Mapping deforestation and urban expansion in Freetown, Sierra Leone, from pre- to post-war economic recovery
- 650 Downloads
Freetown, the capital of Sierra Leone has experienced vast land-cover changes over the past three decades. In Sierra Leone, however, availability of updated land-cover data is still a problem even for environmental managers. This study was therefore, conducted to provide up-to-date land-cover data for Freetown. Multi-temporal Landsat data at 1986, 2001, and 2015 were obtained, and a maximum likelihood supervised classification was employed. Eight land-cover classes or categories were recognized as follows: water, wetland, built-up, dense forest, sparse forest, grassland, barren, and mangrove. Land-cover changes were mapped via post-classification change detection. The persistence, gain, and loss of each land-cover class, and selected land conversions were also quantified. An overall classification accuracy of 87.3 % and a Kappa statistic of 0.85 were obtained for the 2015 map. From 1986 to 2015, water, built-up, grassland, and barren had net gains, whereas forests, wetlands, and mangrove had net loses. Conversion analyses among forests, grassland, and built-up show that built-up had targeted grassland and avoided forests. This study also revealed that, the overall land-cover change at 2001–2015 was higher (28.5 %) than that recorded at 1986–2001 (20.9 %). This is attributable to the population increase in Freetown and the high economic growth and infrastructural development recorded countrywide after the civil war. In view of the rapid land-cover change and its associated environmental impacts, this study recommends the enactment of policies that would strike a balance between urbanization and environmental sustainability in Freetown.
KeywordsFreetown Land-cover Change detection Civil war Population growth GDP growth
Land-cover is one of the most significant issues in global environmental change (Fan et al. 2007). An understanding of land-cover change would provide insights to the patterns of environmental change and the processes underlying them. With land-cover data, managers can be well informed about the nature of land-cover change and its associated impacts on the environment. Such data could provide a basis for the formulation of environmental policies (Misra and Balaji 2015).
Human influences are principal driving forces of land-cover change. An increase in population means more areas will be required for residential land-uses (Yang and Liu 2005; Cohen 2003). Malmir et al. 2015 noted that, urbanization, agricultural expansion, and degradation of natural resources due to migration and over population of cities are the main land-use changes in today’s world. A study by Dewidar 2004 on land-use/land-cover change in the northern Nile Delta of Egypt shows an increase in the residential area from 54.7 km2 in 1984 to 79.9 km2 in 1997. Similar results were recorded in the Nile Delta from 1985 to 2000 (Schlerf et al. 2007). Sun et al. 2009 noted an urbanization growth of 21.3 % in the Datong Watershed in China from 1977 to 2006 which they attributed mostly to rapid urban population growth and economic development. Several studies have explored the relationships between gross domestic product (GDP) and land-use/cover change (Fan et al. 2007; Long et al. 2007; Seto and Kaufmann 2003). A major driver of urban land-cover change is the investment in capital construction per capita, which is a proxy of foreign direct investment (Seto and Kaufmann 2003). Fan et al. 2007 attributed crop land loss to GDP and economic development in Guangzhou, southern China. In the current study, however, no specific GDP data for Freetown exists. We have utilized GDP data for Sierra Leone to make an inference on the possible impact of national GDP growth on land-cover change in Freetown.
Sierra Leone has experienced vast land-cover changes over the past three decades (Forkuor and Cofie 2011; Wilson and Wilson 2013). At the end of the civil war in 2002, Sierra Leone was opened to investments, and rapid infrastructural development was central to economic recovery. The population of Sierra Leone is also increasing rapidly, especially after the civil war (Statistics Sierra Leone 2016). Even though the impact of population and economic development on land-cover in Sierra Leone is very glaring, limited data exist and environmental reports continuously repeat land-cover estimates made by the Food and Agriculture Organization (FAO) in the 1970s.
Freetown is the political and business capital of Sierra Leone. The city undoubtedly experienced the most severe land-cover changes in the country over the past three decades. During the civil war (1991 to 2002), Freetown was much safer than most parts of the country, and hence, there was heavy influx of refugees into the city. In the post-war era (after 2002), Freetown also seems to offer more business and livelihood opportunities which further increases the city’s population.
