Spatiotemporal Analysis, Driving Force, and Simulation of Urban Expansion Along the Ethio–Djibouti Trade Corridor: The Cases of Dire Dawa City, Eastern Ethiopia

Abstract

Urbanization has emerged as one of the most significant global challenges and opportunities of the 21st century, driven by a complex interplay of dynamic processes. In Ethiopia, cities have undergone rapid expansion in recent decades, largely due to state-led economic reforms and infrastructure development. This study aims to investigate the spatiotemporal dynamics, driving forces, and future projections of urban expansion along the Ethio–Djibouti trade corridor, with a focus on Dire Dawa City in eastern Ethiopia. Landsat imagery from 1993, 2003, 2013, and 2023 was utilized to detect land use and land cover (LULC) changes and analyze urban growth patterns. Additionally, maps illustrating the city’s demographic, economic, and topographic characteristics were developed to identify the key driving factors behind land conversion and urban expansion. The spatial matrix and landscape expansion index were employed to examine the spatial patterns of urban growth. Furthermore, the study applied the Multi-Layer Perceptron–Markov Chain (MLP–MC) model to simulate future LULC changes and urban expansion. The results indicate that the built-up area in Dire Dawa has increased significantly over the past three decades, growing from 6.21 km² in 1993 to 21.54 km² in 2023. This urban growth is predominantly characterized by edge expansion, reflecting a pattern of unidirectional, unsustainable development that has consumed large areas of agricultural land. The analysis shows that socioeconomic development and population growth have had a greater influence on LULC conversion and urban expansion than physical factors. Based on these identified drivers, the study projected land conversion and simulated urban expansion for the years 2043 and 2064. The findings underscore the urgent need for context-sensitive urban growth strategies that harmonize local realities with national development policies and the Sustainable Development Goals.

1. Introduction

Urbanization is one of the most transformative and complex global processes of the 21st century, driven by a dynamic interplay of demographic, economic, technological, and environmental factors [1]. Today, urban areas accommodate 58% of the world’s population and are expected to more than double by 2050, at which point nearly 7 in 10 people will live in cities [2]. While urbanization brings opportunities for economic development, innovation, and improved living standards [3], its unregulated expansion, especially in developing countries, poses significant challenges due to its profound effects on land use patterns, infrastructure, environmental sustainability, and social dynamics [4]. In this context, spatially explicit evidence is crucial for understanding urban landscape transformations and designing sustainable urban planning strategies at both regional and local scales.

Urban expansion is a near-universal phenomenon that presents both opportunities and constraints. Numerous studies have investigated the extent of urban expansion and its spatiotemporal patterns across varying spatial scales [5,6,7]. According to [8], China’s urban built-up areas experienced significant growth between 1995 and 2020, with an average expansion rate of 79.78% from 1995 to 2000, 79.06% from 2000 to 2005, 21.62% from 2005 to 2010, 37.00% from 2010 to 2015, and 69.43% from 2015 to 2020. In most developing countries, urban growth tends to be predominantly horizontal and uncontrolled, often at the expense of fertile agricultural land and natural ecosystems [9]. In recent years, significant land consumption has been observed in Ethiopian cities such as Addis Ababa [10], Hawassa [11], and Bahir Dar [12]. These studies generally suggest that the rate and magnitude of urban expansion vary across cities, influenced by socio-economic contexts and local conditions. However, most of the existing research provides limited spatially explicit information on drivers of urban expansion, often relying on qualitative descriptions.

Globally, urban expansion results from a combination of human and physical factors, manifesting differently across regional and local contexts [13]. Human drivers such as population growth, economic development, and state policy reforms have intensified in recent decades [14,15]. Geographic Information Systems (GIS) and remote sensing are increasingly used to address urban challenges due to their potential capability to generate spatially explicit information on drivers of urban expansion and integrate them in a single system [16,17]. More importantly, the combination of geospatial tools with statistical and simulation models has significantly advanced urban studies and improved the ability to solve complex urban issues. Several modeling approaches, such as logistic regression [18], multiple regression, artificial neural networks (ANN) [19,20], Cellular Automata–Markov models [21,22], and the Markov Chain Model [23], have been widely employed to examine the driving forces behind urban expansion and to predict future scenarios. This study applies the Multi-layer Perceptron–Markov Chain model to predict urban expansion due to its ability to model land use transitions based on historical data and estimate future land cover changes [24,25]. Its strength lies in its simplicity, efficiency, and suitability for integration with other geospatial and simulation tools, making it a robust method for forecasting long-term urban growth patterns [24]. While the Multi-layer Perceptron–Markov Chain model has been widely used to simulate urban expansion, little is known about its effectiveness and underlying mechanisms in cities shaped by regional economic corridors.

