This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

Author's personal copy

Research Paper

Measurement and prediction of soil erosionin dry field using portable wind erosion tunnel

Se-Woon Hong a,b, In-Bok Lee b,c,*, Il-Hwan Seo d, Kyeong-Seok Kwon b,c,Tae-Wan Kim c, Young-Hwan Son c, Minyoung Kim e

aDepartment of Biosystems, Division M3-BIORES: Measure, Model & Manage Bioresponses, KU Leuven,

Kasteelpark Arenberg 30, 3001 Heverlee, BelgiumbResearch Institute for Agriculture and Life Sciences, College of Agricultural and Life Sciences, Seoul National

University, 1 Gwanak-ro, Gwanak-gu, Seoul 151-921, Republic of KoreacDepartment of Rural Systems Engineering, College of Agricultural and Life Sciences, Seoul National University,

1 Gwanak-ro, Gwanak-gu, Seoul 151-921, Republic of KoreadCenter for Green Eco Engineering, Institute of Green Bio Science and Technology, Seoul National University,

Republic of KoreaeDepartment of Agricultural Engineering, National Academy of Agricultural Science, Rural Development

Administration, Republic of Korea

a r t i c l e i n f o

Article history:

Received 16 May 2013

Received in revised form

25 October 2013

Accepted 5 November 2013

Published online 8 December 2013

The purpose of this study was to develop a wind erosion prediction model by in situ

measurement using portable wind erosion tunnel. The model has a modified form of the

wind erosion equation (WEQ) to represent short-term wind erosion with fast and simple

measurable factors. To collect the data under controlled wind conditions but on in situ

soils, a portable wind erosion tunnel was designed and utilised during field experi-

ments. Notwithstanding measurements might include any possible error, the multiple

linear regression analysis of repetitive experimental data derived the wind erosion

prediction model, which showed a good agreement with the measured data with

R2 ¼ 0.61. The short-term wind erosion predicted by the model was made available to

CFD simulation by coupling the erosion mechanism with sophisticated wind environ-

ment analysis over complex terrain. The land cover data was linked to the CFD simu-

lation by mapping the virtual porosity and using user-defined functions. The CFD

simulation coupled with the regression model produced useful results concerning

spatial distributions of soil erodibility, erodible area and soil erosion over complex

terrain showing good potential of coupling the experimental model with CFD simulation

technique. It is also a promising method for evaluation of various wind erosion pre-

vention measures as well as for effective planning and decision-making for wind

erosion control.

ª 2013 IAgrE. Published by Elsevier Ltd. All rights reserved.

* Corresponding author. Research Institute for Agriculture and Life Sciences, College of Agricultural and Life Sciences, Seoul NationalUniversity, 1 Gwanak-ro, Gwanak-gu, Seoul 151-921, Republic of Korea.

E-mail address: iblee@snu.ac.kr (I.-B. Lee).

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier .com/locate/ issn/15375110

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2

1537-5110/$ e see front matter ª 2013 IAgrE. Published by Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.biosystemseng.2013.11.003

Author's personal copy

1. Introduction

Although global surface temperatures have warmed 0.74 �Cover the past 100 years, as reported by the fourth assessment

report of the Intergovernmental Panel on Climate Change

(IPCC), temperatures in Korea have increased by 1.7 �C. Wind

erosion due to global warming leading to desertification was

considered as an important issue in arid and semi-arid regions

(Buschiazzo & Zobeck, 2008), but recent continuous abnormal

warming has influenced Korea to prepare against possible

wind erosion. Agricultural land in particular may result in a

loss of crop production function by the removal of valuable

loam as desertification accelerates the soil erosion process in

dry fields. Although the most severe soil is caused by rainfall

rather than wind, methods for supplementing and improving

soil have become a burden to farmers. For example, the

amount of soil supplement applied to fields reached

1770 t ha�1 every two to three years for dry fields in mountain

region of Pyeongchang County where this study was con-

ducted (Park, 2006). Eroded particles, nutrients and dust may

also be scattered and dispersed to nearby habitation causing

residents health problems or to farms causing the decline in

crop and animal productions (Bitog et al., 2009).

The most well-known model to predict a soil erosion by

wind is the WEQ (wind erosion equation) empirically devel-

oped in 1960s (Woodruff & Siddoway, 1965). Based on theWEQ,

revised or new models, such as RWEQ (revised wind erosion

equation, Fryrear, Saleh, & Bilbro, 1998) and WEPS (wind

erosion prediction system, Hagen, 1991), have been suggested.

The later models have supplemented various physical pro-

cesses of soil erosion because the wind erosions predicted by

models do not show significant level of agreement with

measured in situ under certain situations due to varied, non-

uniform and changing climate and soil conditions.

WEQ-based studies have been conducted through field

measurement and numerical simulation targeting mostly

large areas over long time frames using yearly or monthly

units, and daily units in the particular case of theWEPS (Webb

& McGowan, 2009). These long-term approaches give good

predictions by reducing various factors that fluctuating from

moment to moment, but this approach may decrease the ac-

curacy and efficiency of predictions of temporal variation in

soil erodibility caused by changes in wind conditions. For

example, where wind breaks are installed to prevent wind

erosion, the number, location, arrangement and direction of

the breaks needs investigation at a suitable scale to develop

methods that will efficiently prevent the soil erosion over the

wider field.

Laboratory-basedwind tunnelshavebeenusedtoanalyse the

links between soil erodibility and various physical factors to

derive a numerical relationship between them (Gillette, 1978;

Hagen, 1999; Han et al., 2009; Liu et al., 2006). Wind tunnels

provide a controlled environment protecting against variable

field conditions in order to investigate the effects of several

particular factors on soil erosion behaviour. Wind factors, such

as vertical profiles of wind speed and turbulence quantities can

be artificially controlled in the wind tunnel and soil factors

including soil texture, grain size, water content, surface rough-

ness, soil compactness, etc. can be manually adjusted to be

similar to field conditions. However, while each of the soil

properties canbe independentlyvaried it ispossible tovary them

beyondfield conditions. If theproperties of the soil samplesused

for testing are not realistic, the test results may produce errors

and uncertainty despite the advantages of using a wind tunnel.

The use of a portable wind tunnel is an alternative method

to overcome the uncertainty of using artificial soil samples.

Portable wind tunnels have been used by installing them on

the ground of test site thus removing the requirement for

preparing soil samples to investigate soil erosion (Fister,

Iserloh, Ries, & Schmidt, 2012; Fister & Ries, 2009; Gartmann,

Fister, Schwanghart, & Muller, 2011; Leys & Raupach, 1991;

Pietersma, Stetler, & Saxton, 1996). The strength of this

approach is that erosion behaviour can be investigated on real

soil whilst retaining the ability to control wind speed. Because

the erosion area for testing is limited by the test area of the

portable wind tunnel, which is typically a few square metres,

very low soil losses through wind erosion, less than 1 g m�2

10 min�1, have been observed (Fister & Ries, 2009). Therefore

minor errors or losses of tiny soil aggregates during the

Nomenclature

C Climatic factor of the Wind erosion equation

(WEQ)

Cir Inertial resistance of the medium (m�1)

Cr Chain roughness (%)

C0 Correction factor

C1, C2, C3, C4 Exponents for a regression model

E Soil erosion by wind (g m�2 min�1)

I Soil erodibility factor of the WEQ (tonne

ha�1 year�1)

K Soil roughness factor of the WEQ

Kr Ridge roughness value

L Unsheltered field width factor of the WEQ

n! Normal unit vector of each cell in a terrain

P Pressure (kg m�1 s�2)

Rc Soil roughness with respect to ridge orientation

SR Ratio of percentages of soil aggregates to soil

particles

TWC Topsoil water content (%)

u Wind speed (m s�1)

uhor��! Wind unit vector projected onto a horizontal plane

ut Threshold wind speed of wind erosion for dry soil

particles (m s�1)

V Vegetative cover factor of the WEQ

WD Wind angle from the direction perpendicular to

ridges (�)wf Wind value (m3 s�3)

a Permeability of the medium (m2)

q Ground slope (�)r Air density (kg m�3)

[1 Chain length (m)

[2 Horizontal distance between chain ends (m)

m Viscosity of air (kg m�1 s�1)

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 69

Author's personal copy

experiment may sometimes cause significant errors in the

results. Non-uniformity of soil characteristics, such as surface

roughness, water content and grain size distribution, over the

test area of the portable wind tunnel may cause problems

relating the soil characteristics to the measured wind erosion.

