1. Introduction
Mangroves are located in the tropical and sub-tropical countries primarily between 30° N and 30° S latitude. These coastal forests are distributed in the inter-tidal regions, and along river banks and lagoons. This ecosystem is comprised of plant families with specialized adaptations to live in the tidal environment. According to an estimate from 2000, mangrove forests accounted for less than 1% of total tropical forests in the world [
1]. Yet, these are one of the most productive and biologically complex ecosystems that store three to four times more carbon per equivalent area compared to tropical forests [
2]. Besides, mangrove forests provide protection to coastal communities from natural disasters, especially storm surge and small to moderate tsunamis [
3,
4]. However, due to increasing land competition for agriculture, aquaculture, tourism, and infrastructure development, these forests have declined from 18.8 million hectares in 1990 to 15.2 million hectares in 2005 [
5].
Official estimates for the year 2000 suggest that Sierra Leone had 105,300 hectares of mangroves, or roughly 0.007% of the global total [
5]. In this West African country, mangroves are an essential source of wood for the coastal communities and provide a number of indirect services such as fish breeding sites and coastal protection. Despite their undeniable benefits, mangrove forests are under increasing pressure due to urbanization and land reclamation on the flood plains, conversion to rice paddies and unsustainable exploitation for fuelwood and fish smoking. This fragile ecosystem is also sensitive to changing environmental conditions such as increased temperature and sea level rise, as well as water characteristics such as salinity, pollutants, and sedimentation [
6]. The combination of increasing human-induced and environmental stress may lead to unsustainable conditions for mangroves and ultimately their decline. A better understanding of the recent changes in mangrove extent and quality, human pressures, the impact of climate change as well as management practices and opportunities may help sustaining mangroves and their benefits for future generations. However, up-to-date, good quality in situ data about mangroves are not available in Sierra Leone, hindering assessment of changes and the design of sustainable management plans. Remote sensing can be an alternative tool in this context, given the availability of free satellite data dating from the 1970s at spatial and temporal scales suitable for landscape-level monitoring. Taking advantage of these satellite images, this study aims at assessing landscape-level changes in mangrove extents in Sierra Leone during 1990–2016.
Remote sensing tools and techniques have been widely used in mangrove mapping, and have rapidly evolved over the past decade [
7,
8,
9]. One of the most widely used satellite sensors for mangrove mapping is Landsat, due to its spatial and temporal coverage, and ease of accessibility. Many studies have used Landsat images, as well as optical imagery from SPOT, MODIS, ASTER, QuickBird, WorldView, and IKONOS, for quantifying mangrove extents and spatiotemporal changes across the globe [
1,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38,
39,
40]. These studies use a range of classification techniques and machine-learning algorithms such as unsupervised, supervised, hybrid, classification and regression tree (CART), support vector machine (SVM), object-oriented classification among others.
The currently available global mangrove estimates are either directly calculated from optical satellite data, national-level statistics, or are derived from other global datasets [
1,
5,
41,
42,
43]. Most of these studies provide only a snapshot for mangrove extents because of the massive scale of work involved in global mapping efforts. Besides, it is difficult to directly compare estimates from these studies, ranging from 12 to 20 million hectares (see [
5] for details) due to the differences in methods involved. Hamilton and Casey [
41] identify that there are notable differences in mangrove estimates as provided by these global studies, and hence developed a new global dataset, CGMFC-21, that provides annual mangrove estimates for 2000–2012 by compiling other existing datasets [
1,
42,
44] and using statistical techniques to predict estimates for 2013–2014. However, national/regional estimates should be derived from direct observation (either via satellite, or field visits), as these estimates are often used for revising national/regional policies with impacts on local livelihoods.
Many studies have reported spatiotemporal changes in mangroves from different parts of the world, especially the Sundarbans in India and Bangladesh, and West-Central Africa [
12,
13,
14,
25,
26,
31]. Past studies have reported widely variable single-date mangrove estimates for Sierra Leone, mostly owing to the different methodologies involved in these studies (
Table 1). For the year 2000 alone, the estimates range from 655.67 km
2 to 2917.01 km
2 [
41]. Even studies relying solely on Landsat images report a wide range of estimates [
1,
45], likely because different definitions of ‘mangrove forests’ were used in these studies. However, no study, to the best of our knowledge, has focused on long-term changes in Sierra Leone despite the importance of mangroves in providing coastal protection and livelihood opportunities in this vulnerable nation. Moreover, as one of the West African countries selected for the USAID-funded West Africa Biodiversity and Climate Change (WA BiCC) project, a detailed land cover change analysis over the past decades is now required for Sierra Leone in order to develop coastal conservation and climate resilience building activities.
In this study, we provide the first multi-year assessment (1990, 2000, 2010, and 2016) of spatial changes in the mangrove extents in Sierra Leone coastal landscape complex (SLCLC). Since one of the primary objectives of this study is to inform WA BiCC project for effective and sustainable coastal management, we use variable buffers (1 km, 2.5 km, and 5 km from the coastline) to identify potential ‘deforestation hotspots’ that might require immediate attention from policy-makers. The other objective is to develop a landscape monitoring method using freely available data that can be easily deployed for other WA BiCC countries. Hence we take advantage of the freely available Landsat images and recent advancement in the cloud computational techniques. We utilize Landsat-5, 7, 8 and Sentinel-2 images along with field data to quantify and interpret changes in mangrove covers in the SLCLC. As a requirement of the WA BiCC project, we specifically focus on four estuaries (Scarcies River Estuary, Sierra Leone River Estuary, Yawri Bay and Sherbro River Estuary) for addressing the following questions:
How has the mangrove extent changed in the SLCLC over the past 26 years? Is there net mangrove gain or loss?