Land-cover change in Freetown and even for the rest of Sierra Leone has been poorly studied. The most recent land-cover study in Freetown was conducted within the framework of the Cities Farming for the Future (CFF) program (Forkuor and Cofie 2011). In their research, Landsat data at 1974, 1986, and 2000 were employed to investigate land-cover dynamics and their effects on urban agriculture in Freetown. However, considering the rapid infrastructural development and the unprecedented population growth after 2002, such data are now obsolete to reveal the current status of land-cover, or to be used as complementary data for land-use planning in Freetown.
Environmentalists and planners in Sierra Leone have expressed their concerns over the changing environment. However, contemporary statistical evidence to back up this much talked about land degradation is grossly inadequate. There is therefore, an urgent need to conduct more research on land-cover change in the country. Data obtained from such studies could be presented to policy markers for efficient land resource planning and design of environmental management policies.
In view of the aforementioned, this study was conducted to analyze land-cover change in Freetown over the past three decades (1986 to 2015). The study provides an analysis of the potential impacts of socioeconomic factors (population growth and economic development) on land-cover change in Freetown. Multi-temporal Landsat data at 1986, 2001, and 2015 have been used in this study. The primary objectives include the following: the proposition of a simplified land-cover classification scheme for Freetown; the generation of land-cover maps for Freetown at 1986, 2001, and 2015; and the detection of land-cover change and the dominant categorical transitions between time intervals in the light of rapid population growth and infrastructural development in post-conflict Freetown.
Results obtained from this study are expected to fill the land-cover data gaps in Freetown from 1986 to 2015. This study is also envisaged to offer additional basis for applying remote sensing data in land-cover mapping by environmental planning and protection agencies in Sierra Leone. Results obtained could be useful to the current Freetown City Council’s “urban planning project.”
Sierra Leone is located in West Africa between latitudes 7° N and 10° N, and longitudes 10° W and 13° W. The country is bordered to the north and north-east by the Republic of Guinea, to the south and south-east by the Republic of Liberia, and to the west by the Atlantic Ocean. The total land area of the country is about 73,000 km2, and current population is estimated at 7.1 million.
Like the rest of Sierra Leone, Freetown has two major seasons. The rainy season spans from May to October, and the dry season from November to April. Temperatures range from 16 to 37 °C. Freetown experiences the highest rainfall in Sierra Leone, and annual totals exceed 3000 mm. The vegetation consists of forest mostly in highland areas, grasslands with scattered trees mostly in the lowlands, and mangrove vegetation or riverine forests along the coast and tidal estuaries.
The physiography of Freetown consists of the Peninsula Mountains, lowlands, and coastal plains. The topography is undulating in some areas between mountains and coastal plains. The elevation of the city ranges from 50 to 1000 m, and slopes exceed 50 % at some points (Dixey 1922).
The rapid population growth of Freetown as seen in Fig. 2 definitely has implications for the land-cover changes in the city. Residential land-use is apt to increase with increasing population. As land-use such as agriculture cannot compete with residential, they relocate to land with lower rents (Forkuor and Cofie 2011; Mundia and Aniya 2005). Thus, the relocation of displaced land-uses to areas with minimal rents, such as forest and wetlands, exacerbates the land change rate.
At the end of the war in 2002, Sierra Leone received huge amounts of foreign aid for economic recovery. The country was also opened to international investments, and building infrastructure was pivotal in the recovery process. From 2006, Sierra Leone has been one of the fastest growing economies in Africa, until the Ebola scourge in 2014. The construction of residential homes in Freetown has largely increased, apparently in response to the soaring population, high rents, and GDP growth. Moreover, people consider owing real estate in Freetown a very lucrative business.
The major economic activities of most migrants in Freetown are small-scale farming, logging, quarrying, sand mining, and artisanal fishing. These activities, being intensified by the soaring population, have combined with the urban sprawl to trigger marked environmental degradation in Freetown. Studies have noted the impact of deforestation on surface runoff (Haregeweyn et al. 2015). The recent disasters of flooding, landslides, and rock falls after spells of torrential rain in Freetown could provide food for thought for the increasing environmental degradation of the city.
