The application of remote sensing in environmental monitoring and mapping has become very popular over the last few decades. According to J.B Campbell, remote sensing is the practise of deriving information about the earth’s land and water surface using images acquired from an overhead perspective, using electromagnetic radiations in one or more regions of the electromagnetic spectrum, reflected or emitted from the earth’s surface. Majority of the sensors used for remote sensing are passive sensors which means they use thermal energy emitted by the surface or solar energy reflected by the surface. The energy is captured and stored by the sensor as digital numbers which is used to build up an image of the region being captured. In remote sensing, the term resolution is used to characterize different systems and sensors. A pixel represents the smallest area unit in a digital image. Spatial resolution refers to the size of a pixel in a digital image. The Temporal resolution on the other hand describes the revisiting frequency of a satellite for a specific location. Spectral resolution of a sensor specifies the number of spectral bands in which the sensor can collect emitted or reflected radiance.The history of modern remote sensing application dates back to the early 20th century. World war I marked the beginning of the acquisition of aerial photography on a regular basis (jb camp). During the first and second world war, aerial photographs captured by cameras mounted on aeroplanes were used for surveying reconnaissance as well as military surveillance (Colwell,1983). Technological advancements made in the first and second World Wars were continued and improved in the subsequent years especially during the Cold war era. The United States for example launched the CORONA strategic reconnaissance satellites to gather reliable information about the soviet bloc (Dwayne et al., 1988). In the subsequent years, remote sensing applications were broadened from just military application to other areas and fields. In 1960, The first satellite designed for climatological and meteorological observations, the TIROS-1 was launched.
The first civilian satellite programme, the Landsat missions was launched in 1972. This paved the way for the application of remote sensing in several fields by providing free and continuous data. The programme is jointly managed by The United States Geological Survey and NASA. Landsat 1(which was then called Earth Resource Technology Satellite) became operational in 1972. The Landsat 1 as well as the subsequent Landsat 2 and 3 carried the Multi spectral scanner(MSS) sensor. Landsat 1-3 had 4 spectral bands with a spatial resolution of 60 meters The success of these satellites prompted further developments. Landsat 4 and 5 were later launched and had the Thematic Mapper (TM) on board. The spatial resolution of these satellites was 30 meters and had 6 bands. The Landsat 7 carries the Enhanced Thematic Mapper (ETM+) and has 7 spectral bands and a spatial resolution on 30 meters. In February 2013, the Landsat 8 satellite which has on board the Operational Land Imager (OLI) and Thermal Infrared sensor (TIRS) became operational. Landsat 8 images consist of 9 spectral bands and the same resolution as Landsat 7. Landsat 1-3 have a temporal resolution of 18 days while all the subsequent satellites have a temporal resolution of 16 days. The success of the Landsat programme set the pace for other earth observing satellites including Systeme Pour l'Observation de Ia Terre (SPOT) of France, Indian Remote Sensing Satellite (IRS) of India and Environmental Satellite (ENVISAT) operated by the European space agency. The last two decades has also seen the launch of satellites such as GeoEye, Quickbird, WolrdView and IKONOS by private companies. These satellites are able to capture remote sensing data at high resolutions with some new sensors having a spatial resolution of 31 centimetres and a temporal resolution of one day.
Advancements in Remote sensing have made it possible to conduct several studies and research especially with regards to environmental monitoring and natural resource management. One theme which has become very common in remote sensing research is land use/land cover change detection and analysis. Although the term land cover and land use are very similar and sometimes used interchangeably, they actual mean different things. Land cover refers to the biophysical feature on the surface of the earth which may include vegetation, water, bare soil, artificial structures etc. Land use on the other hand describes how a particular land is being used and influenced by human activities such as agriculture, recreation, transportation and so on. Mapping and monitoring of land use/land cover are very important to our understanding of change mechanism and modelling the impact of change on the ecosystem (Chen et al.,2003). It has been noted that, land use changes are cumulatively transforming land cover on a global scale especially in the tropical regions (turner et al.,1994). They further noted that, most of these areas however lack a comprehensive data on the nature, extent and location on such changes. Land use/land cover change analysis has been successfully used to study deforestation, crop land loss, erosion, water quality change and so on. Studies conducted by some researchers indicate land cover and land use change have become important in several fields including ecology(Weng,2001), agriculture, forestry and hydrology (Li & Yeh,1998). Several techniques and models have been developed that use remote sensing data to detect land use/land cover change. Yuan et al. (1998) grouped the methods for change detection into pre-classification and post-classification techniques. Pre classification techniques include image differencing, image ratioing, vegetation indices and principal component analysis. Ridd & Liu (1998) argued that these techniques can be used to generate difference maps but are unable to produce change statistics. The post classification techniques on the other hand use classified images acquired at different times to produce difference maps and change statistics.
