The city of Hawassa is situated in the southern region of Addis Ababa, Ethiopia. Geographically, the city stretches 275 km along the Addis Ababa-Moyal international highway and is embedded between Tikur-wuha River on the North, Alammura Mountain on the South, Chelelaka-Marshy on the East, and Lake Hawassa on the West (Weldesilassie et al., 2017). The city of Hawassa has an average elevation of approximately 1,690 m above sea level and extends on a partially flat plain in the rift valley (Hawassa City Municipality, 2020).

Hawassa borderland geographical coordinates extend approximately between latitude: 7°03′43.38″North and longitude: 38° 28′ 34.86″ East, covering a total area of 8,282.6 sq. km. The borderland contains eight Woredas within the Regional State of Oromia, namely, Shashemene, Arsi Negele, Kofele, and Siraro, and Southern Nations Nationalities and People’s Regional (SNNPR) Hawassa Zuria, Shebedino, Dale, and Boricha (Hawassa City Municipality, 2020).

In the context of Hawassa City, there have been no studies carried out on the physical and socioeconomic driving forces of land use and land cover change area investigating three decades of LULCC (1990–2020). The investigation of changes that occurred on Earth using images of the same scene at different times over the three decades was used to calculate the area of different land covers and observe the changes taking place over time. This analysis is very helpful in identifying various changes occurring in different classes of land use, like changes in urban built-up areas or changes in cultivated land and so on. Moreover, the pattern of geographical features varies from place to place and from time to time.

The main reason for the consideration of regional and peri-urban study in the development plan of the city and the linkage between rural-urban and urban-urban, which plays a vital role in attaining sustainable economic and social development and it requires a proper spatial, functional, technological, and organizational integration of the two spatial components of a region. Their relationship is normally symbiotic; the city serves as the center of provision of public and social services and processed and semi-processed industrial products and is used as an area with better employment opportunities. Rural areas, however, are sources of agricultural products for both food and input for industries, natural resources, and sources of labor forces (Figure 1).

The qualitative method provides details of the data and a view of alternative explanations; using this approach, the data collection and analysis would be carried out on the condition of social reality, allowing for in-depth research. A quantitative approach was used because it is specific, well-structured, and can be defined explicitly (Hoepfl, 1997). A mixed approach was used by the researcher to present the primary and secondary data, which was relevant to the study and used to meet the stated specific goals.

The study area was selected purposely due to the confidence of the researcher who had more information to collect data and information from different sources for the severity of the problem. In the city, there are eight sub-city administrations and 32 kebeles, of which, five kebeles were purposively selected from peri-urban areas, such as Alammura, Dato, Tulla-rural, Guwe-stadium, and Millennium-adeboye. Out of a total of 11,711 households, in those five kebeles, 317 sample households were selected using a sample size calculation developed by Kothari (2004). Data were collected using a simple random sampling technique and each was interviewed for the purpose. A semi-structured questionnaire was designed and pre-tested for its capacity to collect the data.

The researcher used a 95% level of confidence, the corresponding standard normal. In the study area, the population is characterized by similar economic activities, income levels, educational levels, and infrastructures, and a homogeneous population sample size depends on the type of research design to be used, the desired level of confidence, the level of accuracy, and the characteristics of the study population.

The study applied the formula provided by Kothari (2004); which is generally given as follows (mathematical Eq. 1): Z 2 x p x q x N e 2 N − 1 + Z 2 P . q n = (1)

Where: n = stands for estimated sample size.

e = the acceptable error (0.05).

N = size of population under study (11,711).

p = sample proportion of successes (0.5).

q = 1 – p = 0.5.

Z = standard confidence level (1.96) (95% confidence level).

As we have not been given the p-value as a proportion of defectives in the universe, let us assume it to be p = 0.5. If there is no prior information about P available to obtain a conservative estimate of the required sample size, we use p = q = 0.5. The researcher used a 95% confidence level and the corresponding standard normal. z = 1.96 and the desired level of significance was e = 0.05. Thus, p = 0.5 and q = 1-p = 1–0.5 = 0.5. We assumed a confidence level of 95%, N = 2,452, e = 0.05, z = 1.96; p = 0.5, and q = 1–0.5 = 0.5.

