Quantification of land use/land cover dynamics and urban growth in rapidly urbanized countries: The case Hawassa city, Ethiopia

ABSTRACT

Land Use and Land Cover Changes (LULC) are some of the worldwide factors that have the most impact on city growth. The aim of this study could be a quantification of the land use/land cover dynamic over three decades in Hawassa City, Ethiopia. This study used multi-spectral satellite images of the years 1990 and 2020. The six category classifications within the study area were agriculture, bare land, built-up, bushland, forest, and water body. The data for the study was acquired from a satellite image of Landsat 5TM 1990, 2000, and 2010 and Landsat 8 OLI and SWIR 2020. Packages like QGIS version 3.2, ArcGIS 10.3, ENVI 4.2, and ERDAS Imagine 2013 were accustomed to performing image classification. From this finding, it is clear that the study space is underneath serious land use conversion of bare land to built-up space and bush land to agricultural. The general LULCC between the years 1990 and 2020, satellite results show that between 1990 and 2020, the quantify of built-up land increased from 12.76% to 16.85% of the study area as bare land and bush decreased from 4.77% to 2.48% and 14.29% to 10.16%, respectively, from the entire study..

KEYWORDS:

• Change detection
• Geographic information system
• land use/land cover
• satellite image
• Thematic Mapper

1. Introduction

Land is an important natural resource which has numerous social, economic, and biophysical uses. It used to create wealth and employment, grow economies and also use as a source of water, food and energy. It provides services such as conserving biodiversity, storing carbon, purifying and storing water and regulating the Earth’s climate by absorbing the heat from the sun as well. Land use and land cover change are the two interrelated ways of observing Earth's surface. The former represents the manner human population manipulate the biophysical attributes of the land and the purpose for which land is used, while the latter represents biophysical state of the earth’s surface and immediate subsurface. LULC is perhaps the most prominent form of global environmental change and it occurs at spatial and temporal scales (Molla, 2014).

Land Use Land Cover (LULC) changes are parts of world environmental change and have an effect on ecosystem processes and services. For instance, the growing demand for agricultural, industrial, or urban areas compromises the ability of natural forests, water bodies, and grasslands to care for mankind (Goldman-Benner et al., 2013). However, in recent years, a large number of changes in LULC have been found which were caused by different socioeconomic and biophysical drivers, such as population growth, agricultural expansion, and intensive biophysical drivers, such as population growth, agricultural expansion, and intensification, accessibility to infrastructure/markets and water availability or climate (Shiferaw et al., 2019).

Land use changes area unit advanced methods that arise from modifications within the land cover to land conversion process (Aboud, 2002). Despite this complexity, very little is understood concerning how human and environmental factors operate and the way they move to have an effect on land use patterns (U.S EPA, 2009). Consistent with Lambin (2005), land use modification is driven by the interaction in house and time between biophysical and human dimensions. There are potential impacts on physical and social dimensions. Consistent with Gereta et al. (2001), throughout the whole history of human beings, intense human utilization of natural resources has resulted in vital changes in land use and land change (LULC). Since the time of industrial enterprise and fast growth, land use modification phenomena have powerfully accelerated in several regions. Land use changes in area units are often indicated to be one of the most human iatrogenic factors influencing the hydrological system (Akotsi et al., 2006). Agriculture has swollen into forests, savannas, and steppes altogether elements of the globe to satisfy the demand for food. Agricultural enlargement has shifted between regions over time; this followed the final development of civilizations, economics, and increasing population (FAO, 2000).

Although the urbanization method usually suggests accelerated economic performance for a country, the attendant rise in costs of urban land and its conversion from one type to a different affects the natural and cultural resources of the town. The inflow of individuals into the cities difficult the urban conditions difficult through structural growth (Redman, 1999; Xiao et al., 2006). Lateral changes occur once the city expands in geographic boundaries resulting in sprawl and peripheral developments, whereas structural growth relates to the extension in land use density inside urban centers. With a rise in urban population and demand for land use, urban growth will occur. Also, as a lot of land is being regenerated, thanks to rising population and land costs, tenacity areas step by step become subjected to intensive use and so become high-density or medium-density use (Redman, 1999; Xiao et al., 2006).

Globally, cities are urbanizing at unprecedented rates, resulting in profound and unintended impacts on landscapes and urban hydrology (Seto et al., 2011). In 2014, more than half (54%) of the world’s population lived in cities, and by 2050, the urban population is projected to rise by 2.5 billion people, with 90% of this population anticipated to be concentrated in Asia and Africa, where 60% of this population is anticipated to live in slums (Miller et al., 2014). There is now a wide consensus that cities in Asia and Africa are more vulnerable to the effects of flooding due to increased urbanization, climate change, poor planning, weak regulations, and poor adaptive capacity (Li et al., 2014). As a consequence, there is a critical need to understand the historical patterns of land-use change, the extent of future urbanization, and their resulting impacts on urban landscapes in these regions to inform flood risk management strategies.

Urban growth within the case of Hawassa city is not an associate degree exception to different Ethiopian cities but exhibits a peculiar pattern and complexity due to its geographical setting. To an outsized extent, the city represents most of the characteristics, causes, and effects mentioned on top of. Identification of patterns of sprawl and analyses of abstraction and temporal changes would greatly facilitate vastly within the coming up of infrastructure facilities. The current rate of increase and projected figures also signify the prime responsibility of the planning department to plan for a future development that is engaging, affordable, and property socially, economically, culturally, and environmentally (Marshall et al., 2009).