Application of remote sensing in monitoring and mapping urban environments has become very important over the last few decades mainly due to technological advancements and societal needs (Weng & Quattrochi, 2006). Mapping urban landscapes presents several challenges as noted by Ban et al. (2010) and Griffiths et al., 2010. Urban environments are mostly a mixture of different buildings, parking spaces, roads, bare soil, urban vegetation, water and so on. Each of these component have different spectral and radiometric properties. This makes it difficult to accurately map and classify urban landscapes especially when low spatial resolution remote sensing data are used. Several scholars have however been able to successfully use remote sensing data to monitor and map urban spaces in different parts of the world. Most of these studies were conducted using Landsat data as identified in the works of Yang et al., 2003; Lo and Choi, 2004; Lee and Lathrop, 2006; Furberg and Ban, 2012; Zhang et al., 2015. Yuan et al. (1998) were able to use images from the Landsat Thematic Mapper to monitor urban expansion in the Minnesota Metropolitan area. Kolehmainen and Ban (2008) assessed three different change detection techniques to map newly constructed areas in Stockholm from 1986 to 2004. In Ghana, early efforts to map urban expansion using remote sensing data was conducted by Møller and Yankson (1994). They assessed the suitability of using Landsat Thematic Mapper images for urban change studies as well as producing a land cover map of Accra. In their study, they were able to identify newly urbanized and emerging settlements in the Greater Accra Region such as Mallam-Gbawe, Madina, Dome and Ashaley Botwe. Atua and Fisher (2011) were also able to use remote sensing data to detect land cover change in the New Juabeng municipality. Appiah et al. (2015) analysed land use/land cover change in the Bosomtwe district of Ghana. The study indicated that urban areas and bare land had increased in size mainly at the expense of forest and open woodland between 1986 and 2014.
Urbanization
It is a well-known fact that the rate of urbanization has increased over the last few decades and studies suggest that the trend will continue for some time. According to the latest World Urbanization prospect, about 54% of the global population were living in urban centres as at 2014 and this percentage is projected to reach 66% in 2050. The report further indicated that 90% of the increase is expected to occur in Africa and Asia with Nigeria, India and China expected to make up 37% of the increase. Although urbanization has several economic and social benefits, it could also have adverse effects on the environment especially in countries were the rate of urbanization is not properly managed. Consequences of rapid urbanization include urban heat islands, increased waste generation, flooding, landslides, air and water pollution, loss of urban green spaces among others. If not well managed, urbanization could also cause social problems such as shortage of houses, development of slums, inadequate infrastructure and so on.