We obtained the following formula: n = Z 2   × p × q × N e 2 N − 1 + Z 2 × Pq 1.96 × 0.5 × 0.5    11 , 711 0.05 2 13 , 713 − 1 + 1.96 × 0.5 ∗ 0.5 = 11 , 711 35.27 = 332

As stated above, the study involved 332 households, which were selected proportionally from each of the kebeles. Therefore, the sample was arrived at by first selecting a random starting point and then picking up every “K” in the population among households of the independent kebeles of the sampling frame. As shown in Table 1, out of the 332 intended questionnaire respondents, to whom questionnaires were distributed, 317 (95.5%) filled out the questionnaires properly and returned them.

Data processing, which includes editing, coding, data input, cleaning, and consistency checking, is a significant component of the research process. The study researcher was responsible for carrying out all tasks. LULC driving forces are reported using descriptive statistics such as mean, frequency, percentages, and the Landsat 5, 7, and 8 images that were acquired from USGS. In order to determine the main socioeconomic driving forces, LULCC, and the effectiveness of urban growth continuous variables, the statistical significance of the dummy/discrete variables was also examined using the Chi-square test.

Discrete regression models are models in which the dependent variable assumes discrete values. The simplest of these models is that in which the dependent variable Y is binary (it can assume only two values denoted by 0 and 1), as described by Amemiya (1981), Gujarati (2002), and Maddala (1998). According to these authors, the three most commonly used approaches to estimating such models are the linear probability models (LPM), the logit model, and the probit models. The linear probability model is the model that expresses the dichotomous dependent variable (Y) as a linear function of the explanatory variable (X). Because of its computational simplicity, LPM has been used in econometric applications, especially during and before the 1960s. However, as indicated by Amemiya (1981), Maddala (1998), and Gujarati (2002), the linear probability model has an obvious defect in that the estimated probability values can lie outside the normal 0–1 range. The fundamental problem with the LPM was that it is not logically a very attractive model because it assumes that the marginal or incremental effect of explanatory variables remains constant, that is Pi = E (Y = 1/X) increases linearly with X (Maddala, 1998; Gujarati, 2002). The defects of the linear probability model suggest that there was a need for an appropriate model in which the relationship between the probability that an event would occur and the explanatory variables was non-linear (Amemiya, 1981; Madalla, 1998; Gujarati, 2002). These same authors suggested that the sigmoid or S-shaped curve, which very much resembles the cumulative distribution functions (CDF) of random variables, be used to model regressions where the response variable is dichotomous, taking 0–1 values. To explain the behavior of dichotomous dependent variables, we would have to use a suitably chosen cumulative distribution function (CDF).

The dependent variable is the household livelihoods of surrounding displaced farmers due to urban expansion and the independent variables are age, gender, family size, farm size, education, income, TLU, occupation, compensation paid, land use dynamics, socioeconomic activities, culture, environmental change, and social status regarding the expropriation. Since the dependent variable of this objective is the household livelihood of surrounding displaced farmers due to urban expansion, which is dichotomous and takes the value of 0 if relating to a household before displacement due to urban expansion and 1 relating to after displacement, a binary probit model would be used (Gujarati, 2002).

The logit model (Gujarati, 2002) is used to analyze the impact of various socioeconomic factors on peri-urban farmers’ urban expansion. Of the 317 total respondents, the binary response variable is the probability that peri-urban development results in farmer displacement with a value of 1 (yes) or that there is no urban expansion before farmer displacement, with a value of 0 (No). This function is generally given as follows (mathematical Eq. 2): i = I n p i 1 − p i = β 0 + β 1 X 1 + … ± β i X i (2)

The cumulative logistic distribution function is as follows: Pi = E Y = 1 X   i = 1 1 + e − ( β 0 + β 1 X i … … . . . + β i X i = e − z i 1 + e − z i (3)

Where Li is the logit, which is the natural log of the odds ratio,

Pi is the probability that a farmer will sell their land for non-agricultural use; 1 − Pi is the probability of farmer displacement with a value of 1 (Yes) or no urban expansion before farmer displacement, with a value of 0 (No). In general, the model can be βi is the ith parameter of the model to be estimated, zi = β 0 + β 1 X 1 + … … … … … … . … … … … … … … … … … … … … … … … . . + β iXi (mathematical Eq. (3)).

