View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by International Institute for Science, Technology and Education (IISTE): E-Journals

Journal of Natural Sciences Research www.iiste.org

ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

Modelling Soil Degradation in Libya

Gandura O. Abagandura1* Dara Park2 David White3 William C. Bridges Jr4
1.Soil and Water Department, Omar Al-Mukhtar University, Albida, Libya
2.Agricultural and Environmental Science Department, Clemson University, 273 Poole Agricultural Center,
Clemson, South Carolina, 29634
3.Holcombe Department of Electrical and Computer Engineering, Clemson University, South Carolina. 322
Fluor Daniel, Clemson, South Carolina, 29634
4.Mathematical Sciences. Clemson University, South Carolina. Martin O-117. Clemson, South Carolina, 29634

Abstract
Soil degradation is considered one of the most important factors limiting agricultural development in Libya,
however little effort has been taken to identify the distribution of soil degradation occurrence and type for the
country. While the soil degradation for the primary agriculture regions (PAR) has been previously determined as
thirty-three percent (33%), the degradation for the rest of the country was still unknown. For this reason,
polygons representing soil and climate characteristics, landscape feature and soil degradation from the PAR were
converted to raster using ArcGIS (at a resolution of 1000 m2) resulting in 850 points which were then exported as
a table for modelling purposes. The data set was subjected to logistic regression to model the binomial outcome
of soil degradation occurrence (occurrence, no occurrence). A multinomial logistic regression was used to relate
predictor variables to the type of soil degradation since there was more than two outcome options (salinization,
water erosion, and wind erosion). Finally, the prediction models were used to determine the remainder of the
country’s degradation occurrence and type. Results indicated that slope, texture and wind speed are the most
important variables for soil degradation occurrence and type in PAR. When these models are applied to the
reminder of the country, they show that salinization was the primary type of soil degradation (30 %), with water
erosion and wind erosion causing 10 % and 15 % of soil degradation, respectively. The intention is for these
models to assist stakeholders in identifying areas where agriculture is most likely to be successful, while also
applicable to countries with similar climate and soils in North Africa.
Keywords: Agriculture, GIS, Libya, Logistic regression, Soil degradation.

1. Introduction
The primary issue hindering agricultural development in Libya is soil degradation, a condition caused by
salinization, water erosion and wind erosion due to the geology and climate (Nwer et al., 2013; Saad et al., 2013),
and improper use of natural resources (Gebril and Saeid 2012). The low rainfall and high evaporation promotes
soil salinity and subsequently leads to soil instability (Habel, 2013). Approximately 700799 hectares of Primary
Agriculture Regions (PAR) (identified as the Kufrah, Murzuq, Jabal Nafusah, Jabal al Akhdar, and Jifarah)
regions of Libya, regardless of irrigation, are degraded due to salinity (Hachicha and Abdegawed, 2003), with
12 % of the northwestern areas and 23 % of the northeastern areas considered salt-affected (Nwer, 2013).
Increased use of fresh groundwater as a potable water source for a growing population and extensive agriculture
activities increased seawater intrusion into groundwater wells (Nwer et al., 2013) further increasing the soil
salinity problem. Extensive irrigation with this contaminated water source, coupled with poor drainage, resulted
in the proliferation of salt affected soils in the western part of Jifarah (Atman and Habibah, 2013) and in Jabal
Nafusah (Laytimi, 2005).
Rainfall in Libya is spatially and temporally unevenly distributed with deleterious effects. The occasional
heavy showers (Nwer, 2013) accelerates soil erosion by detaching and transporting vulnerable soil either by rain
splash or rill and gully erosion (Pang et al., 2015). Water erosion has degraded 797 ha of the Jabal Nafusah soils
and Jabal al Akhdar soils, collectively (Mahmud, 1995; Ben-Mahmoud et al., 2000). Loss of vegetation covers
from over-grazing and over-cultivation of Libya’s two primarily rain-fed agriculture areas (Jabal Nafusah and
Jabal al Akhdar) resulted in bare soil further exasperating erosion by storm water (Gebril and Saeid, 2012).
Soils of the Jabal al Akhdar region has been degraded the most by wind erosion (Aburas, 2014) since there
is minimal plant vegetative cover protecting the soil (Laytimi, 2005). The soils of Jifarah, one of the most
cultivable areas of the country due to the availability of groundwater experiences both water and wind erosion
due to aridity, poor vegetation cover, and poor landuse decisions during the last half of the 20th century (El-
Tantawi, 2005).
While new strategies may address soil degradation issues (Ben-Mahmoud et al., 2000), more complete
baseline data and information infrastructure to determine the best management strategy are lacking (Khaled,
2001). Over the past 30 years, numerous soil surveys have been conducted in Libya (primarily in Jabal Nafusah,
Jifarah Plain and Jabal al Akhdar in addition to scattered areas in the south) (Nwer et al, 2013). Different
agencies and their interests resulted in varied parameters and geographic extent (Khaled, 2001), thereby limiting

