formulations

The surveyed literature supports several recurrent branch types. A first class is sign-explicit branching, where the drag law changes its analytical expression according to the sign of a relative velocity, as in coronal mass ejection propagation or landslide motion. A second class is history-dependent branching, where the force is split into transient and asymptotic steady parts, or into local and nonlocal contributions, as in unsteady wave drag. A third class is hierarchy- or manifold-based branching, where a high-dimensional drag response is reduced to variation along a one-dimensional ridge and variation across it, or where unresolved morphology is mapped to an equivalent branch order. A fourth class is component-wise partitioning, where total drag is decomposed into viscous and pressure parts, ground and obstacle parts, or additive sub-grid corrections. A fifth class is stability or regime branching, where drag couples branches such as one-fluid versus two-fluid behavior, static versus fluttering motion, or conventionally neutral versus stable boundary layers.

This suggests that “branch-wise” need not mean only piecewise formulas in one scalar variable. In the cited works, branches may be defined by the sign of $v-w$, by the existence of memory kernels, by parameter-space ridgelines, by hierarchy levels such as $n_{\mathrm{eff}}$, by drag partitions such as $\tau=\tau_R+\tau_S$, or by mode transitions in a dispersion relation. The common analytical feature is explicit regime separation with interpretable transition conditions.

2. Sign-explicit kinematic drag laws

In heliospheric propagation, the drag-based model (DBM) assumes that beyond about $R \gtrsim 15\,R_{\odot}$ a CME is governed solely by magnetohydrodynamical drag with constant solar-wind speed $w$ and constant drag parameter $\gamma$. The governing law is

$\frac{dv}{dt}=-\gamma(v-w)|v-w|.$

If $v_0>w$, the fast-CME branch is decelerating,

$v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$

$$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$

If $n_{\mathrm{eff}}$0, the slow-CME branch is accelerating,

$n_{\mathrm{eff}}$1

$n_{\mathrm{eff}}$2

With $n_{\mathrm{eff}}$3 and $n_{\mathrm{eff}}$4, the compact branch-wise form is

$n_{\mathrm{eff}}$5

$n_{\mathrm{eff}}$6

Arrival time at $n_{\mathrm{eff}}$7 follows from the implicit relation

$n_{\mathrm{eff}}$8

and arrival speed is

$n_{\mathrm{eff}}$9

Geometry is handled separately through impact screening such as $\tau=\tau_R+\tau_S$0. The ensemble extension DBEM preserves this branch-wise analytical engine and propagates uncertainty by sampling $\tau=\tau_R+\tau_S$1. In the 2018 evaluation on 25 refined events, DBEM yielded $\tau=\tau_R+\tau_S$2 h, $\tau=\tau_R+\tau_S$3 h, and $\tau=\tau_R+\tau_S$4 h, comparable to ENLIL. In the 2021 DBEMv3 evaluation on 146 CME–ICME pairs, the recommended inputs shifted to approximately $\tau=\tau_R+\tau_S$5 and $\tau=\tau_R+\tau_S$6, with $\tau=\tau_R+\tau_S$7 h and $\tau=\tau_R+\tau_S$8 h. Both evaluations emphasized a systematic early-arrival bias for fast CMEs (Dumbović et al., 2018, Čalogović et al., 2021).

In the extended landslide model, the steady reduction of

$\tau=\tau_R+\tau_S$9

and mass conservation yields

$R \gtrsim 15\,R_{\odot}$0

with $R \gtrsim 15\,R_{\odot}$1. This creates a kinematic branch at

$R \gtrsim 15\,R_{\odot}$2

If $R \gtrsim 15\,R_{\odot}$3, the solution is on an upwind branch; if $R \gtrsim 15\,R_{\odot}$4, it is on a downwind branch. A second branch structure enters through the earth-pressure coefficient

$R \gtrsim 15\,R_{\odot}$5

with the minus sign for expanding motion $R \gtrsim 15\,R_{\odot}$6 and the plus sign for contracting motion $R \gtrsim 15\,R_{\odot}$7. Because

$R \gtrsim 15\,R_{\odot}$8

the drag coefficient is branch-dependent: $R \gtrsim 15\,R_{\odot}$9 The paper’s central claim is that drag is fundamentally different for expanding and contracting motions (Pudasaini, 2022).

