Scienti c African 27 (2025) e02587

Contents lists available at ScienceDirect

Scientific African
journal homepage: www.elsevier.com/locate/sciaf

Computational fluid dynamics assessment of natural ventilation in

three types of large-scale broiler poultry houses in Botswana
Pius Emesu *, Justin Hakgamalang Chepete , Ellen Letsogile Thipe
Department of Agricultural and Biosystems Engineering, Botswana University of Agriculture & Natural Resources, Private Bag 0027, Gaborone,
Botswana

A R T I C L E I N F O A B S T R A C T

Editor: DR B Gyampoh Effects of house design, orientation, and regional location on the natural ventilation of com­
mercial, large-scale broiler poultry houses in the North-eastern and South-eastern farming regions
Keywords: of Botswana were assessed at full scale through computational fluid dynamics (CFD) simulations
Computational fluid dynamics (CFD) using the OpenFOAM CFD code. Local air change effectiveness (ACE), εL was computed in a 0.25
Local air change effectiveness (ACE)
m high animal occupation zone inside gable-, hoop-, and seesaw-roofed houses selected from
Natural ventilation
three different farms per region. Observed airflow patterns and εL values did not show regional
House orientation
Wind incidence angle (WI) dependence, whereas both were influenced by house orientation and, to a lesser extent, house
Broiler poultry house design. Local ACE values obtained from the North-eastern farms ranged from 0.62 to 0.75, while
in the South-eastern region values ranged from 0.56 to 0.90, with hoop- and gable -roofed
structures yielding better εL values than the seesaw-roofed structures from both regions. Simu­
lations of theoretical perpendicular wind incidence (WI) angles for houses oriented to 140 and
135◦ reduced their ACE values by 18.75 % and 0.53 %, respectively, indicating that their current
orientation was optimal and allowing them to experience high WI. Houses oriented to 100◦ and
105◦ exhibited moderate improvements in εL values of 5.84 % and 6.4 %, respectively, while
houses oriented to 95◦ and 90◦ showed the greatest improvement in εL values of 19.9 % and 37.52
%, respectively. Thus, site-specific orientation of poultry houses to 120◦ -150◦ would be recom­
mended as it would allow for higher WI angles and better εL in both geographical farming zones in
Botswana. Furthermore, it may be concluded that design proposals to improve natural ventilation
based on the hoop- and gable-roofed structures – supported by CFD simulations with experi­
mental validation – is warranted.

Introduction

Over the last four decades Botswana’s poultry industry has recorded steady growth, with broiler chicken meat production peaking
at 9360 metric tonnes in the year 2001, after which a steady decline has been recorded to the level of 3696 tonnes reported for the year
2020 [1,2]. The sector is dominated by a handful of large-scale producers who keep flocks of >50, 000 birds per production cycle [3].
Most of the producers raise their chickens in naturally ventilated housing structures, but there is a modest investment in environ­
mentally controlled housing units amongst them. Typically, the broiler houses are >70 m long, having dwarf side walls, with

* Correspondence author at: Department of Agricultural and Biosystems Engineering, Botswana University of Agriculture & Natural Resources,
Private Bag 0027, Gaborone, Botswana.
E-mail addresses: [email protected], [email protected] (P. Emesu).

https://doi.org/10.1016/j.sciaf.2025.e02587
Received 26 September 2024; Received in revised form 14 December 2024; Accepted 10 February 2025
Available online 12 February 2025
2468-2276/© 2025 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
P. Emesu et al. Scienti c African 27 (2025) e02587

Table 1
Physical dimensions of broiler houses.
House Type Length (m) Width (m) Dwarf wall height (m)
North-eastern region

Gable 100.0 9.12 0.48

Hoop 35.6 6.8 0.43
See-saw 90.0 12.0 0.54
South-eastern region
Gable 100 8.9 0.2
Hoop 101.7 9.0 0.3
See-saw 80.2 8.0 0.4

Fig. 1. Clipped isometric views of house models. Top (a) Gable; Middle (b) Hoop; Bottom (c) Seesaw.

mesh-covered side openings having retractable curtains for natural ventilation, and they are roofed with galvanised iron sheets,
usually without insulation.
Historical data from the World Bank Climate Change Data Portal [4] indicates that in the period 1901–1990, Botswana experienced
only four months out of the year with temperatures exceeding 25 ◦ C, but from 1991 to 2016, the number of months experiencing
summer-type temperatures above 25 ◦ C has increased to six, running from October to February. Environment statistics for 2018–2019
published by Statistics Botswana [5] indicated that mean maximum temperatures ranged between 29 and 38 ◦ C. Rearing poultry
intensively – especially broilers which have poor tolerance of heat stress [6] – under this climate regime raises concern not only from a
bird-welfare perspective, but also from a production efficiency standpoint.
According to Department for Environment Food and Rural Affairs [7] key features of broiler housing that would help to protect
birds from hot weather include insulation, house design and location, and the ventilation system in use. For naturally ventilated
structures, adequate ventilation principally depends on the exposure and orientation of the structure relative to prevailing winds [8].
Bruce [9] states that advances in studies of the theory and practice of natural ventilation have shown it to be as predictable,
controllable, dependable, and effective as mechanical ventilation, with a greater advantage of it being cheaper.
Studies on air flows in agricultural housing structures in Botswana, particularly broiler poultry housing at a commercial scale, are
scanty hence the importance of this study to fill this knowledge gap. A full-scale experimental setup is costly, complicated and would
interfere with production workflow in industry. CFD simulations of the environment in agricultural housing structures offers a less-
costly, minimally intrusive, versatile, and valid complementary tool to circumvent the aforementioned hurdles. Best-practice guide­
lines and reports on successful implementation of CFD simulations to improve ventilation performance of agricultural housing
structures are available in the literature [10–12]. The simulations entail solving the fundamental equations describing the processes of
fluid flow and heat and mass transfer to provide spatial dynamics of climate variables such as velocity, temperature, and humidity [13,