In the Ethiopian context, state-led economic reforms and infrastructural investments, such as industrial park developments and the expansion of trade corridors, have played a central role in driving urban growth. These dynamics are further intensified by high natural population growth and internal migration [26]. A prime example is the Ethio–Djibouti trade corridor, where urban centers have rapidly expanded due to enhanced commercial activity and improved connectivity. Dire Dawa, strategically located along this corridor, has experienced significant spatial growth influenced by infrastructure development, population pressure, and regional economic integration. This urban expansion has reshaped the city’s physical structure while also generating socio-economic and environmental challenges, including land use conflicts, the loss of agricultural land, and the proliferation of informal settlements [27,28].

Despite the city’s strategic location and rapid transformation, there is a noticeable gap in comprehensive, spatially detailed studies that analyze the patterns, drivers, and future trajectories of urban expansion in Dire Dawa. This study aims to fill that gap by investigating the spatiotemporal dynamics of urban growth in the city from previous decades to the present, identifying the key forces driving this change, and simulating future urban development scenarios for the years 2043 and 2063. This study provides evidence-based insights to guide sustainable urban planning and land management in Dire Dawa and other fast-growing cities, while integrating theory with practice to establish a transferable framework and methodology for future planning and research.

2. Materials and Methods

2.1. Description of the Study Area

This study was conducted in Dire Dawa Metropolitan City, which is situated in the eastern part of Ethiopia along the Ethio–Djibouti trade corridor. Geographically, it lies between 9°33′30″ and 9°39′00″ N latitude, and 41°45′30″ to 41°53′30″ E longitude (Figure 1), covering an estimated area of 77.05 km². The city’s elevation ranges from 1069 to 1461 m above sea level. Based on agroclimatic zonal classifications, the area is described as having a semi-arid climate and is characterized by a bimodal rainfall type [29]. According to the long-term climate data registered by the Ethiopian National Meteorological Institution (ENMI, 2024), at Dire Dawa Station, the average annual rainfall was 612 mm, while the mean monthly maximum and minimum temperatures were 31.8 °C and 17.9 °C, respectively.

Dire Dawa is the second federal city next to Addis Ababa and shares an administrative boundary with the Somali Regional State to the north and east, and with the Oromia Regional State to the south and west. The city was officially founded in 1902 during the reign of Emperor Menelik II, primarily due to the construction of the Addis Ababa–Djibouti Railway [30]. Dire Dawa has experienced significant population growth in recent decades. In 1950, the population of Dire Dawa was 18,176. As reported by the UN World Urbanization Prospects, its 2025 population is now estimated at 507,101. This reflects an increase of 21,543 over the past year, representing a 4.44% annual change. The key livelihoods of the population in Dire Dawa City and its peri-urban areas include commerce and trade, which dominate the urban economy due to the city’s strategic location. In the peri-urban zones, livestock rearing and small-scale farming remain vital sources of household income. Currently, it provides significant contributions to both the national and regional economy due to its strategic location and serves as the hallmark of economic activity, including a major trade corridor to access the seaport, industries, dry port, free trade zone, commercial trade, and unique cultural heritage and historical sites. These development activities have changed not only the demography and increased the urban population, but also the size of the area occupied by urban settlements.

2.2. Data Source and Processing

This study employed a diverse dataset comprising remote sensing imagery, a Digital Elevation Model (DEM), topographic information, and socio-economic data obtained from various sources (

Table 1. Data types and sources used in this study.

Dataset	Source	Purpose
Landsat 5 (TM), 1993	USGS Earth Explorer	Preparation of LULC maps for the study area for 1993, 2003, 2013, and 2023
Landsat 7 (ETM+), 2003	USGS Earth Explorer
Landsat 8 OLI, (2013)	USGS Earth Explorer
Landsat 9 OLI, (2023)	USGS Earth Explorer
Aerial photograph (1993, 2003, 2013) and Google Earth	Ethiopian Geospatial Information Institute (EGII)	For validation of classified LULC
ASTER (DEM)	Earthdata	Preparation of elevation and slope maps
Road	OpenStreetMap	Preparation of road map
Railway	OpenStreetMap	Preparation of railway map
Population	Worldpop.org	Population density map
Point Location of (Public Institution, Industries, Factories and Airport)	Google Earth and Field Survey	Preparation of Euclidean distance maps

). To analyze urban expansion in Dire Dawa, Landsat images from four different years—Landsat 5 TM (1993), Landsat 7 ETM+ (2003), Landsat 8 OLI (2013), and Landsat 9 OLI (2023)—were utilized. These years were selected based on data availability and key historical and state policy shifts.