Errors due to non-uniform soil characteristics are not related

to how sophisticated is the portable wind tunnel but how

portable it is to make repetitive measurements. To obtain

sufficient experimental data, portable wind tunnels and

experimental procedures should be designed for short time

cycles and easy mobility between repetitive experiments.

In this study, wind erosion was measured using a portable

wind erosion tunnel in a dry field. The design of a portable

wind erosion tunnel and its evaluation in a preliminary lab-

oratory experiment are presented in this paper. The main

objective of the study was to propose a wind erosion predic-

tionmodel based on fast and simplemeasurable factors using

multiple linear regression analysis of various wind and soil

factors and measured values of wind erosion. The regression

model was to be derived from short-term fieldmeasurements.

It was hypothesised that this would allowmore accurate long-

term (yearly or monthly) predictions to help understand the

transient nature of soil erosion.

Short term prediction models can also be utilised for

computational fluid dynamics (CFD) simulation, which uses

numerical methods to solve fluid flows and has a powerful

benefit in modelling wind flow and turbulence around build-

ings and along terrain features (Hong et al., 2011; Lee et al.,

2013; Scargiali, Rienzo, Ciofalo, Grisafi, & Brucato, 2005). The

combination of the CFD and an accurate short-term model

could make precise prediction of transient soil erosion by

complex wind environments feasible and capable of predict-

ing spatial and temporal changes in wind erosion caused by

various wind erosion control measures, such as windbreaks,

vegetative cover, and tillage, and to evaluate the performance

of these control measures. While the CFD researches have

studied on wind forces on the erodible surfaces of earth

buildings and terrains (Hussein & El-Shishiny, 2009; Zhang

et al., 2012) and on transportation of eroded soil particles (Seo

et al., 2010), this study focuses on predicting the amount of soil

eroded by terrain features and complexwind environments in

collaboration with a wind erosion regression model by

showing an illustrative example.

2. Materials and methods

A portable wind erosion tunnel was designed and built in

order to measure in situ soil erosion according to various soil

and wind conditions. The measured data were utilised to

derive the wind erosion prediction model, which has a

modified form of the WEQ to represent short-term wind

erosion. The erosion model was then applied to CFD simula-

tions to investigate the possibility of predicting spatial and

temporal wind erosions over complex terrain.

2.1. Factors for wind erosion prediction model

The basic form of the WEQ (Woodruff & Siddoway, 1965) was

maintained in this model, but the I (soil erodibility factor), K

(soil roughness factor), C (climatic factor), L (unsheltered field

width factor) and V (vegetative cover factor) factors of the

WEQ were modified to reflect short-term prognosis and to be

easily calculated through the field experiment with the

portable wind erosion tunnel.

2.1.1. I factor modificationSoil erodibility factor (I) is the potential soil loss from non-

crusted field surface and computed from the percentage of

soil fractions greater than 0.84 mm in diameter using a stan-

dard dry sieving procedure (Woodruff & Siddoway, 1965). It

represents the maximum possible soil erosion by wind since

sieving examines soil particles and individual soil particles

agglomerate together to form larger soil aggregateswhichhave

higher resistance to the wind erosion. Larger aggregates are

not eroded before broken down, and therefore the measured

amount of wind erosion is smaller than potential estimates.

When wind erosion measurements continue over years or

months, a portion of the soil aggregates must be broken and

eroded, particularly within the measurement area. The effect

of soil aggregates on soil erosion will therefore be less over

time. However, in the shorter term such as hours or minutes,

the soil aggregates are less breakable and they will reduce the

soil erosion. Therefore, the amount of soil aggregates needs to

be considered together with the amount of soil particles to

avoid overestimation of the short-termwind erosion. The size

distribution of the soil aggregates was measured by the dry

sieve procedure but the soil sampleswere carefully handled in

order not to break the soil aggregates that occurred in the

samples. As soil erodibility is computed from particle size

distributions, the potential and predicted soil erodibilities

were estimated from the distributions of soil fractions and soil

aggregates, respectively. Consequently the erodibilities were

included into the short-term prediction model to compute the

wind erosion in combination with other factors of the model.

2.1.2. K factor modificationSoil roughness factor (K) is used todescribe the effect of surface

roughness on reducing soil loss by wind erosion. The WEQ

model used a ridge roughness value, Kr which is expressed in

terms of height of the ridges and ridge spacing (Woodruff &

Siddoway, 1965). To compensate the limitations of Kr, the

RWEQ model included the effects of aggregate (i.e. random)

roughness, surface roughness decay by rainfall and ridge di-

rection while the WEQ considered only the wind direction

perpendicular to the ridge (Fryrear et al., 1998). This study also

used a revised form to reflect the detailed effects of surface

conditions in order to improve prediction of short-term wind

erosion.While the ridge roughnesswas directlymeasured by a

ruler, very small random and aggregate roughness was

measured using the chain method (Saleh, 1993). The chain

method places a chain of the given length ([1) on a surface and

measures the horizontal distance between chain ends ([2). The

chain roughness is calculated using the [2/[1 ratio which de-

creases as the surface roughness increases.

Cr ¼�1� [2

[1

�� 100 (1)

The chain roughness, Cr was combined with Kr of the WEQ

to include various roughness scale ranged from small

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 270

Author's personal copy

aggregate roughness to large ridge roughness (Saleh & Fryrear,

1999).

K ¼ exp�1:88Kr � 2:44K0:934

r � 0:124Cr

�(2)

The effect of ridge orientation on K factorwas also included

according to wind direction raging from perpendicular to

parallel to the ridges;

K ¼ exp�Rc �

�1:86Kr � 2:411K0:934

r

�� 0:124Cr

�(3)

where, Rc is calculated as Rc ¼ 1.0 � (0.00032 WD þ 0.000349

WD2 � 0.00000258 WD3), and WD is the wind angle from the

direction perpendicular to the ridges in degrees.

When the wind is parallel to the ridges (i.e. WD ¼ 90), Rc

becomes zero and ridge effect is negligible while only aggre-

gate roughness predominates. However, if the wind is

perpendicular to the ridges (WD ¼ 0), the ridges fully in-

fluences the wind erosion.

2.1.3. C factor modificationThe climatic factor (C) describes the effect of climate bymeans

of the mean annual wind speed and effective moisture in the

WEQ and the wind speed data for each 15 day and number/

amount of rainfall events in the RWEQ. The wind speed data

used in the WEQ and RWEQ are built-up by yearly or monthly

measurements. Regarding the moisture data, the monthly

rainfall and temperature data for the WEQ and daily evapo-

transpiration, rainfall and irrigation for the RWEQ are also

used for wind erosion prediction. The moisture data is

considered as the soil wetness factor in the RWEQ since the

wind and moisture data by year, month and day units are not

appropriate for a soil erosion prediction using hours as in this

study. This is because the effect of a single rainfall event on

the soil erodibility is not uniformly distributed over the event

period and because rainfall affects the soil water content a few

days after the event. Of course the accumulated amount of

soil erosion for a certain period is correlated with the amount

of rainfall during that period. However, instantaneous soil

erosion and erodibility is more influenced by the soil water

content, especially at the topsoil where the soil fractions are

eroded, thus the topsoil water content (TWC) was measured

and utilised as the moisture factor in this study.