Materials and methods
Three cloud-free Landsat images were collected as follows: Landsat 5 Thematic Mapper (TM5), Landsat 7 Enhanced Thematic Mapper (ETM+), and Landsat 8 Operational Land Imager (OLI), with acquisition dates of January 19, 1986; January 20, 2001; and January 3, 2015; respectively. The 30-m reflective bands of TM5 and ETM+ (bands 1, 2, 3, 4, 5, 7) and their corresponding spatial and spectral ranges of OLI (bands 2, 3, 4, 5, 6, 7) were used in this study. The first four bands in each of the categories above cover the visible near infrared (VNIR) and the last two bands cover the short wave infrared (SWIR) portion of the electromagnetic spectrum. Data were downloaded freely from the United States Geological Survey (USGS) Earth Explorer website, at a World Reference System (WRS) path/row of 202:054. Data were collected at near anniversary dates to minimize the seasonal effects which may affect land-cover change detection (Munyati 2000). January also marks the mid-dry season in Sierra Leone during which skies are clearer and most satellite images are of good quality. All images were projected to the Universal Transverse Mercator (UTM) zone 28 north and in the World Geodetic System (WGS) 84 datum.
A field campaign was conducted from March 4 to March 13, 2015. The period from January to March corresponds to the mid-dry season in Sierra Leone during which changes in vegetation and water bodies are most minimal. Hence, the collection of Landsat data and validation data at different months could have little or no impact on the validity of accuracy assessment. The field investigation was conducted to provide validation data for accuracy assessment of the 2015 land-cover map. A total of 535 points from the various land-cover types were collected using Global Positioning Systems (GPS) with an average positional accuracy of 4.1 m. These points, with each labeled to its corresponding land-cover, were only used for accuracy assessment of the 2015 map.
Landsat data used in this study are all level 1T products, meaning the images have already been corrected for terrain and geometric distortions by the data providers to an acceptable accuracy. However, as remote sensing data are generally prone to unsystematic geo-location errors, there still exists a need for geometric correction (Dewan and Yamaguchi 2009). We therefore used 35 points from the validation data, located at easily identifiable features on Landsat imagery for geometric correction. A root mean square error (RMSE) of 0.42 (<0.5 pixel) was achieved. Relative atmospheric correction was then implemented using the fast line-of-sight atmospheric analysis of spectral hypercubes (FLAASH) module available in the Environment for Visualizing Images (ENVI) software (Owojori and Xie 2005). The VNIR/SWIR bands and a spatial subset corresponding to the Freetown municipality were derived from each of the Landsat images.
Land-cover classification scheme
Outline of the proposed land-cover classification scheme for Freetown
Includes rivers, ponds, lakes, sea, continental shelf or coastal waters
Includes streams, swamps, and intertidal coastal and estuarine areas
Includes residential, industrial, tarmac and other impervious surfaces
Dense forest (DF)
Includes evergreen forest, mostly of tall, hard wood or deciduous trees
Sparse forest (SF)
Includes degraded forest with reduced tree height, canopy and density
Includes areas mostly of grass, shrubs, scattered trees, and annual crops
Includes bare earth, rocks, excavated surfaces, and sandy beaches
Includes coastal and riverine forests, mostly of the Rhizophora species
The maximum likelihood supervised classification algorithm (Soffianian and Madanian 2015) was used in this study. This classification approach has been proven to obtain good results from remotely sensed images if each class has a Gaussian distribution (Bolstad and Lillesand 1991). Selection of areas for training was guided by our knowledge of the study area. Training areas for classification were spread throughout the images to ensure an even representation of the spectral signatures of each land-cover class (Foody and Mathur 2004). At least 17 training areas were collected for each land-cover in each of the multi-temporal images. Each training polygon has over 50 pixels to ensure statistical validity. The training data were collected independently of the validation data via manual on-screen digitization of land-cover classes from the raw images.