Preprocessing of the satellite imagery was performed using ERDAS IMAGINE 2022, which involved layer stacking, false-color composite, subsetting, mosaicking, enhancement, and radiometric correction [31,32]. Ground control points (GCPs) and aerial photographs were used for the classification of images taken at different times. Accordingly, a total of 150 GCPs were collected using stratified random sampling, of which 70% were used for calibration and the remaining 30% for validation [27,33]. A supervised classification using the Maximum Likelihood Classification (MLC) algorithm was applied to generate LULC maps of Dire Dawa City [34].

Based on East African vegetation classification standards [35,36], four primary LULC classes were identified: agriculture, bare land, built-up area, and vegetation (

Table 2. Description of land use/land cover categories used in the study.

LULC Class	Description
Agriculture	Areas dedicated to agricultural production include cultivated fields, grazing lands, fruit orchards, and livestock confinement facilities. This includes both actively farmed land and fallow fields.
Bare land	Areas with minimal vegetation primarily contain exposed earth materials such as stone, gravel, sand, silt, and clay. Examples include sandy areas, barely exposed rocks, and quarries and dried-up rivers.
Built-up	Areas of high-density usage where structures dominate the landscape include urban centers, rural settlements, roadside developments, infrastructure for transportation and utilities, industrial and commercial zones, and institutional facilities.
Vegetation	Areas characterized by natural or partially natural vegetation, such as urban forests, areas dominated by shrubs, and grasslands.

).

The study used overall accuracy and Kappa coefficients to validate the accuracy of the classified map based on the formula used by [37] as indicated in Equations (1) and (2).

$$OAC={\displaystyle \frac{{\sum }_{i=1}^r{\mathrm{x}}_{ii}}{N}}$$

(1)

$$K={\displaystyle \frac{N\sum _{i=1}^r{x}_{ii}-\sum _{i=1}^r(x_{i+}\ast x_{+i})}{N^2-\sum _{i=1}^r(x_{i+}\ast x_{+i})}}$$

(2)

where OAC is overall accuracy, K is the Kappa coefficient, X_ii is the number of correctly classified samples for class i (diagonal element in the confusion matrix), X_i+ is the total number of reference samples (ground truth) for class i (i.e., sum of row i), X_+i is the total number of classified samples as class i (i.e., sum of column i), N is the total number of samples, and r is the number of classes.

Accordingly, the results of accuracy assessments showed overall accuracies of 88.72%, 90.08%, 89.23%, and 87.30% for 1993, 2003, 2013, and 2023, respectively, with corresponding Kappa coefficients of 0.8206, 0.8315, 0.8079, and 0.8068, meeting recommended standards.

On the other hand, previous studies have highlighted both natural and anthropogenic drivers of urban expansion [38,39,40] at different spatial scales.

In this study, ten key variables, namely roads, industries, manufacturing, population, railway, institutions, airport, elevation, slope, and 2023 built-up area, were systematically developed and integrated as drivers of urban expansion through spatial and thematic analysis. Proximity-based layers, such as distance to roads, railways, industries, manufacturing facilities, institutions, and airports, were generated using Euclidean distance tools to reflect accessibility and attraction potential for urban growth. Elevation and slope data were derived from a Digital Elevation Model (DEM) to account for the physical constraints of the landscape. Population density data were generated from worldpop.org, which directly reflects the pressure people place on land and resources. The 2023 built-up area was extracted from classified satellite imagery to represent the existing urban footprint, serving as a baseline for analyzing expansion trends. These factors were standardized, reclassified, and converted to raster format with a spatial resolution of 30 m (Figure 2).

This study employed different geospatial software, such as ERDAS IMAGINE 2022, QGIS 3.40.1, ArcGIS 10.8, and TerrSet, to process and analyze the data used in the study.

2.3. Drivers of Urban Expansion and Hypothesis

Urban expansion is significantly shaped by proximity to key infrastructural and socio-economic elements. These variables serve as critical constraints in urban growth modeling, as areas with higher elevation or steeper slopes tend to be less suitable for development due to construction difficulties and increased cost; however, slope climbing of urban expansion occurs [41].

The railway runs from the capital city of Ethiopia to Djibouti, and the vast road network development has transformed Dire Dawa into an important transport corridor, which has increased land demand and support for urban expansion. Areas near roads and railways tend to grow faster due to enhanced accessibility [42], while flatter slopes are preferred for development because of lower construction costs [43]. Dire Dawa has many industries, manufacturing areas, and business centers; industrial zones and the location of factories stimulate nearby growth by creating job opportunities, drawing both residential and commercial developments [44]. Similarly, public institutions such as schools and hospitals attract settlements and services around them [45]. Expansion is also more likely near existing built-up areas, as development typically radiates outward for continuity of infrastructure [46]. Additionally, proximity to airports fosters growth by boosting economic activity and attracting diverse land uses [47]. These factors collectively guide the direction and intensity of urban sprawl. Accordingly, Figure 3 indicates the spatial distribution of drivers of urban expansion used in the study area.