Wind factor was also replaced with instantaneous wind

speed measured at the portable wind tunnel. Since soil par-

ticles are detached above a threshold wind speed, the wind

force applied to the particles was determined by the magni-

tude of the wind speed above a threshold wind speed cubed

(Fryrear et al., 1998; Youssef, Visser, Karssenberg, Bruggeman,

& Erpul, 2012) as shown in Eq. (4). The threshold wind speed

for dry soil particles was assumed to be 5 m s�1 in this study

(Fryrear et al., 1998; Youssef et al., 2012). According to the

literature, the erosion threshold wind speed increased by

either linear, exponential or logarithmic functions of soil

moisture content due to the cohesive force increasing with

increasing soil moisture (Chen, Dong, Li, & Yang, 1996).

Theoretically, the relationship between the cohesive force and

soil moisture varies depending on the soil texture, structure

and soil class, which is difficult to define conventionally.

Therefore the effect of soil moisture was independently

considered in this study as a form of TWC.

wf ¼ uðu� utÞ2 (4)

where, wf is the wind “value” (m3 s�3), u is the wind speed

(m s�1), and ut is the threshold wind speed (m s�1) for dry soil

particles.

The average wind speed and a wind speed cubed (“wind

value”) were used for the WEQ and RWEQ, respectively and

needed to be included in this study to select which is prefer-

able for the wind factor. Therefore, the climate factors were

expressed as TWC, wind speed (u) and wind value (wf).

2.2. Experimental site

The field experiments were conducted in the Highland Agri-

culture Research Center (HARC) of the Rural Development

Administration located in Pyeongchang County in the Tae-

baek Mountains region along the eastern edge of Korea

(37�4005200N, 128�4304500E, altitude of 770 m). Experimental

fields in the research centre which were being utilised for the

genetic improvement of crops, soil improvement and soil

management provided suitable test sites with various soil

conditions. The field experiments using the portable wind

erosion tunnel were conducted in October and November,

2012.

2.3. Portable wind erosion tunnel

A portable wind erosion tunnel was designed as shown in

Fig. 1 to provide easy assembly, disassembly and mobility for

repetitive experiments. The portable wind tunnel consisted

of flow development section and test section. The test section

was 3 m long and rectangular in shape (1 m � 1 m). The floor

of the tunnel at the test section was open and created a 3 m2

erosion working area. The structure of the wind tunnel was

made of aluminium profile with 30 mm � 30 mm rectangular

hollow section. Sidewalls and ceilings are made of trans-

parent acrylic sheets (thickness ¼ 5 mm) to observe a

movement of eroded soil particles on the surface of the

erosion area. The acrylic sheets were coated with anti-

electrostatic material to prevent the eroded fine particles

from adhering to the inner surface of the wind tunnel. The

tunnel of the test section was divided into three parts, each

1 m long. The first part has a 1 m � 1 m cross-section, and the

cross-sections of second and third parts increased by twice

the thickness of the frame, thus they were 1.06 m � 1.06 m

for the second part and 1.12 m � 1.12 m for the third part. The

three parts can be laid over each other and overlapped for

easy mobility. This slight expansion of the cross-sections also

had the benefit of compensating for the loss of flow area by

boundary layer growth at the wind tunnel inner walls

(Bradshaw, 2002).

The flow development section consists of 1.5 kW high-

pressure fan with 1.27 m in diameter, contraction part and

flow stabilisation part. The contraction part was made of a

tarpaulin fabric shaped in a frustum of quadrangular pyramid

to connect the fan with a 1.4 m � 1.4 m cross-section to the

flow stabilisation part with a 1 m � 1 m cross-section. The

addition of the anti-swirl vanes straightens the flow of air

along the contraction section preventing excessive mo-

mentum loss when the cross-sectional area of the passage is

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 71

Author's personal copy

suddenly contracted. At downstream of the contraction sec-

tion, three short tunnels, each 250 mm long are connected to

stabilise the highly turbulent flow using contraction. The

honeycomb located in the middle short tunnel was 75 mm

long and had hexagon-shaped holes with the longest diagonal

of 7 mm made from aluminium sheets. Upstream and

downstream of the honeycomb wire meshes with 7 stainless

wires per 10 mm and the wire diameter of 0.3 mm were

attached. The middle short tunnel, including the honeycomb,

was easily detached for maintenance.

The eroded soil fractions from the erosion area were

caught in the deposition area. A furrow was made down-

stream of the test section and a deposition net (size ¼ 1.25 m

(width) � 0.2 m (length) � 0.4 m (height)) made of several

layers of wire mesh were placed on the furrow to reduce the

wind speed and hence momentum of air flow in the deposi-

tion area to enable eroded particles to be deposited into the

furrow. To easily measure the weight of the deposited parti-

cles, aluminium foil covered the perimeter of the furrow and

was crumpled up into a small ball after the eroded particles

were completely collected. The weight of the eroded soils was

calculated by subtracting the pre-measured weight of the

aluminium foil from the weight of the used and crumpled

aluminium foil. The T-board is installed as shown in Fig. 1

because some large particles may slip between the

aluminium foil and the soil surface via creeping.

The portable wind erosion tunnel was initially tested for

flow distribution in the test section and performance at the

deposition area before being used in a field experiment. For

the flow distribution inside the test section wind velocities

were logged by the multi-channel how-wire anemometer

(System 6243, Kanomax, Japan) with a frequency of 10 Hz at 36

measuring points: three positions along the flow direction (D1,

D2, D3), four positions along the vertical direction (B, L, M, H)

and three positions along the width direction (L, M, R). The D1,

D2 and D3 indicated the middle of three parts, each 1 m long,

and the B, L, M andH designated the height of 100, 250, 500 and

750 mm, respectively from the floor. The L, M and R are the

positions equally divided across the width of 1 m into each

250 mm sections. In addition, wind velocities at the front,

inside and rear of the deposition net were measured in the

same way to evaluate the reduction of wind speed at the

deposition area.

For the performance of the deposition area, the percentage

of the soil particles captured in the aluminium foil out of the

total eroded soils from the erosion area was measured using

laboratory experiments. Soil samples collected from the test

site and desiccated and classified into three groups by particle

size: group I for 2e0.85 mm, group II for 0.85e0.425 mm and

group III for 0.425e0.074 mm in diameter. The portable wind

tunnel and deposition area were set up on a smooth concrete

floor to prevent entry of particles except those emanating

from the soil samples. The soil samples of each group in turn

were scattered over the erosion area and blown by the wind

until no soil particles remains on the erosion area. The

possible maximum wind speed inside the portable wind tun-

nel was set during the performance evaluation because the

lower percentage of the captured soil was expected at higher

wind speeds.

2.4. Experimental procedures for wind erosionmeasurementsThe portable wind erosion tunnel was set up on test sites as

shown in Fig. 2 and then moved to an experimental lot in the

HARC to collect experimental data. The main parameters

measured were wind speed in the wind tunnel, topsoil TWC,

surface roughness, ground slope, percentage of soil particles

(>0.84 mm), percentage of soil aggregates (>0.84 mm) and the

amount of material collected in the deposition area. The

percentage of soil particles and aggregates were investigated

by carefully moving soil samples to laboratory for the dry

sieving procedure. The remaining factors were measured in

situ immediately during every experiment.