30
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

its practical use (Nwer et al., 2013) beyond the conclusions made in the previous paragraphs. An alternative,
remote approach to classify soil and type degradation across the country must be considered since field
determinations are resource intensive (Mueller et al., 2005). Since soil and climate characteristics are linked to
the development of the soil degradation, empirical models can be developed to identify areas most susceptible to
soil degradation. This empirical model can help identify areas that are either already degraded or most prone to
soil degradation, and protect Libya’s most viable areas for sustainable agriculture development.
The objectives of this study are to:
1. To create a database of soil resources and soil degradation using ArcGIS 10.3 software (ESRI, Redlands,
California). This step will result in a more efficient and economical method for updating soil information and
making it available to a wide variety of stakeholders.
2. To model soil degradation occurrence as a function of soil and climate characteristics within the PAR
using logistic and multinomial logistic regression models.
3. To apply these models to the reminder of the country to predict degradation occurrence and type for the
entire country.

2. Methodology
The paper reported here had three approaches:
1- Because the existing soil data for the country is from different classification systems and no complete
data set was found, creating a baseline data set of soil coverage and degraded areas was the first approach.
2- Develop PAR soil degradation occurrence and type models using the baseline data set of soil and
climate properties.
3- Applying these models to the rest of the country in order to predict the degradation for all Libya.

2.1 Assessment of the Primarily Agriculture Regions 2.1.1 Study Area Description
The study was conducted for the PAR in the country of Libya: Kufrah, Murzuq, Jabal Nafusah, Jabal al Akhdar,
and Jifarah covering 846126 km2 or 52.4 % (Fig. 1a), of the total area of the country. Yields from rainfed
agriculture in Libya are generally low due to the climate, thus the PAR were primarily selected due to the
availability of groundwater aquifers for irrigation (Fig. 1b). In 2005, approximately 22 % of the country’s PAR
depends on groundwater fed irrigation (Food and Agriculture Organization (FAO), 2005). According to
Abagandura and Park (2016), the seasonal rainfall distribution for each PAR varies. Jifarah experiences dry
summers and relatively wet winters. Both Jabal al Akhdar and Jabal Nafusah has a plateau type climate with
greater rainfall (approximately 500 mm in Jabal al Akhdar and 400 mm in the Jabal Nafusah, annually). Murzuq
experiences predesert and desert climatic conditions and Kufrah is characterized as an area with little annual
rainfall (50 to 150 mm). Wind storms occur all year (Libyan Research Center, 2012). The main productions of
PAR are wheat, barley, and the main non-grain agricultural crops include potatoes, onions, tomatoes,
watermelons, oranges, dates, and olives (Abagandura and Park, 2016). Traditionally, a fallow period is used after
a crop production to regenerate its original state of productivity (Omar 2017 pers. comm.).

Fig. 1 The Location of (a) the primary agriculture regions, Kufrah, Murzuq, Jabal Nafusah, Jabal al Akhdar, and
Jifarah; and (b) groundwater aquifers in Libya.

31
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

2.1.2 Creation of Baseline Data (Objective 1)

To complete objective one, ArcGIS was used to integrate data from three existing maps: the African Soil
Classification Map created by the FAO in 2005, the Libyan Primary Agricultural Regions Degradation Map
created by Libyan Government in 2010, and the Libyan Regional Map created by Environmental Systems
Research Institute (ESRI) in 2012.
1. African Soil Classification Map
The FAO generated the Libyan GIS portion of the African Soil Classification Map from Russian and
American topsoil (1 to 5 cm) surveys (Nwer, 2006). However, the governments used different
classification systems and methods of soil analyses: The American surveys were conducted in the
northwestern, central, northeastern, and southern zones over an area of 9 to 4 000 km2, and used the
hydrometer method at a scale of 1:200000 and 1:500000, while the Russian surveys of the northwest
and west zone, determined texture by the feel test at a scale of 1:300000 and 1:500000. The clip and
export data tools in ArcGIS were used to export data from this map to create the Libyan Soil
Classification Map.
2. Libyan Primary Agricultural Regions Degradation Map
In 2010 the Libyan government published the map from assessing soil degradation by the following
methods. Some of the surveys that identified soil texture also measured soil salinity. Electrical
conductivity was determined by laboratory analysis from 1:2.5 soil–water suspensions. Wind and water
erosion deterioration were determined from Landsat satellite imagery between 1960 and 1980 at
1:25000 to 1:60000 scale. The Satellite imagery were used to define the spatial extent of soil erosion.
Segmentation techniques combined with graphic mask were applied over these imageries in order to
obtain the extent of eroded areas for the previous years (Libyan Research Center, 2000).
3. Libyan Region Map
In 2012, ESRI created a map of education infrastructure in the 24 Libyan regions including the five
within the PAR. The layer of the PAR was clipped creating the Libyan Regions Data Map.
To achieve objective one and two, the Libyan Soil Classification Map, Libyan PAR degradation Map and
Libyan Regions map were projected in the same reference system (Africa_Albers_Equal_Area Conic), then
intersected in ArcGIS to relate the soil degradation of the PAR to soil texture. Also intersected were shapefiles of
average annual rainfall, temperature and wind speed derived from twenty observation stations covering Libya
from 2000 to 2008 (maintained by the Libyan Meteorological Administration). New polygons were created by
the intersection of the input polygons and were converted from polygon shapefile to a raster (gridded) data set
with a resolution of 1000 m2. The result was a table of 850 points which was exported to JMP® Pro 12.0.1
software (product of SAS Institute Inc., Cary, NC) for the calculations of the predictive models discussed later in
the paper. The predictive dataset was imported back into ArcMap as a raster and finally converted back to a
polygon dataset.
To study the relationship between soil texture and soil degradation type, Pearson’s Chi-squared test was
used to determine how proportions of the degradation types change across texture levels. The distribution of soil
degradation and each texture is illustrated using a mosaic plot. The vertical length of each rectangle is
proportional to the proportions of the soil degradation in each texture level.