3. Method of Data Analysis for Urban Expansion and Simulation

3.1. Rate of Urban Expansion

In this study, classified satellite images were utilized to quantify both the magnitude and rate of LULC changes across different time periods. To analyze the rate of urban expansion during the study period, Equation (3), as recommended in [37] was used.

$$P=\left({\displaystyle \frac{A_2-A_1}{A_1}}\right){\displaystyle \frac{1}{(T_2-T_1)}}\ast 100$$

(3)

where P is the rate of change (in %), and A₁ and A₂ represent the area of the LULC classes at times T₁ and T₂, respectively.

To analyze the LULC conversion to built-up area and assess the losses, gains, and the net change between two successive periods, the study applied the formula used by [48] as indicated in Equations (4) and (5).

$$L_i={\sum }_{j\ne i}{A}_{i\to j}$$

(4)

$$G_i={\sum }_{j\ne i}{A}_{j\to i}$$

(5)

where L is the losses, G is the gains, A_i→j is the area converted from class i to class j, and A_j→i is the area converted from class j to class i.

3.2. Landscape Expansion Index

This study calculated the Landscape Expansion Index (LEI) to show the urban growth pattern in the study area. Accordingly, a 200-m buffer zone was generated from the center of the city. The value of the index was calculated in Equation (6) as described in [10]. Based on the value of the Landscape Expansion Index, urban growth was classified into infilling, edge expansion, and outlying, where LEI > 50–100, (LEI > 0–50), and (LEI = 0), respectively.

$$LEI=100\ast {\displaystyle \frac{A_o}{A_o+A_v}}$$

(6)

where LEI is the landscape expansion index, A₀ is the area where the buffer around a new urban patch overlaps with pre-existing urban zones, and A_v is the area where the buffer intersects with nonurban or undeveloped land.

3.3. Urban Expansion Direction

The study also analyzed the spatiotemporal patterns of urban expansion in terms of distance and equal-sector analysis [10,40]. Built-up areas from 1993, 2003, 2013, and 2023 were analyzed to capture expansion trends across directions and distances, using Equations (7) and (8). Thus, the city was divided into 16 directional quadrants at 22.5° intervals and six concentric buffer zones, each with a 2-km radius from the core center of the city (Addis Ketema), which is known as the central business district (CBD; X = 814,510.077; Y = 1,060,250.059).

$${DUEI}_i={\displaystyle \frac{A_i}{A_t}}\ast 100$$

(7)

$$UED={\displaystyle \frac{\sum _{i=1}^n\left(d_i\ast a_i\right)}{{\sum }_{i=1}^n{a}_i}}$$

(8)

where DUEI is the directional urban expansion index; A_i is the built-up area change in direction sector i; A_t is the total built-up area change in all directions; UED is the urban expansion direction; d_i is the distance from the urban center to built-up pixel i; and a_i is the area or weight of built-up pixel i.

3.4. Simulation of Urban Expansion

Urban expansion is driven by the interplay of biophysical, socio-economic, infrastructural, and institutional factors, and is used as an input in urban growth models [49]. In this study, the MLP-Markov chain model was used to predict urban expansion for 2043 and 2063. The model predicts land use/land cover change by analyzing past transition probabilities between land cover classes over time. This study incorporates ten driving forces to identify the likelihood of future changes. The Markov probability matrix was used to quantify the likelihood of transitions between different land cover classes over a specified period based on the following Equation (9) as suggested by [50].

$$P_{ij}=\left(\begin{array}{c}\begin{array}{ccc}{p}_{11}& p_{12}\dots & p_{1n}\\ p_{21}& p_{22}\dots & p_{2n}\\ ⋮& ⋮\ddots & ⋮\end{array}\\ \begin{array}{ccc}{p}_{n1}& p_{n2}\dots & p_{nn}\end{array}\end{array}\right)\left(0\le P_{ij}\le 1\right)$$

(9)

where P is the Markov probability matrix, and P_ij stands for the probability of converting from current state i to another state j in the next time period. Low transition probabilities will have a value near (0), and high transition probabilities near (1).

Finally, the study used the Kappa coefficient to validate the accuracy of the predicted urban expansion using Equation (10).

$$K={\displaystyle \frac{P_o-P_e}{1-P_e}}$$

(10)

where K is the Kappa coefficient; P₀ is the probability that the predicted result is consistent with the actual result (overall accuracy); and P_e is the probability that the predicted results are consistent with the real ones by chance.

