Analysis of Urban Expansion and Modeling of LULC Changes Using Geospatial Techniques: The Case of Adama City

The urbanization is playing a vital role in global development. Increased population and land requirement for settlement, economic and industrial growth, as well as its environmental effects, have also posed concerns in many countries. The world has great exceeding in city development in accordance to the United Nations (UN, 2000) and the Population Comparison Bureau (PRB, 2000). In 20^th century, population in metropolitan cities was very quickly expanded from 0.22 billion to 2.8 billion. An enormous rate of urban development in the developing world will occur within the next few decades (Mesfin, 2009). In Asia and Africa countries, population becomes double in single century. Through 2030, cities and towns in the developing world will make up 81% of the urban population (Mesfin, 2009). The population growth rate in Africa was the highest in the world and it was 4.4% per year (Antonio, 2000). Ethiopia is still largely under-urbanized, taking into account the norms of Africa (Haregeweynet et al., 2012).

In the middle of 1930s, the urbanization rate will be faster (Antonio, 2000). Urbanization was increased between 1950 and 1965 with an average growth rate between 5.4 and 5.6 % per year indicate a doubling of the population in just 13 years. Urban population increase in the last 15 years has again been more than 5% and a high in the five-year period of 1985 and 1990, when the average rate of change was 5.93% (Mesev, 2008). Mostly, the rapid growth of the urban population is due to the difference between rural and urban development and significant rural-to-urban migration up to 1975, while the land reform program of 1975 created incentives and opportunities for peasants and other potential migrants to live in rural areas (Antonio, 2000).

Urbanization has been a product of industrialization in the countries of Europe and North America and has been linked to economic progress, while urbanization has arisen in the developing countries of Asia, Africa and Latin America as a result of population growth and large rural-to-urban relocations (Mesfin, 2009). According to Haregeweyn et al. (2012), many fields have been transformed from rural to urban. According to the 2007 census, Ethiopia has the second largest population in Africa with a total population of over 80 million. It has an annual growth rate of 2.3% and an overall annual population growth rate of 4.6% (Haregeweyn et al., 2012). According to Zubair (2006), the rapid rise of urbanization, the economic gains and the unplanned growth of cities have been resulted many negative effects. The majority of cities are horizontally spreading along the rail and road networks. Such enthusiasm of urban land cover may be caused by social, economic or industrial influences. There are already significant environmental issues that need to be closely examined and tracked for successful land use regulation and planning. In addition, fast population development and improvements related to urban land have attracted many scholars to study and track developments (Mesfin, 2009).

The unexpected urban development is now a major challenge, especially in developing countries where there is a shortage of consistent and accurate data, including spatial data. According to Mesfin (2009), the study of drastic shifts in LULC at continental, national and local level, and additional exploring the scale of potential changes, the existing geospatial expertise on land use dynamics and innovations already plays an important role. Remote Sensing (RS) and Geographic Information System (GIS) are now providing new tools for sophisticated landscape management. The remote sensing analysis allows patterning, system activity, adjustment at regional and local. This data provide a significant connection between localized ecological studies and intensive and international, state and regional conservation (Zubair, 2006). The urbanization, as a deep and also ongoing worldwide occurrence that has been going on for hundreds of years, has been researched by several scientists, concentrating on city growth, city sprawl, and the complexities and trends of urbanization (Antrop and van Eetvelde, 2000) noted that urbanization involves diverse and complex, changes to the rural environment in the, area of towns and cities, leading to very heterogeneous ecosystems. Gerten et al. (2019) have established the trajectory of urbanization expected to remain steady in the future, and will influence trends of land use in several respects. In addition, Keil (2018) claimed that the Earth is the ‘suburban planet’ and predicted that people will want to ‘make the world urban from the outside in’. The European, Environmental Agency (EEA, 2018) cautioned that urbanization, is unsustainable, contributes to soil waste, divides the landscape, and limits the ability of habitats to offer essential ecological services. Henderson and Turner (2020) shows that Europe, Latin and North America and the Caribbean as well as Western Asia tended to be entirely urbanized. Yang et al. (2018) reported that Chinese scientists have observed and marked uneven urban developments in various locations in the world, particularly in Benjin. In comparison, there is a limited research on urban development and its impacts using geospatial data. In order to evaluate land cover change and growth in Adama Region, a change, detection study was undertaken to evaluate the magnitude and scale of land-use change and estimation over time and space. The findings measure LULC change patterns of change in the city and show available satellite data to provide, reliable, cost-effective means of measuring and assessing the land cover changes through a spatial-temporal context that can, be used as guides to land form management, and policy decisions on various themes relevant to land cover, such as agricultural areas, croplands, forest lands.