The specific model form for the econometric analysis is given by Li = In p i 1 − p i = β 0 + β 1 SEXi + β 2 AGEi + β 3 FAMSIZEi + β 4 FLANSIi + β 5 EDUi + β 6 COMPONSi + β 7 INCOMEi + β 8 TLUi + β 9 OCCUPIi + β 10 SOCRSHIi + μ i (4)

Where

Y = dependent variable.

X₁ …X_n = independent variables.

β_o = Intercept.

β 1   … . .   β n = Coefficient of independent variables; and.

e_i = error term.

Because our data is at the individual level, we used the maximum likelihood (ML) method to estimate the parameters (mathematical Eq. (4)) (Aldrich et al., 1984; DeMaris, 1992). The independent variables used in the model include gender, age, family size, farmland size, education, compensation paid, total livestock unit, and occupation. These variables are discrete or binary and are defined in Table 2.

The satellite image data covers a very large area but the study area is small. The spatial resolutions of images were enhanced using a resolution merge technique that integrates images of different spatial resolutions or pixels. Generally, the image was processed to minimize the error during analysis due to atmosphere, sensor, cloud, bad lines, or other related errors. The preprocessing techniques employed include radiometric correction, geometric correction image enhancement, and that of a sub-setting area of interest before further analysis takes place. Therefore, the image was preprocessed to show a clear and easily interpretable image. Creating subsets that are areas of interest (AOI) is one of the preprocessing steps. The Area of Interest was delineated/a subset from the images of the whole period for the respective years and scenes. Finally, the satellite image that was delineated or masked using the boundary of Hawassa City was ready for further analysis.

The classification describes spectral attributes for each feature in the area of interest. The spectral qualities of pixels in the dataset are numerically compared to each category and labeled with the name of the land cover. Image classification was undertaken to extract thematic information from the images. This technique is reliant on the three thematic maps of different dates to detect changes; with assistance from the reference data and the personal knowledge about the study area. The reference data include the Global Positioning System (GPS) for currently existing LULC, Google Earth images, and aerial photos for other periods.

The LULC detection was based on quantitative analysis and the analysis of the driving forces of the change was conducted using qualitative analysis. Therefore, to see any changes that took place within these selected periods, the study year was selected based on factors such as what the LULC pattern looked like in 1990 and 2020 and what is happening in the study area nowadays. QGIS version 3.2 and ArcGIS 10.3 were used to perform image preprocessing and analysis which include the following processes: pre-processing, image classification, accuracy assessment, and production of a change map.

The images of different years were classified using the supervised classification method to identify LULC types within Hawassa City. The supervised classification with a maximum likelihood classifier was used for image classification and the preparation of base maps for change detection (Chipman et al., 2004). Supervised classification was chosen because it classifies land uses based on training sites that are assigned by Classifier. This is done for all of the images according to their arrangement with the help of reference data to have accurate classification.

Successful use of remote sensing for LULCC detection largely depends on an adequate understanding of the study area, the satellite imaging system, and the various information extraction methods for change detection to achieve the aim of the present study (Yang and Lo, 2002). In change detection analysis, remotely sensed data were used to monitor LULCC (Singh, 1989). After the image classification, the post-classification change detection was performed and evaluated with “from–to” change information (Macleod and Congalton, 1998).

Finally, the table matrix was generated, which holds overall information about the change matrix in the study periods from 1990 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 calculated as follows (Addis, 2010; Abate and Angassa, 2016). CA = TA t 2   −  TA  t 1 CE = CA / TA  t 1 *   100 CR = CE / t 2   – t 1

The Maximum likelihood used to categorize the Landsat 5 TM 1990, 2000, and 2010; Landsat 8 OLI; and SWIR 2020 produced a map showing the distributions of seven prevalent LULC classifications. The ArcGIS software and spatial analyst tools demarcated the boundary of the study area. The linear relationships among the variables were described. The detailed methodology adopted is given in Figure 2.

