TURB-Smoke. A database of Lagrangian pollutants emitted from point - sources and dispersed in turbulent flows

Luca Biferale, Fabio Bonaccorso, Niccolò Cocciaglia, Robin A. Heinonen, Lorenzo Piro · 2025

Read TURB-Smoke. A database of Lagrangian pollutants emitted from point - sources and dispersed in turbulent flows by Luca Biferale, Fabio Bonaccorso, Niccolò Cocciaglia, Robin A. Heinonen, Lorenzo Piro (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-22749.txt.

Introduction

Luca Biferale and co-authors release a large data set on how pollution spreads in churning air, in a 2025 paper. Finding a leak or smoke source from a few sensor hits is hard, since gusts scatter the traces. The team built clean test data that search methods can be graded against.

The data come from direct simulations of full three dimensional flow. They track hundreds of millions of tracer particles released from five point sources, with and without a mean wind. Each record holds a particle path and the local flow speed along it. The set also gives smoothed density maps for quick tests of source finding methods.

Catalog record

CREATORLuca Biferale, Fabio Bonaccorso, Niccolò Cocciaglia, Robin A. Heinonen, Lorenzo Piro
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-22749

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

TURB-Smoke. A database of Lagrangian pollutants emitted from point - sources and dispersed in turbulent flows Luca Biferale, Fabio Bonaccorso, Niccolò Cocciaglia, Robin A. Heinonen, Lorenzo Piro Abstract Identifying the location and characteristics of pollution sources in turbulent flows is challenging, especially for environmental monitoring and emergency response, due to sparse, stochastic, and infrequent cue detection. Even in idealized settings, accurately modeling these phenomena remains highly complex, with realistic representations typically achievable only through experimental or simulation-based data. We introduce TURB-Smoke, a cutting-edge numerical dataset designed for investigating odor and contaminant dispersion in turbulent environments with and without mean wind. Generated via direct numerical simulations of the fully resolved three-dimensional Navier–Stokes equations, TURB - Smoke tracks hundreds of millions of Lagrangian particles released from five distinct point sources in fully developed turbulence, thus providing a reliable ground-truth framework for developing and evaluating source-tracking strategies using stationary sensors or mobile agents in realistic flows. Each particle’s trajectory is continuously tracked on many characteristic turbulence timescales, recording both the position and the local flow velocity. Additionally, we provide coarse-grained concentration fields in 3D and in quasi-2D slabs containing the source, ideal for quickly testing and optimizing search algorithms under varying flow conditions. 1 Background & Summary Atmospheric [ 1 , 2 ] and aquatic [ 3 , 4 ] pollutants pose grave threats to ecosystems and human health [ 5 , 6 ] . Predicting the dynamics of pollutants, containing their spread, and locating their sources are difficult and critical tasks that are extremely sensitive to the physics governing their mixing into the background flow. In both oceanic and atmospheric contexts, this flow is generically turbulent, with profound and complex consequences for the dispersion of pollutants and other chemicals such as odors [ 7 ] . Physically, chemicals immersed in a fluid can be modeled as passive scalars [ 8 ] and are transported by a combination of flow advection and molecular diffusion. Chemicals organize into a statistically steady but constantly fluctuating landscape [ 9 ] consisting of a correlated, evolving ensemble of puffs and filaments. Such landscapes exhibit strong spatiotemporal intermittency and nontrivial, non-Gaussian statistics such as power-law distributions of the timings between consecutive large-concentration events. Even in idealized scenarios, the precise mathematical modeling of these phenomena is extremely challenging, and realistic descriptions are currently only accessible through data from experiments and simulations. Beyond its relevance to environmental science, the dispersion of chemicals in turbulent flows also enjoys important applications in biology. Numerous animals have evolved strategies to track sources of chemical odors in geophysical flows, in the service of mating, foraging, or predation — a problem usually referred to as “olfactory search” [ 10 ] . These include aquatic animals such as plankton [ 11 ] and lobsters [ 12 ] , flying insects such as moths and fruit flies [ 13 ] (which instead search in an atmospheric setting), and mammals such as