The main objective of this study was quantification of the LULC dynamic over three decades of Hawassa space by group action geographic information system (GIS) and remote sensing techniques. Our analysis provides an important contribution to understanding the LULC maps of the study area for three decades, evaluating judge accuracy, looking at changes in land use, landscape metrics, and the tremendously rapid urbanization within the study area. The results of this study may offer a reliable guide for the design department of Hawassa city Government and property development of the study area; analysis may be a reliable guide for decision-makers and city planners towards the advancement of the standard of life in Hawassa city.

2. Study area

2.1. Description of the study area

This study was conducted in Hawassa City, Ethiopia. Hawassa city is the capital of Sidama Regional State and Southern Nations, Nationalities, and Peoples Regional State (SNNRPS). Moreover, it is the executive center of the Sidama Regional State. It was shaped on 18 June 2020. It is a city Administration standing with eight sub-cities, 21 urban and 12 Rural Kebeles. It is set on the international road that connects Addis Ababa with the national capital at a distance of 275 km south of Addis Ababa. It lies on a comparatively flat plain within the vale topographical region having a mean elevation of around 1690-m higher than the water level with approximate geographical coordinates of the town is found between Latitude: 7°03’43.38“North and Longitude: 38° 28’ 34.86” East (see Figure 1). The area that corresponds to the watershed is encircled by steep escarpments; however, the area itself the landscape is either flat or gently undulating (more than 50% of the slopes are flat to gentle (0–8%) with an extra 33% moderately sloping (8–30%) and solely 5% steep to terribly steep (>30%) (Hawassa city Municipality, 2020).

Figure 1. Location of study area. Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

2.2. Climate

The average mean monthly rainfall in Hawassa is 89.60 mm whereas the annual mean rainfall is 1075 mm. (maximum mean rainfall occurs in the month of April). The amount of rainfall decreases from west to east. On the basis of the rainfall coefficient values, months in the water year are classified as Dry months (January, February, November, and December), distinctly rainy months (March and October), and big rainy months (April, May, June, July, August, and September). The area receives adequate sunshine hours of 100–200 hours per month from March to October and 200–300 hours every month in the dry season.

2.3. Data sources

To analyze 30-year LULC four Landsat satellite imagery of 1990, 2000, 2010, and 2020 were downloaded from the United States Geological Survey (USGS) official website (earthexplorer.usgs.gov). All the satellite imaging acquired represents the season and Landsat 5 TM Landsat8 and OLI sensing element information were taken into account. The LULC mapping for the realm was supported by Landsat 5 Thematic mapper (TM) of December 1990, January 2000, and November 2010, Landsat 8 Operational Land Imager (OLI), and short wave Infrared (SWIR) 2020 (Campbell & Wynne, 2011).

3. Materials and methods

3.1. Land use land cover classifications

Landsat images from four different years (1) Thematic Mapper of 1990, (2) Enhanced Thematic Mapper Plus of 2000 and 2010, (3) Operational Land Imager or Thermal Infrared Sensor of 2020 were used for LULC classifications. These images were classified by using supervised classification with a maximum likelihood algorithm. The LULC types of the study area were classified into a built-up area, agriculture, bare land, built-up, forest, and water body.

All the images were taken with little cloud cover in order to analyze LULC change and urban expansion. Besides satellite images, field observation and Google Earth were used for generating ground truth points (training sites) for LULC mapping. The source and description of data used in this study are presented in Table 1.

Table 1. Details of acquired satellite imageries.

Satellite	Sensor	Spatial resolution	Path/Row	Acquisition Date
Landsat 5	TM	30m*30m	168/055	18, 12, 1990, 28, 012000 & 07,11,2010
Landsat 8	OLI & SWIR	30m*30m	168/055	19, 01, 2020

Source: [60] US Geological Survey (USGS) (http://glovis.usgs.gov).

OLI= Operational Land Imagery SWIR = Short Wave Infrared.

The LULC data are obtained from the multi-band formation imageries through the method of image interpretation and classification (Li et al., 2014). Image classification (supervised or unsupervised) is meant to Associate for automatic categorization of pixels with a standard coefficient range into specific LULC classes (Chica-Olmo & Abarca, 2000; Lille & Kiefer, 1994). Supervised classification could be a user-guided approach that involves the choice of coaching sites as a reference for the categorization (Campbell, 1996; Lille & Kiefer, 1994). There are several ways obtainable that are being employed to implement the supervised classification like, parallelepiped classification, K-nearest neighbor, minimum distance classification, and so on (Zhu et al., 2006). In the present study, we adopted a commonly used maximum likelihood classifier (Platt & Goetz, 2004) for LULC classification using Arc GIS 10.3 and ERDAS IMAGINE 2013 software.

3.2. Accuracy assessment

The reference data used for accuracy assessment were obtained from GPS points during fieldwork and the original mosaic image. The GPS points used in the classification accuracy assessment were independent of the ground truths used in the classification. Based on the error matrix, overall accuracy and kappa statistics were used to illustrate the classification accuracy. Therefore, an overall accuracy of 86%, 87% and 89% was achieved for the Land sat TM of 1984. Land sat ETM+ of 2001 and Land sat ETM 2018, respectively. These imply excellent classifications of Land sat images.