The Kappa coefficient quantifies the agreement between predicted and actual land use, with values >0.8 indicating very high accuracy and <0.4 denoting poor reliability [48].

4. Results

4.1. LULC Map of the Study Area

From 1993 to 2023, Dire Dawa City underwent significant land use and land cover (LULC) changes characterized by a substantial increase in built-up areas and a marked decline in agricultural land. In 1993, agriculture dominated the landscape, covering 54.02% of the area, while built-up land accounted for only 8.06%. Over the next three decades, built-up areas expanded more than threefold, reaching 27.96% by 2023. This urban growth came largely at the expense of agricultural land, which shrank to 38.07% by 2023. Bare land, after an initial decline from 19.72% in 1993 to 5.90% in 2003, rose again to 14.92% in 2023, likely due to land degradation and development activities. (

Table 3. Area cover of LULC in Dire Dawa City over three decades (1993–2023).

LULC Class	1993		2003		2013		2023
	Area (km2)	%	Area (km2)	%	Area (km2)	%	Area (km2)	%
Agriculture	41.62	54.02	40.28	52.23	43.50	56.46	29.33	38.07
Bare land	15.20	19.72	4.55	5.90	6.14	7.97	11.49	14.92
Built-up	6.21	8.06	12.55	16.27	13.76	17.86	21.54	27.96
Vegetation	14.02	18.20	19.75	25.60	13.64	17.70	14.68	19.06

and Figure 4).

4.2. Accuracy Assessment for the Classified Maps

Table 4. Accuracy assessment of the LULC maps generated for Dire Dawa City from 1993 to 2023.

LULC Class	Accuracy (%)
	1993		2003		2013		2023
	Producer’s	User’s	Producer’s	User’s	Producer’s	User’s	Producer’s	User’s
Agriculture	92.96%	90.41%	96.05%	93.59%	94.94%	92.59%	89.06%	90.48%
Bare land	92.86%	81.25%	70.00%	87.50%	83.33%	83.33%	92.31%	85.71%
Built-up	78.57%	91.67%	84.21%	88.89%	80.77%	84.00%	83.33%	92.59%
Vegetation	75.00%	93.75%	84.62%	81.48%	78.95%	83.33%	84.21%	72.73%
Overall Accuracy	88.72%		90.08%		89.23%		87.30%
Kappa Coefficient	0.8206		0.8315		0.8079		0.8068

summarizes the accuracy of LULC classifications for 1993, 2003, 2013, and 2023 using producer’s accuracy, user’s accuracy, overall accuracy, and the Kappa coefficient. Agriculture consistently achieved high accuracy, peaking in 2003 and slightly decreasing by 2023. Bare land accuracy varied, with a notable drop in 2003 but a recovery by 2023. Built-up areas improved in user’s accuracy over time, while vegetation showed the most fluctuation, especially a decline in 2023. Despite some variation, overall accuracy remained strong, ranging from 87.30% to 90.08%, with Kappa values above 0.80, reflecting reliable classification amid increasing land-use complexity.

4.3. Rate of Urban Expansion from 1993–2023

The analysis of the rate of change in LULC in the study area from 1993 to 2023 revealed notable transformations in land use and land cover, characterized primarily by the rapid expansion of built-up areas, which grew at an annual rate of 24.69%, indicating intense urban growth and infrastructure development. This urban sprawl came at the cost of agricultural land, which declined by 2.95% per year, suggesting significant conversion of farmland to urban uses. Bare land exhibited inconsistent changes, initially decreasing and then increasing sharply, leading to an overall annual decline of 2.44%, possibly reflecting both land degradation and redevelopment. Vegetation cover showed only a slight net increase of 0.47%, with alternating periods of gain and loss. These dynamics are illustrated in Figure 5.

Between 1993 and 2023, the land use and land cover (LULC) of the study area underwent notable changes marked by dynamic gains and losses. Agricultural land, despite consistent losses totaling 17.36 km², also experienced gains of 20.48 km², resulting in a net positive change of 3.11 km², likely due to land recovery or expansion. Bare land showed the most significant decline, particularly in the first decade, with a 9.66 km² loss, and remained relatively unchanged thereafter, resulting in a total net loss of 9.66 km². Built-up areas expanded steadily, with a 9.71 km² gain and a net increase of 6.1 km² over the period, primarily at the expense of agricultural and bare lands. Vegetation cover initially increased but later declined sharply, especially after 2003, leading to a net loss of 5.63 km² over the study period. Transitions to built-up land were mainly driven by agricultural and bare land, especially in the first decade, while their contributions stabilized in later years. Over the entire period, the contributions to built-up areas were 12.19 km² from agriculture and 3.25 km² from bare land, indicating a gradual yet sustained urban growth fueled by modest land conversions. These patterns reflect the influence of urban expansion, land reclamation, and environmental shifts (Figure 6 and Figure 7).