The wind speeds were adjusted in three steps, 6, 7.5 and

9.5m s�1 considering fan performance. The surface roughness

was measured by the chain method to produce the Cr value.

The Kr value was set to zero because all experiments were

conducted on dry fieldswith no ridges. The effect of ridgeswill

be considered in future study. The five levels of the ground

slope (�20�, �12�, 0�, 12�, þ12�) were determined by utilizing

artificially prepared slope lands (Fig. 2). The positive ground

slope means the upward slope.

The TWC was measured using the alcohol method

(Bouyoucos, 1931). By putting a small quantity of soil in a small

can and pouring somemethyl alcohol into the can, the alcohol

absorbed the moisture of soils. The moisture content of soils

Fig. 1 e Schematic view of the portable wind erosion tunnel.

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 272

Author's personal copy

was then determined by burning the alcohol out more quickly

and conveniently than by the conventional drying oven

method.

Ten minute duration was used in every experiment. This

time was considered long enough for all erodible soil particles

to be transported (Fister & Ries, 2009). By moving the portable

wind tunnel to various test sites which have different soil

conditions, experimental data sets were produced so that the

relationship between the soil and wind factors and the wind

erosion could be analysed. The results were also used for a

multiple linear regression analysis to derive the wind erosion

prediction model.

2.5. Modelling procedures for wind erosion prediction

The short-termwind erosion regressionmodel was connected

with a three dimensional CFD model to incorporate the

topographical features and thereby model complex wind en-

vironments. The CFD simulation was conducted using the

commercial FLUENT package (ANSYS, Inc., Pennsylvania,

USA) following the terrain modelling procedure used by Hong

et al. (2011), which suggested suitablemodelling procedures to

build computational domains and meshes from geographical

information.

The target area for the simulation was 2 km � 2 km region

including the test sites, HARC and its surroundings. The

satellite image and CAD contourmap provided by theNational

Geographic Information System (NGIS) of Korea are shown in

Fig. 3. Firstly, the contour map was imported into SketchUp

(Ver. 6, Google Inc., CA, USA) and modified to create a TIN

(triangulate irregular network) structure. Then, the TIN

structure was moved to Rhinoceres (Ver.4, Robert McNeel &

Associates, Seattle, WA, USA) to create a solid surface object

over the TIN structure. Finally, GAMBIT (Ver. 2.4, Fluent Inc.,

Lebanon, NH, USA) imported the solid surface object to create

terrain model and two-dimensional triangular meshes with

the size of 10 m, and TGrid (Ver. 5, Fluent Inc., Lebanon, NH,

USA) builds three-dimensional computational domains and

triangular prismatic meshes. The heights of the prismatic

meshes were 0.5 m on average near the floor and gradually

increased upward with the ratio of approximately 1.115e1.17

up to totally 37 layers in order to save the number of meshes.

The cumulative height of the computational domain was

about 600 m and the number of 3D meshes was 1,555,332. The

equiangular skewness value for mesh quality evaluation was

0.147 on average and maximum 0.82. Considering the equi-

angular skewness value closed to zero indicates an excellent

quality and a value of 1 indicates a completely degenerate cell

(Fluent Inc., 2007), three-dimensional meshes built in this

study showed a reliable quality.

Thewholedomainhadaboxshape.Thefoursideboundaries

were set as inlet or outlet according to wind direction and the

Fig. 3 e Satellite image (left) and CAD contour map (right) of the test site.

Fig. 2 e Wind tunnel (left) and sloping lands with a slope of 12� and 20� (right) on test site.

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 73

Author's personal copy

ceiling boundary was set as a symmetry wall. The floor was set

as a wall with the non-equilibrium wall function (Fluent Inc.,

2007). The FLUENT (ver. 6.3, Fluent Inc., NH, USA) was used for

a main processing of a steady state simulation. The pressure-

based solver was used and the SIMPLE algorithm was used for

pressureevelocity coupling. The well-known realisable k-ε

model was utilized for turbulence modelling. The well-known

power law with 15 m s�1 at 10 m height was designed at inlet

boundaries for the wind profile of the atmospheric boundary

layer and four wind directions, such as north, east, south and

west winds were set up for comparison of wind erosion.

The area under consideration in Korea includes forests in

much of the region. The forest changes vertical wind profiles

within the atmospheric boundary layer and sometimes

functions as a windbreak to reduce a leewardwind speed. The

effects of the forest mostly have been neglected or embodied

by applying surface roughness in simulation models (Folch,

Costa, & Hankin, 2009; Seo et al., 2010), although the surface

roughness can reflect only the former effect. On the other

hand, two effects can be realised by including geometric

feature of the forest in simulation models. Bitog et al. (2011)

and Lin, Barrington, Choiniere, and Prasher (2007) simplified

the forest or tree as a porous box or cone containing source

terms for momentum and turbulence quantities. However,

because of irregular shape and distribution of the forests in

the target area it is difficult to construct geometry and three-

dimensional meshes of the forests together with the com-

plex terrain features. Moreover, in the case of planning a

windbreak the addition of a windbreak necessarily requires

reconstructing the geometry and meshes. As one of the so-

lutions to this problem, this study focused on new method

which models a forest as a virtual porous media without

creating any extra geometry. This virtual porosity method

maps the actual position of the forest recognized by the user’s

land cover data to corresponding coordinates in the compu-

tational domains and assign proper porosity values to the

computational cells of the coordinates as shown in Fig. 4. It

does not requiremodelling geometry and therefore changes of

forest shape and distribution or additions of a windbreak are

easily implemented in a single simulation model. The pro-

cedures for the virtual porosity method were programmed

using user defined functions (UDFs) coded in C language and

connected to the main processing code by FLUENT.

In this study the height of forests in the target area was

assumed as 5 m and the DarcyeForsheimer equation in Eq. (5)

was used to describe the variation of pressure drop inside the

forests. The resistance term expressed by the permeability of

the medium can be ignored when the flow is turbulent with

the Reynolds number of 5000 or more. On the assumption that

the flow inside the windbreaks is turbulent, the porous char-

acteristics of the trees were realized by only inputting the

inertial resistance value, Cir. The inertial resistance of trees

was obtained as 0.46 from wind tunnel experiment by Bitog

et al. (2011). Regarding the HARC the inertial resistance and

its effective height were assumed to be 0.1 and 2 m, respec-

tively by expecting low porous effects.

VP ¼ ��m

au!þ Cir

12rjuj u!

�(5)

Where, P is the pressure, m is the viscosity, a is the perme-

ability of the medium, Cir is the inertial resistance of the me-

dium, and r is the density of the fluid.

While most factors related to soil characteristics of the

wind erosion regression model were assumed by users, the

factors of wind speed, wind value and ground slope were

directly computed by the simulation model. The wind speed

and wind value were obtained at each cell and calculated by

Eq. (4). The ground slope was determined not only by topo-

graphical feature but also by wind direction because it in-

dicates the slope along the flow direction. Therefore the

ground slope, q is expressed as Eq. (6) by geometric relation

between two vectors: the normal unit vector at each cell of the

ground, and the wind unit vector projected on the horizontal

plane at each cell of the ground.

q ¼

8>><>>:

if uhor��!$ n!� 0 ; �cos�1 uhor

��!$ n!uhor

��!�j n!j

if uhor��!$ n!< 0 ; cos�1 uhor

��!$nuhor

��!�j n!j � 90

(6)

Where, q is the degree of the ground slope (�), uhor��! is the wind

unit vector projected onto a horizontal plane, and n! is the

normal unit vector of each cell in a terrain.