2.2 Modelling Climatic and Soil Characteristics to the Occurrence and Type of Soil Degradation Within PAR
(Objective 2)
2.2.1 Modelling PAR Soil Degradation Occurrence
Logistic Regression (LR), which is commonly used in environmental and ecological studies (Dai et al., 2001;
Lee and Min, 2001; Lee et al., 2013), is a statistical model that is used to predict the binomial (Yes/No) outcome
of a response (dependent) variable using one or several predictor (independent) variables (Bennett, et al., 2008;
Hagan et al 2014). Several factors must be considered for soil degradation including topography (Liu et al., 1994;
Liu et al., 2015), soil temperature and moisture (Wei et al., 2014), and soil texture (Fecan et al., 1998; Li et al.,
2013). Rainfall and air temperature also affects degradation. For example, the greater the intensity and duration
of a rainfall, the higher the soil degradation potential (Wang et al., 2013). In addition, wind intensity is one of the
factors inducing movement of soil (Borrelli et al., 2014).
The relationship between soil degradation (y) (occurrence, no occurrence) and the seven independent variables
(soil texture, soil moisture, soil temperature, slope, rainfall, air temperature and wind speed) was evaluated using
LR model.
The form of the LR model was
=1/1+exp-( + =1 - )
where
P(y) is probability of soil degradation being 1, β0 is an intercept of the model,
βi (with 1< i < 7) are the model coefficient to associated with the independent variables used in the specific

32
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

model evaluated. A positive regression coefficient means that the variable increases the probability of the
outcome, while a negative regression coefficient decreases the probability (Agresti, 2002).
The variables (denoted independent variables, Xi) related to the probability of soil degradation are listed in Table
1. All possible combinations of including and excluding the seven independent variables resulted in 128 different
models to evaluate in JMP.
Table 1. Independent and dependent variables used in logistic regression model for the primary agricultural
regions in the Libya.
Variable Description
I Dependent
Soil degradationa Degraded soil (1—degraded, 0—not)e
Salinization
Water erosion
Wind erosion
II Independent
Soil moistureb I Class I “dry” (1—class I, 0—other classes)d
b
Soil moisture II Class II “xeric” (1—class II, 0—other classes)d
b
Soil temperature I Class I “aridic” (1—class I, 0—other classes)d
b
Soil temperature II Class II “thermic” (1—class II, 0—other classes)d
b
Soil temperature III Class III “moderate” (1—class III, 0—other classes)d
b
Soil temperature IV Class IV “warm” (1—class IV, 0—other classes)d
b
Soil texture I Class I “coarse loamy” (1—class I, 0—other classes)d
b
Soil texture II Class II “loamy” (1—class II, 0—other classes)d
b
Soil texture III Class III “silty loam” (1—class III, 0—other classes)d
b
Soil texture IV Class IV “sand” (1—class IV, 0—other classes)d
b
Soil texture V Class V “ loamy very fine sand” (1—class V, 0—other classes)d
c
Slope Slopee (%)
c
Climate (temperature) Temperaturee (c°)
c
Climate (rainfall) Rainfalle (cm)
a
Wind speed Wind speede (m s-1)
a
Libyan Government (2010)
b
FAO (2005)
c
ESRI Landscape (2014)
d
categorical variable
e
continuous variable
2.2.1.1 Logistic Regression Statistical Analysis
The LR model development approach included:
1- Examining the overall significance of the 128 LR models using the overall Chi-square test and p-values.
When the p-value of the LR was ≤ 0.05, the model was kept.
2- The goodness-of-fit was determined for each kept LR model using the Akaike Information Criterion
(AIC), where the optimal fitted model is identified by the minimum AIC value.
3- Fisher’s F-tests (with associated p-values) tested each independent variable considered in the optimal
fitted model.
To further validate the LR model developed in this study, the Leave-One-Field-Out Validation approach was
used (Pike et al., 2009). A series of analyses were performed in which one PAR was left out and used as a test
case. The coefficients for the LR model were used to estimate the probability of a degradation occurrence in the
test case (Bishop, 2002). If the probability of degradation for a test case was < 0.5, it assumed to have no
degradation and if the probability was ≥ 0.5 it was assumed to have degradation. (Mueller et al., 2005; Pike et al.,
2009). The predicted degradation was compared to the actual degradation. This was repeated with each PAR
being a test case. Misclassification (disagreement between actual and predicted [degradation) totals were used to
examine prediction errors. To display results of the LR validation, data results were imported into ArcMap as a
table with respective latitude and longitude, and then converted into polygons using the shapefile tool.
2.2.2 Modeling PAR Soil Degradation Type
The LR developed in this study modeled the soil and climate characteristics with the soil degradation occurrence
(occurrence or no occurrence) within the PAR. Because the soil degradation had three types (salinization, water
erosion and wind erosion) which were determined already for the PAR, the next step was to develop a model to
predict the three types. For this step, a Multinomial Logistic Regression (MLR) model was developed. The MLR
was used because it is a simple extension of the LR that allows for more than two categories of the dependent
variable (Lin et al., 2014; Starkweather and Moske, 2011). Only significant variables in LR became part of the