4.4. Directional Concentration of Urban Expansion

Between 1993 and 2023, Dire Dawa experienced significant and directionally biased urban expansion, predominantly toward the N, WNW, NW, NNW, W, and NNE.

In 1993, the built-up area was compact, concentrated within a 2 km radius of the city center, totaling around 3.34 km². Expansion was notably higher in the WNW (1.36 km²), N (1.13 km²), and NNW (1.06 km²) directions, while expansion in southern and eastern areas was minimal. As distance increased beyond 2 km, the extent of built-up land diminished significantly, indicating a tightly clustered urban core (Figure 8i and Figure 9a).

By 2003, the urban area expanded to 12.55 km², with intensified development along the NW (1.64 km²), N (2.34 km²), and especially the WNW (3.27 km²) corridors. Urban growth radiated primarily along the NNW, NW, N, and WNW axes, with significant built-up area within 4 km, which is 4.72 km^2, and tapering off beyond 6 km except in the W and WNW directions, where linear expansion extended up to 13 km. (Figure 8ii and Figure 9b).

In 2013, the built-up area grew further to 13.76 km², maintaining the strong west-northwest to north-northwestward expansion. The NW (2.05 km²), N (2.56 km²), and WNW (3.20 km²) directions remained dominant. Development within a 2–4 km radius continued to intensify, while growth beyond 6 km persisted, especially along roads and in favorable terrains. Conversely, the south, east, southwestern, and east-southeastern areas again showed limited development, indicating sustained spatial imbalance (Figure 8iii and Figure 9c).

By 2023, the built-up area reached 20.54 km², with the WNW corridor emerging as the most heavily urbanized (7.50 km²), followed by NW and N. Expansion extended notably beyond 6–8 km, showing a ribbon-like growth pattern likely shaped by flat topography, the industrial zone, and transportation infrastructure. Although moderate growth was recorded within the 2 km urban core across all directions, peripheral spread was concentrated toward the WNW and NW. Southern and eastern sectors like SW (0.02 km²), ENE (0.05 km²), and ESE (0.06 km²) remained minimally developed, reinforcing a long-standing trend of asymmetrical urban sprawl. Overall, the city’s growth trajectory over three decades illustrates a consistent expansion along the WNW to NNE axis, driven by geographic, infrastructural, and socio-economic factors (Figure 8iv and Figure 9d).

4.5. Urban Growth Pattern

Between 1993 and 2023, the urban landscape expansion of the study area exhibited dynamic changes in spatial growth patterns, characterized by three primary growth typologies: outlying growth, edge expansion, and infilling. During the initial decade (1993 to 2003), urban expansion was predominantly driven by edge expansion, which accounted for 89.58% (11.11 km²) of the total growth, indicating a compact and contiguous development pattern. Outlying growth contributed 10.03% (1.24 km²), while infilling was minimal at only 0.39% (0.05 km²) (Figure 10a).

In the subsequent period (2003–2013), the trend remained similar, with edge expansion still dominant at 80.44% (10.98 km²). However, there was a noticeable rise in infilling, which increased to 9.21% (1.26 km²), and a slight increase in outlying growth at 10.36% (1.41 km²), suggesting a gradual densification and outward sprawl (Figure 10b). From 2013 to 2023, edge expansion intensified dramatically, comprising 97.86% (21.08 km²) of the total growth, while outlying growth declined sharply to just 0.57% (0.12 km²), and infilling rose slightly to 1.57% (0.34 km²) (Figure 10c). This pattern highlights a significant increase in lateral urban expansion with minimal leapfrog development and moderate internal densification over the 30-year period.

4.6. Prediction of Urban Expansion for Dire Dawa City

The prediction of urban expansion was conducted using a robust modeling framework trained with 3403 samples per class over 10,000 iterations, achieving an overall accuracy of 80.62% and a skill measure of 0.8725, indicating strong predictive reliability. The model demonstrated high proficiency in simulating land cover transitions, particularly for agriculture to built-up areas, with a skill value of 0.581, while transitions from vegetation to built-up areas, 0.2547, and bare land to built-up areas, 0.4844, showed moderate predictive strength. Persistence values indicated that agriculture 0.3120 was the least stable, whereas vegetation 0.4029 and bare land 0.4151 exhibited relatively higher stability over the modeled period. Sensitivity analysis revealed that proximity to industries, existing built-up areas, manufacturing facilities, transportation networks, elevation, and population density were significant drivers of urban growth, exerting a notable influence on model performance. In contrast, distance from public institutions was found to be the least influential factor. Interestingly, the distance from the airport skill measure, with a value of 0.4616 and a slope of 0.4932, played a moderate role in shaping the predicted spatial patterns of expansion. Following this, the comparison between the actual built-up area in 2023 (21.5409 km²) and the predicted value (21.5442 km²) shows a high level of consistency (Figure 11a,b).