Classification accuracy assessment is necessary as it shows the agreement between the classified map and reference data (Lillesand et al. 2014; Yousefi et al. 2015). However, we could not find good ground-truth or reference data for 1986 and 2001. Therefore, accuracy assessment was only performed for the 2015 land-cover map based on reference data obtained from field work in March 2015. The 535 GPS (ground-truth data) points were coded to their respective land-cover types. These validation data were overlain on the 2015 land-cover map, and a classification error matrix (contingency table) was generated. From this error matrix, the producer’s accuracy, user’s accuracy, conditional Kappa for each land-cover type, and the overall classification accuracy and Kappa statistic (index of agreement) of the 2015 land-cover map of Freetown, were all reported.
In the absence of an accuracy assessment for the 1986 and 2001 maps, the authors assume that the 2015 classification accuracy may not have deviated too far from those that would have been achieved for the historical maps if reference data were available. This assumption is based on the premise that the land-cover categories were persistent across the time points, the images were collected at near anniversary dates, and the same classification methods were used for all images.
Post-classification change detection (Yuan et al. 2005) was used to quantify land-cover changes in Freetown over the past three decades. Change detection involves quantifying land changes across a time period. The persistence, gain, and loss of each land-cover, and the overall change at each time interval were computed. The net change of each land-cover and selected conversion categories between 1986 and 2015 were analyzed. Furthermore, some basic intensity analyses of land-cover change (Aldwaik and Pontius 2012; Huang et al. 2012) were also applied in our study.
Results and discussion
Land-cover maps of Freetown
Accuracy assessment report
Accuracy assessment report (error matrix) for the 2015 land-cover map of Freetown
From the error (confusion) matrix in Table 2, an overall classification accuracy of 87.3 % and an overall Kappa statistic of 0.85 were achieved for the 2015 land-cover map of Freetown.
Analyses of land-cover change
Statistics of land-cover change in Freetown from 1986 to 2015
Time points and area (km2)
Category interval change (%)
It can be seen in Fig. 7 that, across the three time points, the land area covered by water, wetland, sparse forest and grassland have been fluctuating. For instance, water declined at 2001 and increased at 2015. Wetland increased at 2001 and decreased at 2015. Sparse forest increased at 2001 and decreased at 2015. Grassland increased at 2001 and decreased at 2015. On the other hand, the land categories of built-up and barren have constantly increased, whereas the land categories of dense forest and mangrove have constantly been on the decline. This can be ascribed to the heavy pressure of humans on dense forest and mangroves. Dense forest has been on constant degradation to sparse forest, grassland, and to a lesser extent, built-up. Also depleted are the mangroves. They are a vital source of fuel wood for homes and industrial fish smoking.
From Fig. 8, at least 60 % of the area of each land-cover remains unchanged between 1986 and 2001. The most persistent land-cover types were water, wetland, built-up, and dense forest. Wetland, sparse forest, grassland, and barren recorded the highest percentage gains (>40 %). Sparse forest, grassland, barren, and mangrove recorded the highest percentage loses (>35 %). It is also observed that sparse forest, grassland, and barren recorded both significant gains and significant loses. This phenomenon is apparently due to the following: sparse forest actively gaining from dense forest, and actively losing to grassland; grassland actively gaining from forests, and actively losing to barren and built-up; and barren actively gaining from grassland, and actively losing to built-up. Water, dense forest, and mangrove were net losers as their losses were higher than their gains. The remaining land-cover categories recorded more gains than loses.
In Fig. 9, only water, built-up, dense forest, and mangrove recorded persistence of up to 60 %. The remaining categories had below 45 % persistence. Built-up, sparse forest, grassland, and barren are the fastest gaining categories (≥40 %). Wetland, sparse forest, grassland, and barren are the fastest losing categories (>40 %). Dense forest also recorded a substantial loss greater than 35 %. As observed at 1986–2001, sparse forest, grassland, and barren were also both significant gainers and significant losers at the 2001–2015 interval. This is equally due to sparse forest actively gaining form dense forest and actively losing to grassland, grassland actively gaining from sparse forest and actively losing to barren and built-up, and barren actively gaining from grassland and actively losing to built-up. Between 2001 and 2015, only water, built-up, and barren were net gainers; the rest of the land-cover categories were recorded as net losers.