To determine the positive and negative consequences, as well as potential causes, of these expansions, peri-urban farmers were questioned about how urban growth has affected their family’s quality of life and that of the neighborhood. A three-level Likert scale (Low (0), Moderate (1), and High (2)) was used to obtain the descriptive statistics of mean and standard deviation values. A mean value of 0–1.5 is low, 1.5–2.5 is average, and above 2.5 is considered high.

The results in Table 4 show that the majority of respondents, 171 (55.7%), described the city expansion rate as high and a significant number, 143 (45.1%), expressed it as a moderate rate. However, only 3 (1.1%) respondents expressed the expansion rate as low. The mean score for this item was 2.55, with a standard deviation of 0.522. This implies that there is strong agreement among respondents about the rapid expansion of Hawassa City to the surrounding rural Kebele Administration.

The Hawassa City urban expansion is not properly managed by the concerned people and the urban land is taken from the peripheral people, but the rule and the regulation of how to take the urban land must need respect ion and proper implementations of the urban land, and there is a need for a proper compensation system for those losing their land. Urban land is not properly utilized. For this situation, assigning sufficient budget and professional human resources is critical to identifying the societies exposed to losing their land and reinforcing the restoration policy of the urban land management program in Hawassa City. Proper land utilization for urban expansion requires the participation of all concerned stakeholders in the city administration. Therefore, conducting further research on modern utilization of land in urban centers and the need to respect the rules and regulations of urban land expansion programs is helpful. Urban land policies regarding how to apply urban expansion and providing sufficient compensation to the land loss societies in the urban areas should be followed. Following all the listed actions makes the urban land efficient, effective, and suited to society’s need for the proper utilization of the urban land, leading to a better economic, productive, and social situation for the city dwellers.

According to the current study, urban growth has a considerable influence on the socioeconomic activities of the neighborhood. This finding is related to that of Dociu and Dunarintu (2012), who claimed that urbanization has a significant impact on places due to several factors, including the degree of use of land or infrastructure, which manifests as contact with the labor market established in urban areas, the population’s adjustment to the labor market, specializing in the needs of the existing brand and, thus, framing and adaptation, and gaining money to cover expenses. As one of the socioeconomic elements of society, a change in fertility has also been observed following urban expansion in peripheral areas.

As part of the GTP II (Growth, 2016), the government of Ethiopia has been implementing industrial parks in Addis Ababa and other regional cities. The parks are specialized in textile and garment production, and their implementation has been coordinated by the Industrial Parks Development Corporation (IPDC); Hawassa Industrial Park (HIP) was the first to start its operations in 2016. Currently, 20,000 workers, plus 2,000 indirect workers are employed by the project. They plan to have 40,000 direct workers by the end of 2020, reaching their peak by 2025, with 84,000 workers.

According to the HIP Tenants Association, out of the 20,000 workers, approximately 16,000 are basic operators, 3,000 are mid-level technical and managerial Ethiopian staff, and approximately 1,000 are expat technical and managerial staff. This number should double over the next 2 years, with a shrinking share of expat staff; however, absolute numbers might not change so much. The operators are almost exclusively young women, mainly from rural areas, who moved to Hawassa City in search of employment. They are between the ages of 18–30 (although on average, at the lower end of this age range). Only a small number of the operators brought dependent family members with them (Mohr et al., 2016). The mid-level Ethiopian staff, mostly men, are recent university graduates. By 2025, the size of the workforce at the HIP is expected to reach 84,000, almost four times the current workforce of 20,000. This dramatic growth in the number of industrial workers is bound to put increasing pressure on the city of Hawassa and the manufacturers to find lasting solutions to the housing crisis (Figure 3).

In Hawassa City, urban land is not more affordable to the residents due to its limited capacity. In Hawassa City, almost the price of urban lease law serves higher-income groups of people within society rather than middle- and lower-income groups. This shows the higher income groups’ advantages over other society members in gaining land leases. According to the interviewed respondents, middle- and lower-income groups buy illegal land in peri-urban areas because the price of land is cheaper than land leases (Ling et al., 2018). Additionally, the studies conducted by World Bank (2019) and Adamu (2014) reported that urban land is used for various purposes, but they found that it had several flaws, including corruption, a lack of openness, and other issues. The municipality made a public announcement and encouraged various society members to participate in an open lease competition. Therefore, people with higher incomes were competing more frequently for each parcel of land by taking advantage of these opportunities. The participation of wealthy individuals and other privileged groups raised the ceiling price over the initial amount allotted.