3.3. Land use land cover change detection

One of the most powerful advantages of remote sensing images is their ability to capture and preserve a record of conditions at different points in time, to enable the identification and characterization of changes over time. Therefore, after doing supervised classification, change detection was done to see which land use was changed to which one and in what amount or percentage. These help to decide whether the change is positive or negative, and the amount and rate of change from time to time during the selected study period. Finally, the table matrix was also generated which holds overall information about the change matrix between study periods from 19,990 to 2020. Total area (TA), changed area (CA), change extent (CE), and annual rate of change (CR) variables were used to determine the magnitudes of change in terms of LULC. The variables were computed as equation 1 (Daniel et al., 2012).

$\mathrm{C}\mathrm{A}=\mathrm{T}\mathrm{A}\left({\mathrm{t}}_2\right)-\mathrm{T}\mathrm{A}\left({\mathrm{t}}_1\right)$

(1) $\mathrm{C}\mathrm{E}=\left(\mathrm{C}\mathrm{A}/\mathrm{T}\mathrm{A}\left({\mathrm{t}}_1\right)\right){ }^{\ast }100$(1)

$\mathrm{C}\mathrm{R}=\mathrm{C}\mathrm{E}/\left({\mathrm{t}}_2--{\mathrm{t}}_1\right)$

3.4. Evaluating classification error matrices

First, we can begin to appreciate the need to consider overall, producer’s, and user’s accuracies simultaneously. In this example, the overall accuracy of the classification is 65%. This would potentially lead one to the conclusion that although the overall accuracy of the classification was poor (65%), it is adequate for the purpose of mapping class. In fact, if the number of classes is small, such a random assignment could result in a surprisingly good apparent classification result: a two-category classification could be expected to be 50% correct solely due to random chance. The k ̂ ‘kappa’ or ‘KHAT’ statistic is a measure of the difference between the actual agreement between reference data and an automated classifier and the chance agreement between the reference data and a random classifier conceptually, and k ̂ can also be defined as the statistic that serves as an indicator of the extent to which the percentage correct values of an error matrix are due to ‘true’ agreement versus ‘chance’ agreement. The k ̂statistic is computed as equation 2 as follows:

(2) $\hat{k}\frac{N{\sum }_{i=1}^r{x}_{ii-{\sum }_{i=1}^r\left(x_{i+}.x_{+i}\right)}}{N^2-{\sum }_{i=1}^r\left(x_{i+}.x_{+i}\right)}$(2)

where

r= number of rows in the error matrix

x_ii= number of observations in row i and column i (on the major diagonal)

x_i+= total of observations in row i (shown as marginal total to right of the matrix)

x_{+i= total} of observations in column i (shown as marginal total at bottom of the matrix)

N= total number of observations included in the matrix. Equation 3(3) $k1=\frac{UB-UA}{UA}\frac{1}{T}\times \hspace{0.17em}100\mathrm{\%}$(3) has been computed.

(3) $k1=\frac{UB-UA}{UA}\frac{1}{T}\times \hspace{0.17em}100\mathrm{\%}$(3)

Where K1 is the land use dynamics index, measuring the rate of target LULCC t; UA and UB are the areas of the target LULCC type at the beginning and end of the study period, respectively; and T is the length of the study period. K1 can concisely express the overall characteristics of certain during LULCCs a study period but usually misses some information on the spatial process of the LULCC. Thus, another index for the rate of change of a single LULCC type, K2, was proposed as shown in equation 4 (Peng 2008).

(4) $K=\frac{\mathrm{\Delta }Uin-\mathrm{\Delta }Uout}{Ua}\times \frac{1}{T}\times 100\mathrm{\%}$(4)

Where K2 is a land use dynamics index measuring the rate of change of the target LULCC type; U in and U out are the areas of the target LULCC type conversion from or to other LULCC types, respectively; and T is the length of the study period. When focusing on the process of LULCC, K2 reflects the area ratio of the conversion from and to the target LULCC type. In general, K2 measures the total ‘loss or gain’ conversion of the target LULCC type, while K1 reflects the algebraic summation of conversion loss and gain of the target LULCC type. The absolute value of K1 expresses the relative difference between conversion loss and gain, with a positive sign representing the dominance of conversion gain from other LULCC types to the target LULCC type and a negative sign representing the dominance of conversion loss (Pieterse, 2008).

Various types of spectral indices extract different types of LULC. The major LULC types in the urban land are green vegetation, water, wetland, barren land, and settlement. Thus, MNDWI (Xu 2006), NDBI (Zhao & Chen, 2005), NDVI (Tucker, 1979), and NDBI (Zha et al., 2003) were the chosen spectral indices as these can extract water body, bare land settlement, and vegetation by applying their numerical thresholds (Chen et al., 2006). Based on the analysis of the unique spectral responses of built-up areas and other land covers in seven Landsat TM bands, the original NDBI approach developed by Zha et al. (2003)

The Normalize Difference Build-up Index value lies between −1 to +1. Negative value of NDBI represents water bodies, whereas higher value represents build-up areas. The NDBI value for vegetation is low. The built-up areas were extracted by the following equation 5

(5) $\mathrm{B}\mathrm{U}=\mathrm{N}\mathrm{D}\mathrm{B}\mathrm{I}-\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}$(5)

Where BU b is the resultant binary image with only the built-up and barren pixels having positive value, thus allowing built-up areas to be mapped automatically (Zhaet al. 2003).