3. Results and discussion

3.1. Performance test of portable wind erosion tunnel

The portable wind erosion tunnel was built and then tested

through laboratory experiments. The wind speed

Fig. 4 e Schematic diagram of UDF for setting virtual porous media from user’s land cover data.

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 274

Author's personal copy

distributions inside thewind tunnel were shown in Fig. 5 (left).

The distributions at D1, D2 and D3 along the flow direction

were almost similar to each other. Because the cross-section

area of the third parts (D3) was increased by the thickness of

the frame, the measured wind speeds at D3 were slightly

lower than those of D1 and D2. However, the cross-sectional

distributions of wind speed showed the difference to be

approximately 2e2.5 m s�1 between the lowest value at the

centre point (MeM) and the highest value at BeL point. The

drop in wind speed at the centre was due of the typical flow

pattern of rotary fan (Hong et al., 2012) where this feature was

not completely remedied. To enhance the cross-sectional

uniformity of wind speed, the length of the flow develop-

ment section needed to be increased and more wire meshes

needed to be installed in the section enough to distribute the

pressure more evenly over the cross-section. However, this

method of improving uniformity cannot avoid reducing

working pressure and increasing the risks of structural

modifications.

By shifting the focus from the whole wind tunnel to the

floor where the wind erosion occurs, the wind speeds at 9

points near the floor (BeL, BeM and BeR at D1, D2 and D3

each) were shown in Fig. 5 (right) by the average marked by

dots and max/min values marked by lines. While the wind

speeds gradually increased according to fan performance

rate, the max/min values were observed to be within

w1 m s�1 from the average value. This wind speed distri-

butions inside the wind tunnel may leave much to be

desired. However, the wind conditions of the portable wind

tunnel were judged to reach reasonable levels because we

focused on experimentally conducting wind erosion mea-

surements in the field. During the design process, emphasis

was placed on the capability and mobility of the tunnel in

order to quickly repeat experiments as with other portable

wind tunnels (Fister & Ries, 2009; Gartmann et al., 2011).

The reduction of wind speed caused by the deposition net

is shown in Fig. 6. The wind speed at the end of test section

was initially reduced in the deposition net to 65e70%. It also

decelerated to around 40% after passing through the deposi-

tion net. By using the two-step reduction process, the wind

speed at rear of the deposition net was below the assumed

threshold wind speed of 5 m s�1 at all operating wind speeds

while it exceeded the threshold wind speed inside the depo-

sition net. Therefore the deposition net might be expected to

deposit soil particles by impacting themwithin the net and by

settling in the rear of the net.

The performance of the deposition area was evaluated

under the wind speed of 9.5 m s�1 below which the portable

wind tunnel operated stably according to the preceding

measurements of wind speed. Table 1 presents the averaged

results of each soil sample group which was tested twice.

The collecting ratios, the percentage of the soil particles

captured out of the total eroded soils, were considerably high

with 92.6% and 86.7% for group I and group II, respectively.

However group III comprised of smaller-size particles than

group I and group II showed lower collecting ratio of 53.1%.

According to observation during the lab experiment a large

amount of the eroded particles did not deposit within or

around the deposition net but blew laterally away from

aluminium foil as shown in Fig. 7 because of complex swirl

inside furrow. When the wind flowed into the furrow which

has a rectangular cross section, it bumped and bounded

against the rear wall of the furrow creating large eddies. As

the wind continuously flowed into the furrow, the eddies

travelled bilaterally along the furrow. The large-size parti-

cles in group I and group II had enough mass and inertial

force to escape from swirling flow. However, the small-sized

particles could not deposit around the deposition net due to

strong eddies and were transported by the wind along the

furrow.

Fig. 5 e Wind speed distributions inside the portable wind erosion tunnel: (left) wind speeds at all measurement positions

at 60% fan performance, and (right) average (marked by dots), maximum and minimum (marked by lines) of wind speeds at

bottom height according to fan performance rate.

Fig. 6 e Wind speed reductions by deposition net: inside

the deposition net (1st reduction) and after passing

through the deposition net (2nd reduction).

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 75

Author's personal copy

Some modifications in the furrow were made to enhance

the collecting ratio for small-sized soil particles. Unfortu-

nately they could not be applied to the preliminary laboratory

experiment, which was conducted on the concrete floor and

furrow. Three modifications for a field experiment were

considered: i) trapezium shaped furrow, ii) deeper furrow, and

iii) lateral extension of aluminium foil. The trapezoidal cross-

section as shown in Fig. 1 can avoid a strong eddy and the deep

furrow can reduce any escaping particles from the furrow. In

case of laterally travelling flows along the furrow, the

extended aluminium foil is an efficient measure.

3.2. In situ wind erosion measurements

The results, including the main factors and wind erosion

measurements, are summarised in Table 2. Soil factors like

the percentage of soil fractions, TWC and surface roughness

were within limited ranges because all test sites have been

used for a dry field. The percentage of soil particles greater

than the size of 0.84mmwas in the range of 36.59e47.84% and

43.52% in average. The percentage of soil aggregates greater

than 0.84 mm in size was in the range of 67.75e79.02% and

73.40% in average. While the percentages of soil aggregates

were 1.69 times higher than those of soil particles in overall,

the ratio of the two percentages varied from1.49 to 1.85, which

represented the test sites had various degree of soil particle

aggregation.

TWC content was observed to be in the range of 1.16e8.96%

and the average was 5.61%. The surface roughness Cr was

6.85% on averagewith range of 3.20e11.16%. Thewind erosion

measured was observed in the range of 0.04e21.11 g and

averaged 2.82. The minimum amount of erosion (0.04 g)

occurred when the percentage of soil particle, the percentage

of soil aggregates, TWC and roughness Cr value were 45.35%,

76.45%, 8.96% and 8.40%, respectively, revealing that all soil

parameters were higher than their average values. Minimum

erosion occurred at the lowest of the three wind speeds

(6 m s�1) when the slope of the ground was þ12�, the highest

value of the 5 slope conditions. The largest wind erosion

(21.11 g) occurred when most parameters were lower than

their average but when the wind speed was at its highest

value, 9.5 m s�1, and the ground slope was at its lowest, 20�.When the soil was highly erodible all factors operated in

opposite directions compared to when the soil was least

erodible.

This trend is not shown only at the highest and lowest

results but also similar throughout the data. For the tests with

four largest wind erosions (21.11 g, 16.91 g, 5.87 g, 5.00 g), the

values of soil factors were also lower than their averages and,

except for one case, ground slopes were �20�. For the tests

with the five lowest wind erosions (0.04 g, 0.06 g, 0.18 g, 0.19 g,

0.20 g), TWCswere higher than the average values, the surface

roughness, Cr values were higher than the average except for

one case, ground slopes were flat and the lowest wind speed

was 6 m s�1, both except for one case.

To summarise, soil factors, i.e. the percentage of soil par-

ticles, percentage of soil aggregates, ground slope, TWC and

surface roughness were in inverse proportion to the wind

erosion while increasing wind speed proportionally increased

the amount of wind erosion. The TWC was the most influen-

tial on the highest or lowest wind erosions, and surface

roughness also had a considerable effect on extreme values.

3.3. Linear empirical model for wind erosion prediction

To derive an empirical model to predict wind erosion, corre-

lations between all factors and the wind erosion were ana-

lysed. Before the analysis the percentage of soil particles was

replaced with I factor following the conventional procedure of

the WEQ (Woodruff & Siddoway, 1965). The surface rough-

ness, Cr value was also replaced with K factor using Eq. (3).

