Vegetation enhances curvature-driven dynamics in meandering rivers

Meandering rivers are among the most dynamic landforms and strongly influence Earth’s biogeochemical cycles. Classical hypotheses posit that riparian vegetation stabilizes riverbanks via root reinforcement and promotion of cohesive muds, facilitating the formation and persistence of single-thread meanders. Experimental work supports vegetation enabling transitions from braided to single-thread channels, and the emergence of meander indicators broadly coincides with the rise of land plants in the Paleozoic. Yet, modern observations and modeling show meanders can form without vegetation given other cohesive agents (mud, ice) and appropriate slopes, including in barren landscapes and on Mars. This raises the central research question: to what extent does vegetation, or its absence, directly and indirectly affect meander planform morphology and curvature–migration dynamics? The study aims to quantify vegetation’s influence by comparing planform metrics and migration–curvature relationships across rivers spanning a gradient of riparian vegetation density worldwide.

Prior work links vegetation to bank strength, mud production and retention, and the development of single-thread meanders; laboratory experiments demonstrated vegetation-maintained channels. Stratigraphic studies note the rarity of classic point-bar deposits before vascular plants, supporting a vegetation–meandering connection in Earth history. However, field evidence and physics-based models show meanders in unvegetated settings, including arid fans, permafrost terrains, and Martian analogs, with comparable geomorphic units (levees, splays, point bars). Additional literature documents curvature–migration relationships in vegetated systems, their breakdown at high curvature due to nonlinear hydrodynamics, and the roles of sediment supply, hydrological regime, and floodplain heterogeneity in meander dynamics. Recent studies highlight vegetation’s potential to slow migration, while aridity, flashy flows, and frequent avulsions in drylands promote chute cutoffs and limit meander maturity.

• Dataset: 54 single-thread meandering rivers across diverse biomes and climates, spanning three orders of magnitude in width and most continents. Inclusion criteria: at least 35 consecutive bends; located away from coastal backwater/tidal influence; minimal human modification.
• Imagery and digitization: Hand-digitized riverbanks from georeferenced aerial/satellite imagery (Google Earth Pro, SASPlanet, USGS Earth Explorer, QuickMapService in QGIS). Bank positions defined by vegetation boundaries in vegetated rivers; in unvegetated settings, outer banks from sharp cutbanks and inner banks from debris lines on point-bar tops. Verified channel width measurements against high-resolution topography in selected sites (r^2=0.98).
• Centerline and morphometrics: Derived smoothed centerlines (Savitzky–Golay filtering, spline resampling at ~0.1× average width). Computed local curvature C= dθ/ds and identified inflection points (C=0), apexes (local curvature maxima), half/full bends. Extracted morphometrics: intrinsic wavelength l, Cartesian wavelength L, amplitude A, radius of curvature R, sinuosity x=l/L, asymmetry θ=(l_u−l_d)/l. Normalized dimensional variables by average bend width B′ (L′/B′, l′/B′, A′/B′, R′/B′, C′/B′).
• Vegetation density and aridity: Derived multi-annual NDVI from Landsat 7 8-day composites (1999–2019), using 90th-percentile NDVI per pixel (NDVI90) within a buffer of 10× average width around centerlines; snow filtered via NDSI>0.4. Classified rivers: unvegetated (NDVI≤0.2; n=16), semi-vegetated (0.2<NDVI≤0.4; n=11), vegetated (NDVI>0.4; n=27). Computed median Aridity Index (AI) in the same buffer (Global Aridity v3, 1970–2000).
• Statistical analyses: Examined power-law scaling of morphometrics with width (z=a B^b) using linear regressions on log-transformed data. Compared classes via two-sample Wilcoxon Rank Sum and Kolmogorov–Smirnov tests at 5% significance. Conducted PCA on selected 17 morphometric variables (first–fourth moments of distributions of sinuosity x, asymmetry θ, width-adjusted curvature C, and intrinsic wavelength l) to assess multivariate separation by vegetation class.
• Migration analysis: Subset of 32 actively migrating rivers with image pairs at sufficient resolution. Computed lateral migration M using Dynamic Time Warping (DTW) alignment of centerlines (distance matrix includes spatial and curvature terms; curvature weighting parameter λ), masked cutoffs, smoothed M with Savitzky–Golay filter, and computed width-adjusted migration rate M_r = M/(Δt B). Accounted for curvature–migration spatial lag by shifting M by reach-average Δ_CM. Analyzed relationships between C and M_r for C≤0.3 (linear regime) and C>0.3 (saturation), using 50th and 95th percentiles of binned data. Further normalized migration by reach-average M_r to obtain m = M_r/⟨M_r⟩ to control for hydrologic/sedimentologic variability.

