Scribd
Upload
41 views

Assignment4-2023 For Numerical Analysis

1. The document describes modeling atmospheric pollution using the finite element method. It presents the advection-diffusion equation governing pollutant transport, defines relevant parameters, and outlines the weak formulation and numerical implementation in FEniCSx. 2. The assignment asks students to complete code implementing the finite element method, run simulations, and analyze results. This includes refining the mesh, recording solutions for different time steps, and comparing cases with different wind velocities. 3. Students are to write a report discussing their results, including plots of the pollutant concentration over time and an explanation of how the solution varies with problem parameters.

Uploaded by

Pratyush Kumar Singh
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
41 views

Assignment4-2023 For Numerical Analysis

1. The document describes modeling atmospheric pollution using the finite element method. It presents the advection-diffusion equation governing pollutant transport, defines relevant parameters, and outlines the weak formulation and numerical implementation in FEniCSx. 2. The assignment asks students to complete code implementing the finite element method, run simulations, and analyze results. This includes refining the mesh, recording solutions for different time steps, and comparing cases with different wind velocities. 3. Students are to write a report discussing their results, including plots of the pollutant concentration over time and an explanation of how the solution varies with problem parameters.

Uploaded by

Pratyush Kumar Singh
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Assignment 4

Modeling atmospheric pollution with the


Finite-Element Method

Numerical Analysis For PDE’s (WI4014TU-2023/2024)

Figure 1: TROPOMI image of the atmospheric pollution above the Netherlands.


The TROPOspheric Monitoring Instrument (TROPOMI) is an instrument on board
the Copernicus Sentinel-5 Precursor satellite, launched on 13 October 2017. It pro-
vides daily data with 5 km spatial resolution about various pollutants: aerosoles,
carbon monoxide, formaldehyde, methane, nitrogen dioxide, ozone, and sulfur diox-
ide.

We want to model the spread of a polluting gas, such as NO2 , in the


atmosphere, see Figure 1 for an illustration. The computational domain Ω
is a (Lx × Ly ) rectangle with the lower-left corner at (xc , yc ). The time
interval of interest is t ∈ [0, T ]. The gas is emitted by two localized sources.
The emission rate density is given by the function:
(
0, t < 0;
f (x, y, t) = P2 2 2
 (1)
i=1 A exp −α(x − xi ) − α(y − yi ) , t ≥ 0.

The horizontal transport of the pollutant u(x, y, t) in the atmosphere is

1
described by the advection-diffusion equation:
∂u
− D∆u + ∇ · (uw) = f, t > 0, (x, y) ∈ Ω, (2)
∂t
where D is the effective diffusivity characterizing the (approximately) diffu-
sive transport of gas caused by the atmospheric turbulence, and w(x, y, t) =
⟨wx , wy ⟩ is the vector field of the mean wind velocity with its components
given by:
y
wx (x, y, t) = B − p , t ≥ 0,
x + y2
2
x (3)
wy (x, y, t) = p , t ≥ 0,
x + y2
2

where B is a constant. We assume that no pollutant was present in Ω prior


to the emission by the sources.

1. Formulation of the problem and its weak form

(a) Is this compressible or incompressible flow? To find out, calculate


∇ · w (show intermediate steps) and simplify the third term of
the equation (2), if possible.
(b) Assuming that the concentration of pollutant at the boundary of
Ω is zero at all times, write down, in formal mathematical terms,
the boundary value problem to be solved.
(c) Apply the Backward-Euler time integration method to your PDE,
writing down the recurrence relation. State the mathematical
properties of the chosen time integration method (without proofs).
(d) Derive the standard weak form of the Galerkin Finite-Element
Method for your time-discretized problem, showing intermediate
steps. Clearly indicate the bi-linear and linear forms. Define the
solution and test function-spaces.
(e) Briefly discuss what stability and other numerical issues can be
expected when solving these types of problems, and how you
intend to deal with these issues.

2. Implementation in FEniCSx and numerical experiments.

(a) The FEniCSx code structure for this is provided in the file
Assignment4-code-strycture.py. Some of the code parts have
been written explicitly, others are missing. Finish the code by
implementing the missing parts. You can use the tutorial at
https://jsdokken.com/dolfinx-tutorial/chapter2/heat equation.html

2
variable value
Lx 6
Ly 6
T 5
A 5
B 1.5
α 10
(xc , yc ) (−3, −3)
(x1 , y1 ) (−2.4, 0)
(x2 , y2 ) (0, 1)
D 0.0025

Table 1: Problem parameters.

(b) Set the problem parameters as in Table 1 and run the code with
nt = 10 and nx = 51, ny = 51. Insert the figure with the cal-
culated u(x, y, T ) in your report and discuss it.
(c) Refine the spatial discretization by changing nx and ny until you
are satisfied with the calculated result. Due to the structure of
the wind function, it is better to work with nx = 51, 61, 71,
... etc. Insert the figure of u(x, y, T ) at the final appropriate
spatial resolution and explain what criterion did you use to settle
on these values of nx and ny.
(d) Calculate the numerical approximation of u(x, y, T ) at the appro-
priate spatial resolution and with the time discretization set as
nt = 10, 20, 50, 100, 200, 300, recording the value of
u(x, y, T ) obtained with different nt at some chosen spatial lo-
cation. You can use a single value from the array unow obtained
by the code at the end of a run for this purpose. For example,
find the location of the maximum of unow at some nt and record
the value of that element of unow at other nt’s. Plot that value
as a function of nt and discuss it.
(e) Set B = 0.5 and run the code with the appropriate spatial reso-
lution established in the previous steps and with nt = 50. Insert
the figure of the calculated u(x, y, T ) in your report. Find a sim-
ple and visual way to explain the difference between the B = 1.5
and B = 0.5 cases.

You might also like