2
P. Emesu et al. Scienti c African 27 (2025) e02587

14]. From the simulations, ventilation may be described and visualized using air velocity and temperature distribution patterns or flow
field data may be extracted to calculate parameters such as air change rate and air change effectiveness [15]. In their comprehensive
review of mechanisms of natural ventilation in livestock buildings, Rong et al. [16] stated that CFD has been applied mostly to improve
designs of ventilation systems and to study the effects of parameters on air flow rate.
In the present study, air flow in three types of large-scale commercial broiler poultry houses was analysed at full scale using
OpenFOAM, an open source CFD code [17], to establish the effects of house type, orientation, and geographical location on the
effectiveness of wind-driven natural ventilation in the current structures. The aim was to provide a basis for proposals to improve their
designs for better thermal performance. Simulations were based on the fully coupled simulation approach, wherein outdoor wind flow
and indoor air flow were computed simultaneously within the same computational domain, after van Hoof et al. [18] and Norton et al.
[19].

Materials and methods

Description of physical poultry houses

Large-scale broiler poultry houses situated on farms owned by two of the leading poultry producers were selected from both the
North-eastern and South-eastern agricultural production regions of Botswana for long-term monitoring of indoor and exterior envi­
ronmental variables to obtain the baseline data which was then used to complement and validate the CFD simulations. Three house
types differing by their roof designs (gable, hoop and seesaw) were used in the study and their physical dimensions are presented in
Table 1, while their isometric models are depicted in Fig. 1 for clarity.

On-site weather monitoring

Wind speed and direction were measured over three summer-time production cycles. At each site, an HP 1000 (Shenzhen Fine
Offset Electronics Co., Ltd, P.R. China) portable weather station was mounted in an unobstructed location outside the house at 1.8 m
height and it logged data at hourly intervals throughout the growth cycle of each of the flocks. Wind speed and direction data was
captured and used to plot wind roses using the OpenAir library in the R Statistical Environment [20]. Internal air flows were measured
using three vane anemometers (Model AZ8905, AZ Instrument Corp., Taiwan) lined up in the middle of the broiler houses in the
South-North direction, and, for the seesaw roof type, one additional device was mounted at the ridge opening.

CFD modelling setup

Computational setup
An HP Z440 (HP, USA) workstation was used for the study. It featured an 8-core, Intel Xeon E5–1660v3 CPU, with frequency and
RAM of 3.00 GHz and 20 MB, respectively. The machine also had a 2GB NVIDIA Quadro graphics processor and 128 GB of DDR4 RAM.
The open-source Scientific Linux (Version 7.5) operating system was installed within which other open-source applications were
deployed for the CAD generation and numerical simulations. OpenFOAM Version 8.0 [17] was used for the CFD simulations.

Computational grid generation and verification

Based on the house dimensions and satellite imagery from Google Maps [21], FreeCAD (Version 0.18) was used to produce the CAD
geometry, which was exported in the “Standard for the Exchange of Product Data” (STEP) format for grid generation. Salome Platform
(Version 8.4.0) was used to construct the computational grid which included the houses being monitored and any other that were
within a distance <10 times the height of the structure in question [16]. The in-built Gmsh 3D meshing algorithm was employed for
surface triangulation and the resulting mesh was exported in the stereolithography (stl) format for reading into the OpenFOAM
environment.
In OpenFOAM, a computational domain measuring 300 × 300 × 100 m in the x, y, and z axes was generated using the in-built
blockMesh utility and partitioned into a coarse background mesh with 10 m sized cells. Next, using the snappyHexMesh utility, the
house geometries were snapped to the background mesh created using blockMesh. Feature refinement was conducted at level 10 since
levels below this led to some structures being poorly modelled. Surface refinement was set to level 8 while region-wise refinement
levels ranged from 0 to 4. For all cases, further refinement at the edges of the structures was achieved through addition of three layers.
Expansion ratio was set to 1.2 and the thickness of the smallest cell was set to 1.0E-3 m. The physical model was simplified by omitting
birds, architectural elements such as columns, and equipment [19].
To verify the quality of the grid, the grid convergence index (GCI) of Roache [22] recommended by Rong et al. [12] and reviewed
by NASA [23] was computed using Eqs. (1) to (4) as follows:
F s |ε |
GCIfine = (1)
(rp − 1)

Fs |ε|rp
GCIcoarse = (2)
(rp − 1)

3
P. Emesu et al. Scienti c African 27 (2025) e02587

(f2 − f1 )
ε= (3)
f1

h2
r= (4)
h1

where Fs is a safety factor set to Fs = 1.25; ε is the relative error; f1 is the variable value at a given point in a fine grid, and f2 is the
variable value at the same point in the coarse grid, respectively; r is the grid refinement ratio, with h1 being some measure of grid
spacing or representative grid size for the fine grid and h2 is the representative grid size for the coarser one; p is the order of grid
convergence, which was taken as 2 for second order numerical schemes.
Richardson extrapolation was used to compute the variable f0, i.e. variable value at grid spacing of h = 0, using Eq. (5), after which
variable values f were plotted against the representative grid size to graphically confirm grid independence [23].
f1 − f2
fh=0 ≅ f1 + (5)
rp − 1