2.1 Study Area

Adama city is located 100 km from southeast of Addis Ababa about 8º 35' 00" and 8º 36' 00" N and 39º 11' 57" to 39º 21' 15" E (Figure 1) at an average altitude of 1620m above the sea level. Adama is situated in the Eastern Shewa region which is part of the central plateau. According to the 2007 census, the total population of Adama city was 222023. Besides, the population growth rate of the city (5.4% per annum) was estimated to be higher than the national average (2.5% per annum) (CSA, 2007). CSA’s 1994 Population and Housing Census and 1999 National Labour Force Survey (NLFS) indicates that there were about 187,000 people living in Adama City in 2004. Nevertheless, the year 2002-2003 estimates of the City Administration reveals that the City population is known to be well over 200000. The proportion of male to female during the period was about 48.5 to 51.5 percent revealing further that the sex ratio was 94.0 percent, which implies little excess of women over the men (about 100 females for each 94 males). Furthermore, according to the CSA’s 1994 census, about 33.6 percent of Adama residents were below the age of 15. Similarly, the data further reveals that populations within the working age group (15-64) and above 64 years of age were 63.5 and 2.9 percent, respectively.

As a result, various studies [CSA, 1984, 1994, 1999 and Adama Project Office’s (APO) socio-economic surveys, 2004], Adama has been one of the areas both in the region and the country that receives heavy influx of migrants each year. For instance, the 1984 Census indicates that 53.7% of the City populations were migrants. Similarly, after a decade, in 1994 the second census showed that the extent of migrants were 53.6%. The finding of the 1999 NLFS neither contradicts with the above mentioned two census outcomes. According to the results of the NLFS, hence, 52.1% of the populations of Adama City were contributed from migrants in 1999. The findings of many socio-economic surveys undertaken by APO in 2003-2004 too, confirm that immigration stand out as one of the major issues in Adama City. In the Poverty and Socioeconomic Problem survey undertaken in Adama, about 68.0% of the poor households were migrants. Similarly, in the same study it was found that about 83.0 percent and 87.0 percent of the commercial sex workers and street occupants were migrants, respectively (APO, 2004). The census data further points that Adama households were estimated to have about 4.8 persons on the average. Thus, it could be possible to estimate that about 39.0 thousand households were residing in the city in July 2004.

2.2 Data and Processing

In this research, geospatial technology was, used to detect the urban changes in a particular timeframe and, to examine patterns and trends of transition. Satellite imageries are almost ideal to provide details on the numerous geospatial patterns on the surface of the earth. However, attention must be, given to the degree of, resolution, the degree of accuracy of the image, through processing. However, similar to Mas (1999), it was suggested that it would be impossible to collect multi-date data at the same time of year. According to Gluch (2002), satellite image, MSS, ETM+ and, ETM+ satellite data, are typically used, for their multispectral and environmental, significance (Table 1). The pre-processing and post-image processing analysis was carried out to increase the images accuracy, and the readability of the applications using GIS and ERDASS software. The spheroid and the date are referred to in WGS84. All images were, geometrically, co-registered using ground control, points in the UTM projection. Ground Control Points (GCPs) obtained from identified ground points, the coordinates of which can be precisely placed on digital, imagery.