33
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

MLR models. One soil degradation type served as the reference category.
2.2.2.1 Multilogistic Regression Statistical Analysis
The MLR model development approach included:
1- Fisher’s F tests and p-values were used to determine if the independent variables used in MLR were
related to the soil degradation type.
2- The coefficients in the MLR were interpreted based on odds ratio (OR). The OR indicated the amount
by which the odds of the soil degradation types (with respect to the reference type) changed as the independent
variables changed by one unit (with respect to the reference category) (Institute for Digital Research and
Education, 2014). Although the magnitude of odds ratio of a category changes with the reference category, its
relative trend and the fit of the overall model is not affected. An OR ≥ 1 indicated a positive relationship between
the independent variable and the probability of soil degradation type existence, while < 1 indicated negative
correlation (Debella-Gilo and Etzelmüller, 2009). With wind erosion as the selected reference, a model was
developed for soil salinization and water erosion.

2.3 Determining Soil Degradation Occurrence and Type for Libya (Objective 3)
After developing LR and MLR (modelling the soil and climate characteristics with the soil degradation
occurrence and type) for the PAR, the next step was applying these models to the areas outside the boundaries of
the PAR to predict degradation occurrence and type for the reminder of Libya. Dunes, salt flats and rocks were
omitted from the models. Although dunes play a very important role in preventing and delaying intrusion of
waters into inland areas (Gómez-Pina et al., 2002), and rocks increase hydraulic roughness and friction,
decreasing the overland flow speed and thus decreasing soil erosion (Jomaa et al., 2012), they would not be
considered for agriculture purposes. The degradation results were imported into ArcMap and converted to a
polygon shapefile to display the results as a degradation map for the entire country.

3. Results and Discussions

3.1 Baseline Soil Data Map
Intersecting the Libyan soil degradation, classification and region maps in ArcGIS created existing data baseline
maps for soil texture and landscape feature distribution (Fig. 2). Calculations included areas (km2) of each soil
texture and land feature present in each PAR (Table 2), and the area (km2) of the degradation type for each
texture within each PAR in Libya (Table 3) were performed.

Fig.2 Soil and land features of the primary agriculture regions in Libya, (a) Jifara, (b) Jabal al Akhdar, (c) Jabal
Nafusah, (d) Murzuq, and (e) Kufrah as created by FAO, 2005.

34
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

Table 2. Area (km2) of different soil textures and land features for the primary agriculture regions in Libya.
Region Soil Land feature
Silty Loamy Coarse Loamy very Sandy Sand Rock Salt Dunes
loam loamy fine sand clay flats
Kufrah 296893 0 0 18289 0 0 825 0 108923