Numerical schemes
The finite volume approach was used to solve the steady state Reynolds Averaged forms of the Navier-Stokes equations for the
conservation of mass and momentum [13,24], presented as Eqs. (6) and (7):
∂ρ
+ ∇.(ρV) = 0 (6)
∂t

DV
ρ = − ∇.ρ + μ ∇2 .V + ρf (7)
Dt

where ρ is the fluid density (kg m− 3) and V is the fluid velocity (ms− 1), t represents time (s), f represents net body forces acting on the
mass of the fluid, and μ is the dynamic fluid viscosity (m2 s− 1).
Within the control volumes, the governing equations were discretized and solved linearly for each relevant variable using different
solvers provided in the OpenFOAM installation [17]. The pressure equation was solved using the generalised geometric-algebraic
multi-grid (GAMG) solver, while the velocity (u), turbulent kinetic energy (κ), turbulent kinetic energy dissipation (ε) and age of
air (θ) were solved using the preconditioned bi-conjugate gradient (PBiCG) solver. Second order discretisation of convective and
divergent terms of the derivatives was done using the bounded Gauss upwind schemes.
The simpleFoam solver, an implementation of the semi-implicit method for pressure-linked equations (SIMPLE) algorithm, was
used to couple the equations for momentum and mass conservation, with convergence criteria for all residuals set to 1e-6. Turbulence
effects were modelled using the standard κ-ε model, reported to perform well in simulations of flows in agricultural buildings [25].

Initial and boundary conditions

The boundary condition within the atmospheric boundary layer (ABL) was based on a logarithmic wind profile. A power law model
was used to compute the inlet reference wind velocity (URef) at the domain inlet patch using Eq. (8) after Flores & Muñoz [26] and Liu
et al. [27], as follows:
( )n
z
U(z) = URef (8)
zRef

where U(z) is velocity at height z, URef is the wind velocity at reference height zref (m) and n was taken as 0.25.
Other ABL parameters included: surface roughness length, z0 of 0.1 m and the minimum coordinate value in z direction, zGround was
0 m. Pressure at the inlet and outlet boundaries was constant at a value of 0. Wall boundary conditions were implemented on the
ground and solid building surfaces, while slip boundary conditions were specified at the domain top and side patches. To model viscous
effects on flow at the wall boundaries, epsilonWallFunction, kqRWallFunction, nutkAtmRoughWallFunction which are standard
OpenFOAM wall functions [17], were implemented for the ε, κ, and ν (turbulent viscosity) flow fields. Initial values for ε, κ, and ν were,
respectively, 0.01, 1.3, and 0.

Local air change effectiveness computation

According to ASHRAE [28] and Cehlin et al. [29], air change effectiveness (ACE), ε is a measure of an air distribution system’s
ability to deliver ventilation air to a building, zone, or space.
Ventilation conditions with perfect mixing where outdoor air is well distributed in the ventilated space corresponds to ACE = 1,
with lower magnitudes reflecting less-than-ideal air distribution. According to ASHRAE [28], local air change effectiveness, εL shows
the effectiveness of outdoor air delivery to one specific point in space. Thus, an εL value of 1 indicates that the air distribution system
delivers air to the space in a manner equivalent to that of a system with perfectly mixed air, with lower magnitudes representing less
than perfect mixing. ACE may also be defined as the ratio of the average age of air in a room when the air is fully mixed, compared to
the average age of air at breathing height, and may be used to indicate how well air is distributed within the breathing height [30,31].
Local ACE, εL was, therefore, calculated using Eqs. (9) and (10), as follows:

4
P. Emesu et al. Scienti c African 27 (2025) e02587

Table 2
House orientation and measured wind data.
House Type House orientation azimuth (◦ ) Prevailing wind direction (◦ ) Wind incidence to house ridge (◦ ) Maximum wind speed (m s− 1)
North-eastern region

Gable 95 60 35 7.9
Hoop 100 60 40 6.5
See-saw 140 60 80 10.4
South-eastern region
Gable 105 60 45 6.1
Hoop 135 30 105 6.0
See-saw 90 90 0 8.5

Table 3
Mesh quality parameters for all computational simulation grids.
Roof Number of structures in Region refinement Total number of cell Hexahedral Max Aspect Max Average NO
Type domain level volumes cells (%) ratio Skewness *

North-eastern region
Gable 2 1 12,705,140 65.92 9.69 4.85 15.91
Hoop 4 2 6318,130 63.90 7.91 5.94 15.28
See-saw 1 1 6846,408 72.36 10.58 5.06 13.90
South-eastern region
Gable 2 1 12,416,822 66.80 7.76 3.08 16.07
Hoop 2 3 12,722,396 68.93 8.84 4.46 14.52
See-saw 3 1 17,439,064 70.65 7.57 4.54 13.87
*
NO = nonorthogonality.

Table 4
Grid convergence study parameters and results.
Case number Region refinement level Grid spacing (m) Number of cells in computational domain ACE GCI (%)

1 0 10 17,435,081 0.554 –
2 1 5 17,439,064 0.558 0.97
3 2 2.5 17,475,107 0.559 0.28
4 3 1.25 17,775,169 0.555 0.24
5 4 0.625 20,187,383 0.559 0.25
– – 0 – 0.560* –
*
ACE value computed using Eq. (5).