The transition probability matrix for 2003–2013 quantifies the likelihood of land use/land cover (LULC) changes among agriculture, bare land, built-up, and vegetation classes, serving as a foundation for future projections using the MLP–Markov Chain model (

Table 5. Transition probability matrix for 2003–2013.

Given	Probability of Change
	Agriculture	Bare Land	Built-Up	Vegetation
Agriculture	0.312	0.0515	0.581	0.0555
Bare land	0.0481	0.4151	0.4844	0.0524
Built-up	0.0328	0.0153	0.9044	0.0475
Vegetation	0.0791	0.2633	0.2547	0.4029

). This model assumes that future land cover states depend solely on the current state and transition probabilities. The diagonal values represent the probability of persistence, while off-diagonal values indicate transitions to other classes. Agriculture exhibits a moderate persistence probability of 31.2% but a high likelihood of 58.1% converting to built-up areas, indicating significant urban encroachment into farmland. Bare land has a 41.51% chance of remaining unchanged, with a nearly equal probability of 48.44% transitioning into built-up areas, suggesting that bare land is a prime target for development. Built-up areas are the most stable category, with a very high persistence rate of 90.44% and minimal likelihood of reverting to other classes, reflecting the permanence of urban infrastructure once established. Vegetation has a 40.29% probability of persistence but faces notable transitions to bare land (26.33%) and built-up (25.47%), highlighting pressures from both land degradation and urban expansion.

Following the validation and acceptance of the model, the Markov chain model was used to predict the LULC for 2023, 2043, and 2063 using the same reclass and filter file assumptions.

The Markov model was applied to determine the transition area and potential matrices for the periods 2023–2043 and 2043–2063. According to the transition probability matrix, classes such as agriculture, vegetation, and bare land are transitory categories that are subject to more changes over time, primarily shifting to built-up areas. The suitability file for 2023 was then created to train the model and to understand the changes from 1993, 2003, and 2013. Finally, the Markov Chain model was run with 2023 LULC as the base image to forecast the next 20 and 40 years (Figure 12 and Figure 13).

The predicted LULC changes in Dire Dawa City from 2023 to 2063 indicate significant urban expansion. In 2023, agriculture was the dominant land use, followed by built-up areas, vegetation, and bare land. By 2043, agricultural and vegetative lands are expected to decline, while built-up areas are expected to increase moderately. By 2063, built-up areas will have nearly doubled, reaching 38.98.44 km², primarily at the expense of agriculture and vegetation. Bare land is expected to remain relatively stable throughout the period (

Table 6. Areal extent of simulated LULC scenarios for Dire Dawa in 2023, 2043, and 2063.

LULC Class	Predicted for 2023 (km2)	Predicted for 2043 (km2)	Predicted for 2063 (km2)
Agriculture	29.333	25.078	21.899
Bare Land	11.492	9.085	6.863
Buit-up	21.544	30.385	38.854
Vegetation	14.683	12.504	9.436

).

5. Discussion

This study reveals that the expansion of Dire Dawa City along the Ethio–Djibouti corridor has significantly intensified between 1993 and 2023. During this period, the built-up area increased from 6.21 km² in 1993 to 21.54 km² in 2023, while agricultural and bare land declined from 41.62 km² to 29.33 km² and 15.20 km² to 11.49 km², respectively. The observed trends and patterns are consistent with urbanization processes across many parts of sub-Saharan Africa, including Ethiopia, underscoring the urgency of addressing the challenges associated with such expansion [51,52,53,54]. The conversion of agricultural land into built-up areas has serious implications for local ecosystems, livelihoods, and food security in surrounding communities.

The findings align with previous studies in other Ethiopian cities. For instance, Ref. [10] reported a 320% increase in Addis Ababa’s built-up area alongside an 82.1% reduction in agricultural land between 1990 and 2023. Similarly, the built-up areas of Hawassa and Bahir Dar increased by 284% and 148%, respectively, between 2000 and 2015. According to [55], the built-up area of Addis Ababa and its surrounding towns is expanding into the peri-urban region, leading to high losses of farmland. Other researchers [56,57,58] have explained that the urban expansion was at the expense of agricultural land.