Murzuq 205031 877 0 37566 0 186 5305 0 81376

Jabal 59451 1128 15499 461 0 0 0 336 0

Nafusah
Jabal al 6568 666 0 4082 0 0 0 0 0
Akhdar
Jifarah 0 0 2664 0 0 0 0 0 0

Total 567943 2671 18163 60398 0 186 6130 336 190299

Libya 1012990 22427 81134 109602 8434 10312 38502 676 330\366

Table 3. Areas (km2) of the degraded type for each soil texture in the primary agriculture regions in Libya based
of existing data.
Region Silty loam Loamy Loamy very Coarse loamy Sand
fine sand
SA WA WI SA WA WI SA WA WI SA WA WI SA WA WI
Kufrah 53964 0 10000 0 0 0 0 0 0 0 0 0 0 0 0
Murzuq 49627 0 2057 0 0 0 7851 0 6064 0 0 0 0 0 187
Jabal 50261 3028 4800 671 457 0 254 206 0 381 15118 0 0 0 0
Nafusah
Jabal al 5173 0 1000 0 0 0 1607 0 0 0 0 0 0 0 0
Akhdar
Jifarah 0 0 0 0 0 0 0 0 0 562 2102 0 0 0 0
Total 159025 3028 17857 671 457 0 9712 206 6064 943 17220 0 0 0 187
SA refers to salinization, WA refers to water erosion and WI refers to wind erosion.
Silty loam soils dominated in all the regions except for Jifarah, which is dominated by coarse loamy soils
(Fig. 2 and Table. 2). Dunes and rocks covered 25.6 % of Kufrah and 24.6 % of Murzuq. Salt flats, which
developed from groundwater evaporating and developing a salt pan (Schulz et al. 2015), covers 0.55 % of Jabal
Nafusah.
The quantity of degradation measured at 215370 km2 (33.16 %) of the total area of the PAR (Table 3). Jabal
Nafusah, Jabal al Akhdar, and Jifarah are the most degraded regions (98.2 %, 68.5 % and 99.4). Salinization is
the greatest type of degradation in all the PAR with exception of Jifarah which has soils degraded primarily from
water erosion (Table 3). Laytimi (2005) reported that irrigation with saline groundwater led to soil salinization
in some areas of the PAR. Water erosion is also a significant source of degradation in the Jabal Nafusah region,
probably because the terrain includes steep slopes coupled with a very high content of very fine sand particles
and very low clay content. Sandy soils lack the ability to aggregate leading to weaker physical resistance to
water erosion (Khaled, 2001; Aboufayed, 2013). The wind erosion affected Murzuq, Jabal Nafusah, Jabal al
Akhdar and Kufrah soils (Table 3).
3.1.1 Relationship of Climatic and Soil Characteristics to Soil Degradation Occurrence and Type
The proportions of degraded type (salinization, water, and wind erosion) changes across soil texture (p < 0.001)
(Fig. 3) and reflected what is found in each region (Obj. 2). For example, salinization occurs in all soil textures
except sand. Water drains freely from sand soils with very little to no capillary rise to be expected (Li et al.,
2013). In comparison, the other finer textured soils have micropores in which capillarity resulted in evaporation
of the water from the soil surface and concentrated dissolved salts precipitated at the soil surface (Osman, 2014).
No degradation existed from water erosion and salinity in the sandy texture, most likely because sands contain
macropores that allow water to drain freely producing little runoff (Adekalu et al., 2007). Weakly cohesive soils
(dominated by sand and silt) are more susceptible to wind erosion, especially when desiccated (Geological
Society of London, 2012). In this study soil degradation from wind erosion occurred in all sand dominated areas,
as well as associated with silty loam, loamy, and very fine sand textures. The Libyan desert is a significant
source of mineral dust due to sandblasting from intense (yet not so frequent) wind storms (Laurent et al., 2008).
However, without more sophisticated equipment to collect more complex and detailed wind data, it is most likely
that the predictions of wind erosion found in the present model are low.

35
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

Fig.3 Soil degradation type’s proportions across texture levels in the primary agriculture regions in Libya. The
thickness of each texture represents the percentage of observations for each texture compared to the total data.

3.2 Logistic Regression Development for the Distribution of Soil Degradation Occurrence
From the 128 models with all possible combinations of seven independent variables, only ten possible models
were found significant for predicting soil degradation occurrence (p-values < 0.05) (data not shown). Of these
ten models, the model that had slope, soil texture and wind speed had the lowest AIC values, (AIC = 28.13, p-
value < 0.001). The equation for this model is:
Logit (Soil degradation) = 11.4+ (0.22*Slope) + (2.63*Texture [loamy very fine sand]) + (-3.12*Texture
[Coarse loamy]) + (-7.03*Texture [Loamy]) + (-15.2* Texture [Sand]) + 0.42*Wind speed.
The F-tests with associated p-values of this model variables suggested that slope (p = 0.0142), soil texture
(p <0.0001 for all texture categories), and wind speed (p = 0.0321) had significant relationships with soil
degradation.
The influence of slope on soil degradation, especially from water erosion is well documented (Liu et al.,
1994; Liu et al., 2014; Sensoy and Kara, 2014). In the present study, slope exhibited a positive relationship with
the extent of soil degradation. One-degree increase in slope results in a 0.22 increase in the logit of soil
degradation. This relationship is best seen in the degradation map of the Jabal Nafusah region with slopes
ranging from 4 % to 28 % and having the largest degraded area due to water erosion (30 %). This result supports
Liu et al. (2014) who documented that variations in slope can affect soil erosion.
Salako (2003) reported that soil susceptibility to degradation is influenced by small differences in texture.
The chance of soil degradation increases most when a silty loam soil is present compared to other textures. The
present regression model agreed with the existing degradation data of the PAR, showing that silty texture soil
had the highest degradation occurrence compared to other textures (Table 3).
Wind speed had a positive relationship with soil degradation. One-degree increase in wind speed results in a
0.42 increase in the logit of soil degradation. That agrees with Nwer (2015), who reported that wind erosion is
one of the most important threats to agriculture development in Libya.
Model performance tests included the leave-one-field-out validation analyses and completed by utilizing the
existing data in the PAR (Fig.4, Table 4). Jifarah has no previously determined non-degraded areas so it was not
included in the validation. The validation identified that 66 % to 76 % of the model degradation occurrence
predictions as correct (Table 4) and thus this model was used to predict degradation occurrence in the remaining
areas of Libya.