Table 5
Experimental validation calculation.
Validation site Measurement Experimental maximum Predicted maximum REb, Bounded relative REm, Mean error,
No. point velocity (ms− 1) velocity (ms− 1) error (%) bounded (%)

1 1 6.5 3.72 40 39
2 3.8 2.09 42
3 1.8 1.16 34
4 2.9 1.65 41
2 1 6.1 2.79 49 47
2 3.2 1.74 43
3 3.9 0.99 63
4 2.8 1.91 31
3 1 6.0 3.46 40 58
2 7.6 1.26 68
3 3.3 0.82 64
4 6.2 1.75 62
1 8.5 6.14 27
2 3.9 2.16 42
4 3 4.6 1.70 56 44
4 4.8 2.51 44
5 2.5 3.90 51

τn
εL = (9)
LMAAOZ

5
P. Emesu et al. Scienti c African 27 (2025) e02587

Fig. 2. Dependence of ACE on grid spacing, h (m).

Fig. 3. Visualization of air flows in Gable-roofed structure, North-Eastern Region. Background coloured by age of air (s) field, vectors coloured by
velocity magnitude (m/s). (a) Streamline and velocity vector plots at z = 0.25 m, 35◦ WI. (b) Velocity vectors at x = 0 m, 35◦ WI. (c) Streamline and
velocity vector plots at z = 0.25 m, theoretical house orientation for 90◦ WI. (d) Velocity vectors at y = 0 m, theoretical house orientation for 90◦ WI.

V
τn = (10)
Q

where τn (s) is the nominal time constant, or the shortest time air would take to traverse the domain from inlet to outlet; LMAAOZ, (s) is
the local mean age of air (LMA) computed in an animal-occupied zone (AOZ); V (m3) is the volume of the domain; and Q (m3 s− 1) is the
volumetric flow rate at the inlet patch. LMA is defined as the average time for air to travel from the supply inlet area to a given location

6
P. Emesu et al. Scienti c African 27 (2025) e02587

Fig. 4. Visualization of air flows in Gable-roofed structure, South-Eastern Region. Background coloured by age of air (s) field; vectors coloured by
velocity magnitude (m/s). (a) Streamline and velocity vector plots at z = 0.25 m, 45◦ WI. (b) Velocity vectors at x = 0 m, 45◦ WI. (c) Streamline and
velocity vector plots at z = 0.25 m, theoretical house orientation for 90◦ WI. (d) Velocity vectors at y = 0 m, theoretical house orientation for 90◦ WI.

in the domain [32]. It was computed from the age of air, θ (s) defined as the time taken for a particle to convect from an inlet to a given
location in the flow. As of OpenFOAM version 8, age of air is included as a function object which is solved as a flow field [17]. The AOZ
was taken as the entire floor area up to a height of 0.25 m which was deemed as average breathing height of broiler chickens after
Küçüktopcu & Cemek [33]. In this zone, 1500 – 2100 data points were sampled using the plotOverLine filter in ParaView, and their
average calculated to yield LMAAOZ which is deemed to provide a measure of the total residence time of air in the local environment at
the house centre as would be experienced by the birds [15].

Post-processing of results
Simulation results were post-processed and evaluated using the in-built functions, namely, age and flowRatePatch[17]. Visual­
isation and plotting of age of air and velocity flow fields was accomplished using ParaView, version 5.11.0 [34].

CFD simulation validation

Experimental measurements of air flows from four sites in both the Northern and Southern regions were compared to the simulation
results and the mean relative error bounded (REm) was calculated per site, according to Hong et al. [25] and Kat & Els [35] using Eqs.
(11) and (12) as follows:
∑
N REb
REm = (11)
N
⃒p − m⃒
⃒ ⃒
REb = tanh⃒ ⃒ (12)
m

where p is the predicted or simulated velocity value, m is the measured velocity value and REb is the bounded relative error.

Results and discussion

Measured wind profile data

Wind profile data measured during summer growing seasons is presented in Table 2, which shows prevailing winds in both farming

7
P. Emesu et al. Scienti c African 27 (2025) e02587

Fig. 5. Visualization of air flows in Hoop-roofed structure, North-Eastern Region. Background coloured by age of air (s) field; vectors coloured by
velocity magnitude (m/s). (a) Streamline and velocity vector plots at z = 0.25 m, 40◦ WI. (b) Velocity vectors at x = 0 m, 40◦ WI. (c) Streamline and
velocity vector plots at z = 0.25 m, theoretical house orientation for 90◦ WI. (d) Velocity vectors at y = 0 m, theoretical house orientation for 90◦ WI.

regions to range from being Easterly to North-easterly, with maximum wind speeds ranging from 6.0 to 10.4 ms− 1. Considering house
orientation with respect to prevailing winds, it is evident that wind incidence angles ranged from 0 to 105◦ , influences of which on
natural ventilation are described in further detail in subsequent sections hereafter.

Physical modelling and spatial discretisation

Mesh quality parameters for cases simulating the current housing configurations at all farms studied are presented in Table 3. As
shown, the meshes for sites with hooped roofs needed higher region-wise refinement levels of 2 and 3 for the North and South lo­
cations, for successful modelling of the roof curvature. Maximum aspect ratios of the grids ranged between 7.57 and 10.58, while
maximum skewness ranged from 3.08 to 5.94. Zikanov [13] has recommended a range of 0.2–5 for aspect ratio for both structured and
unstructured grids, while Hong et al. [25] recommended a maximum internal skewness of four. Although some grids reported herein
exceeded these recommendations, it did not adversely affect the simulation results as the set convergence criteria were met. Average
nonorthogonality values ranged between 13.87 and 16.07, which are generally low and acceptable. The ideal requirement for this
parameter is full orthogonality, and a low value of nonorthogonality is thus recommended [13]. Furthermore, all meshes used in this
study were of high topological quality as they consisted predominantly of hexahedral cells as recommended by Franke et al. [36].