3. Memory, asymptotic, and threshold branches

For a disturbance moving on an inviscid, irrotational, incompressible fluid of infinite depth, the unsteady surface elevation obeys

$w$0

with dispersion relation

$w$1

The exact wake is the history integral

$w$2

and the exact unsteady drag is

$w$3

For straight-line motion with axisymmetric pressure, this reduces to

$w$4

The resulting branch structure is explicit. If motion reaches constant velocity in finite time, the unsteady wake and drag converge to the steady Havelock solution; this defines a transient versus asymptotic steady split. The same formulas also distinguish local/quasi-static versus nonlocal/history-dependent drag, because $w$5 is only an approximation to the exact functional of trajectory history. Additional branches arise from the dispersion relation: pure gravity versus capillary-gravity behavior; below-radiation-threshold versus above-threshold asymptotics in capillary-gravity flow, with

$w$6

and subcritical-transient versus supercritical-transient decay, separated by

$w$7

The paper reports that if $w$8 the final steady wave drag is zero, whereas if $w$9 it is nonzero; it also reports exponential decay of oscillations for $\gamma$0 and algebraic $\gamma$1 decay for $\gamma$2. For pure gravity waves, quasi-static replacement by steady Havelock drag becomes reasonable when $\gamma$3, but in capillary-gravity cases significant oscillatory memory may persist even for slow ramps (Gierczak et al., 2020).

4. Hierarchy- and manifold-based branch reductions

In turbulent boundary-layer drag reduction by spanwise traveling transversal surface waves, the response surface

$\gamma$4

was reduced to a one-dimensional ridgeline

$\gamma$5

with ridge response

$\gamma$6

For $\gamma$7, the fitted ridge coordinates are

$\gamma$8

and the along-ridge drag-reduction law is

$\gamma$9

With normalized cross-ridge coordinates

$\frac{dv}{dt}=-\gamma(v-w)|v-w|.$0

the full self-similar model becomes

$\frac{dv}{dt}=-\gamma(v-w)|v-w|.$1

The model was trained on 71 LES cases over $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$2, $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$3, $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$4, achieved $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$5 for the SVR interpolant and $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$6 for the reduced self-similar model, and extrapolated successfully to $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$7 with relative errors of $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$8 at the on-ridge point $\frac{dv}{dt}=-\gamma(v-w)|v-w|.$9 and $v_0>w$0 at the off-ridge point $v_0>w$1 (Fernex et al., 2019).

In porous-media tree modeling, the analogous branch variable is not a trajectory branch but an equivalent hierarchy level. The local porous sink is

$v_0>w$2

with a cell-wise closure

$v_0>w$3

Here

$v_0>w$4

and

$v_0>w$5

Performance was evaluated with aerodynamic porosity

$v_0>w$6

using planes at $v_0>w$7 and $v_0>w$8. Over grid resolutions $v_0>w$9 and $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$0, the proposed model reduced the standard deviation of $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$1 across grids to about $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$2, compared with about $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$3 for a general conventional constant-drag model and about $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$4 for an advanced constant-drag model (Tokiwa et al., 24 May 2026).

A direct branch-wise force model appears in fractal trees modeled as assemblies of cylindrical segments. For each branch,

$v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$5

Whole-tree coefficients are formed from branchwise sums,

$v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$6

Because branch Reynolds number scales with diameter, larger branches enter drag crisis first while smaller branches remain subcritical. For the tested L-system trees, the generation scaling laws are

$v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$7

The analysis indicates a whole-tree drag-crisis transition near $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$8 under uniform inflow and near $v(t)=w+\frac{v_0-w}{1+\gamma (v_0-w)(t-t_0)},$9 when inflow turbulence with $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$0 is imposed. Increasing structural complexity smooths the transition because smaller branches remain subcritical (Tokiwa et al., 30 Mar 2026).