Here, the Kr value and Rc value were assumed to be zero

because there were no ridges in the test sites. Regarding wind

factor, wind value of Eq. (4) and the wind speed were included

in the correlation analysis to select which of the two is the

more closely related to wind erosion.

To examine the significance of the parameters which

contribute to the wind erosion, the correlation coefficients

between parameters were calculated. They are presented in

Table 3. The correlation coefficients between the wind erosion

and the rest indicate that the percentage of soil fraction,

ground slope and TWC were negatively correlated with the

wind erosion. The I and K factors, calculated from the per-

centage of soil fraction and surface roughness Cr, respectively,

showed positive correlations with wind erosion. Wind speed

and wind value were also positively correlated with wind

erosion.

The I factor and the percentage of soil aggregates showed

similar correlation coefficients with wind erosion, but the

correlation coefficient of 0.622 between them revealed they

were not equivalent to each other. Although they had similar

tendency, the difference between them indicated the degree

Table 1 e Collecting efficiency of deposition net accordingto soil sample groups.

Sample group Used (g) Collected (g) Collectingratio (%)

I 30 27.77 92.6

II 10 8.67 86.7

III 15 7.97 53.1

Fig. 7 e Schematic view of preliminary experiment to

measure soil collecting efficiency.

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 276

Author's personal copy

of soil aggregation. Here another factor, SR, the ratio of per-

centages of soil aggregates to soil particles, can be introduced

into the regression analysis to reflect the effect of both rather

than selecting the better parameter. Therefore, the effect of

soil size distribution in the model was incorporated using the

well-known I factor and a new SR factor. A SR value of one

indicates that soil particles are perfectly separated and that as

the SR value increasesmore soil particles agglomerate to form

larger aggregates.

Wind speed and wind value had similar correlation co-

efficients with wind erosion but, unlike the size-distribution

factors, the correlation coefficient between them was very

high (0.977) indicating that only one of these factors is

required. The regression model will be satisfactory if the

better of the two is selected.

The regression model for wind erosion prediction was

developed by selecting independent variables, such as I factor,

SR, ground slope, TWC, K factor and one of thewind speed and

wind value. The wind speed and wind value were each

included into the regression model and compared by assess-

ing the accuracy of themodels to each other. The independent

variables were then combined in Eq. (7) to maintain a form of

the conventional WEQ. The conventional factors, I factor and

K factor, were multiplied in the same manner but the

Table 2 e In situ experimental data sheet through portable wind erosion tunnel.

% Of soilfractions >0.84 mm

Undisturbed %of >0.84 mm

Slope (�) Windspeed(m s�1)

Topsoil watercontent (%)

Roughness(Cr, %)

Soil erosion(g)

47.84 79.02 0 7.5 5.39 6.25 0.78

9.5 7.35 0.39

47.84 71.46 0 9.5 4.17 4.74 1.53

1.61 6.23 3.42

2.22 11.16 1.67

3.65 5.48 5.87

6.10 9.01 4.70

41.94 69.46 0 6.0 7.91 7.50 0.06

8.5 0.19

9.5 0.33

6.0 7.58 7.50 0.54

8.5 0.96

9.5 1.08

6.0 6.70 7.50 0.20

8.5 0.44

9.5 0.62

41.58 76.28 0 6.0 6.92 3.20 0.18

8.5 0.34

9.5 0.50

36.59 67.75 20 6.0 4.10 7.67 0.64

8.5 2.60

9.5 3.27

�20 6.0 4.33 4.74 5.00

8.5 16.91

9.5 21.11

45.35 76.45 12 6.0 8.96 8.40 0.04

8.5 1.84

9.5 4.87

�12 8.5 7.24 6.53 1.97

9.5 2.52

Table 3 e Correlation matrix to show significance between variables.

I factor Undtb(%)a

Slope(�)

Wind speed(m s�1)

Wind value(m3 s�3)

TWC(%)b

KFactor

Soil erosion(g)

I factor 1 �0.622 �0.023 �0.283 �0.298 �0.143 0.206 0.391

Undtb (%)a �0.622 1. 0.047 0.080 0.056 0.401 0.213 �0.312

Slope (�) �0.023 0.047 1. �0.066 �0.059 0.102 �0.428 �0.517

Wind speed (m s�1) �0.283 0.080 �0.066 1 0.977 �0.185 �0.041 0.242

Wind value (m3 s�3) �0.298 0.056 �0.059 0.977 1 �0.211 �0.047 0.236

TWC (%)b �0.143 0.401 0.102 �0.185 �0.211 1 �0.121 �0.372

K factor 0.206 0.213 �0.428 �0.041 �0.047 �0.121 1 0.239

Soil erosion (g) 0.391 �0.312 �0.517 0.242 0.236 �0.372 0.239 1

a Undtb: Undisturbed particles>0.84 mm diameter.b TWC: Topsoil water content.

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 77

Author's personal copy

remainder, except for the ground slope were modified using

the involution form. The ground slope was applied in an

exponential form to set the factor to one when the ground

slope was zero.

E ¼ C0$I$K$eC1q$TWCC2$SRC3$

huC4 or wC4

f

i(7)

Where, E is the soil erosion bywind in an area of 1m2 for 1min

(g m�2 min�1); TWC is the soil water content in the topsoil (%);

SR is the ratio of the percentage of soil fractions greater than

0.84 mm by wet sieving to the percentage of soil fractions

greater than 0.84 mm by dry sieving (%); C0 is the correction

factor; and C1, C2, C3 and C4 are the exponents to be calculated

by regression analysis.

The exponents in Eq. (7) were calculated by the multiple

linear regression analysis using 30 data points in Table 2

and presented in Table 4. The use of wind speed or wind

value as an independent variable produced the different

values for exponent C4 while exponents, C1, C2 and C3, for

the ground slope, TWC and SR were similar regardless of the

use of the wind speed or wind value. The exponents for q,

TWC and SR had negative values and had the same trend

which is in line with the results of field experiments. Wind

erosion was inversely related to SR, which meant that more

aggregation hindered erosion and transportation of soil.

Figure 8 a) shows a comparison of the linear regression

models and the measured data. The calculated wind ero-

sions using the model and wind speed u rather than the

wind value wf were slightly lower with higher R2 (coefficient

of determination) values. However, in the both models, two

data points where the wind erosion was 3.42 g and 1.67 g

(see Table 2) considerably deviated from the regression

lines. These two data points were possibly affected by error

during the field experiment and considered as outliers. The

remaining 28 data points were used to improve the regres-

sion models.

The revised regression models, using 28 data points, and

their results are presented in Table 4 and Fig. 8 b). Other two

data points still deviated from the regression lines, but much

improvement is shown in Fig. 8 b). In this analysis the results

of the model using wf showed lower values but improved R2

values compared to those using u for the regression. Therefore

q, TWC, SR value and uwere used as the independent variables

of the final regression model.

The exponents for the final regression model were calcu-

lated as C1 ¼ �0.02768, C2 ¼ �2.78927, C3 ¼ �1.59853 and

C4 ¼ 0.46187 as shown in Table 4. The correction factor C0 was

determined to be 0.06252. Figure 9 presents the wind erosion

measurements from the field experiments and those pre-

dicted by the final linear regressionmodel. The predictedwind

erosion quantities were slightly higher than the measured

Table 4 e Exponents calculated by multiple linearregression analysis according to the use of u and wf for avariate.