Grid convergence study and simulation validation

Results of spatial discretisation verification based on the GCI of Roache [22] are presented in Table 4, while the validation of
simulation accuracy is presented in Table 5. GCI values obtained in this study were generally low, ranging from 0.24 to 0.97 %, which
suggested that the grids were of good quality and did not negatively impact the results obtained. Hong et al. [25] have reported GCI
values ranging from 7.3 to 38.2 % showing that the values obtained in the current report are good and acceptable. Because the dif­
ference in GCI values between region refinement levels 1 and 2 was only 0.69 %, while further refinement to level 3 produced a
marginal difference of 0.04 %, results from all three levels were deemed to be similar and hence all levels were utilised as needed in
subsequent simulations.
Relative error as reported in Table 5 was high, being attributable to low wind speeds, but was similar to results reported by Hong
et al. [25] who documented values ranging between 30 and 45 % in the majority of simulations they carried out on natural ventilation
in agricultural buildings. The metric is useful, however, as it avoids division by zero error.

8
P. Emesu et al. Scienti c African 27 (2025) e02587

Fig. 6. Visualization of air flows in Hoop-roofed structure, South-Eastern Region. Background coloured by age of air (s) field; vectors coloured by
velocity magnitude (m/s). (a) Streamline and velocity vector plots at z = 0.25 m, 105◦ WI. (b) Velocity vectors at y = 0 m, 105◦ WI. (c) Streamline
and velocity vector plots at z = 0.25 m, theoretical house orientation for 90◦ WI. (d) Velocity vectors at y = 0 m, theoretical house orientation for
90◦ WI.

Furthermore, a plot of ACE against grid spacing is presented in Fig. 2, showing that as the grid spacing reduced, the simulated ACE
values gradually approached the zero grid spacing ACE value calculated as 0.56, indicating that results from region refinement levels
1–3 would yield acceptably comparable results.

Air flow distribution analysis

Graphics depicting air flow vectors, streamline traces and age of air in the horizontal (z-normal) and vertical (x- or y-normal) planes
are shown in Figs. 3–8, for both the extant house configurations and theoretical orientation of the houses to achieve 90◦ wind inci­
dence. In all grids, air flow into the computational domains is in the -x direction, the buildings having been rotated appropriately
during grid construction in the Salome environment.
The domains with gable-roofed structures from both the North-eastern (Fig. 3) and South-eastern (Fig. 4) regions which experi­
enced oblique wind incidences (Table 2) of 35◦ and 45◦ relative to the houses ridge axes, respectively, exhibited similar recirculating
flow patterns situated on the windward sides of the buildings, a pattern reportedly typical for such structures [15]. The South-eastern
site exhibited two counter-rotating vortices at the building’s eaves with air flow directed vertically downwards, which could poten­
tially improve air mixing at the AOZ. Streamlines at z = 0.25 m were deflected slightly from the windward side of the building
channelling air currents along the length of the building. Also, examination of the age of air colour profile shows that air stagnated in
regions close to the longitudinal ends of the houses. A similar observation of vorticity was seen in the hoop-roofed building from the
North-eastern region (Fig. 5) which experienced wind incidence of 40◦ but was not observed in the South-eastern region (Fig. 6) where
the wind was incident to the building ridge axis at 105◦ . At this site, streamlines at z = 0.25 m reveal unidirectional airflow across the
building width, with minimal deflection by the windward dwarf wall. Also, air replenishment was uniform along the length of the
building as revealed by the uniform colour profile of the age of air field.
At the sites with gable- and hoop-roofed buildings from both regions (Figs. 3–6), theoretical orientation of the buildings to achieve
perpendicular flow resulted in visible change in flow patterns as seen from the straight streamlines and evenly distributed air velocity
vectors. To achieve this practically would require increasing building orientation azimuths from the current (Table 2) 95, 105, and

9
P. Emesu et al. Scienti c African 27 (2025) e02587

Fig. 7. Visualization of air flows in Seesaw-roofed structure, North-Eastern Region. Background coloured by age of air (s) field; vectors coloured by
velocity magnitude (m/s). (a) Streamline and velocity vector plots at z = 0.25 m, 80◦ WI. (b) Velocity vectors at y = 0 m, 80◦ WI. (c) Streamline and
velocity vector plots at z = 0.25 m, theoretical house orientation for 90◦ WI. (d) Velocity vectors at y = 0 m, theoretical house orientation for 90◦ WI.