Urbanization in Greater Accra Region and GAMA
Ghana is fast urbanizing and the trend is expected to continue in the coming years. According to the 2010 Population and Housing census, about 50.9% of the population are living in urban centres. The proportion of people living in urban centres is however uneven across the country with some regions having a higher rate of urbanization. The Greater Accra region has the highest proportion of urban dwellers in the Country. As at 1960. About 72.6% of the population in the Greater Accra region were living in urban centres (GSS,2010). The percentage increased to 85.3 in 1970 and later to 83% in 1984. The latest population census which was held in 2010 puts the proportion of urban dwellers in the region at 90.5%. The economic hub of the region, The Greater Accra Metropolitan area located within the Greater Accra region has one of the highest population density among any density across the country. GAMA has almost reach its carrying capacity with its population spilling into other adjoining districts. Kasoa and Prampram located on the western and eastern sides of the region respectively have become important settlements absorbing much of the physical expansion in region. Peripheral areas such as Madina, Ashiaman and Afienya have also sprung up to life in the last couple of years as the city core continues to grow. In 1960, GAMA had a population of approximately 450,000 people. This figure increased to 1,922,898 people in 1984 and later reached 2,715,805 in 2000(Yankson and Bertrand,2012). The total population of GAMA as at 2010 was projected to be around 4 million people (Future proofing cities,2016) The Greater Accra metropolitan area is the most economically vibrant district in the region, encompassing Accra and Tema which are the economic nuclei of the region. As the regional capital of Ghana, the region attracts a lot of people from different parts of the country as well as other countries mainly due to availability of economic opportunities. The implementation of modernization strategies based on import substitution in the post-independence era provided the impetus for the rapid urbanization of the region. Part of these strategies included the creation of the port city of Tema and the establishment of industrial estates in Accra(Yankson,2012). The concentration of the head offices of major banking and commercial companies, manufacturing and industrial firms as well as other political activities further strengthened the economic status of the region. Between 1994 and 1999, the Greater Accra Metropolitan area alone attracted 80% of all foreign direct investment in Ghana.
Land use change modelling and prediction
Land use change has become a major issue over the last few decades as different land use types compete with each other for the same space. Thus, it has become imperative for urban planners to know how the urban space and its functions are going to change in the future as well as the potential impact of urban expansion on the environment. According to Pijanowski et al., (2009) urban expansion is determined by three main factors namely physical geography of the area, the demand for households by people in the area and policies that determine land use and spatial interaction. Several models have been designed to model urban expansion. However, one problem is the lack of spatial explicit data. The incorporation of Remote sensing and GIS together with stochastic models provides a predicting tool to monitor land use change over time. One model which is commonly used for predicting land use change is Artificial Neural Network (ANN). Li & Yeh (2002) noted that although Markov chain can be used to predict land use change, it is unable to produce predictions with spatial details. The Multi-Layer Perceptron(MLP) is a type of ANN which has widely been used to predict land use change. Pijanowski et al.,(2009) used Artificial Neural Network to project urban expansion in the Tehran Metropolitan area. The Multi-Layer Perceptron is a system of interconnected neurons or nodes. The system consists of input and output nodes with several hidden nodes between them. The output of a node is scaled by connecting weight and fed forward to be an input to the nodes in the next layer of the network. (Gardner & Dorling, 1998). This process implies the direction of information in the system. The Multi-layer perceptron is therefore known as a feed-forward neural network.
Advancements in GIS technology has provided an effective tool for spatial and statistical analysis. GIS software such as ArcGIS consist of several tools which can be used to incorporate digital images, spatial data as well as spatial and statistical models. The Land Use Change modeler extension for ArcGIS developed by Clarks lab provide a set of statistical tools and models which can be used to perform land use change analysis and perdiction. The MLP algorithm in this software allows the incorporation of explanatory variables referred to as drivers. The MLP algorithm uses the drivers to train on and to develop a multivariate function that can predict the potential for transition based on the values at any location (TerrSet tutorial). Several drivers have been considered in analysing urban expansion. Studies have shown that urban centres tend to grow outward with places closer to urbanized areas more likely to transition into urban landscape than places far away. Proximity to urban centres increases job opportunities and offers a higher level of medical treatment and education. According to Liao & Wei (2012), Transportation is one of the most important factors behind urban development. The construction of new roads and the enlargement of existing roads influences the location of new houses as well as real estates (Rui & Ban,2011). In most countries, water bodies such as lakes, rivers and streams are important for transportation and recreational purposes. Schnaiberg et al., (2002) have noted that the distribution of water bodies influences the choice of residential locations. Tain et al., (2002) also noted that geophysical feature such as elevation and slope can impact the size, distribution and density of settlements thus influencing urban expansion. In their study, Pijanowski et al (2009) incorporated exclusionary zones in their analysis to avoid reserved areas, parks and wet lands from becoming urbanized.