Entering the data obtained into STATA version 14, the probit model was analyzed. Before the analysis, the explanatory variables were checked for multicollinearity and heteroscedasticity. Before running the model, all the hypothesized explanatory variables were checked for the existence of a multicollinearity problem. Two measures are often suggested to test the existence of multicollinearity: VIF for association between the continuous explanatory variables and Contingency Coefficient for dummy variables. The multicollinearity indicator, VIF, ranged from 1.116 to 3.755 and the Tolerance index ranged from 0.896 to 0.266. It therefore suggested that regression coefficients did not suffer from multicollinearity among the independent variables. The overall goodness of fit for the model parameter estimates was assessed based on several criteria. First, it was found that the correlations between the dependent and independent equations were significantly different from zero the (ρ ≠ 0), which confirmed that the application of probit was valid and it was statistically a true stochastic specification. The distributions are independent if and only if ρ = 0. The relationship between the dependent and independent equations can be estimated consistently with the single-equation probit method. However, such a commonly used approach was inefficient because it ignored the correlation between the error terms of the underlying stochastic production functions (Maddala, 1998; William, 2003). Finally, the validity of the constraint was examined empirically by usual statistical test statistics as well as the LR or Wald test. The overall fit of the equations indicated that LR: the χ2 (10) statistics = 87 was significant at less than a 1% significance level through the livelihoods of surrounding displaced farmers. Based on the outcome of this test, the study yielded the basis to reject the null hypothesis at a reasonable level of significance. Binary probit model analysis was conducted to determine the impacts of urban expansion on the livelihoods of surrounding displaced farmers. The probit estimates the likelihood of urban expansion before and after the displacement of the farmers and statistically significant variables were identified in order to measure the impact of their relatives on the farmers’ decisions (Table 5).

However, the linkage between the city’s sustainability and land use is not only a matter of proportions. Thus, how natural resources and non-built-up land are spatially distributed determines equity (equitable distribution of benefits provided by natural areas) and the proper functioning of the natural areas. For example, having only one huge park within the city provides natural benefits that are different (in quantity and quality) than the same surface area of green space split into many smaller parks. For this reason, the natural areas of the city cannot only be the random negative of the built-up land and there is a need to think about the green infrastructure of Hawassa City.

As seen in Figure 4, the size of the farmland of displaced farmers is the main determining factor, but due to the scarcity of land in the city, the majority of farmers, 132 (41.6%) respondents, had a land size of below 0.5 ha, followed by 112 (35.3%) respondents with between 0.5 and 1.5 ha, 56 (17.7%) respondents with between 1.5 and 2.0 ha, 14 (4.4%) respondents with between 2.0 and 2.5 ha, and 3 (1.0%) respondents with between 2.5 and 3.0 ha. All in all, 244 (76.9%) respondents had a land size of below 1.5 ha.

Land Liquidity, deposit mobilization, and investment linkage stem from an estimate that the capital value of land and property constitutes one-half to three-quarters of a country’s wealth. The proportion is higher in less developed countries with limited domestic capital. This is not only observed at the country level but also at the individual and family level. Land and property are, therefore, likely to form by far the largest class of assets in most developing economies. As a result, the efficient use and management of land and property is the key to promoting inclusive growth that targets the country’s bottom 40% in terms of per capita income. Land registration projects contribute to a more comprehensive and accurate collection of property taxes (World Bank, 2019).

Poorly functioning land management leads to several problems, including land speculation, the creation of slums and squatter settlements, environmental deterioration, and an inefficient urban development pattern, which increases the cost of doing business in the cities/towns and adversely affects urban land management (World Bank, 2019). This results in the creation of slums and squatter settlements, environmental deterioration, urban land-related conflict, an inefficient urban development pattern, and the violation of the structural plan of the area.