100◦ for the North-east gable, South-east gable and North-east hoop sites, respectively, to 150◦ given that prevailing winds at these
sites blow at 60◦ azimuths, which could be recommended for new poultry house sitings in these locations. Such a recommendation is in
line with the report of Barrington et al. [8] who found that orientation of livestock buildings to 120 and 150◦ gave significantly better
summer ventilation performance for temperatures greater than or equal to 20 ◦ C compared to orientations of 0, 30, 60 and 90◦ .
The seesaw-roofed structures in both regions exhibited smooth flow patterns as seen by the unidirectional streamlines and evenly
distributed velocity vectors (Fig. 7 and Fig. 8). In the North-eastern region, wind was almost perpendicular (80◦ ), whereas it was
parallel to the house ridge axis at the South-eastern farm (Table 2). Since prevailing wind in the North-eastern site was North-easterly
(60◦ azimuth) and the building was oriented at ca. 140◦ , this produced almost perpendicular wind incidence, resulting in crossflow
through the building. Age of air analysis at z = 0.25 m, however, reveals significant stagnation on the leeward side of the building,
probably caused by shielding from the significantly tall (0.54 m) side walls (Table 1). In contrast, however, the prevailing wind in the
South-eastern site was Easterly (90◦ azimuth), with an East-West building orientation which resulted in parallel flow of air along the
length of the building, thus delaying its evacuation from the animal-occupied zone (z = 0.25 m).
Assessment of the effect of changing the house orientation to achieve perpendicular wind incidence in the seesaw-roofed structure
from the North-eastern site reveals minor change in air flow pattern and age of air distribution. This is because the current house
orientation gives almost perpendicular wind incidence (80◦ ), given that the house is oriented at ca. 140◦ azimuth and would need only
a slight alteration of 10◦ . This, therefore, suggests that the present house orientation lies at an optimal azimuth, which could be
recommended for broiler poultry housing at a location with a similar prevailing wind pattern [8]. In the South-eastern site, however,
changing the house orientation to achieve perpendicular wind incidence significantly improves the air flow pattern and age of air
distribution, but would require orientation of the house in the North-South direction since the prevailing winds are Easterly (90◦
azimuth). This is not practicable, however, as the large side openings would admit direct sunlight for prolonged hours as the sun
traversed the building overhead, which would stress the exposed birds.

Air change effectiveness analysis

Results showing air change effectiveness (ACE) as a measure of natural ventilation are presented in Table 6. In both regions studied,
the hoop-roofed structures exhibited higher ACE values, with the structure from the South-eastern region exhibiting the highest ACE
value, while the seesaw-roofed structure from the South-eastern region had the lowest value, suggesting independence of ACE on

10
P. Emesu et al. Scienti c African 27 (2025) e02587

Fig. 8. Visualization of air flows in Seesaw-roofed structure, South-Eastern Region. Background coloured by age of air (s) field; vectors coloured by
velocity magnitude (m/s). (a) Streamline and velocity vector plots at z = 0.25 m, 0◦ WI. (b) Streamline and velocity vectors at x = 0 m, 0◦ WI. (c)
Streamline and velocity vector plots at z = 0.25 m, theoretical house orientation for 90◦ WI. (d) Velocity vectors at y = 0 m, theoretical house
orientation for 90◦ WI.

Table 6
Air change effectiveness (ACE) for current and simulated perpendicular wind incidence.
House Type Current conditions Simulated conditions Change in ACE (%)

HO ( )◦
ACE HO ( )◦
ACE

North-eastern
Gable 95 0.62 150 0.74 19.9
Hoop 100 0.75 150 0.79 5.84
See-saw 140 0.712 150 0.708 − 0.53
South-eastern
Gable 105 0.74 150 0.78 6.4
Hoop 135 0.90 120 0.73 − 18.75
See-saw 90 0.56 180 0.77 37.52

HO: House ridge orientation azimuth.

geographical region. Also, scrutiny of the results shows that ACE was influenced by wind incidence angle, rather than wind speeds
experienced, as seen from the seesaw-roofed structure from the South-eastern region with a wind incidence of 0◦ which had the lowest
ACE value of 0.56. This is corroborated by the report of Norton et al. [19] on natural ventilation of livestock buildings who found that
distribution of local mean age of air showed increasing levels of mixing as WI approached 90◦ . Furthermore, in their analysis of
ventilation efficiencies of different designs of greenhouses in Korea, Hong et al. [37] reported increasing ventilation rates with
increased wind incidence from 0 (parallel flow) to 45 and 90◦ for designs with relatively small roof vents, while ventilation rates for
designs with large roof openings were not negatively affected by changes in wind direction as the large roof openings allowed them to
have comparatively large air inflows.
Imposing perpendicular wind incidence angles of 90◦ at the hoop-roofed structure in the South-eastern region and the seesaw-
roofed structure from the North-eastern region which experienced, 105 and 80◦ wind incidence angles, respectively, caused their
ACE values to reduce, indicating that their current orientations with respect to the prevailing winds experienced locally were optimal.
On the other hand, the structure which recorded the greatest improvement in ACE was the seesaw-roofed house from the South-eastern
region which experienced parallel WI and, therefore, its orientation could potentially be increased by 90◦ . However, as previously
stated, this is not a practicable proposition due to excessive exposure of the house interior to solar radiation during the day, therefore
increasing the danger of thermal stress to the birds.

11
P. Emesu et al. Scienti c African 27 (2025) e02587

These results suggest that orientation of houses at azimuths between 105 and 150◦ would allow for high wind incidence of between
45 and 120◦ in both production regions studied herein. The results also suggest that the roof designs studied herein did not significantly
affect the air change effectiveness, but rather that it was influenced by the angle of house orientation which would present the highest
possible wind incidence normal to the sides of the buildings [8]. Furthermore, it has been reported elsewhere by the authors [38] that
birds reared in the hoop structures studied herein had better production performance in terms of feed consumption, average daily
weight gain and water consumption, while birds reared in the gable-roofed structure in the North-eastern region showed better
economic performance, suggesting that the two house designs may be considered for further design optimisation studies.

