Printer Friendly

Review and combination of recent remote sensing based products for forest cover change assessments in Cameroon/Analyse et combinaison de produits derives de la teledetection pour l'evaluation des changements du couvert forestier au Cameroun/Revision y combinacion de productos recientes basados en teledeteccion para la evaluacion del cambio en la cubierta forestal en Camerun.


Central African forests provide ecosystem services in various economic, social, climatic and ecological domains. Essential roles played by forests are central to multilateral environment agreements such as the United Nations Framework Convention on Climate Change (UNFCCC) (Mayaux et al. 2013; Desclee et al. 2014). In developing countries specifically, initiatives for Reducing Emissions from Deforestation and forest Degradation (REDD+) are being currently implemented and require countries to provide accurate information about their forest reference emission levels (UNFCCC 2014).

To assess the state of forest cover in tropical countries, earth observation data are principally used thanks to their synoptic and frequent views (Mayaux et al. 2013). Nowadays, the assessment of forest cover change at regional scales, e.g. over Central Africa, is realised with satellite images at 2030 m spatial resolution (provided by sensors like Landsat, SPOT, DMC). To provide consistent statistics at the globalscale, sample-based studies have been privileged (FAO and JRC 2012; Achard et al. 2014). Until recently, wall to wall global-scale studies were using coarser spatial resolution (Hansen and DeFries 2004). A change in US policy in 2008, giving free access to all archived Landsat imagery, facilitated in the last years the development of methodologies using wall to wall Landsat imagery to estimate deforestation (Hansen et al. 2013; Ickowitz et al. 2015).

The objective of this paper is to evaluate the potential of existing remote sensing based products to present an overview of the recent dynamics of forest cover changes in Cameroon and to place it in the regional context of the Congo Basin.

Forest mapping initiatives exist over the territory of the Congo Basin, however, they often present maps for a limited area or a single country and at different epocs (Desclee et al. 2014). In Cameroon, local-scale remote sensing studies have been realised such as the GSE-FM project which produced estimation of forest change in the Eastern region of Cameroon for the periods 1990-2000 and 2000-2005 based on Landsat and DMC satellite data (Hirschmugl et al. 2014). But at the scale of the Congo Basin, only two remote sensing products can be considered to provide consistent estimates of forest cover changes at both the national scale of Cameroon and the regional scale. The first one is the tropical forest changes assessment of the European Commission's Joint Research Centre (JRC) based on the analysis of a systematic sample of sites at three baseline years (1990, 2000 and 2010) (Mayaux et al. 2013; Achard et al. 2014). The second is the wall-towall Global Forest Change (GFC) product from year 2000 to 2012 delivered by Hansen et al. (2013).

While both JRC and GFC products rely mainly on Landsat imagery, their comparison is not straightforward as different methodologies and class definitions have been used. However, their respective specificities can be exploited to improve the description of the forest change dynamics in Cameroon and the Congo Basin. The possibility to derive rates of forest cover changes from 1990 to 2012 in Cameroon combining the JRC and GFC products will be explored. First, the products will be compared in term of forest cover changes statistics for the Congo Basin, highlighting their agreements and discrepancies. Second, a closer analysis of the spatial and temporal pattern of deforestation in Cameroon will be carried out. The potential and limitations of the current approaches to monitor known drivers of forest changes in Cameroon will be discussed.


The focus of the paper is set on forest cover changes occurring in the humid domain of the Congo Basin delimited by the Guineo-Congolian vegetation class of White (1983). According to a recent regional map (Verhegghen et al. 2012), the dense moist forest class covers an area of 19.12 million ha in Cameroon. This represents 11% of the total humid forest spreading in Cameroon, Republic of Congo (RoC), Gabon, Central African Republic (CAR), Equatorial Guinea and Republic Democratic of Congo (DRC). The majority (62%) of the total Congo Basin dense humid forest is to be found in DRC. On the other end CAR and Eq. Guinea present a very limited extent of dense humid forest (3% and 1% respectively). With respectively 12% and 11%, Gabon and RoC present a dense humid forest cover similar to Cameroon. In Cameroon, the dense humid forest covers more than 40% of the territory. Dense humid forests are mainly located in the southern part of the country while dry forests and woodlands are found in the central and northern parts (Verhegghen et al. 2012).