Ground Controls Points (GCPs) were collected by using Global Positioning System (GPS). Hence, Arc Map 10.5, ERDAS Imagine 2014, and IDRISI 32 software tools were used to make land-use modeling and future prediction. Besides, IBM SPSS statistics 20 and Microsoft Excel for tabulating and graphical representation were used.

2.3 Land Cover Nomenclatures

In order to make sample collection and classification easy, land cover nomenclatures are required to create and define the possible LULC classes first. Although the focus of the paper is on built-up areas, LULC map of the study areas was first generated using land cover classes presented in table 2. The LULC classes applied in this study were adopted from the classification used by European  Environment  Agency  (EEA),  Coordination  of  Information  on  the Environment (CORINE), which describes LULC (and partly land use) according to a nomenclature of  44  classes  organized  hierarchically  in  different  levels. Moreover, due to the geographical location of the study area in Tropics, the CORINE land classes adopted was not entirely applied. It is a land cover class which is more applicable for Temperate and cold Polar Regions. In order to solve this challenge,

After a supervised classification has been carried out, the land cover maps as well as post-classification algorithms are performed as given in figure 3 to 5 and table 4 to 6. The accuracy of a categorized map must be measured and compared to the referenced data using an error matrix. The accuracy measurement in this analysis was rendered using the original mosaic image for the years 1973, 2000 and 2010.

3.1 User’s Accuracy

The findings of user’s accuracy show that in 1973, the highest class accuracy (96%) for forest land and bare land and the minimum accuracy was estimated for grass lands (68.42%) (Table 4). Class precision ranged from 76.47% to 94.2% in 2000, although it ranged from 82.35% to 95.5% in 2010. The lowest, values of class accuracy were reported due to spectral similarity with other classes. As seen in table 4, 5 and 6, the lowest value was observed for grass land during the time of 1973 (68%) and the rural settlements for the years 2000 and 2010 with 76.47% and 82.35%, respectively. In addition, the resolution of Landsat data may have an impact on the classification of the image.

Findings of Markov Chain Model simulation are based on a transfer probability matrix of land-use shifts from t₁ (1973) to t₂ (2000). For another time, this has been the basis for projection (2010). Figure 9 (c) and (d) are the real and virtual maps of Adama City for the year 2010. This was predicted as the historical transition cycles of 1973-2000 and 2000-2010 may not be the same in the Markov Chain analysis. As seen in figure 9c, a significant part of the crop land has been turned into a rising field. While there has been a great deal of expansion of built-up areas, the transfer into other land cover groups has been constrained (Table 12).

In the most optimal case, the change of area should be 0. The more the deviation, the less sophisticated is the model. In terms of area, it is clear that urban areas, rural settlements, crop land and bare land at almost the same. But forestland has decreased while grass land has increased. It means forest land and grass land show unexpected results in the case of quantitative analysis. The statistics presented in the validation output allow the user to distinguish the error of quantification from the error of position. The quantification error occurs where the number of cells in a category on a map varies from the number of cells in that category on the other map. Place error arises where the location of the division on one map is different from the location of the category on the other map. Thus, in addition to calculating the conventional standard Kappa index of agreement (KIA, also denoted K standard in cross table), validate offers more statistics: Kappa for no information (K no), Kappa for location (K location), Kappa for quantity (K quantity), value of perfect information of location (VPIL) and value of perfect information of quantity (VPIQ). All of these statistics are linear functions of the nine points given in the validation output (Pontius, 2000) (Table 13).

The result of Kappa Index of Agreement (KIA) was very good for overall kappa for urban areas, crop land, and bare land moderate for rural settlement, fair for forest land and grass land. This variation occurs due to the temporal variations of satellite data. Kappa value <0.2 is poor, 0.21-0.4 is fair, 0.41-0.6 is moderate, 0.61-0.8 is good and 0.81-1 is very good (Landis and Koch, 1977). In comparing, the map of reality to the alternative map: K no indicates the overall agreement. K location indicates the extent to which the two maps agree in terms of the location of each category given the specified quantities. VPIL shows the additional percentage right to be achieved if the alternate map was