Perseus Sepahi

I study turbulence in the night-time atmosphere, and I write software: simulation tools, engineering systems, and apps for Mac, iPhone and Apple Watch.

GitHub Google Scholar LinkedIn ResearchGate Email

A night over open ground

Atmosphere Stable(clear night) Neutral(overcast, windy) Unstable(sunny day)

Geostrophic wind 16.5 m s⁻¹ Surface cooling 0.25 K h⁻¹

ColderPotential temperature θWarmer

How this is drawn

About 65,000 tracers are moved by your GPU through a sketch of the atmospheric boundary layer. Stable air (a clear night, the ground cooling) has a low-level jet, Kelvin–Helmholtz billows in the shear above it, and eddies that thin out and flatten as the ground cools. Neutral air (overcast and windy, neither heated nor cooled) has only turbulence made by the wind. Unstable air (a sunny day, the ground heating) has thermals rising from the warm ground, which a strong wind stretches into rolls. Wind-made turbulence grows with the wind; thermals do not depend on it. The mean wind is drawn to scale (about nine-tenths of the geostrophic wind aloft); eddies and thermals are drawn about four times stronger than in the real atmosphere so that they can be seen, keeping their proportions between conditions. Each tracer keeps the potential temperature of the height it started at, so mixing shows as folded colour, and the sky changes with the atmosphere. Zero cooling is the neutral case, so a stable night starts at 0.25 K h⁻¹. My thesis cases are winds of 8.25 and 16.5 m s⁻¹, either neutral or stable with cooling of 0.25, 0.5 or 1 K h⁻¹. The picture is an illustration; the results come from the simulations.

Research

The atmospheric boundary layer is the lowest part of the atmosphere, the part that feels the ground. Its turbulence carries heat, moisture, momentum and pollution between the surface and the air above, and it changes character over every day. Sunshine heats the ground and drives a convective layer a kilometre or two deep; after sunset the ground cools, the air above it becomes stably stratified, and the turbulent layer shrinks to a few tens or hundreds of metres.

I study how that turbulence responds to stratification, above all at night. With enough wind it stays continuous and follows similarity theory. With weak wind and strong cooling it turns patchy and intermittent, the air near the ground can decouple from the flow above, and the usual models stop working. Large-eddy simulation, which resolves the energy-carrying eddies and models only the smallest ones, is the main tool for studying that transition.

The flow at the top of this page follows the same physics. The resolved wind obeys the filtered Navier–Stokes equations in the Boussinesq approximation, (1)–(3), and the kinetic energy of the turbulence obeys the budget (4).

∇·ũ = 0

∂ũ∂t + (ũ·∇)ũ = −1ρ₀∇p̃ + g θ̃ − θ₀θ₀ k̂ − f k̂ × (ũ − U_g) − ∇·τ

∂θ̃∂t + ũ·∇θ̃ = −∇·q

∂e∂t = −u′w′ ∂U∂z − v′w′ ∂V∂zP + gθ₀ w′θ′B − ∂∂z(w′e + 1ρ₀w′p′)T − ε

Here ũ is the resolved wind, θ̃ the potential temperature, f the Coriolis parameter, U_g the geostrophic wind, and τ and q the stress and heat flux of the eddies too small for the grid. In (4), primes are fluctuations about the mean and bars are averages. Shear in the wind produces turbulence (P); buoyancy adds energy when the ground heats the air and removes it when the ground cools it (B); transport and pressure work move it up and down (T); and dissipation turns it into heat (ε). The contest between P and B is measured by the gradient Richardson number, Ri = N²/S²: where it is large, stratification wins and the turbulence fades.

In the flow above Stable (clear night): the heat flux at the ground is downward, so B < 0 and buoyancy takes energy from the turbulence.

Read my thesis

Software

Apps I design, build and release myself, from network tools to a video player. All free.