Several land conversions have occurred over the past three decades in Freetown. Due to the existence of a complex variety of land change trajectories, we have selected only a few conversion categories. These land conversions have been analyzed for the two time intervals (1986–2001 and 2001–2015). Special attention was paid to forests and built-up as these land categories seem to be most influenced by the rapid population growth and infrastructural development in Freetown.
Statistics of selected land-cover conversions in Freetown
Selected conversion types
Conversion statistics at time intervals
In Fig. 10, except for the conversion of dense forest (DF) to sparse forest (SF), all other conversion types were more intensive at 2001–2015, a further proof of the interval’s larger overall change. In order to define which land conversions have been active or dormant relative to uniform intensity, a threshold value was employed. This threshold was calculated by averaging the conversions between the two intervals. A threshold value of 0.7 km2 was obtained (shown as green line in Fig. 10). A conversion at a given time interval is considered dormant when it falls below the threshold and active when it extends beyond the threshold. Based on this threshold, it is observed that the conversion of dense forest to sparse forest was active at the first interval (1986–2001) and dormant at the second interval (2001–2015). Conversion of dense forest to grassland (GL) was active at both time intervals. Conversion of dense forest to built-up (BT) was both dormant at the two intervals. Conversion of sparse forest to grassland was dormant at the first time interval, and active at the second time interval. Conversion of sparse forest to built-up was dormant at both intervals. Conversion of grassland to built-up was active at both intervals.
In developing countries like Sierra Leone where the availability of finance is a major obstacle to the acquisition of land-cover data through detailed ground-based studies, satellite remote sensing presents a cost- and time-efficient alternative to not only provide contemporary land-cover data, but also reveal the dynamics of land-cover over a period of time. It is with this in mind, that we have conducted this study to provide updated land-cover data for Freetown, while quantifying apparent transitions over the past three decades with minimal finance, time, and human resources.
Reference to Fig. 2, the population of Freetown increased by 123.6 % between 1985 and 2015. The rate of population increase between 1985 and 2004 is similar to that between 2004 and 2015. However, land-cover change was more pronounced between 2001 and 2015. As population increases, the demand for residential land also increases. As built-up areas expand into areas previously used for other purposes, such displaced land-uses relocate to areas where competition is relatively less (such as forests), thereby intensifying the rate of land-cover change. Built-up areas increased by 101 % at 2001–2015, more than triple that recorded at 1986–2001 (32.5 %).
Another important driving factor of land-cover change in Freetown is the post-war economic recovery. Sierra Leone between 2006 and 2014 was rated as one of the fastest growing economies in Sub-Saharan Africa (GDP >7 %). During the war, the country had one of, if not the lowest GDP growth rates in the world. Even though the Freetown population has been on a constant increase, urban expansion became rapid after the end of the war. This urban expansion is a manifestation of the nation’s strides to economic recovery. The slower growth of built-up relative to population at 1986–2001 could be accounted for by the economic stagnation during the civil war (1991–2002), in contrast to the infrastructural development seen after the war. Forest degradation was most recoded during the war compared with a post-war era where urban expansion is dominant.
Even though a linear relationship between two variables does not necessarily indicate causality, it is quite evident that the population increase in Freetown has had a huge impact on land-cover. Migrant populations are mostly unskilled. They depend mainly on primary economic activities, and forests are one inevitable target of exploitation. Between 1986 and 2015, dense forest and sparse forest declined by 48 and 24.2 %, respectively. The mountainous forests of the Freetown Peninsula have become sources of timber and charcoal. Mangroves have also been tremendously depleted due to their increasing economic importance as sources of fuel wood and agricultural land. From 1986 to 2015, the area covered by mangroves declined by about 25.3 %. Wood from mangroves are used for energy in homes. The local artisanal fishery industry also uses wood from mangroves during preservation (smoking) of fish. It could therefore be said that when the growing population engages mainly in primary economic activities, forest resources become the inevitable targets of exploitation. Additionally, land values in Freetown have been increasing exponentially. Urban and peri-urban farming were normally practiced in the lowland grasslands. As residential and industrial demand for land increase, peasant farmers can no longer afford the rents on grasslands. Some have therefore, embarked on crop cultivation on mangrove swamps, thus, depleting mangrove ecologies. Mangroves have been known to be very effective in carbon sequestration, and in this era of global warming, their protection is of utmost essence.