The Normalized Difference Vegetation Index (NDVI) is the most commonly used vegetation index for observing greenery globally. Other commonly used vegetation indices Enhanced Vegetation Index (EVI), Perpendicular Vegetation Index (PVI), Ration Vegetation Index (RVI) equation 7.

(6) $\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}=\left(\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R}\mathrm{e}\mathrm{d}\right)/\left(\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R}\mathrm{e}\mathrm{d}\right)$(6)

3.5. Data preprocessing

Image preprocessing could embody the detection and restoration of bad lines, geometric rectification or image registration, radiometric standardization, atmospheric correction, and topographic correction. Correct geometric rectification or image registration of remotely perceived information could be a necessity for a mix of various supply information in an exceeding classification method. Several textbooks and articles have delineated this subject thoroughly (Jensen, 1996; Toutin, 2004).

Once multi-temporal or multi-sensing element information is used, atmospheric standardization is necessary. This can be very true once multi-sensing element information, like Landsat TM and SPOT, and measuring system information arentegrated for image classification. A range of strategies, starting from easy relative standardization and dark object subtraction to standardization approaches supported by advanced models, are developed for radiometric and atmospheric standardization and correction (Chavez, 1996; Gilabert et al., 1994; Hadjimitsis et al., 2004; Heo & FitzHugh, 2000; Markham & Barker, 1987; Song et al., 2001, McGovern et al. 2004; Stefan & Itten, 1997; Tokola et al., 1999; Vermote et al., 1997). Geographic correction is another necessary side if the study space is found in rugged or mountainous regions (Civco, 1989; Colby, 1991; Gu & Gillespie, 1998; Hale & Rock, 2003; Meyer et al., 1993; Rannikko, 1999; Teillet et al., 1982).

The Maximum likelihood classified to categorized the Landsat 5 TM 1990, 2000 and 2010, Landsat 8 OLI and SWIR 2020 has produced map showing the distributions of six prevalent LULC classification. The ArcGIS software and spatial analyst tools and demarcated the boundary of the study area. The detailed methodology adopted is given in Figure 2.

Figure 2. Data processing flow chart. Source: Author’s construct, 2020.

4. Results and discussion

4.1. Land use/land cover change of Hawassa city

In this study, LULC types of study areas were classified from Landsat images of 1990, 2000, 2010, and 2020. The LULC classes of the study period (1990, 2003, and 2020) were classified into agriculture, bare land, built-up area, and bushland, forest, and water body (Figure 3). Water body and agricultural land dominated LULC classes in the study area but the agricultural land declined during the last decade of 2010 to 2020 by 1.35% this is due to an increment of built-up area in 1.91% in the same period.

Figure 3. Classified LULC maps of Hawassa city for years 1990, 2000, 2010 and 2020.

Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

4.2. Trend of land use/land cover change of Hawassa city

The land cover classes in Hawassa City from 1990 to 2020 are shown in Table 2 and Figure 3. The supervised classification result revealed the same land use type; however, the degree of coverage varied depending on the kind of cover. Other contributing aspects to the LULCC include built-up regions with the spatiotemporal dynamics of urban expansion, land use change processes linked to population increase and national economic transformation programs.

Table 2. Area statistics of LULC of years 1990, 2000, 2010 and 2020.

		Year 1990		Year 2000		Year 2010		Year 2020
S.N	LULC Class	Area (km^2)	%	Area (km^2)	%	Area (km^2)	%	Area (km^2)	%
1	Agriculture	69.12	26.51	73.00	28.00	78.50	30.11	74.97	28.76
2	Bare land	12.44	4.77	9.68	3.71	8.17	3.13	6.47	2.48
3	Built up	33.26	12.76	37.15	14.25	38.96	14.94	43.93	16.85
4	Bush land	37.26	14.29	28.90	11.08	27.44	10.53	26.50	10.16
5	Forest land	21.89	8.40	21.63	8.30	20.20	7.75	17.52	6.72
6	Water body	86.72	33.26	90.34	34.65	87.42	33.54	91.31	35.03
Total		260.69	100	260.69	100.00	260.69	100	260.69	100
1990–2000		2000–2010		2010–2020
−3.88		−5.5		3.53
2.76		1.51		1.7
−3.89		−1.81		−4.97
8.36		1.46		0.94
0.26		1.43		2.68
−3.62		2.92		−3.89

Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

Agriculture expanded from 26.51% to 28% of the total area between 1990 and 2000, according to the highest incremental LULC classes in the study area, whereas built-up areas increased from 12.76% to 14.25%. In general, throughout this time, the total amount of bushland, bare land, and forest decreased substantially by 3.21, 1.06, and 0.10, respectively.

According to the data from Indicated Landsat TM, the built-up area increased at the greatest rate in the study region from 2010 to 2020, at 1.91%. The structure grew mostly in the northeasterly to southeasterly direction at a pace of 0.136%. Per annual rate of change. Additionally, the percentage of bare land converted to agriculture increased from 1990 to 2010, increasing by 26.5%, 28%, and 30.11%, respectively. However, the area of agricultural land from 2010 to 2020 decreased by 78.50 km²2 (30.11%) to 74.97 km² (28.76%) with a 10-year rate of 1.35% due to a number of mega projects that increased built-up that was established in the study area. This implies that built-up areas have tremendously increased the expense of agriculture and bare land in the study area. The shrinking of both agricultural and bare land is mainly due to urban expansion to surrounding rural communities and the related sale of land for residence, commerce, administrative institutions, services, roads, and other urban functions.

