pith. sign in

arxiv: 2511.22430 · v2 · submitted 2025-11-27 · 📊 stat.AP

Spatial constraints improve filtering of measurement noise from animal tracks

Alexandre Delporte , Susanne Ditlevsen , Adeline Samson This is my paper

Pith reviewed 2026-05-17 04:59 UTC · model grok-4.3

classification 📊 stat.AP
keywords animal movementspatial constraintsstochastic differential equationKalman filterparticle filtermeasurement noiseArgos telemetrybowhead whale
0
0 comments X

The pith

Incorporating spatial boundaries into a movement model sharpens filtering of noisy animal position data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a latent movement model based on an underdamped Langevin stochastic differential equation augmented with an extra drift term that keeps the animal inside a known fixed spatial domain. This constraint is incorporated to improve the accuracy of recovering true positions from noisy satellite observations such as Argos telemetry. The authors implement both an extended Kalman filter and a particle filter, using splitting schemes to solve the SDE, and demonstrate the gain on real data from a bowhead whale. A sympathetic reader would care because more accurate tracks support better understanding of habitat use and movement patterns in bounded environments like seas or fenced areas.

Core claim

The central claim is that adding an additional drift term to the underdamped Langevin SDE to enforce a known spatial domain improves the accuracy of filtering noisy observations of the positions; this is shown by comparing filtered estimates to unconstrained versions on both simulated data and a real Argos track of a bowhead whale in Foxe Basin.

What carries the argument

Underdamped Langevin SDE with an added drift term that encodes the hard spatial boundary constraint, solved via splitting schemes to enable filtering.

If this is right

  • Filtered position estimates more closely recover the true paths that respect the domain boundaries.
  • The particle filter option effectively handles heavy-tailed non-Gaussian measurement errors.
  • The method applies directly to aquatic animals in water bodies or terrestrial animals in fenced or natural restricted zones.
  • Practical implementation is achieved through splitting schemes that approximate the latent dynamics for both Kalman and particle filters.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This approach could reduce bias in downstream ecological estimates such as home-range size or energy budgets derived from tracks.
  • It might be tested on other boundary-constrained systems like vehicles on road networks to check generalizability.
  • Direct comparison of error reduction against existing barrier-aware movement models would quantify the specific gain from the drift term.

Load-bearing premise

The animal remains strictly inside a known fixed spatial domain at all times and the added drift term enforces this without distorting the base movement dynamics.

What would settle it

Generate simulated tracks strictly inside a known domain, add realistic non-Gaussian noise, then compute root-mean-square error of filtered positions; if the constrained model shows no lower error than the unconstrained model, the accuracy improvement claim is falsified.

Figures

Figures reproduced from arXiv: 2511.22430 by Adeline Samson, Alexandre Delporte, Susanne Ditlevsen.

Figure 1
Figure 1. Figure 1: Part of a trajectory simulated from model ( [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Trajectory from (2.1) with a mixture of Gaussian potentials using the Lie-Trotter approximation scheme. Parameters are h = 1/3600, λ = h 0.8 , τ = 1, ν = 5, ω = 0.1. 5 Filtering algorithms In this section, we present algorithms for estimating the latent states of the movement process from noisy observations. We first consider the classical Extended Kalman Filter for Gaussian measurement errors and then gen… view at source ↗
Figure 3
Figure 3. Figure 3: Results of filtering for 50 simulated trajectories with additive Gaussian noise. (a) [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Results of filtering for 20 simulated trajectories with additive Student noise. (a) [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Results of filtering for 20 simulated trajectories with additive Argos X-shaped noise. [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
read the original abstract

Advances in tracking technologies for animal movement require new statistical tools to better exploit the increasing amount of data. Animal positions are usually calculated using the GPS or Argos satellite system and include potentially non-Gaussian and heavy-tailed measurement error patterns. Errors are usually handled through a Kalman filter algorithm, which can be sensitive to non-Gaussian error distributions. We introduce a latent movement model through an underdamped Langevin stochastic differential equation (SDE) that includes an additional drift term to ensure that the animal remains in a known spatial domain of interest. This can be applied to aquatic animals moving in water or terrestrial animals moving in a restricted zone delimited by fences or natural barriers. We demonstrate that the incorporation of these spatial constraints into the latent movement model can improve the accuracy of filtering for noisy observations of the positions. We implement an Extended Kalman Filter as well as a particle filter adapted to non-Gaussian error distributions. Our filters are based on solving the SDE through splitting schemes to approximate the latent dynamic. We illustrate the approach on a real Argos telemetry track of a bowhead whale in Foxe Basin, Canada.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes augmenting an underdamped Langevin SDE with an additional confining drift term to enforce known spatial boundaries in a latent movement model for animal telemetry. It develops an Extended Kalman Filter and a particle filter (both using splitting-scheme discretizations of the SDE) to handle non-Gaussian measurement errors, and claims that the spatial constraints improve filtering accuracy for noisy position observations, illustrated on a single real Argos track of a bowhead whale in Foxe Basin.