• Planform scaling: Meander wavelength (Cartesian and intrinsic), amplitude, and radius of curvature scale approximately linearly with channel width across all vegetation classes (power-law exponents ≈1). Prefactors differ slightly among classes.
• Class differences: Compared to vegetated rivers, unvegetated rivers exhibit longer width-adjusted wavelengths, larger width-adjusted radii of curvature, lower width-adjusted curvature, and lower sinuosity. For vegetated rivers, 26% of meanders have sinuosity x>2 and 13% have curvature C>0.3; in unvegetated rivers these fractions are 7% and 5%, respectively. Vegetated and semi-vegetated rivers show stronger upstream-skewed bends (more negative asymmetry), while unvegetated rivers tend to be more symmetric.
• Statistical significance: KS tests reject that unvegetated and semi-vegetated morphometrics are drawn from the same distributions as vegetated rivers; WRS tests show different medians in most parameters (exceptions: amplitude in unvegetated; wavelengths and curvature in semi-vegetated).
• Multivariate separation: PCA shows clustering separation between vegetated and unvegetated rivers, driven by higher sinuosity, curvature, and upstream skewness in vegetated systems; semi-vegetated rivers are intermediate and more dispersed.
• Migration rates: Width-adjusted migration rates are higher in unvegetated systems (median ± SD M_r = 0.019 ± 0.022 yr⁻¹) than semi-vegetated (0.012 ± 0.051 yr⁻¹) and vegetated (0.010 ± 0.058 yr⁻¹), indicating vegetation stabilizes banks and slows migration.
• Curvature–migration relationship: For C≤0.3, M_r scales linearly with curvature with positive slopes; coefficients of proportionality are higher in vegetated rivers than in unvegetated and semi-vegetated. For C>0.3, migration decreases or saturates with increasing curvature (negative/flat slopes), consistent with nonlinear hydrodynamic constraints, and differences among classes diminish.
• Normalized dynamics: Normalizing by reach-average M_r strengthens linearity and reduces scatter, especially for semi-vegetated rivers; intercepts are larger in semi-vegetated and unvegetated rivers, indicating relatively high migration even at low curvature. For C>0.3, all classes show plateauing behavior.
• Process interpretation: Vegetation stabilizes low-curvature reaches (pinning inflections), enabling growth in curvature and sinuosity toward morphodynamic maturity and neck cutoffs. In barren settings, reduced bank resistance, enhanced chute cutoffs, flashy hydrology, and frequent avulsions disrupt meander growth, elevate migration at low curvature, and weaken curvature–migration coupling.

The analyses demonstrate that vegetation modifies curvature-driven meander dynamics primarily at low curvatures (C≤0.3), strengthening the proportionality between curvature and migration while reducing baseline migration at low curvature. This stabilizing effect allows bends to continue growing in sinuosity and curvature, consistent with higher upstream skewness and more mature planforms in vegetated rivers. At high curvature (C>0.3), migration becomes insensitive to curvature across all classes due to nonlinear flow dynamics in sharp bends, so vegetation-related differences are minimal. Although planform morphologies overlap and confounding environmental gradients (e.g., aridity, hydrologic regime, sediment supply) co-vary with vegetation, the normalized curvature–migration relationships remain distinct across vegetation classes, indicating that vegetation exerts a robust control on meander dynamics independent of these factors. These dynamics influence floodplain reworking and channel mobility, with implications for carbon storage and fluxes, as well as for interpreting stratigraphic records and reconstructing paleo-hydrology on Earth and Mars.

Vegetation stabilizes riverbanks, reduces lateral migration rates, and enhances curvature-driven migration proportionality at low curvature, promoting sustained meander growth to higher sinuosity and curvature prior to neck cutoff. Unvegetated and sparsely vegetated rivers exhibit higher migration at low curvature, more frequent chute cutoffs, and avulsions that curtail meander maturation. These findings quantify vegetation’s morphodynamic imprint and should inform depositional models for unvegetated single-thread rivers and the interpretation of ancient fluvial deposits, including on early Earth and Mars. Future work should integrate grain-size, bank erodibility, and floodplain heterogeneity data; explicitly account for hydrologic regime and sediment supply; and extend analyses to longer timescales and diverse environmental contexts to resolve causality and feedbacks between vegetation, curvature dynamics, and avulsion regimes.

• Causality cannot be unequivocally established due to confounding factors: unvegetated rivers commonly occur in more arid climates with flashy hydrology, higher avulsion frequency, and greater propensity for chute cutoffs that limit meander growth.
• Remote sensing constraints: analyses do not explicitly quantify sediment grain size, bank material properties/erodibility, or floodplain heterogeneity, which can influence curvature–migration relations.
• Semi-vegetated class shows high variability; vegetation patterns (species, rooting depth, stem diameter) and channel size can modulate dynamics but are not resolved in detail.
• Some reaches include inactive paleochannels (e.g., Uzboy), with uncertain past vegetation states; though care was taken to ensure recency for others, temporal representativeness may vary.
• Curvature–migration relationships derived from two-image pairs per site and smoothed signals; while DTW improves alignment, methodological choices (e.g., curvature weighting, filtering) may affect estimates.