Parameters C1 C2 C3 C4

q, TWC, SR, u

(using 30 data points)

�0.03251 �1.99764 �3.26727 0.61459

q, TWC, SR, wf

(using 30 data points)

�0.03053 �1.98321 �3.22799 0.29699

q, TWC, SR, u

(using 28 data points)

�0.02873 �3.25977 �4.39525 2.04296

q, TWC, SR, wf

(using 28 data points)

�0.02768 �2.78927 �1.59853 0.46187

Fig. 8 e Comparison of the measured and computed E/(I K)

according to the use of u and wf for a variate. The solid and

dotted lines represent trend line of each case.

Fig. 9 e Relationship between the measured and computed

wind erosion (E).

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 278

Author's personal copy

wind erosions but showed acceptable predictions with

R2 ¼ 0.61. Thus, the model derived from the multiple linear

regression analysis makes possible reliable prediction of wind

erosion in a dry field over the short term (e.g. 10 min). It also

has the advantage predicting wind erosion using common

factors that are rapidly and simply measured in situ. At this

stage, the exponents and correction factor for the prediction

model were derived from very limited data and thus the

model is valid only for a limited range of soil characteristics.

Further research in a variety of soil characteristics is required

to determine the appropriate exponents of the wind erosion

prediction model.

3.4. CFD modelling for wind erosion prediction

The CFD model combined with the wind erosion prediction

model of Eq. (7) was used to analyse the distribution of soil

erodibility over region around the HARC in the light of the

effect of complex topographical features andwind conditions.

The factors I, K, TWC and SR value were assumed homoge-

neous with I ¼ 100, K ¼ 0.5, TWC ¼ 6.5 and SR ¼ 1.65 over the

target area while the ground slope q and the wind value wf

were calculated by the UDF based on the computational re-

sults of the CFD model.

Figure 10 presents distributions of the soil erodibility

computed by the CFD simulation according to four wind di-

rections. The erodible areasmarked by contours differ slightly

depending on thewind directions but aremostly concentrated

in ‘field’ areas shown in the land cover data of Fig. 4. However,

the north wind condition showed fewer erodible areas

compared to other wind directions. Because the ‘field’ areas

were located mainly in the south of the target area, the

mountains of the north seemed to protect the wind erosion

against the north wind. For ‘forest’ areas where the porous

characteristics decreased the wind speed, some erosion was

predicted on the northeastmountain in northerly and easterly

wind conditions. There were also few wind erosions in ‘test

site’ area due to the porous characteristics but very little

erosions were observed in northerly, easterly and westerly

wind conditions.

Erodible areas according to wind directions are presented

in Fig. 11. The high erodibility above 0.2 g m�2 min�1 was

shown in very small areas less than 5 ha. The erodible areas

with low erodibility below 0.003 g m�2 min�1 were also very

Fig. 10 e The CFD computed soil erodibility distributions with logarithmic scale of colour bar according to four wind

directions: a) north, b) east, c) south and d) west.

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 79

Author's personal copy

limited compared to total erodible areas. The greater part of

the erodible area ranged between 0.003 and 0.2 g m�2 min�1 of

the soil erodibility. The total erodible areas were 27.3, 21.8,

12.5 and 8.1 ha for west, east, south and north winds,

respectively in order of size. The largest areawas shownwhen

the wind came from the west, and it was 3.4 times larger than

the smallest area under the north wind.

Figure 12 presents the histogram of the wind erosion in

percentage of total wind erosion. Compared to the erodible

area distribution in Fig. 11, wind erosion at higher soil erod-

ibility increased because the amount of wind erosion in each

soil erodibility class was calculated by multiplying the soil

erodibility and the erodible area. Therefore, while most of the

erodible areas ranged between 0.003 and 0.2 g m�2 min�1, a

large amount of wind erosion occurred in the higher classes of

the soil erodibility. In case of the north wind, 70.2% of total

wind erosion was concentrated in the range of

1e10 g m�2 min�1. In case of the east wind 81.0% of the total

wind erosion was distributed in the range of

0.1e4.6 g m�2 min�1. For the south and west winds, 76.5% and

81.0% of total wind erosion were occurred in the areas with

the soil erodibility of 0.1e2.2 g m�2 min�1 and

0.046e2.2 g m�2 min�1, respectively. On the whole, the range

of the soil erodibility where the bulk of the wind erosions

occurred was from 1 to 2.2 g m�2 min�1 in all wind directions.

The north wind generated the largest wind erosion over the

target area, followed by the east, south and west winds. The

largest wind erosion under the north wind was 32.7 kg min �1

which was 1.7 times larger than the smallest in the west wind

while the eroded areawas the largest under thewestwind and

the smallest under the northwind. Therefore, it was predicted

that a north wind has strong erodibility over the small areas

while a west wind generates weak wind erosion over a wide

area.

The CFD simulation, combined with the wind erosion

prediction model, provided spatial distributions on the

Fig. 11 e Cumulative erodible area curves versus soil erodibility according four wind directions.

Fig. 12 e Histogram of wind erosion for each soil erodibility class as % of total wind erosion according four wind directions.

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 280

Author's personal copy

erodible area, soil erodibility and thereby the amount of wind

erosion. The CFD simulation revealed its strong advantage in

predicting turbulent flows over complex terrains despite the

limitation of assuming soil factors in the CFD simulation. It

could provide a useful modelling approach in erosion studies

by representing the surface shear stress in target fields.

Although this was not analysed in this study, it has been

found to be a useful property to explain wind erosion in many

studies (Webb & McGowan, 2009). The CFD results can also be

improved by refining the soil factors. Fortunately detailed soil

characteristics, such as soil texture and land cover data are

currently available in GIS services and can be introduced into

the CFD simulation by utilising virtual porosity and the

mapping method used in this study. In addition, simulations

could help optimise the use of wind erosion prevention

measures such as windbreaks and shelterbelts, crop, tillage,

watering, commercial materials. Their advantages and dis-

advantages have been well described (Nordstrom & Hotta,

2004), but their quantitative performances are not yet well

understood. However, simulation studies could suggest

effective decisions on selecting, locating and constructing

wind erosion measures leading to more effective and eco-

nomic plans for wind erosion control.

4. Conclusions

A wind erosion prediction model was developed by using in

situ measurement data under various soil and wind condi-

tions. Themodel was a form of theWEQmodified to represent

short-term wind erosion by replacing existing factors with

common factors which can be measured quickly and easily in

situ. A portable wind erosion tunnel was utilised to collect

measured data under controlled wind conditions. The wind

erosion predictionmodel derived bymultiple linear regression

analysis, as provided in Eq. (7), showed a good agreement with

the measured data.

The model suggested in this study still has some limita-

tions as follows:

� The model is valid within limited ranges of soil factors

because all experimental data for deriving the model were

obtained from a dry field.

� Measurements may include any possible error during field

experiment and thus reliable model demands quite a

number of measurements.

However, the model has definite advantages.

� Results of the model are more reliable than those by lab-

oratory or wind tunnel experiment which generally use

artificial soils.

� Themodels are easily and quickly developedwith regard to

new regions or soil conditions because the portable wind

erosion tunnel has merit in collecting a lot of data by re-

petitive experiments.

� The short-term wind erosion predicted by the model is

available to other numerical simulations like CFD by

coupling the erosion mechanism with sophisticated wind

environment analysis.

This study also showed the potential of coupling the

experimental model with a CFD simulation technique. It

produced notable results concerning spatial distributions of

soil erodibility, erodible area and soil erosion over complex

terrain notwithstanding the limitation of assuming simplified

soil factors as an illustrative attempt. Virtual porosity and a

mappingmethod could be used to incorporate land cover data

into CFD simulations without any geometric modifications.

Thus, the provision of detailed soil characteristics by GIS

services as an input data will improve the accuracy of the

wind erosion prediction in future studies. The model coupled

with a CFD simulation may also be a valuable tool to evaluate

and design various wind erosion prevention measures by

simulating their performances as well as helping establish

effective and economic plans for wind erosion control.