The JRC and the GFC products are the only recent remote sensing derived products which are consistent at the regional scale. The JRC study covers the whole tropical belt, and is a component of the Remote Sensing Survey for the FAO Forest Resources Assessment 2010 (FAO and JRC 2012). The JRC approach is based on a systematic sample (distributed along the geographical grid). Satellite imagery from medium resolution (circa 20-30 m) sensors (Landsat TM or ETM+, SPOT, DMC) was acquired for target years of 1990, 2000 and 2010 to cover the sample units of 10 by 10 km size. For the humid domain, a regular sampling scheme of half degree (longitude --latitude) was chosen. The raw satellite data were segmented to a minimum mapping unit of 5 ha and classified for the three target years. The legend definition complies with the FAO FRA requirements with the following land cover classes: 'tree cover (at least 70% tree cover portion in segment)', 'tree cover mosaic (30-70% tree cover portion)', 'other wooded land (at least 70% shrubs, forest regrowth)', 'other land' and 'water'. For each sample unit land cover transition matrices are produced for the two respective periods (1990-2000 & 2000-2010). Deforestation was defined as the conversion of tree cover and half of the tree cover mosaic into one of the other land-cover classes: reforestation and/or afforestation were defined as the opposite (Mayaux et al. 2013). These sample estimates are extrapolated statistically to produce regional estimates using appropriate sampling weights (to account for sampling intensity differences by latitude). Linear temporal extrapolations are used at sample unit level to harmonise estimates to the target dates (mid-June for each target year). Cloud coverage and limitations in image quality prevented the processing of all sample sites. For Cameroon, after a quality check of available satellite images for each of the three baseline years, only 45 sample units were covered with good quality imagery and further analysed (out of a total of 86 potential sample units in the humid forest domain of Cameroon) (Desclee et al. 2014). For the present study, the JRC dataset is used to derive rate of deforestation and reforestation for each country of the Congo Basin for the period 1990-2000 and 2000-2010.

The GFC product of Hansen et al. (2013) consists in a baseline tree cover percent product for the year 2000 and annual maps of global tree cover loss and gain for the period 2000-2012. This wall-to-wall product is derived from automated processing of Landsat imagery at 30m resolution. Forest loss was defined as a stand-replacement disturbance and forest gain as the inverse of loss, or the establishment of tree canopy from a non-forest state. Results for tree cover loss were disaggregated by percent tree cover strata (0-25%, 25-50%, 50-75%, 75-100%) and by year. Forest gain is only available for the overall period 2000-2012. The GFC product allows to derive gross tree cover loss estimates per year. Net tree cover loss is only possible to compute for the full period 2000-2012 (by combining gain and loss for the full period). Estimates of the tree cover extent in year 2000 are also divided in the four tree cover strata (defined hereabove) that are used to derive estimates of tree cover losses (areas and rates) per countries.

However, the baseline tree cover percent product was not found to be the most adequate definition for forest in our study. For instance, when considering the 50-75 and 75100% tree cover strata (considered in Hansen et al. (2013) for the forest loss and gain estimation) it appears that they do not match well the classes of humid forest defined by the regional map of vegetation from Verhegghen et al. (2012) (Figure 1). There is a better spatial agreement between the forest classes from Verhegghen et al. (2012) and a 70-100% tree cover stratum over Central Africa (dark green in Figure 1c). However, a 70-100% tree cover stratum includes areas that are not forest but are classified as Rural Complex, and correspond to a mixture of shifting cultivation and young tree regrowth (yellow areas in Figure 1c). Also, a 70% tree cover threshold is missing some areas of dense forest along the Atlantic Coast. The persistent heavy cloud cover along the Atlantic coast is affecting the GFC tree cover percent product. Cloud-affected imagery are automatically processed and in some areas, too few cloud free images are available to reliably represent the tree cover in 2000. In Cameroon, the GFC tree cover product exhibits artefacts unrelated to actual variation in the tree cover percentage.