Where did mangrove forests undergo the most changes—closer to the coastline or further away?
Are there spatial differences in mangrove changes, e.g., northern SLCLC vs. southern SLCLC?
4. Discussion and Conclusions
This study, based on a discrete classification of Landsat pixels, provides an estimate of 1526.42 km
2 mangrove extent for 2000 that is similar to the estimates provided by [
1]. While a continuous classification approach can, as argued by [
41], provide more accurate estimates, it is almost impossible to apply for historical land cover analyses in the absence of existing maps with known accuracy or extensive field data from earlier years. Therefore, it could not be applied in our study aimed at estimating the evolution of the mangrove extents in Sierra Leone. Our study relies solely on dominant spectral signature within a single pixel (30 m × 30 m) for a discrete classification, i.e., presence/absence of a certain class. While it is possible that our study overestimates mangrove coverage in the SLCLC because of the way discrete classification works, our approach allows us to provide important insights into decadal changes in spatial patterns in mangrove forests. Besides, our findings agree with those from the past studies that showed an increase in mangrove extents after 2000 [
5,
41].
As shown in
Figure 5 and
Figure 6b, all four focus areas witnessed an increase in mangrove extents in all three buffer zones between 2000 and 2010. However, only SLRE witnessed an increase in mangrove area post-2010, most probably as a result of reforestation efforts. Considering the overall changes during 1990–2016, mangroves have declined only in the farthest buffer in the Sherbro and Yawri Bay regions, pointing to the possible reclaiming of the land and expansion of agriculture by the communities living inland. The Scarcies region underwent the most extensive mangrove loss among all the areas (
Table 5), but in different locations for pre-2000 and post-2010 time periods (
Figure 6). In both cases the most probable cause is the conversion of mangroves to rice paddies. However, the rapid recovery of the mangroves over a period of 10 years between 2000 and 2010 in some areas is more indicative of degradation or ‘thinning out’ during the periods prior to 2000 rather than a complete deforestation. While causality is difficult to attribute, the decade of the 1990s was also a period of civil war, which may have heightened dependence on mangrove forests for fuel wood, charcoal production, and construction, and/or weakened conservation and protection measures. These findings, especially the overall patterns in change, can particularly benefit regional and/or national policy-makers in drafting coastal conservation policies.
It is important to note that it is challenging to distinguish between wetlands and sparse mangrove forests, because of the inherent limitations of optical remote sensing in terms of the spectral and spatial resolution, especially in tropical countries with frequent cloud cover. Hence, it is possible that a pixel with sparse mangrove cover will record a spectral signature of the underlying water that is dominant within that pixel and thus be classified as water/wetland. A validation check with high-resolution images from 2016 indeed confirms that the area along and surrounding the Scarcies river hosts young and sparsely distributed mangroves, rather than a complete clear-cut of dense mangrove forests. Even though high-resolution satellite images do not exist for earlier years for validation purposes, it is possible that mangrove forests in the Scarcies region witnessed degradation or ‘thinning out’ during 1990–2000 rather than a complete deforestation. Thus, the fluctuations in mangrove extents as shown in
Figure 6 possibly represent alternate thinning and reforestation. This is consistent with the high regeneration potential observed on field visits to the Scarcies region.
To be useful at large scale, satellite-based assessment of mangrove cover needs to be complemented by in-situ evaluation of the quality of the forests. The field trips during summer 2016 indicate that, despite degradation, the remaining mangroves in the Scarcies region show higher species diversity (relative to other regions in Sierra Leone), and high regeneration level; thus indicating human pressure on the forests, but also high regeneration potential should human pressures be lowered or better managed. This is further supported by the fluctuations in mangrove forest cover, from dense to sparse to dense, estimated in this study. The Sherbro area is on the opposite side of the spectrum, with lowest species diversity, highly dominated by Rhizophora racemosa. These mangroves are also the oldest among the four areas and have the lowest regeneration rates with little disturbance in the forests. These forests, while with high commercial potential, exhibit low adaptation potential to future, potentially altered climatic conditions. The SLRE has the smallest trees with the lowest basal area pointing to youngest forests, sign of past and current exploitation of the forest as well as recent reforestation efforts. The Yawri Bay has fewer adult trees but the highest number of seedlings, both showing signs of good potential for regeneration and sustainability.
Overall, despite a noticeable overall long term decrease of the mangrove cover in Sierra Leone a closer look at mangrove evolution shows a good potential for conservation if properly monitored and managed. Conserving the fluctuating mangroves in the SLCLC would require a carefully designed management plan involving current and alternative livelihood strategies, sustainable resource management schemes at local and national levels and sensitization of the populations about mangrove services and their value under climate change, beyond responding to communities’ immediate needs. Among the coastal populations, fishing communities are the ones that will be the most affected by the decline in mangroves, yet are also the ones who currently rely on mangroves the most for their livelihoods. These populations exhibit low livelihood diversification mostly revolving around fishing and fish processing with some contribution from agriculture and petty trade [
52]. The SLRE has slightly higher diversification of livelihood strategies, largely due to the proximity of Freetown and slightly higher grounds. In the Scarcies and Sherbro regions alternative or supplementary livelihood opportunities are quasi non-existent. In most of the communities visited during field trips, mangroves are also open-access and individuals can access ownership by ‘adding value’ to the land, which in most cases translates into clearing the mangroves for rice cultivation. Existing community-based management initiatives seem to be mostly geared towards short term economic benefits from lands covered by mangroves, rather than driven by holistic, long term plans that include a wide range of services provided to a wide range of populations that would insure sustainability of the resource and fishing livelihoods. While projects aimed at mangrove conservation could benefit from the existing management structures, significant change to the management goals would be required.