Acknowledgements

The authors would like to acknowledge the financial assis-

tance provided by the Rural Development Administration

(RDA) in Korea.

r e f e r e n c e s

Bitog, J., Lee, I.-B., Hwang, H.-S., Shin, M.-H., Hong, S.-W., Seo, I.-H., et al. (2011). A wind tunnel study on aerodynamic porosityand windbreak drag. Forest Science and Technology, 7(1), 8e16.

Bitog, J. P., Lee, I.-B., Shin, M.-H., Hong, S.-W., Hwang, H.-S.,Seo, I.-H., et al. (2009). Numerical simulation of an array offences in Saemangeum reclaimed land. AtmosphericEnvironment, 43, 4612e4621.

Bouyoucos, G. (1931). The alcohol method for determiningmoisture content of soils. Soil Science, 32(3), 173e180.

Bradshaw, P. (2002). Hypertext document on wind-tunnel design: Testsection (version August 2002). Available at: http://navier.stanford.edu/bradshaw/tunnel/testsec.html Accessed 4 April2013.

Buschiazzo, D. E., & Zobeck, T. M. (2008). Validation of WEQ,RWEQ and WEPS wind erosion for different arable landmanagement systems in the Argentinean Pampas. EarthSurface Processes and Landforms, 33(12), 1839e1850.

Chen, W., Dong, Z., Li, Z., & Yang, Z. (1996). Wind tunnel test ofthe influence of moisture on the erodibility of loessial sandyloam soils by wind. Journal of Arid Environments, 34, 391e402.

Fister, W., & Ries, J. B. (2009). Wind erosion in the central EbroBasin under changing land use management. Fieldexperiments with a portable wind tunnel. Journal of AridEnvironments, 73, 996e1004.

Fister, W., Iserloh, T., Ries, J. B., & Schmidt, R.-G. (2012). A portablewind and rainfall simulator for in situ soil erosionmeasurements. CATENA, 91, 72e84.

Fluent Inc.. (2007). Fluent 6.3 documentation. Lebanon,N.H.: Flent, Inc.Folch, A., Costa, A., & Hankin, R. K. S. (2009). TWODEE-2: a shallow

layer model for dense gas dispersion on complex topography.Computers & Geosciences, 35(3), 667e674.

Fryrear, D. W., Saleh, A., & Bilbro, J. D. (1998). A single event winderosion model. Transactions of the American Society ofAgricultural Engineers, 41(5), 1369e1374.

Gartmann, A., Fister, W., Schwanghart, W., & Muller, M. D. (2011).CFD modelling and validation of measured wind field data in aportable wind tunnel. Aeolian Research, 3(3), 315e325.

b i o s y s t em s e ng i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 2 81

Author's personal copy

Gillette, D. (1978). A wind tunnel simulation of the erosion of soil:effect of soil texture, sandblasting, wind speed, and soilconsolidation on dust production. Atmospheric Environment, 12,1735e1743.

Hagen, L. J. (1991). A wind erosion prediction system to meet userneeds. Journal of Soil and Water Conservation, 46(2), 106e111.

Hagen, L. (1999). Assessment of wind erosion parameters usingwind tunnels. In International soil conservation organizationconference proceedings, May 24e29, 1999 at Purdue University,West Layfayette, IN (pp. 742e746).

Han, Q., Qu, J., Zhang, K., Zu, R., Niu, Q., & Liao, K. (2009). 2009.Wind tunnel investigation of the influence of surface moisturecontent on the entrainment and erosion of beach sand bywind using sands from tropical humid coastal southern China.Geomorphology, 104, 230e237.

Hong, S., Lee, I., Hwang, H., Seo, I., Bitog, J., Kown, K., et al. (2011).CFD modelling of livestock odour dispersion over complexterrain, part II: dispersion modelling. Biosystems Engineering,108, 265e279.

Hong, S.-W., Lee, I.-B., Seo, I.-H., Kwon, K.-S., Ha, T.-H., &Hwang, H.-S. (2012). Evaluation and CFD modelling of flowbehind livestock ventilation fan for small-scale wind powergeneration. Korean Society of Agricultural Engineers, 54(5), 79e89(In Korean).

Hussein, A. S., & El-Shishiny, H. (2009). Influences of wind flowover heritage sites: a case study of the wind environment overthe Giza Plateau in Egypt. Environmental Modelling & Software,24, 389e410.

Lee, I.-B., Bitog, J. P., Hong, S.-W., Seo, I.-H., Kwon, K.-S.,Bartzanas, T., et al. (2013). The past, present and future of CFDfor agro-environmental applications. Computers and Electronicsin Agriculture, 93, 168e183.

Leys, J. F., & Raupach, M. R. (1991). Soil flux measurements using aportable wind erosion tunnel. Australian Journal of Soil Research,29, 533e552.

Lin, X.-J., Barrington, S., Choiniere, D., & Prasher, S. (2007).Simulation of the effect of windbreaks on odour dispersion.Biosystems Engineering, 98, 347e363.

Liu, M., Wang, J., Yan, P., Liu, L., Ge, Y., Li, X., et al. (2006). Windtunnel simulation of ridge-tillage effects on soil erosion fromcropland. Soil & Tillage Research, 90, 242e249.

Nordstrom, K. F., & Hotta, S. (2004). Wind erosion from croplandin the USA: a review of problems, solutions and prospects.Geoderma, 121, 157e167.

Park, H. K. (2006). Environmental damages and control measures byhighland agriculture e Study on system improvement. Ph.D. thesis.Korea: Kangwon National University (In Korean).

Pietersma, D., Stetler, L. D., & Saxton, K. E. (1996). Design andaerodynamics of a portable wind tunnel for soil erosion andfugitive dust research. Transactions of the ASAE, 39(6), 2075e2083.

Saleh, A. (1993). Soil roughness measurement: chain method.Journal of Soil and Water Conservation, 48(6), 527e529.

Saleh, A., & Fryrear, D. W. (1999). Soil roughness for the revisedwind erosion equation (RWEQ). Soil and Water ConservationSociety, 54(2), 473e476.

Scargiali, F., Di, Rienzo E., Ciofalo, M., Grisafi, F., & Brucato, A.(2005). Heavy gas dispersion modelling over a Topographicallycomplex mesoscale: A CFD based approach. Process Safety andEnvironmental Protection, 83(3), 242e256.

Seo, I.-H., Lee, I.-B., Shin, M.-H., Lee, G.-Y., Hwang, H.-S., Hong, S.-W., et al. (2010). Numerical prediction of fugitive dustdispersion on reclaimed land in Korea. Transactions of theASABE, 53(3), 891e901.

Webb, N. P., & McGowan, H. A. (2009). Approaches to modellingland erodibility by wind. Progress in Physical Geography, 2009,1e27.

Woodruff, N. P., & Siddoway, F. H. (1965). A wind erosion equation.Soil Science Society of America Proceedings, 29(5), 602e608.

Youssef, F., Visser, S., Karssenberg, D., Bruggeman, A., & Erpul, G.(2012). Calibration of RWEQ in a patchy landscape; a first steptowards a regional scale wind erosion model. Aeolian Research,3(4), 467e476.

Zhang, Z., Wieland, R., Reiche, M., Funk, R., Hoffmann, C., Li, Y.,et al. (2012). Identifying sensitive areas to wind erosion in thexilingele grassland by computational fluid dynamicsmodelling. Ecological Informatics, 8, 37e47.

b i o s y s t em s e n g i n e e r i n g 1 1 8 ( 2 0 1 4 ) 6 8e8 282