Significance. If the improvement in accuracy is substantiated through controlled validation, the approach would provide a useful way to incorporate domain knowledge into state-space models for telemetry data, particularly for aquatic or fenced animals. The adaptation of filters to non-Gaussian errors and the splitting-scheme approximation of the SDE are technical strengths that could be of interest to the movement ecology community.

major comments (2)
  1. [Results] Results section (the bowhead whale illustration): the manuscript presents filtered tracks with and without the spatial constraint but supplies no quantitative error metrics, no simulation study in which latent paths are generated from the model, corrupted with known noise, and scored against ground truth, and no statistical comparison of accuracy; consequently the central claim that the constraints 'improve the accuracy of filtering' rests on an unquantified visual demonstration.
  2. [Methods] Model formulation (the additional drift term): the assumption that the confining drift encodes a hard spatial constraint without distorting the underlying Langevin dynamics is stated but not tested; no sensitivity analysis, no derivation showing preservation of the original friction/diffusion parameters, and no controlled experiment isolating the drift's effect on position recovery are provided.
minor comments (2)
  1. [Abstract] Abstract: states that the constraints 'can improve the accuracy' but gives no quantitative results, error metrics, or comparison details, which is inconsistent with the level of evidence actually shown.
  2. [Model] Notation: the precise functional form of the confining drift (strength, functional dependence on position) is introduced without an equation number or explicit definition that can be directly referenced in the filtering recursions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments and for recognizing the potential utility of our approach for incorporating domain knowledge into state-space models for telemetry data. We address each major comment below and describe the revisions we will make.

read point-by-point responses

  1. Referee: [Results] Results section (the bowhead whale illustration): the manuscript presents filtered tracks with and without the spatial constraint but supplies no quantitative error metrics, no simulation study in which latent paths are generated from the model, corrupted with known noise, and scored against ground truth, and no statistical comparison of accuracy; consequently the central claim that the constraints 'improve the accuracy of filtering' rests on an unquantified visual demonstration.

    Authors: We agree that the current real-data illustration relies on visual comparison and lacks quantitative error metrics or a controlled simulation study. For the bowhead whale Argos track, the true latent path is unknown, which precludes direct quantitative scoring against ground truth. To strengthen the central claim, we will add a simulation study to the revised manuscript. Latent paths will be generated from the underdamped Langevin model with the confining drift, corrupted with synthetic non-Gaussian measurement errors calibrated to Argos characteristics, and then filtered both with and without the spatial constraint. Accuracy will be assessed using metrics such as root mean squared error and continuous ranked probability score, with statistical comparisons across multiple replicates. This will provide quantitative evidence isolating the benefit of the spatial constraints. revision: yes

  2. Referee: [Methods] Model formulation (the additional drift term): the assumption that the confining drift encodes a hard spatial constraint without distorting the underlying Langevin dynamics is stated but not tested; no sensitivity analysis, no derivation showing preservation of the original friction/diffusion parameters, and no controlled experiment isolating the drift's effect on position recovery are provided.

    Authors: The confining drift is constructed to be negligible in the interior of the domain and to activate only near the known boundaries. We acknowledge that the manuscript does not include a formal derivation or sensitivity analysis. In the revision we will add a derivation showing that the friction and diffusion coefficients of the original Langevin dynamics are preserved away from the boundaries. We will also include a sensitivity analysis varying the strength and functional form of the confining term, together with a controlled simulation experiment that isolates the drift's contribution to position recovery accuracy while holding all other model components fixed. revision: yes

Circularity Check

0 steps flagged

No circularity: spatial constraint is an external modeling choice, not derived from data or self-referential inputs

full rationale

The paper introduces an underdamped Langevin SDE augmented with an additional drift term as a modeling assumption to enforce a known fixed spatial domain. This is presented as an external choice applicable to aquatic or fenced animals rather than derived from the observations or fitted in a self-referential way. Filters (EKF and particle filter) are implemented via standard splitting schemes for the SDE, with no indication that predictions reduce to inputs by construction, no load-bearing self-citations for uniqueness theorems, and no renaming of known results. The illustration on real Argos data serves as demonstration rather than a tautological validation loop. The derivation chain is self-contained and independent of the target filtering improvement.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The model rests on the assumption that the animal's movement can be represented by an underdamped Langevin SDE augmented with a confining drift whose functional form is chosen by the authors.

free parameters (2)
  • friction and diffusion coefficients in the Langevin SDE
    Standard parameters of the base SDE that must be chosen or estimated from data.
  • strength and form of the additional confining drift
    New term introduced to enforce the spatial boundary; its scaling is not specified in the abstract.
axioms (2)
  • domain assumption The true position process obeys an underdamped Langevin SDE inside the domain.
    Invoked to define the latent movement model.
  • domain assumption The animal never leaves the known spatial domain.
    Required for the confining drift to be valid.

pith-pipeline@v0.9.0 · 5491 in / 1358 out tokens · 22246 ms · 2026-05-17T04:59:27.126399+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

  1. [1]

    Th´ eo Michelot, Richard Glennie, Catriona Harris, and Len Thomas

    URLhttps://arxiv.org/abs/2406.15195. Th´ eo Michelot, Richard Glennie, Catriona Harris, and Len Thomas. Varying-Coefficient Stochastic Differential Equations with Applications in Ecology.Journal of Agricultural, Biological and Environmental Statistics, 26(3):446–463, September

    work page arXiv

  2. [2]

    We use the following formulas for matrix and vector differentiation

    25 Filtering for penalized SDEs A Derivation of the gradient and Hessian of the potential We derive the gradient and Hessian matrix of the potential H(x) =− JX j=1 Hj(x) withH j(x) =α j exp(−(x−x ∗ j)⊤Bj(x−x ∗ j)). We use the following formulas for matrix and vector differentiation. For anyϕ:R→R, a:R d →R,v:R d →R d,B∈M d,d(R), ∇ϕ(a(x)) =ϕ ′(a(x))∇a(x) (A...

    work page 2019