Fig. 4 Soil degradation occurrence in the primary agriculture regions in Libya as (a) previously determined
(existing), and (b) predicted by LR model using the combination of slope, texture and wind speed.

36
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

Table 4. Actual degradation, frequency of correctly determining degradation occurrence, and percent of
degradation observations that were correctly classified by model predictions from the leave-one-field-out
validation analyses.
Region Predicteda
Actual Degradation Degraded Not degraded Percentage correctb
(# of Observations) (%)
Kufrah Degraded 70 30 70.0
Not degraded 32 70 68.8
Murzuq Degraded 35 18 66.6
Not degraded 24 77 76.2
Jabal Nafusah Degraded 45 16 73.7
Not degraded 9 20 68.0
Jabal al Akhdar Degraded 44 22 66.6
Not degraded 13 28 68.2
a
Degraded if probabilities ≤ 0.5 and non-degraded if probabilities were > 0.5.

3.3 Multilogistic Regression Development for the Soil Degradation Type

To develop the models of predicting soil degradation type, wind erosion was the reference dependent variable
and sand texture was the reference category for the independent variable. JMP software automatically selects the
reference category to be the last in alpha-numeric order. Slope and soil texture, which were significant variables
in predicting the soil degradation distribution, are again significant (F test = 5.23 to 6.33 for salinization and 6.21
to 7.42 for water erosion, p-values < 0.05 for all variables in salinization and water erosion models) in predicting
the soil degradation type. Interpreting the OR identified that silty loam texture (OR= 14.07) resulted in
salinization the most and slope (OR= 13.11) is the leading factor in soil degradation from water erosion.

3.4 Prediction Distribution of the Soil Degradation Occurrence and Type Throughout Libya
Applying the LR and MLR models developed in this study to the areas outside the boundaries of the PAR
created a map (Fig. 5) of the degradation occurrence and type for all Libya. This map identified that 55 % of
Libya’s soils are degraded, with the greatest type of degradation being salinization (30 %) and the least being
water erosion (10 %).

Fig. 5 The probability degradation soil map for all the country developed in this study using LR and MLR
models.

37
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

4. Conclusions
In Libya, the PAR degradation is a result of salinization, water erosion and wind erosion, with impacts to
agricultural development in the country. The combination of slope, soil texture and wind speed (using LR and
MLR models) successfully predicted the spatial distribution of soil degradation and the type of degradation.
Overall 55 % of Libyan soils are degraded, 30 % due to salinization, 10 % due to water erosion, and 15 % due to
wind erosion. However, due to the nature of the wind speed data used, wind erosion is most likely under-
predicted in the model. Hopefully this model will assist in substantiating the need for better equipment and data
collection to develop more detailed models to predict soil degradation from wind erosion.
Additional parameters, which were not obtainable for this work, may enhance the model performance and
allow for other types of soil degradation to be assessed. For example, long term evapotranspiration and drainage
data may also influence soil salinization. Long term changes in organic matter (or organic carbon) content,
nutrient depletion, and microbial activity can all be considered metrics to gauge soil degradation. Different
management practices at the individual farm scale may also affect the occurrence of soil degradation (Garen et
al., 1999). Future research is needed to collect data to assist in determining soil erodibility factors, the cropping
and land-cover factor, and the support practice factor to integrate the Revised Universal Soil Loss Equation
(RUSLE) to estimate soil loss.
Until more detailed data can be collected and organized, the simple models developed in this study can
assist stakeholders in management of Libyan soils, and be applied to neighboring countries that have similar
geography, climate conditions, and data availability.

Acknowledgments
The Libyan Government provided financial support for this research. We are grateful to Brian Ritter, Doctoral
Candidate at Clemson University, for his help with using ArcMap software.