In contrast to some prior studies, this research found minimal change in the extent of urban green spaces or vegetation cover in Dire Dawa during the study period. Earlier studies have shown that urban expansion can negatively affect vegetation through land conversion [59,60]. However, in this study, vegetation cover increased from 14.02 km² to 14.68 km², which had positive effects by improving green space management through activities such as irrigation and fertilization [38,40]. Notably, Ref. [10] reported a 60% reduction in urban green space in Addis Ababa over three decades (1990–2024), largely due to extensive construction in peripheral areas.

The spatial growth pattern of Dire Dawa is predominantly characterized by edge expansion, which accounts for approximately 95% of urban growth during the study period. This pattern, which involves the outward growth of built-up areas along the edges of existing urban patches, signifies low-density sprawl and unsustainable development. Similar patterns are also observed in other Ethiopian cities like Addis Ababa, Adama, and Hawassa [10,61]. This form of uncontrolled expansion, while driven by economic growth, poses significant threats to peri-urban environments and communities due to its strong association with the loss of agricultural land and vegetation [62,63]. During the study period, edge expansion consumed a total of 21.08 km², outlying growth 0.12 km², and infilling 0.34 km², together accounting for the built-up area in 2023 (21.54 km²), much of which was previously used for agriculture and essential to local livelihoods.

The study further highlights the directional and spatial heterogeneity of the city’s expansion. Sector and concentric circle analyses reveal that the city has mainly expanded westward and west-northwest. This unidirectional urban growth presents major challenges for effective resource allocation, infrastructure development, service provision, and social equity. These results align with previous studies in Ethiopia. For example, Ref. [10] identified southeast, south, and south-southwest as the main expansion directions in Addis Ababa, often accompanied by increasing irregularity. Similarly, Ref. [21] noted varying sprawl rates in Addis Ababa, with the most significant expansion occurring along the Mojo (south) and Jimma (southwest) corridors, leading to unequal service access. Ref. [61] also reported that Hawassa has expanded predominantly toward the northeast, east, and southeast, exerting pressure on local resources and service systems.

The study further reveals that rapid urban expansion in Dire Dawa is primarily driven by human factors, particularly economic development and population growth, rather than physical constraints. Due to its strategic location along the Ethio–Djibouti trade corridor, Dire Dawa has become a national hub for development projects, including the establishment of manufacturing industries, highways, and railway infrastructure. This has made the city a focal point for internal migration and has intensified population growth. Previous studies have also shown that state-led economic initiatives often act as a catalyst for urban expansion by triggering other proximate causes [64,65,66].

The spatial pattern of urban growth observed in this study closely follows major transportation infrastructures such as highways and railways, as well as the locations of key public institutions and industrial facilities like the Dire Dawa Cement Factory. In contrast, physical factors such as topography appear to play a minimal role in shaping the direction of urban expansion. Notably, although the southern, southeastern, and southwestern parts of the city are characterized by steep slopes and highlands, urban growth in these areas has remained minimal. This finding contrasts with studies from other Ethiopian cities where topography significantly influenced urban development. For instance, Ref. [61] highlighted that the Entoto Mountains in northern Addis Ababa constrained the city’s growth in that direction. Likewise, Hawassa’s expansion was limited to the northeast, east, and southeast due to the presence of Lake Hawassa and hilly terrain in the west and south, respectively. The findings of this study suggest a complex interplay between socio-economic, infrastructural, and physical-geographic variables in shaping the city’s future urban form.

6. Conclusions

The Ethio–Djibouti trade corridor has emerged as a major hotspot for urban expansion in eastern Ethiopia over the past decade. This research integrated the spatiotemporal dynamics, driving factors, and projected trends of urban growth in Dire Dawa City from 1993 to 2023, with forecasts extending to 2043 and 2063. The findings revealed major land use and land cover (LULC) transformations, most notably an increase in built-up areas from 6.21 km² in 1993 to 21.54 km² in 2023. This expansion occurred largely at the expense of agricultural land, which declined from 54.02% to 38.07%, alongside reductions in bare land and vegetation. Spatial analysis showed that urban growth was primarily concentrated toward the west, northwest, and west-northwest corridors, strongly influenced by proximity to transportation networks, industrial zones, and areas with low elevation and gentle slopes. The Landscape Expansion Index identified edge expansion as the dominant growth typology across the study period, particularly between 2013 and 2023. Model projections further indicate continued and substantial growth in built-up areas by 2043 and 2063, accompanied by additional declines in agricultural and vegetative lands if current trends persist. Overall, the study highlights that both historical and future LULC trends, particularly the expansion of built-up areas and the loss of agricultural land, pose significant challenges for natural and human systems. These include threats to food security, resource-based conflicts, the intensification of urban heat islands, and broader environmental stresses. Consequently, there is an urgent need for proactive, integrated urban planning and management strategies, supported by appropriate policy measures, to balance urban development with the conservation of vital land resources.