Fractional Brownian Motion¶
Model a Gaussian path with stationary increments whose scale and dependence are governed by a fixed Hurst parameter.
Core Idea¶
Fractional Brownian motion (fBm) is a centered Gaussian random path with stationary increments and a fixed Hurst parameter \(H\in(0,1)\) governing self-similar scaling. After unit-variance normalization, an increment over lag \(\tau\) has variance \(|\tau|^{2H}\); time rescaling by \(a\) rescales amplitude by \(a^H\) in distribution. At \(H=1/2\) it is ordinary Brownian motion. For \(H>1/2\) separated increments are positively correlated; for \(H<1/2\) they are negatively correlated. The path values themselves are not stationary even though increments are.[^ref-8190658e85cc]
The model is a statistical law, not a jagged image or a universal claim about any empirical time series. Gaussianity and the full covariance structure matter as much as an apparent scaling exponent. A particular terrain generator or numerical simulation is not part of the definition.[^ref-8190658e85cc]
Scope of Application¶
Mandelbrot and Van Ness proposed fBm partly to model persistent scaling in hydrological cumulative-flow fluctuations; an observed Hurst exponent is motivation to check the model, not proof of an exact fit. In computer graphics, a related spatial fractional-Brownian field can provide rough synthetic terrain, but Musgrave's original work explains why naive homogeneous fields omit real terrain's spatial heterogeneity and erosion. The field extension and rendered landscape are not literally the same object as a one-dimensional fBm path.[ref-8190658e85cc][ref-ba2d7c693c52]
Leland and colleagues discuss fractional Gaussian noise—the increment process related to fBm—as one model for self-similar Ethernet traffic. Neither their data nor every high-\(H\) trace is automatically an fBm trajectory. The frozen Wikipedia candidate Fractal landscape, and its proposed variant fractal surface, remain distinct unresolved source identities rather than aliases or proven coverage of this reframed process.[^ref-e089e6ad6502]
Clarity¶
The centered Gaussian law plus covariance fixes the process's joint distributions; a rough appearance or one slope estimate does not. The standard covariance is \(\tfrac12(|s|^{2H}+|t|^{2H}-|t-s|^{2H})\). It yields stationary increments and their lag-variance scaling. Independence is the special Brownian \(H=1/2\) case, not a property of every fBm. The terms “stationary process” and “stationary increments” must not be exchanged.[^ref-8190658e85cc]
Manages Complexity¶
One \(H\) and one variance normalization replace a large table of pairwise dependence values with a covariance family. This compresses model comparisons across scales, but can conceal finite-scale behavior, non-Gaussian shocks or changing terrain roughness. A fitted model must specify the scales and evidence that support it; ideal mathematical self-similarity is not a guarantee that a river or mountain range follows one law everywhere.[ref-8190658e85cc][ref-ba2d7c693c52]
Abstract Reasoning¶
Start with centered Gaussian finite-dimensional distributions, fixed \(H\), and a zero-origin normalization. Test whether increments over equal lags share a distribution and whether their variance follows the $2H$ exponent. Derive correlation signs from the covariance rather than declaring persistence from a picture. For data, separate the proposed law from goodness of fit; for synthetic terrain, name the spatial field and any later shaping or erosion steps. The process is defined by its distribution, not by midpoint displacement or spectral-synthesis software.[ref-8190658e85cc][ref-ba2d7c693c52]
Knowledge Transfer¶
Hydrological modeling and spatial terrain synthesis can share a Gaussian scale-law idea without sharing the same index domain or proving their observed data exact. The original fBm is time-indexed; a terrain height model requires a spatial-field extension. Fractional Gaussian noise in network traffic is a related increment model, not a synonym. The staged strict parent is the live Continuous-time stochastic process, which already descends from the Stochastic Process prime; no edge is inferred merely from overlap with Fractal Geometry or other surface methods.[ref-8190658e85cc][ref-ba2d7c693c52][^ref-e089e6ad6502]
[^ref-8190658e85cc]: Benoit B. Mandelbrot and John W. Van Ness, “Fractional Brownian Motions, Fractional Noises and Applications,” SIAM Review 10(4) (1968), 422–437, original article, §§1–3 and 5. [^ref-ba2d7c693c52]: F. Kenton Musgrave, Methods for Realistic Landscape Imaging, Yale dissertation, 1993, indexed Chapter 1 Figure 1.1 discussion and Chapter 2 Terrain Modelling introduction; full PDF not fetched in this review, so terrain claims remain bounded for later reference clearance. [^ref-e089e6ad6502]: Will E. Leland, Murad S. Taqqu, Walter Willinger and Daniel V. Wilson, “On the Self-Similar Nature of Ethernet Traffic,” SIGCOMM 1993, original paper, §§3.2 and 5.4.
Relationships to Other Abstractions¶
Current abstraction Fractional Brownian Motion Domain-specific
Parents (1) — more general patterns this builds on
-
Fractional Brownian Motion is a kind of Continuous-time stochastic process Domain-specific
Fractional Brownian motion is a Gaussian continuous-time stochastic process with a fixed H-indexed law.
Hierarchy path (1) — routes to 1 parentless root
- Fractional Brownian Motion → Continuous-time stochastic process → Stochastic Process
Neighborhood in Abstraction Space¶
Fractional Brownian Motion sits in a sparse region of the domain-specific corpus (85th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2551 abstractions)
Nearest neighbors
- Hurst Exponent — 0.83
- Lorden's Inequality — 0.82
- Exponentially Modified Gaussian Distribution — 0.82
- Linnik distribution — 0.81
- Brownian Skorokhod Embedding — 0.81
Computed from structural-signature embeddings · 2026-10-08