References
Abagandura, G.O., and D. Park. 2016. Libyan agriculture: a review of past efforts, current challenges and future
prospects. Journal of Natural Sciences Research 6(18): 57-67.
Aboufayed, A.F. 2013. Measuring the Amount of Eroded Soil and Surface Runoff Water in the Field. World
Academy of Science, Engineering and Technology 7(12): 825-828.
Aburas, M.M. 2014. Soil erosion, crusting and degradation in the South of Al-Jabal al Akhdar, Libya.
International Conference of Agricultural Engineering. Retrieved 1 November, 2017, from
http://www.geyseco.es/geystiona/adjs/comunicaciones/304/C01300001.pdf.
Adekalu, K.O., I.A. Olorunfemi, and J.A. Osunbitan. 2007. Grass mulching effect on infiltration, surface runoff
and soil loss of three agricultural soils in Nigeria. Bioresource Technology 98: 912–917.
Atman, S.A., and L. Habibah. 2013. A study on soil salinity in Al-Jafarah, Libya using remote sensing
technology. World Applied Sciences Journal 23: 430-434.
Ben-Mahmoud, R., S. Mansur, and A. Al-Gomati. 2000. Land degradation and desertification in Libya, Land
Degradation and Desertification Research Unit, Libyan Center for Remote Sensing and Space Science,
Tripoli, Libya.
Bennett, F. M., S.C. Loeb, M.S. Bunch, and W.W. Bowerman. 2008. Use and selection of bridges as day roosts
by Rafinesque's big-eared bats. Am. Midl. Nat 160(2): 386-399.
Bishop, C.M. 2002. Neural networks for pattern recognition. Oxford Univ. Press, Oxford, England. Retrieved 1
November, 2017, from
http://cs.du.edu/~mitchell/mario_books/Neural_Networks_for_Pattern_Recognition_-
_Christopher_Bishop.pdf.
Borrelli, P., C. Ballabio, P. Panagos, and L. Montanarella. 2014. Wind erosion susceptibility of European soils.
Geoderma 232-234: 471-478.
Dai, F.C., C.F. Lee, and Z.W. Xu. 2001. Assessment of landslide susceptibility on the natural terrain of Lantau
Island, Hong Kong. Environmental Geology 40: 381–391.
Debella-Gilo, M., and B. Etzelmüller. 2009. Spatial prediction of soil classes using digital terrain analysis and
multinomial logistic regression modeling integrated in GIS: Examples from Vestfold county, Norway.
Catena 77(1): 8-18.
El-Tantawi, A.M.M.T. 2005. Climate change in Libya and desertification of Jifara Plain using geographical
information system and remote sensing techniques. Dissertation. PhD Thesis. Mainz, Germany: Johannes
Gutenberg-Universität Mainz.
Fécan, F., B. Marticorena, and G. Bergametti. 1998. Parametrization of the increase of the aeolian erosion
threshold wind friction velocity due to soil moisture for arid and semi-arid areas. Annales Geophysicae. 17:
149-157.
Food and Agriculture Organization. 2005. Libyan Arab Jamahiriya Nutrition Profile – Food and Nutrition

38
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

Division. Retrieved 1 November, 2017, from ftp://ftp.fao.org/es/esn/nutrition/ncp/lby.pdf.

Garen, D., D. Woodward, and F. Geter. 1999. A user agency's view of hydrologic, soil erosion and water quality
modelling. Catena 37: 277–289.
Gebril, A.O., and A.G. Saeid. 2012. Importance of pastoral human factor overloading in land desertification:
case studies in northeastern Libya. World Academy of Science, Engineering and Technology 6: 10-26.
Geological Society of London. 2012. Hot Deserts: Engineering, Geology and Geomorphology: Engineering
Group Working Party Report. M.J. Walker, ed. The Geological Society Publishing House, Bath, UK. 424.
Gómez-Pina, G., J.J. Muñoz-Pérez, J.L. Ramírez, and C. Ley. 2002. Sand dune management problems and
techniques, Spain. Journal of Coastal Research 36(36): 325-332.
Habel, A.Y. 2013. The role of climate on the aggregate stability and soil erodibility of selected El- Jabal Al-
Akhdar soils-Libya. Alexandria Journal of Agricultural Research 58: 261-271.
Hachicha, M., and G. Abdegawed. 2003. Aspects of salt-affected soils in the Arab world. Retrieved 21
November, 2017, from
http://www.nerc-wallingford.ac.uk/research/cairoworkshop/cairoworkshop.pdf.
Hagan, D.L., E.A. Mikhailova, T.M. Shearman, P.T. Ma, J.S. Nankaya, S.K. Hart, H.E. Valdetero, W.C. Bridges,
and H. Yun. 2014. The role of soil and landscape factors in Chinese privet (Ligustrum sinense) invasion in
the Appalachian piedmont. Invasive Plant Sci. Manage 7(3): 483-490.
Institute for Digital Research and Education. 2014. Annotated SPSS output, multinomial logistic regression.
Retrieved 21 November, 2017, from http://www.ats.ucla.edu/stat/spss/output/mlogit.htm.
Jomaa, S., D.A. Barry, B.C.P. Heng, A. Brovelli, G.C. Sander, and J.Y. Parlange. 2012. Influence of rock
fragment coverage on soil erosion and hydrological response: Laboratory flume experiments and modeling.
Water Resources Research 48(5).
Khaled, B. 2001. Land resources information systems in the Near East. Report on World soil resources. FAO,
Report 99, Rome, Italy.
Laurent, B., B. Marticorena, G. Bergametti, J. F. Léon, and N. M. Mahowald. 2008. Modeling mineral dust
emissions from the Sahara desert using new surface properties and soil database. Journal of Geophysical
Research: Atmospheres 113(D14).
Laytimi, A. 2005. Market and trade policies for Mediterranean agriculture: the case of fruit/vegetable and olive
oil. Retrieved 1 November, 2017, from
http://medfrol.maich.gr/documentation/view/reports/wp1-asr/Libya.pdf.
Lee, K., H. Ahn, H. Moon, R.L. Kodell and J.J. Chen. 2013. Multinomial logistic regression ensembles. Journal
of Biopharmaceutical Statistics 23(3): 681.
Lee, S., and K. Min. 2001. Statistical analysis of landslide susceptibility at Yongin, Korea. Environmental
Geology 40: 1095–1113.
Li, X., S.X. Chang, and K.F. Salifu. 2013. Soil texture and layering effects on water and salt dynamics in the
presence of a water table: a review. Environmental Reviews 22(1): 41-50.
Libya Agricultural Center. 2000. The soil erosion in the Tripoli region. 35 (1).
Libyan Research Center. 2012. Wind control in Libya (in Arabic). Albida, Libya.
Lin, Y., X. Deng, X. Li, and E. Ma. 2014. Comparison of multinomial logistic regression and logistic regression:
which is more efficient in allocating land use. Frontiers of Earth Science 8(4): 512-523.
Liu, B. Y., M.A. Nearing, and L.M. Risse. 1994. Slope gradient effects on soil loss for steep slopes. Transactions
of the ASAE 37(6): 1835-1840.
Liu, Q. J., J. An, L.Z. Wang, Y.Z. Wu, and H.Y. Zhang. 2015. Influence of ridge height, row grade, and field
slope on soil erosion in contour ridging systems under seepage conditions. Soil and Tillage Research 147:
50-59.
Liu, Q.J., Z.H. Shi, X.X. Yu, and H.Y. Zhang. 2014. Influence of microtopography, ridge geometry and rainfall
intensity on soil erosion induced by contouring failure. Soil and Tillage Research 136: 1–8.
Mahmoud, K.B. 1995. Libyan Soils. First Edition ed. Tripoli: National Research Scientific Organization.
Mueller, T. G., H. Cetin, R.A. Fleming, C.R. Dillon, A.C. Karathanasis, and S.A. Shearer. 2005. Erosion
probability maps: Calibrating precision agriculture data with soil surveys using logistic regression. Journal
of Soil and Water Conservation 60(6): 462-468.
Nwer, B. 2015. Techniques to control wind erosion in Libya. Retrieved 30 November 2017, from
https://www.researchgate.net/publication/290004132_Techniques_to_Control_Wind_Erosion_in_Libya.
Nwer, B.A.B. 2013. Soil resources in Libya. In Soil Resources of Mediterranean and Caucasus Countries, ed. Y.
Yigini, P. Panagos, and L. Montanarella, 175- 188. Office for Official Publications of the European
Communities: Luxembourg.
Nwer, B., A. Ali, and H. Zurqani. 2013. Soil salinity mapping model developed using RS and GIS in Libya.
Global Science & Technology Forum.
Nwer, B.A.B. 2006. The application of land evaluation technique in the north-east of Libya. PhD Thesis.