As seen in Figure 5, regarding compensation paid to displaced farmers based on permanent agricultural plantations as well as house constructions on the plot of land they owned before land expropriation, one-time compensation payments received by the respondents varied from 100,000 to 500,000 Birr. The majority of farmers, 193 (60.9%) respondents, were paid compensation below 100,000 Birr; 65 (20.5%) respondents were paid lower compensation, between 100,000 and 200,000 Birr; 40 (12.6%) respondents were paid compensation between 200,000 and 300,000 Birr; only 7 (2.2%) respondents were paid higher compensation, between 400,000 and 500,000 Birr; and 7 (2.2%) respondents were paid the highest level of compensation, above 500,000 Birr. On average, each displaced farmer was paid a compensation of 161,075 Birr for the loss of agricultural production. The total compensation paid for the sampled 317 respondents is estimated to be 49,450,000 Birr for the loss of agricultural production. This implies that the farmers who earned more income before displacement and non-displaced farmers received better compensation compared to displaced farmers.

The most frequent complaints among displaced farmers were that they were not offered adequate compensation and they were not given a reasonable time to leave the area. The interview discussion regarding compensation payment showed that many of the displaced farmers got 500 square meter plots of land to construct their houses in the city and the other land was taken from them for an unknown rate of payment per parameter. This paves the way for corruption and maladministration.

Expropriation represents both the most serious infringement of private property rights and the manifest exercise of state sovereignty. It is a formal withdrawal of property rights for the benefit of the state or for private persons designated by the state. There are two types of expropriation: direct and indirect. Direct expropriation usually involves formal processes and is explained in an expropriation decree or law. Expropriation of this type is undertaken against one or several investments. Expropriation, or nationalization, can also be against several investments in one economic sector. The second involves indirect expropriation. This type of expropriation may result from measures that a state takes to regulate economic activities within its territory, even where such regulation is not directly targeted at an investment (Nikièma, 2015).

Generally, it is assumed that landowners may be compensated fully using other approaches, especially where the property is not shown to be both unique in nature and location and (Dfid, 1999) indispensable to the conduct of the landowners; business operations on the site from which a part is taken. So, for the most part, buildings of a unique character are valued using this method. This approach can be used in countries where the market value of property is not determined. The method determines the value in terms of current labor and materials required in assembling a similar asset of comparable utility (Ndjovu, 2003).

Respondents were asked to express their levels of social relationships and values before and after land expropriation on a five-level Likert scale: very low (0), low (1), moderate (2), high (3), and very high (4).

Data collected from sampled respondents expressed in Table 6 show that the mean score for the level of social relationship and values after land expropriation was 1.7 (moderate), with a standard deviation of 0.852 in comparison with the mean score for the level of social relationship and values before land expropriation, which was 3.21 (high) with a standard deviation of 0.79. This shows that the land expropriation program negatively affects farmers’ levels of social relationships and values, and their social relationship scores before displacement and after displacement lay between low and moderate.

Similarly, respondents were asked to express their average annual income before and after land expropriation on a five-level Likert scale: very low (0), low (1) moderate (2), high (3), and very high (4).

The results show that the mean score for the level of social relationships and values after land expropriation was 1.032 (low), with a standard deviation of 1.244, in comparison to the mean score for average annual income before land expropriation, which was 2.16 (moderate), with a standard deviation of 1.23. This shows the land expropriation program negatively affects farmers’ economic levels (see Table 7).

As seen in Table 8, the respondents’ main source of income before and after land expropriation drastically changed according to the transition matrix presented in the above table. From a total of 206 (67.1%) respondents who were dependent only on agriculture before land expropriation, 187 (60.9%) turned to non-agricultural activities after land expropriation, while 19 (6.2%) turned to both agricultural and non-agricultural activities after land expropriation. However, none of them remained in agricultural activities.

As previously indicated, to see the effect of urban expansion on the community, the researcher employed a multivariate analysis of variance alongside descriptive statistics. The correlation between dependent variables was analyzed using Statistical Package for Social Science (SPSS) version 20, using a Pearson Correlation coefficient. The results of the relationships among the variables used in the questionnaires are indicated.