Agriculture has the highest area share in the study region, accounting for 73 km² (28%) of the total area, while bare land has the lowest proportion at 9.68 km² (3.71%), according to the LULC data for the year 2000.    Forest land made up 21.63 km2 (8.3%) of the area, built-up land made up 37.15 km2 (14.25%), and water bodies made up the remaining land. This indicates that barren areas made up the least amount of land in the city and that agricultural land had the largest land share.

The largest area coverage in the Land use and the land cover of 2010 in the study area shows agriculture, which contains 78.50 km² (30.11%) of the total area and the least land class area was bare land which contains 8.17 km² (3.13%) of the total area, built-up 38.96 km² (14.94%) and bushland 27.44 km² (10.53%) from the total area of the land classes. This implies that there are increments of the agriculture and built-up area, and bare land area decrease because of the growth of population in the city after the year 2000 up to 2010.

Table 2 indicates the land category within the years 1990 and 2020 study area. Constant land use type was shown from the result of supervised classification, but the amount of coverage was different for different cover types. The built-up area, which totaled 33.26 km² in 1990, increased to 37.15 km², 38.96 km² and 43.93 km² in the years 2000, 2010, and 2020, respectively. Despite fluctuation water body during different years, has gained an overall increase of 30 years of periods, water occupied 86.72 km² in 1990 it increased to 90.34 km² in 2000, and decreased to 87.42 km² in 2010, but the coverage increase to 91.31 km² in 2020. The water body of the land cover category stays nearly constant study amount.

4.3. Rate of land use/land cover changes of the Hawassa City from 1990 to 2020

During 1990–2000, the bush land was decreased by 10.56 km², bare land 5.97 km² and forest was declined by 4.37 km². Agriculture and built-up area were increased by 5.85 km² and 10.67 km², respectively. Over the study period 1990–2020, bush land, bare land, and forest cover classes were decreased (see Figure 4).

Figure 4. Net gain and loss of each LULC classes from other land uses for the period of 1990 and 2020. Source: [60] US Geological Survey (USGS) ().

They have emphasized that since all of the newly incorporated rural Kebeles of land were taken from the periphery, a significant number of the newly added plots of land are planned to be used for residence, followed by special function and urban agriculture. Furthermore, a substantial portion of this area is expected to be used for the growth of public service sectors, including roads, schools, health facilities, as well as administrative centers like Kebeles and sub-city centers. In relation to how the land was previously used before the insertion of additional territory, the municipality and city administration are very comparable. However, as a result of the growth in municipal government, the city has changed how it uses its territory.

4.4. Accuracy assessment

For this study, accuracy assessment of LULC for 1990, 2000, 2010, and 2020 years produced the overall classification accuracy of 88.41%, 93.48%, 88.41% and 89.86%, respectively. The overall LULC classification Kappa statistics for the study periods were 0.8609, 0.9217, 0.8609, and 0.8783, respectively (see Table 3).

Table 3. Accuracy assessment of classified LULC maps of years 1990, 2000, 2010 & 2020.

Years	Land use/cover Class	Reference Totals	Classified Totals	Number Correct	Producers Accuracy	Users Accuracy
	Agriculture	25	23	21	84.00%	91.30%
	Bush land	22	23	21	95.45%	91.30%
	Bare land	22	23	18	81.82%	78.26%
1990	Built-up	19	23	19	100.00%	82.61%
	Forest	28	23	23	82.14%	100.00%
	Water	22	23	20	90.91%	86.96%
	Totals	138	138	122
Overall Classification Accuracy =88.41% Overall Kappa Statistics = 0.8609
	Agriculture	25	23	23	92.00%	100.00%
	Bush land	25	23	23	92.00%	100.00%
	Bare land	22	23	19	86.36%	82.61%
2000	Built-up	21	23	21	100.00%	91.30%
	Forest	24	23	23	95.83%	100.00%
	Water	21	23	20	95.24%	86.96%
	Totals	138	138	129
Overall Classification Accuracy =93.48% Overall Kappa Statistics = 0.9217
	Agriculture	22	23	20	90.91%	86.96%
	Bush land	22	23	21	95.45%	91.30%
	Bare land	25	23	18	72.00%	78.26%
2010	Built-up	20	23	19	95.00%	82.61%
	Forest	26	23	23	88.46%	100.00%
	Water	23	23	21	91.30%	91.30%
	Totals	138	138	122
Overall Classification Accuracy =88.41% Overall Kappa Statistics = 0.8609
	Agriculture	21	23	20	95.24%	86.96%
	Bush land	23	23	21	91.30%	91.30%
	Bare land	26	23	21	80.77%	91.30%
2020	Built-up	24	23	21	87.50%	91.30%
	Forest	24	23	21	87.50%	91.30%
	Water	20	23	20	100.00%	86.96%
	Totals	138	138	124
Overall Classification Accuracy = 89.86% Overall Kappa Statistics = 0.8783

Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

The accuracy assessments were performed for classified images of 2020, a minimum of 23 random points were generated per class using a stratified random sampling approach for efficient accuracy assessment. The corresponding reference category for every

The growth of the built-up area is mostly responsible for the change in the LULC in the study region, according to Table 3 conversion matrix for the years 1990 to 2020. Over the past 30 years, LULCC trends have indicated a tendency for more land to be developed. These provided statistics explicitly say that the conversion of bare land and bushland to urban built-up areas as a result of population growth mostly leads to the destruction of trees and other natural vegetation.