This study laid more emphases on terrestrial land-cover change with respect to the degradation of forests and expansion of built-up areas. Our analyses show that, built-up expansion had mostly targeted grasslands, the class to which most forests have been lost. Thus, the major trajectory of terrestrial land change in Freetown follows: from forests to grassland; from grassland to built-up. Built-up areas are expanding exponentially at a huge cost to the environment. Lowland areas of the city that were more suitable for residential and industrial activities have been exhausted, and people have resorted to the forested hills and marginal wetlands for housing and peasant farming.
The high rate of forest degradation on the mountains of Freetown could provide clues as to the causes of recently experienced environmental disasters, prominent among which are increased gully erosion, landslides, mud flow, rock fall, and flooding. These have most often occurred after spells of torrential rain, claiming many lives and property. This study therefore, recommends the enactment of regulations that strike a balance between urbanization and environmental quality. The remnant forest resources must be conserved, and government could also reduce urbanization trends through rural empowerment programs that optimize the livelihoods of potential migrants.
This study has attempted to fill some gaps in land-cover data in Freetown, at least from 1986 to 2015. The results may be used as statistical evidence by those advocating for urban planning and adoption of stricter environmental protection measures. Data presented here may be essential for updating the country’s environmental reports such as “the annual state of the environment report.”
This study was financially supported by the Sierra Leone Agricultural Research Institute (SLARI) within the institutional research priority area of: “appropriate mechanisms for managing, sharing and up-scaling environmental information and technologies established and operationalized.” The authors wish to express their sincere thanks and appreciation to two anonymous reviewers whose comments and advice were useful in shaping the contents and structure of this manuscript.
- A. Owojori, & H Xie (2005). Landsat image-based LULC changes of San Antonio, Texas using advanced atmospheric correction and object-oriented image analysis approaches. In 5th International symposium on remote sensing of urban areas, Tempe, AZ.Google Scholar
- Bolstad, P. V., & Lillesand, T. M. (1991). Rapid maximum likelihood classification. Photogrammetric Engineering and Remote Sensing, 57(1), 67–74.Google Scholar
- F Dixey (1922). The physiography of Sierra Leone. Geographical Journal, 41–61.Google Scholar
- Haregeweyn, N., Tesfaye, S., Tsunekawa, A., Tsubo, M., Meshesha, D. T., Adgo, E., & Elias, A. (2015). Dynamics of land use and land cover and its effects on hydrologic responses: case study of the Gilgel Tekeze catchment in the highlands of northern Ethiopia. Environmental Monitoring and Assessment, 187(1), 1–14.CrossRefGoogle Scholar
- J R Anderson (1976). A land use and land cover classification system for use with remote sensor data (Vol. 964). US Government Printing Office.Google Scholar
- Lillesand, T., Kiefer, R. W., & J. Chipman (2014). Remote sensing and image interpretation. John Wiley & Sons.Google Scholar
- Malmir, M., Zarkesh, M. M. K., Monavari, S. M., Jozi, S. A., & Sharifi, E. (2015). Urban development change detection based on multi-temporal satellite images as a fast tracking approach—a case study of Ahwaz County, southwestern Iran. Environmental Monitoring and Assessment, 187(3), 1–10.CrossRefGoogle Scholar
- Schlerf, M., Borens, S., Seeling, S., & J Hill (2007). Land use and land cover change analysis in Rhineland-Palatinate based on multi-temporal image classification. Mitteilungen aus der Forschungsanstalt fur Waldokologie und Forstwirtschaft Rheinland-Pfalz, pp., 237, 245.Google Scholar
- Sierra Leone population and housing census; provisional results, Statistics Sierra Leone, 2016.Google Scholar
- World Bank Report 2014 http://www.tradingeconomics.com/sierra-leone/gdp