All software

Papers

2026
Stratification and Vertical Turbulence Structure in the Atmospheric Boundary Layer: A Large-Eddy Simulation Study

Perseus Sepahi

Master of Mathematics thesis, Department of Applied Mathematics, University of Waterloo

UWSpace

The stably stratified atmospheric boundary layer forms over land at night as the surface cools, and the buoyant suppression of turbulence makes it one of the hardest parts of the atmosphere to observe, simulate and parameterise. This thesis uses large-eddy simulation with the Advanced Research WRF model to study how the vertical structure and intensity of boundary-layer turbulence respond to increasing surface cooling, and how that response depends on mechanical forcing. Eight idealised simulations cross two geostrophic wind speeds (16.5 and 8.25 m/s) with four surface cooling rates (0, 0.25, 0.50 and 1.00 K/h). The strongly forced cases respond monotonically, consistent with the classical quasi-steady stable boundary layer. The weakly forced cases depart from that regime: the surface heat flux saturates near 10 W/m² despite a fourfold increase in imposed cooling, the surface friction velocity collapses, and the flow aloft decouples from the surface.
```
@mastersthesis{sepahi2026thesis,
  title   = {{Stratification and Vertical Turbulence Structure in the Atmospheric Boundary Layer: A Large-Eddy Simulation Study}},
  author  = {Perseus Sepahi},
  year    = {2026},
  school  = {University of Waterloo},
  address = {Waterloo, Ontario, Canada},
  url     = {https://hdl.handle.net/10012/24382},
}
```
2023
Obstacle-induced lateral dispersion and nontrivial trapping of flexible fibers settling in a viscous fluid

Ursy Makanga, Mohammadreza Sepahi, Camille Duprat, Blaise Delmotte

Physical Review Fluids, 8(4), 044303

DOI

We study the sedimentation of flexible fibers through arrays of obstacles in a viscous fluid, combining experiments and numerical simulations. Obstacles induce lateral dispersion of the fibers and, depending on fiber flexibility and obstacle geometry, non-trivial trapping states.
```
@article{makanga2023obstacle,
  title   = {{Obstacle-induced lateral dispersion and nontrivial trapping of flexible fibers settling in a viscous fluid}},
  author  = {Ursy Makanga and Mohammadreza Sepahi and Camille Duprat and Blaise Delmotte},
  year    = {2023},
  journal = {Physical Review Fluids},
  volume  = {8},
  number  = {4},
  pages   = {044303},
  doi     = {10.1103/PhysRevFluids.8.044303},
}
```

2023

Sedimentation of Flexible Fibers in Structured Environments

Ursy Makanga, Mohammadreza Sepahi, Camille Duprat, Blaise Delmotte

SIAM Conference on Computational Science and Engineering (CSE23)

HAL record

@inproceedings{makanga2023siam,
  title     = {{Sedimentation of Flexible Fibers in Structured Environments}},
  author    = {Ursy Makanga and Mohammadreza Sepahi and Camille Duprat and Blaise Delmotte},
  year      = {2023},
  booktitle = {SIAM Conference on Computational Science and Engineering},
  url       = {https://hal.science/hal-04438332},
}

All papers

About

I hold an M.Math in Applied Mathematics from the University of Waterloo (Fluid Mechanics Group). My thesis used large-eddy simulation with WRF to study how the stable night-time boundary layer responds to surface cooling and wind forcing.

Before Waterloo I completed an M.Sc. in Mechanical Engineering at Amirkabir University of Technology and spent a summer at LadHyX (École Polytechnique) on flexible fibers settling in viscous fluids, which led to a paper in Physical Review Fluids.

I now work as a quality engineering contractor at Westport Fuel Systems, where I also built the quality team’s reject-tracking and NCMR software. On my own time I write native apps for macOS, iOS and watchOS. Everything I publish here is free.

Now: Quality engineering at Westport Fuel Systems, and shipping my own apps in the evenings.
Interests: Atmospheric boundary layers and large-eddy simulation
Stratified turbulence and geophysical fluid dynamics
Wind energy and offshore wind turbines
Scientific machine learning for fluid problems
Native Apple platform development
Tools: WRF / LES, OpenFOAM, COMSOL, Gmsh, Python, MATLAB, Swift / SwiftUI, JavaScript, SQL, Linux, LaTeX

Open the CV