4.5. Normalized Difference Built-up Index (NDBI)

Built-up Index (BU). Build-up Index is the index for analysis of urban pattern using NDBI and NDVI. Built-up index is the binary image with only higher positive value indicates built-up and barren thus, allows BU to map the built-up area automatically.

BU = NDBI – NDVI.

Image classification technique (supervised classification and unsupervised classification) is lengthy and complex process. It requires competitive band & apply numbers of operation for the final result. The accuracy derived from image classification technique depends on the image analyst & method followed by analyst. However, NDBI calculation is simple and easy to derived. NDVI can be calculated by following formula.

NDBI = (SWIR – NIR)/(SWIR + NIR).For Landsat 7 data, NDBI = (Band 5 – Band 4)/(Band 5 + Band 4). For Landsat 8 data, NDBI = (Band 6 – Band 5)/(Band 6 + Band 5). Also, the Normalize Difference Build-up Index value lies between −1 to + 1. Negative value of NDBI represent water bodies where as higher value represent build-up areas. NDBI value for vegetation is low. Figures 5 Show the spatial distribution (1990–2020) of NDBI values is higher in the inner and outer peripheries of the core city. However, in 2020, the central part also shows a high NDBI value. Dry season images show higher NDBI than in the summer. From 1990 to 2020, the NDBI values increase rapidly. Except for the central part of the study area, every corner witnesses a high NDBI value. This is due to higher vegetation and water coverage in the middle part of the core city. A very high NDBI value is observed along the outskirts of the city where the soil is dry and the land is comparatively bare.

Figure 5. Spatial distribution of _{NDBI 1990–2020.} Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

From 2000 to 2020 the city grew at a speed of 1 km² per year (from 12.2 km² to 19 km² of built-up land). From 2010 to 2020 the city grew at a speed of 4.3 km² per year (from 19 km² to 44.9 km² of built-up land). Addis Abeba, which is the capital city of almost 4 million people and with a completely different scale compared to Hawassa, is growing at the speed of 4 km² per year. Meanwhile, the agricultural land within the administrative boundary of the city is growing as well (from 90 km² in 2000 to 104 km² in 2020). Thus, the sum of forests, bare land, forest, and key ecosystems for the proper functioning of the environment and human well-being decreased from 71 km² in 2000 to 64 km² in 2010, to 19 km² in 2020. This value is the leftovers of the administrative boundary of the city (175 km²), taken off the built-up areas and agricultural land the natural capital is a necessary and finite capital of the city. The decrease of forest within the city and at the city boundary is not only putting at risk the availability of key resources but also creating costs or disadvantages that can be quantified. For example, 14% forest coverage within the city improves the air quality by up to 4% during the year and up to 10% during traffic hours (Escobedo et al., 2011). A garden or a park with 60% tree coverage can lower the temperature by up to 4°C (Bowler et al., 2010). Avoiding urban sprawl and enabling urban agriculture can limit the transportation of vegetables from the countryside to the point that 50 km² of urban agriculture can reduce CO₂ emission of 100.000 tons per year (Gwan Gyu et al., 2015).

The existing built-up and future expansion directions, major land use categories, and other major elements. In addition, the findings and recommendations of the research have helped develop various scenarios that can be considered. Two scenarios were developed for the conceptual plan: Expansion within the Planning Boundary (Compactness within the Planning Boundary) and Expansion Beyond the Planning Boundary (Compactness and Sustainable Provision of Expansion beyond the Planning Boundary). Based on the findings and justifications, the different scenario approaches are in terms of the direction of growth and expansion of the city, among other major elements.

Scenario One: Expansion within the Planning Boundary (Compactness within the Planning Boundary). UN-Habitat recommends the adoption of the compact city model to address the current challenges the city of Hawassa is confronting, in particular, the housing challenge. Compacting the city encourages the maximization of the urban infrastructure and service, social integration, environmental protection, and economic efficiency for sustainable urban development. The scenario below reflects on the compact city approach and aligns with the goals of the Hawassa City Administration.

Scenario Two: Expansion beyond the planning boundary (Compactness and Sustainable Provision of Expansion beyond the Planning Boundary). This scenario also builds on UN-Habitat’s recommendation of the adoption of the compact city model, however, from the findings; the projected population growth of the city of Hawassa has increased the rate of urban growth and has placed pressure on the demand for affordable houses. If the Hawassa City Administration does not provide a strategy for the management of rapid urbanization over two decades, the informal occupation of land will only increase uncontrollably. Therefore, the second scenario proposes the expansion growth and direction the city must consider for the long term, to accommodate the projected challenge of responding to urbanization.

According to Figure 6, conversion of the matrix for the years 1990–2020, the change in the LULC in the study area was largely attributed to the expansion of the built-up area. In the last 20 years, LULCC patterns have shown a tendency towards more land settlements. This dramatic change might have been driven by several economic and social factors including the population and economic growth, which initiate a massive demand for housing and the subsequent expansion in housing construction, including high-density road-density construction, urban building, and cooperative-based housing development in the city (Kassa et al., 2011) including the study area. These changes also contributed to the massive land degradation observed in the study area. This case study clearly shows that rapid population growth could trigger a chain of environmental problems including land degradation and loss of fertile land due to forced expansion of urban centres. People will be forced to move into protected areas (e.g. forests), productive land (e.g. agricultural land), and unsuitable marginal range lands in order to satisfy their demand for housing and associated infrastructure. A detailed study of the area to prioritize land among desired LULCC is therefore very crucial for environmental-friendly economic and social development of the area where conflicting interests abound.