Conclusions

Natural ventilation in three types of commercial broiler poultry houses commonly used in Botswana has been studied using
computational fluid dynamics simulations backed with minimal experimental measurements. The study has made available for the
first time, an objective visualisation and quantification of natural ventilation in these broiler houses, paving the way for objective
considerations for improvement in their design for better bird comfort and improved production. Analysis of airflow patterns and ACE
as a measure of natural ventilation did not show regional dependence, but both parameters were influenced by house orientation and,
to a lesser extent, house design.
Sites which experienced low WI (0–35◦ ) yielded low ACE values, while those which experienced WI between 40 and 105◦ produced
higher ACE, hence better natural ventilation. At high WI (80–105◦ ), unidirectional transverse airflow was observed with uniform age
of air distribution throughout the houses. Theoretically increasing the house orientation azimuth to achieve perpendicular wind
incidence greatly improved the ACE at two sites with house orientations of 90 and 95◦ which experienced prevailing winds blowing at
60 and 90◦ azimuths, respectively.
It may, therefore, be concluded that orientation of houses at farm sites in Botswana should lie between 120 and 150◦ considering
local prevailing winds to allow for high WI angles that produce crossflows through the buildings for uniform air distribution and high
air change effectiveness in the animal-occupied zone. Furthermore, it may be recommended that proposals for improvement in house
designs can be based on the hoop- and gable-roofed designs studied herein, as their superior production and economic performance
have already been reported elsewhere by the authors [38].

CRediT authorship contribution statement

Pius Emesu: Methodology, Investigation, Data curation, Formal analysis, Visualization, Software, Writing – original draft. Justin
Hakgamalang Chepete: Project administration, Resources, Conceptualization, Methodology, Investigation, Data curation, Formal
analysis. Ellen Letsogile Thipe: Methodology, Investigation, Data curation, Formal analysis.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgements

The team is grateful for the cooperation and support of the management and staff of Tswana Pride (Pty) Ltd, Kwena Pride (Pty) Ltd,
and Oistins Farms (Pty) Ltd, for availing their farmhouses and production data. Also, the support of the Botswana Poultry Producers
Association is duly acknowledged and appreciated.

References

[1] Food and agriculture organization of the united nations, ‘Production /crops and livestock products - Metadata’, Rome, 2022. Accessed: Apr. 27, 2022. [Online].
Available: https://www.fao.org/faostat/en/#data/QCL/metadata.
[2] Roman. Grynberg, Masedi. Motswapong, Competition and Trade policy : the Case of the Botswana poultry Industry, BIDPA, 2011.
[3] J.C. Moreki, T. Montsho, A review of egg production in Botswana, J. World’s Poultry Res. J. World’s Poult. Res 1 (1) (2011) 4–6.
[4] World Bank Climate Change Knowledge Portal, ‘Botswana - Climatology’, Washington, 2021. Accessed: Apr. 27, 2021. [Online]. Available: https://
climateknowledgeportal.worldbank.org/country/botswana/climate-data-historical.
[5] Statistics Botswana, Botswana Environment Statistics: Climate Digest, Statistics Botswana, Gaborone, 2019 [Online]. Available, www.statsbots.org.bw.
[6] S.K. Maloney, Heat storage, not sensible heat loss, increases in high temperature, high humidity conditions, Worlds. Poult. Sci. J. 54 (4) (1998) 350–352,
https://doi.org/10.1079/WPS19980024.
[7] Department for Environment Food and Rural Affairs, ‘Heat stress in poultry solving the problem’, London, 2005. [Online]. Available: www.defra.gov.uk.
[8] S. Barrington, N. Zemanchik, Y. Choiniere, Orienting livestock shelters to optimize natural summer ventilation, Trans. ASAE 37 (1) (1994) 251–255.
[9] J.M. Bruce, Environmental Control of livestock housing, in: E.H. Bartali, A. Jongebreur, D. Moffitt (Eds.), CIGR Handbook of Agricultural Engineering Volume II -
Part I Livestock Housing and Environment, vol. II, American Society of Agricultural Engineers, St Joseph, Michigan, 1999, pp. 54–67, ch. 2.
[10] I.-B. Lee, S. Saaase, S.-H. Sung, Evaluation of CFD accuracy for the ventilation study of a naturally ventilated broiler house, Jpn Agric Res Q 41 (1) (2007) 53–64,
https://doi.org/10.6090/jarq.41.53.
[11] I.-H.H. Seo, et al., Improvement of the ventilation system of a naturally ventilated broiler house in the cold season using computational simulations, Biosyst.
Eng. 104 (1) (2009) 106–117, https://doi.org/10.1016/j.biosystemseng.2009.05.007. Sep.
[12] L. Rong, P.V. Nielsen, B. Bjerg, G. Zhang, Summary of best guidelines and validation of CFD modeling in livestock buildings to ensure prediction quality,
Comput. Electron. Agric. 121 (2016) 180–190, https://doi.org/10.1016/j.compag.2015.12.005. Feb.