mice [ 14 ] and dogs [ 15 ] that integrate odor signals from the ground and air during search [ 16 ] . In all these settings, the properties and physics of the large-scale structure formed by the advected odor, called a plume, have crucial effects on the efficacy of various search strategies and, therefore, are intimately connected to the behavior of animals [ 17 , 18 ] . Motivated and inspired by ethology, a wide range of heuristic search strategies have been developed. A paradigmatic example is infotaxis [ 19 ] , a strategy framed in the context of Bayesian inference that has shown robust performance in a variety of environments for both single and multi-agent search [ 20 , 21 , 22 ] . These approaches typically operate under the assumption that the “agents” can use a model of the environment to interpret their encounters with the contaminant, which are assumed to be statistically independent, thus neglecting the strong spatiotemporal correlations inherent to turbulent transport that can substantially alter search efficiency [ 23 ] . To bypass the need for such assumptions, recent data-driven methods based on reinforcement learning (RL) have been developed, enabling individual agents to learn effective search policies directly from experience in complex and turbulent landscapes [ 24 , 25 ] . Thus, access to high-quality data is crucial for the serious study of the various problems associated with chemical dispersion in turbulent flows. To our knowledge, no such data are currently available to the public. Therefore, in this work, we present TURB-Smoke [ 26 ] , a database generated from high-fidelity direct numerical simulation (DNS) of a turbulent flow with fixed sources of (passive) pollutants. The chemicals are modeled as Lagrangian tracer particles (neglecting the relatively unimportant effect of molecular diffusion), whose positions and velocities are tracked in time. Both raw Lagrangian data and coarse-grained Eulerian concentration fields are provided in the database. To model oceanic currents and atmospheric wind, we include data with and without a background mean flow with several different magnitudes. In addition, each flow contains five independent particle sources, allowing the study of problems involving mixing of multiple contaminants. These data have been used in several publications devoted to developing and testing olfactory search strategies [ 20 , 23 , 27 , 22 ] , but the (physical) framework built for such investigations also offers a unique idealized test bed for many realistic atmospheric or oceanic environments contaminated by polluting substances or hazardous gas leaks. Our database has great potential utility for environmental applications, one of many being the design of robot strategies to detect and locate source fluid leakages [ 28 , 29 ] . 2 Methods All the data presented here are computer generated, so no experimental procedure is involved. The background turbulent flow is reproduced in a cubic, periodic domain with a pseudospectral simulation of the Navier-Stokes equation. The flow is seeded with particles, emitted in “puffs” that are passively advected by fluid motion and are further displaced by a constant mean wind. The 3D and 2D concentration fields are then obtained by counting particles on coarser grids. The following subsections present more detailed descriptions of the methods, along with all numerical parameters. 2.1 Simulation of the turbulent flow The first step of our procedure is to reproduce a turbulent environment into which the Lagrangian particles (i.e. particles transported by the underlying flow) will then be inserted. To this end, we performed a direct numerical simulation (DNS) of the incompressible Navier-Stokes (NS) equations. The NS equations read: \begin{cases}\partial_{t}\bm{u}+(\bm{u}\cdot\nabla)\bm{u}=-\bm{\nabla}p/\rho_{0}+\nu\Delta\bm{u}+\mathbf{F}\\ \bm{\nabla}\cdot\bm{u}=0,\end{cases} (1) where \bm{u}=\bm{u}(\bm{r},t) is the velocity field, p is the pressure, \rho_{0} the mean density of the fluid and \nu its kinematic viscosity. \mathbf{F} is the homogeneous and isotropic forcing that drives the system to a non-equilibrium steady state. The divergence-free condition in ( 1 ) enforces incompressibility. The domain is a cube of size L=2\pi , with periodic boundary conditions. The code follows a standard pseudospectral approach [ 30 ] , in which multiplications are performed in a real (or physical) space while derivatives are computed in Fourier space. To prevent aliasing errors from higher wave numbers - a common problem when dealing with Fourier transforms - a dealiasing procedure with the 2/3 rule is used. Time integration has been implemented with

Read the complete text