39
Journal of Natural Sciences Research www.iiste.org
ISSN 2224-3186 (Paper) ISSN 2225-0921 (Online)
Vol.7 No.24 2017

Bedford, UK: Cranfield University.Osman, K.T. 2014. Wind Erosion. In Soil degradation, conservation and
remediation, ed. Osman, K.T, 103. Dordrecht: Springer.
Omar, Salah. Employee in the Libya Secretariat of Agriculture. pers.comm., 6 November 2017.
Pang, Y., L. Liang, Y. Tang, and Z. Zhu. 2015. Numerical analysis and experimental investigation of combined
erosion of cavitation and sandy water erosion. Proceedings of the Institution of Mechanical Engineers
230(1): 31.
Pike, A.C., T.G. Mueller, A. Schörgendorfer, S.A., Shearer, S.A., and A.D. Karathanasis. 2009. Erosion index
derived from terrain attributes using logistic regression and neural networks. Agronomy Journal 101(5):
1068.
Saad, A.M.A., N.M. Shariff, and S. Gairola. 2013. Nature and causes of land degradation and desertification in
Libya: Need for sustainable land management. African Journal of Biotechnology 10(63): 13680-13687.
Salako, F.K. 2003. Susceptibility of coarse-textured soils to soil erosion by water in the tropics. Retrieved 2
November, 2017, from
http://www.iaea.org/inis/collection/NCLCollectionStore/_Public/38/100/38100125.pdf.
Schulz, S., M. Horovitz, R. Rausch, N. Michelsen, U. Mallast, M. Köhne, and R. Merz. 2015. Groundwater
evaporation from salt pans: Examples from the eastern Arabian Peninsula. Journal of Hydrology 531: 792-
801.
Sensoy, H., and O. Kara. 2014. Slope shape effect on runoff and soil erosion under natural rainfall conditions.
iForest-Biogeosciences and Forestry 7.2: 110.
Starkweather, J., and A.K. Moske. 2011. Multinomial logistic regression. Retrieved 2 November, 2017, from
http://www.unt.edu/rss/class/Jon/Benchmarks/MLR_JDS_Aug2011.pdf
Wang, L., J. Huang, Y. Du, Y. Hu, and P. Han. 2013. Dynamic assessment of soil erosion risk using Landsat TM
and HJ satellite data in Danjiangkou Reservoir area, China. Remote Sensing 5(8): 3826-3848.
Wei, W., F. Jia, L. Yang, L. Chen, H. Zhang and Y. Yang. 2014. Effects of surficial condition and rainfall
intensity on runoff in a loess hilly area, China. Journal of Hydrology 513: 115–126.

40