12
P. Emesu et al. Scienti c African 27 (2025) e02587

[13] O. Zikanov, Essential Computational Fluid Dynamics, John Wiley and Sons Inc., Hoboken, 2010.
[14] F. Rojano, P.E. Bournet, M. Hassouna, P. Robin, M. Kacira, C.Y. Choi, Assessment using CFD of the wind direction on the air discharges caused by natural
ventilation of a poultry house, Environ. Monit. Assess. 190 (12) (2018), https://doi.org/10.1007/s10661-018-7105-5. Dec.
[15] T. Norton, J. Grant, R. Fallon, D.W. Sun, Assessing the ventilation effectiveness of naturally ventilated livestock buildings under wind dominated conditions
using computational fluid dynamics, Biosyst. Eng. 103 (1) (2009) 78–99, https://doi.org/10.1016/j.biosystemseng.2009.02.007.
[16] L. Rong, B. Bjerg, T. Batzanas, G. Zhang, Mechanisms of Natural Ventilation in Livestock buildings: Perspectives on Past Achievements and Future Challenges,
Academic Press, 2016, https://doi.org/10.1016/j.biosystemseng.2016.09.004. Nov. 01.
[17] C. Greenshields, OpenFOAM v8 User Guide, The OpenFOAM Foundation, London, UK, 2020. Accessed: Jul. 22, 2020. [Online]. Available, https://doc.cfd.
direct/openfoam/user-guide-v8.
[18] T. van Hooff, B. Blocken, On the effect of wind direction and urban surroundings on natural ventilation of a large semi-enclosed stadium, Comput. Fluids. 39 (7)
(Aug. 2010) 1146–1155, https://doi.org/10.1016/j.compfluid.2010.02.004.
[19] T. Norton, J. Grant, R. Fallon, D.W. Sun, Assessing the ventilation effectiveness of naturally ventilated livestock buildings under wind dominated conditions
using computational fluid dynamics, Biosyst. Eng. 103 (1) (2009) 78–99, https://doi.org/10.1016/j.biosystemseng.2009.02.007.
[20] R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2021. Nov. 014.1.2.
[21] Google, ‘Google Maps’. Accessed: apr. 20, 2018. [Online]. Available: https://maps.google.com.
[22] P.J. Roache, Perspective: a method for uniform reporting of grid refinement studies, J. Fluids. Eng. 116 (1994) 405–413.
[23] NASA, ‘Examining Spatial (Grid) Convergence’, NPARC Alliance CFD verification and validation website. Accessed: jul. 22, 2024. [Online]. Available: https://
www.grc.nasa.gov/WWW/wind/valid/tutorial/spatconv.html.
[24] J.C. Tannehill, D.A. Anderson, R.H. Pletcherm, Computational Fluid Mechanics and Heat Transfer, Taylor & Francis, Philadelphia, 1997.
[25] S.-W. Hong, et al., Validation of an open source CFD code to simulate natural ventilation for agricultural buildings, Comput. Electron. Agric. 138 (2017) 80–91,
https://doi.org/10.1016/j.compag.2017.03.022. Jun.
[26] F. Flores, R. Garreaud, R.C. Muñoz, CFD simulations of turbulent buoyant atmospheric flows over complex geometry: solver development in OpenFOAM,
Comput. Fluids. 82 (2013) 1–13, https://doi.org/10.1016/j.compfluid.2013.04.029. Aug.
[27] S. Liu, W. Pan, H. Zhang, X. Cheng, Z. Long, Q. Chen, CFD simulations of wind distribution in an urban community with a full-scale geometrical model, Build.
Environ. 117 (2017) 11–23, https://doi.org/10.1016/j.buildenv.2017.02.021.
[28] ASHRAE, 2017 ASHRAE Fundamentals (SI Edition), ASHRAE, Atlanta, 2017.
[29] M. Cehlin, U. Larsson, H.J. Chen, Numerical investigation of air change effectiveness in an office room with impinging jet ventilation, in: 4th International
Conference On Building Energy, Environment, 2018, pp. 641–646.
[30] M.F. King, et al., Investigating the influence of neighbouring structures on natural ventilation potential of a full-scale cubical building using time-dependent
CFD, J. Wind Eng. Indust. Aerodynamics 169 (2017) 265–279, https://doi.org/10.1016/j.jweia.2017.07.020. Oct.
[31] K. Gungor, Guide to air change effectiveness, Ecolibrium (2013) 32–39. MarAccessed: Dec. 04, 2024. [Online]. Available, https://theecolibrium.com/.
[32] K.S. Kwon, I.B. Lee, Assessment of ventilation performance in agricultural facilities using “age of air” concept and CFD technology. Acta Horticulturae,
International Society for Horticultural Science, 2013, pp. 191–200, https://doi.org/10.17660/actahortic.2013.1008.25. Oct.
[33] E. Küçüktopcu, B. Cemek, Evaluating the influence of turbulence models used in computational fluid dynamics for the prediction of airflows inside poultry
houses, Biosyst. Eng. 183 (2019) 1–12, https://doi.org/10.1016/j.biosystemseng.2019.04.009. Jul.
[34] U. Ayachit, The ParaView Guide: A Parallel Visualization Application, Kitware Inc., 2015.
[35] C.J. Kat, P.S. Els, Validation metric based on relative error, Math. Comput. Model. Dyn. Syst. 18 (5) (2012) 487–520, https://doi.org/10.1080/
13873954.2012.663392. Oct.
[36] J. Franke, A. Hellsten, H. Schlünzen, B. Carissimo, Best Practice Guideline For the CFD Simulation of Flows in the Urban Environment: COST Action 732 Quality
Assurance and Improvement of Microscale Meteorological Models, COST Office, Brussels, 2007.
[37] S.-W. Hong, et al., Numerical simulation of ventilation efficiencies of naturally ventilated multi-span greenhouses in Korea, Trans. ASABe 51 (4) (2008)
1417–1432, https://doi.org/10.13031/2013.25235.
[38] J.H. Chepete, E.L. Thipe, P. Emesu, B. Sebolai, K. Kgosikoma, Effect of housing design and location on production and economic performance of broiler chickens
during summer in Botswana, Revista Brasileira de Ciencia Avicola /Brazilian J. Poultry Sci. 25 (3) (2023), https://doi.org/10.1590/1806-9061-2022-1756.

13