Figure 6. Spatial distribution of NDVI _1990–2020. Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

4.6. Normalized Difference Vegetation Index (NDVI)

The Normalized Difference Vegetation Index (NDVI) is the most commonly used vegetation index for observing greenery globally. Other commonly used vegetation indices are Enhanced Vegetation Index (EVI), Perpendicular Vegetation Index (PVI), and Ratio Vegetation Index (RVI). In general, healthy vegetation is a good absorber of the electromagnetic spectrum for visible reasons (see Figure 6).

Chlorophyll contains in greeneries highly absorbs Blue (0.4–0.5 μm) and Red (0.6–0.7 μm) Spectrum and reflects the Green (0.5–0.6 μm) spectrum. Therefore, our eye perceives healthy vegetation as green. Healthy plants having high reflectance in near infrared (NIR) between 0.7 and 1.3 μm. This is primarily due to the internal structure of plant leaves. High reflectance in the NIR and high absorption in the red spectrum, these two bands are used to calculate NDVI. So, the following formula gives the Normalized Difference Vegetation Index (NDVI).

NDVI = (NIR – Red)/(NIR + Red)

For Landsat 7 data, NDVI = (Band 4 – Band 3)/(Band 4 + Band 3)

For Landsat 8 data, NDVI = (Band 5 – Band 4)/(Band 5 + Band 4)

The NDVI value varies from −1 to 1. The higher the value of NDVI reflects the higher Near Infrared (NIR), means dense greenery. Generally, we obtain following result:

• NDVI = −1 to 0 represent Water bodies

• NDVI = −0.1 to 0.1 represent Barren rocks, sand, or snow

• NDVI = 0.2 to 0.5 represent Shrubs and grasslands or senescing crops

• NDVI = 0.6 to 1.0 represent Dense vegetation or tropical rainforest

Furthermore, all NDBI and NDVI rates can be calculated using the raster calculator in ArcGIS v.10.71.

4.7. Land use/land cover change of map from 1990–2020

According to Figure 7, conversion of the matrix for the years 1990–2020, the change in the LULC in the study area was largely attributed to the expansion of the built-up area. In the last 20 years, LULCC patterns have shown a tendency towards more land settlements. This dramatic change might have been driven by several economic and social factors including the population and economic growth, which initiate a massive demand for housing and the subsequent expansion in housing construction, including high-density road-density construction, urban building, and cooperative-based housing development in the city (Kassa et al., 2011) including the study area. These changes also contributed to the massive land degradation observed in the study area. This case study clearly shows that rapid population growth could trigger a chain of environmental problems including land degradation and loss of fertile land due to forced expansion of urban centres. People will be forced to move into forests, agricultural land, and unsuitable marginal rangelands in order to satisfy their demand for housing and associated infrastructure. A detailed study of the area to prioritize land among desired LULCC is therefore very crucial for environmental-friendly economic and social development of the area where conflicting interests abound.

Figure 7. Land use/land cover conversion matrix from 1990 to 2020. Source: US Geological Survey (USGS) (http://glovis.usgs.gov).

4.8. Business land use zoning

The existing land uses in the city indicate the availability of different types with varying character, density, and/or cluster. Low-density residential uses are the dominant land use type in the city. Except for some condominium buildings of G + 2 types, located in different areas of the city, there are no other high-density buildings in the city.

Most commercial activities are concentrated along some major roads (see Figure 7) with different levels of density of employment. However, these areas are not supported by adequate parking spaces and thus the city needs to consider parking spaces in such busy commercial corridors, namely Arab Sefer, Piazza, Old Bus Station, Sefere Selam, Atote Membo/Hawassa College of Teacher Education, Hawassa Industrial Park and Hawassa University, and the New Bus Station surroundings. The back roads in some of these corridors are very proximate to slums and dilapidated kebele houses.

4.9. Future expansion direction

Hawassa City has developed very rapidly, growing from a surface coverage of 4.98 km² in the 1983s to built-up area coverage of around 48.29 km² in 2018. These values demonstrate how the city has expanded exponentially beyond the planning boundary 37 square kilometers occupying agricultural land and encroaching natural systems. Due to the current uncontrolled development pattern and expansion of informal settlement, the city could not possibly expand towards the north, east, and west direction Figure 8.

Figure 8. Business land use zoning. Source: hawassa city Municipality.

Land grading is a regulation that is believed to describe the rank of locations based on their importance within a city. It also shows strategic investment areas and identifies the investment focus area of the city. The city has four types of land grades: Grade One, Grade Two, Grade Three and Grade Four. Grade one refers to the most valuable land, while Grade Four is the least. Each type of grade also has three sub-grades depending on the location and adjacency of main corridors and back roads (see Figure 9). Grading is simply made starting from the main city centers and moving outwards. This makes lakefront areas less valuable than Grade One areas. In Principle Lake front areas should be among the highest grades.

Figure 9. Map of existing land grading in 1983, 2003, 2011 and 2018. Source: hawassa city Municipality.