5. Partition and sub-grid branch decompositions

For rough surfaces, the Transpiration–Resistance framework augments slip-length modeling with two constitutive drag-partition parameters: the shear correction factor $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$1 and the pressure correction factor $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$2. In the flat-interface case,

$$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$3

with

$$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$4

For curved interfaces such as a rough cylinder,

$$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$5

The same microscale cell problem used to compute slip parameters supplies $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$6, $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$7, $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$8, and $$r(t)=r_0 + w(t-t_0) + \frac{1}{\gamma}\ln\!\left[1+\gamma (v_0-w)(t-t_0)\right].$$9. In Couette flow, corrected viscous- and pressure-drag errors dropped from order-$n_{\mathrm{eff}}$00 uncorrected levels to values such as approximately $n_{\mathrm{eff}}$01 and $n_{\mathrm{eff}}$02 for ellipse roughness and $n_{\mathrm{eff}}$03 and $n_{\mathrm{eff}}$04 for square roughness at $n_{\mathrm{eff}}$05, $n_{\mathrm{eff}}$06, over $n_{\mathrm{eff}}$07 (Jain et al., 2022).

For canopy-like roughness, the Raupach partition is

$n_{\mathrm{eff}}$08

An isolated obstacle is characterized by shelter area and shelter volume,

$n_{\mathrm{eff}}$09

and the array density is expressed through

$n_{\mathrm{eff}}$10

In the infinite-area limit,

$n_{\mathrm{eff}}$11

If $n_{\mathrm{eff}}$12 and $n_{\mathrm{eff}}$13, the partition simplifies to

$n_{\mathrm{eff}}$14

The revisiting analysis interprets the classical multiplicative sheltering law as a stochastic renormalization rather than literal geometric subtraction of wake area. Mean fitted values differ by roughness type: plants gave approximately $n_{\mathrm{eff}}$15 and $n_{\mathrm{eff}}$16, whereas cubes gave approximately $n_{\mathrm{eff}}$17 and $n_{\mathrm{eff}}$18 (Buono et al., 2024).

In filtered Eulerian gas–solid drag, the branch structure is additive and sub-grid. A second-order Taylor expansion of the microscopic drag coefficient yields

$n_{\mathrm{eff}}$19

and, in the low-Re reduction,

$n_{\mathrm{eff}}$20

Here the gas drift branch is primarily responsible for drag reduction, the solid drift branch tends to attenuate that reduction, the scalar variance term increases filtered drag over the full solid-volume-fraction range, and the third-order term has the same coefficient as the variance term but mostly decreases drag because it is typically negative. This is one of the clearest examples of a formally additive analytical branch decomposition (Chen et al., 2018).

6. Thermodynamic and gravitational branch structures

In rapid solidification, the key branching variable is

$n_{\mathrm{eff}}$21

Using a phase-field-crystal-based amplitude model and a tanh approximation for the order parameter profile, the liquid concentration correction is

$n_{\mathrm{eff}}$22

with

$n_{\mathrm{eff}}$23

The segregation coefficient is

$n_{\mathrm{eff}}$24

If $n_{\mathrm{eff}}$25, then $n_{\mathrm{eff}}$26 never vanishes and complete trapping occurs only asymptotically as $n_{\mathrm{eff}}$27; this is the diffusive Aziz-like branch. If $n_{\mathrm{eff}}$28, then $n_{\mathrm{eff}}$29 decreases and vanishes at the finite trapping speed

$n_{\mathrm{eff}}$30

which defines the inertial or Sobolev-like branch. Solute drag is then represented thermodynamically through

$n_{\mathrm{eff}}$31

with the paper treating $n_{\mathrm{eff}}$32 as velocity-independent for a given material setting and assigning the velocity dependence primarily to $n_{\mathrm{eff}}$33 (Humadi et al., 2014).