Bivariate Correlation tests whether the relationship between two variables is linear (as one variable increases, the other also increases, or as one variable increases the other variable decreases). In addition to this, Pearson’s product-moment correlation coefficient is a measure of the linear correlation between two variables, X and Y, giving a value between +1 and −1 inclusive, where 1 is a total positive correlation, 0 is no correlation, and −1 is a total negative correlation (Pedhazur, 1982). When Pearson’s r is close to 1, there is a strong relationship between the two variables. This means that changes in one variable are strongly correlated with changes in the second variable. When Pearson’s r is close to 0, there is a weak relationship between the two variables. This means that changes in one variable are not correlated with changes in the second variable (Malhotra and Briks, 2017).

The result of the multivariate test in Table 9 indicates that the effect of urban expansion on the combined dependent variables, namely, on land use, socioeconomic activities, culture, and environmental change is statistically significant; F(8, 616) = 12.704, p = 0.000, Wilk’s Lambda (λ) = .737, partial eta squared(partial η2) = 0.142, observed power = 1.000.

As indicated in Table 10, the effects of urban expansion on socioeconomic and cultural aspects and the land use of nearby farmers were found to be significant. Since the results of the multivariate analysis of variance show the existence of significant mean differences in land use, cultural influence, and socioeconomic influence across the levels of perceived urban expansion by respondents, the researcher conducted a post hoc analysis of the difference within groups as shown.

The multiple comparisons using Least significant difference (LSD) post hoc analysis indicate that there is a significant difference in terms of perceived socioeconomic effect between those who perceive that there is low urban expansion and those who perceive there is medium urban expansion (M = −3.1442, SEr = .71400; LB; −4.5491, UB = −1.7393), and between those who perceive there is low town expansion and those who perceive there is high urban expansion (M = −3.1545, SEr = .77860; LB; −4.6865, UB = −1.6225).

The post hoc analysis also indicated that there is a significant difference in terms of the cultural effect between those who perceive that there is low urban expansion and those who perceive that there is medium urban expansion (M = −3.3679, SEr = .91248; LB; −5.1633, UB = −1.5725), and between those who perceive there is low urban expansion and those who perceive that there is high urban expansion (M = −3.7530, SEr = .99503; LB; −5.7109, UB = −1.7952).

Furthermore, the post hoc analysis also indicated that there is a significant difference in terms of the perceived effect of urban expansion on land use between those who perceive that there is low urban expansion and those who perceive that there is medium urban expansion (M = −2.6846, SEr = .76579; LB; −4.1914, UB = −1.1779), and between those who perceive there is medium urban expansion and those who perceive that there is high urban expansion (M = 3.6466, SEr = .77321; LB; 2.1253, UB = 5.1680) (see Table 11).

The environmental impact of urban expansion was not observed to have a statistically significant independent influence on the adjacent environment in the current investigation. The authors of Seifollahi et al. (2022) noted that the relevant contribution to land degradation caused by urban expansion was demonstrated to have been derived from socioeconomic drivers, the most significant of which were population growth and urban sprawl. This finding contrasts with their findings. These elements were recognized as producing complex environmental syndromes that are fueled by urbanization. This perspective views peri-urban regions as socio-environmental systems adjusting to strong socioeconomic transitions.

The bi-directional or two-way influence of the urban and rural ways of life on the community is another finding in the literature on how urban growth affects the socioeconomic elements of the adjacent village. This occurs when urban and rural areas are altered via commerce and mutual interactions between the physical and practical aspects of existence, as has been seen by the general public. According to Torquati et al., 2020, the peri-urban environment is increasingly regarded as a unique, hybrid environment that combines urban and rural characteristics.

In the current study, the effect of displacement on the socioeconomic conditions is not uniform across the educational levels of heads of households as well as the personal diligence of the whole members of the family, indicating that the better educated and those with more manageable family sizes tend to lead better lives. This finding relates to that of Kasa et al. (2011) who stated that households with relatively older, male, literate heads, whose families received relatively better transfer income, and who are located near transport stations had an increased likelihood of participating in non-agricultural livelihood strategies. In their study, they stressed that regardless of the diverse livelihoods of peri-urban households, the propensity score matching estimation indicated that fully displaced households received lower compensation than partially displaced and non-displaced households, whereas, their average per capita expenditure was found to be too big, signifying that it would be difficult to lead their life. This shows that regardless of compensation, fully displaced people often fail to establish a comparable means of income and they pursue an asset-depleting consumption style.