5. Conclusions

Software packages like ArcGIS for image analysis and map preparation, ERDAS (Earth Resources Data Analysis System) Imagine 2020 for RS application in order to process satellite images including image enhancement, pre-processing, and LULC classification were used for this study. Based on the satellite images and field observation, the land use land cover classes to be analyzed for changes were categorized. ERDAS IMAGINE software was used in image processing, i.e. land use land cover classification, and ArcGIS software was used in, vectorization, area calculation, and thematic map preparation of the study area. In order to detect the land use land cover changes, both remote sensing data and field survey were applied for interpreting the four Land sat satellite images

This study has quantified the dynamics of LULC and explored the drivers over the past 30 years (1990 − 2020) within the study area. Analyses of LULC dynamics over three decades of victimization GIS and remote sensing tools created six types of LULC classes. Land use categories intimate distinguished land cover dynamics throughout the study period. Quantitative spatio-temporal proof created through interpretations of satellite images showed that the study area has undergone important LULCCs since 1990.

LULCCs have a big selection of consequences at all spatial and temporal scales. Due to these effects and influences, it has become one of the main issues for environmental modification such as natural resources management. Characteristic interaction between changes and its drivers over house and time is vital to predict future developments, set decision-making mechanisms and constructing scenarios.

Land use and land cover (LULC) changes are one of the world-wide variations which have the most significant effects on the natural environment and ecosystem due to human activities. Hawassa is facing enormous changes in LULC patterns without any good arrangement from the last few years. The main objective of this study is to identify the LULC changes in Hawassa city on four different years (1990, 2000, 2010, and 2020) using various Landsat images and to detect the reasons for the variations. Supervised classification was used to identify the LULC changes in Hawassa city, as it explains the maximum likelihood algorithm using ERDAS imagine 2013 software. The total population in Hawassa city was 36, 200 in 1984 that increased by 69,200 during the 1994 national census and to 157,100 during the 2007 national census. Projections of urban population of Hawassa for July 2017 by the Central Statistical Agency (CSA) estimate the urban population to be 335,000 excluding 120,000 rural populations which resided within the administrative boundary of the City Administration). Thus, the total population within the administrative boundary of the city was estimated to be 455,658.

This study has been conducted by integrating GIS, remote sensing and spatial modeling tools. In order to sight and analyze changes in land cover categories, these techniques were enforced. Within the first sectionis satellite information for the study periods of 1990, 2000, 2010, and 2020. GIS and remote sensing techniques were applied to come up with land cover maps through the most chance supervised image classification. The accuracy assessment and change detection processes have additionally been done. The overall accuracy of LULC maps generated during this study had got an appropriate worth on top of the minimum threshold. The knowledge of image classification accuracy has become necessary for the modeling procedure. Within the last section, land use modeling was applied to investigate dynamic changes in built-up areas as a result of different driving factors. Bare land converted into agriculture has raised by percentage 1990, 2000, and 2010, 26.5%, 28%, and 30.11%, respectively but the area of agriculture land from 2010 to 2020 in comparison to years 1990–2000 and 2000–2010 where as the area decreases 78.50 square kilometer (30.11%) to 74.97 Square kilometer (28.76%) with 10 years rate of 1.35% dropping due to a number mega projects that increasing built-up that established in the study area.

Rapid urbanization which is happening at an alarming rate in major cities of Ethiopia puts high demand on infrastructure and the housing provision. Hawassa is witnessing this situation, being the hub of a big-scale economic activity, like the Hawassa Industrial Park. The population of the city is expected to double in 10 years’ time from its current number, which came to double in less than 10 years. Currently, quite half the population within the city lives in a housing condition that is crowded, inadequate, and/or informal. These living conditions dominantly give rental tenure rather than possession. The city has not been ready to meet the requirement for housing, particularly for the low- and middle-income inhabitants, because of principally inefficient policies, lack of finance, and absence of land within the designing boundaries of the city. These are increased by the shortage of capability in the municipality. The result has been visible in the insufficiency of affordable housing within the city, and the enlargement of informal settlements, which makes the present unplanned physical growth and high level of land speculations. The longer-term demand is even overwhelming, which needs correct designing methods to guide the city development in a more sustainable manner.

The same land use type was shown from the result of supervised classification, but the amount of coverage was different for different cover types. From 1990 to 2020, built-up and agricultural land increment and bare land, bushland, and forest land showed a decrement in the year compared. Satellite images indicated that an increase during this period was built up, and agricultural land was linked with different factors. Built-up areas with the Spatio-temporal dynamic of urban growth land use change processes that are related to population growth, and economic transformation policies of the country are also another responsible factor of the LULCC.

UN-Habitat recommends the adoption of the compact city model to address the current challenges the city of Hawassa is confronting, in particular, the housing challenge. Compacting the city encourages the maximization of the urban infrastructure and service, social integration, environmental protection, and economic efficiency for sustainable urban development. The scenario below reflects the compact city approach and aligns with the goals projected by the Hawassa City Administration.

6. Recommendations

Based on the findings in the preceding pages, the researcher forward the following conclusive remarks.

• Proper urban planning for management of landscapes and natural resources and sustainable urban development should be implemented.

• Detailed study of the area to land among desired land use/land cover classes ensures environment-friendly economic and social development of the area with conflicting interests.

• The kebele administrations of the nearby community, which have not yet been incorporated into the city administration, as well as those that have already been incorporated, need to launch mitigating works against the alarming land use land cover changes in the study area. This is possible through designing essential programs and implementing them through the urban safety net programs and those related to them aimed at reducing urban poverty.

Disclosure statement

No potential conflict of interest was reported by the author(s).