In dusty self-gravitating discs, drag acts as a branch-coupling parameter between gas and dust. The control parameters are

$n_{\mathrm{eff}}$34

The drag-free two-fluid system has a quartic dispersion relation,

$n_{\mathrm{eff}}$35

where

$n_{\mathrm{eff}}$36

With drag but without backreaction, the problem becomes fifth order,

$n_{\mathrm{eff}}$37

and the branch interpretation is one-fluid versus two-fluid behavior with an additional drag-relaxation mode. In the weak-drag limit $n_{\mathrm{eff}}$38, the model recovers the two-fluid threshold; in the strong-drag limit $n_{\mathrm{eff}}$39, it approaches one-fluid behavior. The gas-to-dust transition is fitted by

$n_{\mathrm{eff}}$40

and, using

$n_{\mathrm{eff}}$41

the critical Stokes number is approximated as

$n_{\mathrm{eff}}$42

The main physical consequence is a dust-driven short-wavelength branch with Jeans masses much smaller than in standard gas-only GI; the paper states that the predicted mass is about $n_{\mathrm{eff}}$43 Earth masses in a representative regime (Longarini et al., 2022).

7. Boundary-layer and compliant-structure branches

In atmospheric boundary-layer modeling, the drag law is branch-wise in both vertical structure and stability class. The total stress magnitude is taken as

$n_{\mathrm{eff}}$44

and the spanwise stress is written piecewise: $n_{\mathrm{eff}}$45 with

$n_{\mathrm{eff}}$46

The inner branch is MOST-consistent,

$n_{\mathrm{eff}}$47

with $n_{\mathrm{eff}}$48 in the conventionally neutral branch and $n_{\mathrm{eff}}$49 in the stable branch. Matching the inner and outer solutions yields a self-consistent geostrophic drag law,

$n_{\mathrm{eff}}$50

The paper validates the model against 41 LES cases and reports excellent agreement for CNBL and good overall agreement for SBL, with the main discrepancy near the SBL top where LES retains non-zero stress above $n_{\mathrm{eff}}$51 (Narasimhan et al., 2023).

For side-by-side flexible blades, branch structure separates a static reconfiguration regime from a flutter regime. The governing nondimensional parameters are

$n_{\mathrm{eff}}$52

and

$n_{\mathrm{eff}}$53

For a bunch of $n_{\mathrm{eff}}$54 overlapping identical blades, the equivalent reduction is

$n_{\mathrm{eff}}$55

In the static branch, drag reduction is described by the reconfiguration number

$n_{\mathrm{eff}}$56

The experiments show that $n_{\mathrm{eff}}$57 decreases in the static regime starting at

$n_{\mathrm{eff}}$58

and settles to an almost constant value in the flutter regime at high $n_{\mathrm{eff}}$59. Increasing $n_{\mathrm{eff}}$60, $n_{\mathrm{eff}}$61, or $n_{\mathrm{eff}}$62 stabilizes the system and delays flutter. The paper’s further claim is that the reactive-force framework, which includes

$n_{\mathrm{eff}}$63

is superior to traditional Morison inertia because the convective reactive terms are necessary to capture both flutter onset and flutter-limited drag reduction (Wei et al., 6 Jan 2025).

Taken together, these works indicate that branch-wise analytical drag modeling is best understood as a general strategy of explicit regime separation. A common misconception is that such models are merely piecewise curve fits. The surveyed literature instead shows branch definitions derived from sign structure, dispersion relations, wake memory, hierarchy mappings, self-similar ridges, component partitions, and stability thresholds. Another common misconception is that branches are only kinematic; several of the cited models show branches generated by thermodynamic coupling, unresolved morphology, or sub-grid covariance structure. The main limitation, also repeated across domains, is that branchwise tractability is usually purchased by strong assumptions—constant background states, linear theory, independent-branch superposition, equivalent-element reductions, or precomputed constitutive tables—so the utility of any given branch model depends on whether those assumptions remain valid in the target regime.