Considering the city’s administrative boundary as a basis for land use classifications, the proposed land use exhibits a different percentage. In this regard, urban agriculture takes the major share (41.87%), followed by recreation and environmental uses (21.74%), housing (16.15%), and road and transport (10.85%) (Figure 6).

Hawassa City is the largest city in the southern region. The city serves as the main political, economic, and employment center for the regional state and within its surrounding context. It is one of the most popular tourist destinations in the country. The presence of Hawassa Lake bounded the city in the west has a significant impact on attracting international and domestic tourists, attracting investment, and, consequently, contributing to local economic development.

Hawassa City has a total area of 178.3019 sq. km made up of residential areas, commercial space, manufacturing centers, roads, recreation, agriculture, etc., of which more than 60% is rural land. The city has eight sub-cities and 32 kebeles. Of the total land area of Tula, 77% constitutes the Tula sub-city while the remaining 23% is shared by the other sub-cities.

The northern part of the city is predominantly flat, while the southern part is predominantly hilly, and the highest peak is located at the southernmost part of the city, at 2,140 m above sea level. Land adjacent to the lake and the oldest settlement is level to very gently sloping, with a slope of 0%–5% and a land area of 86% of the total land of the city. Land in the south and around the Tabour Mountain and Alammura Mountain areas is hilly and has slopes of greater than 10% covering 7% of the total land of the city. These lands are not used for settlement purposes and they are forested suited for recreation and greenery.

Hawassa City is located in the Hawassa sub-basin drainage system and has seasonal and perennial rivers, such as the Tukur-Wuha river, which collect water from the nearby highland areas and drain it into Lake Hawassa. In the city, except in the eastern and southern areas and very flat/lower parts of the city, the run-off flows toward the lake.

The studied area showed a real growth rate and a pattern of land change, and this tendency was supported by image classification and a growth pattern. Figures 10, 11 show how the classification of LULC has changed throughout the study period in the respective areas. The various urban regions contained the seven classifications of LULC. Based on field observation and general historical information gained from bare land, built-up areas, bushland, cropland, farmland, forest land, and water bodies., It was decided to focus on the following seven major LULC classes identified using the field data and satellite images: bare land, built-up areas, bushland, cropland, farmland, forest land, and water bodies.

Conversion change maps were analyzed for each period to clearly show the source and destination of the selected land use land cover class. Analysis of the conversion map was computed in ArcGIS desktop 10.8 using spatial analyst extension by overlaying classified images covering 10 years, from 1990 to 2020. Moreover, all change maps show the gross gain and loss of each land cover category during the study periods. The level of conversion varies from one type of LULC to another (Figures 9, 10).

The results of the research, the built-up area increased dramatically from 5% (12.56 sq. km) in 1990 to 18.71% (47.07 sq. km) in 2020. With this percentage growth, even though it decreased from 11.21 sq. km in 1990 to 4.5% in 2020, bushland remained the most common land use category. Between 1990 and 2020, bare land, farmland, forest land, cropland, and water bodies all significantly decreased. The percentage coverage and geographical distribution of land use categories for the years 1990, 2000, 2010, and 2020 are shown in Table 12 and Figures 9, 10.

The high population expansion in the study region can be linked to the significant growth in built-up areas. Between 2010 and 2020, the population of Hawassa increased as a result of Hawassa Industrial Park (HIP), which was the first to open its doors in 2016. The project now employs 20,000 direct employees and 2,000 indirect employees. They plan to have 40,000 direct workers by the end of 2020, reaching the peak by 2025, with 84,000 workers. This can be partly attributed to the architectural changes currently occurring in the region, with more people favoring bungalow-type (single-tenant) houses over compound (multiple-tenant) houses (Grant, 2009). Additionally, economic growth has led to a massive boost in the housing sector. Market forces, rather than urban planning, shape spatial expansion in the region (Figure 12) (Alliance, 2016).