Delivered in separated layers, the GFC product remains flexible in its definition of forest baseline. For this paper, the baseline forest cover is defined as the area that corresponds to the five dense humid forest class of Verhegghen et al. (2012) (Dense moist forest, Submontane forest, Mountain forest, Edaphic forest, Mangrove) or any pixel that presents a GFC tree cover percentage higher than 70% and intersect with mixed forest classes of the Verhegghen et al. (2012) map (Forest-Savanna Mosaic, Closed to Open Deciduous Woodland and Savanna Woodland-Tree Savanna). To take into account the specificity of the rural complex, a class corresponding to any GFC pixel that presents a tree cover percentage higher than 50% and intersects with the Rural Complex class is created. Estimates in term of areas affected and rates for 2000-2012 of forest losses and gains from the GFC dataset are extracted for each country of the Congo Basin inside the forest and rural complex classes.

The estimates derived from GFC and JRC are compared for Cameroon and the Congo Basin. For Cameroon specifically, the spatial distribution of the forest losses over the 2000-2012 period is analysed using the spatial information provided by the GFC dataset. The spatial analysis of deforestation is further examined by crossing the GFC layer of forest loss with the land use information extracted from the WRI forest atlas of Cameroon (Mertens et al. 2012).


Cameroon within the Congo Basin perspective

The annual average rates of forest change for the periods 1990-2000 (JRC results), 2000-2010 (JRC results), and 2000-2012 (derived from the GFC product) are presented for Cameroon, RoC, Gabon, Equatorial Guinea, CAR, DRC and the whole area in Table 1. Annual rates derived from the GFC product were found very similar for the periods 2000-2010 and 2000-2012 and therefore only the estimates for the period 2000-2012 are presented in the rest of the paper to provide consistent information about net deforestation (forest gains being only provided for the full period).

The two datasets provide information about forest cover changes for different period of times, but annual rates can be compared for the 2000s decade (i.e. from year 2000 to year 2010 or 2012). For the 2000s, gross and net deforestation rates are very similar in both datasets for Cameroon and Gabon. For Cameroon, estimates of annual gross deforestation rates from JRC and GFC are very close: 0.08% and 0.07% respectively, as well as estimates annual reforestation rates: 0.02% and 0.01% respectively. For the RoC, the gross deforestation rate (0.07% +/- 0.02) is in the same order of magnitude that the forest loss rate estimated with the GFC product (0.05%) but the estimates don't agree on the reforestation estimates (null by JRC and 0.01% by GFC). Regarding DRC, the estimates of gross deforestation are higher in the JRC (0.2 +/- 0.04%) than in the GFC product (0.15%). The same issue as for RoC is observed for gross reforestation estimates (null by JRC and 0.03% by the GFC). The biggest divergences in the estimations are found for CAR and Equatorial Guinea. For CAR, the JRC annual rate of gross deforestation (0.06%) is half of the GFC-derived rate (0.12%). However, concerning reforestation rates (forest regrowth), very low rates (0.01%) are found for CAR in both datasets. Estimates for Equatorial Guinea are also diverging with rates for gross deforestation of 0.04% for JRC versus 0.08% for GFC and rates of reforestation of 0.05% for JRC versus 0.01% for GFC. Both CAR and Equatorial Guinea represent a limited proportion of forest cover within the region which can explain some of the discrepancies in the results, due to higher variance in estimates of rates over smaller areas.

These differences stated, some findings about forest change dynamics can be highlighted from the two datasets for the period 2000-2010/12. First, both datasets agree that the rate of gross deforestation for the DRC is the highest one. According to the JRC product, DRC presents the highest rates of annual gross deforestation for both decades, 1990-2000 and 2000-2010. At the other end, Gabon has the lowest rates of gross deforestation. The RoC, Cameroon and, Equatorial Guinea and CAR have rates that are in the same order of magnitude. The estimates of rates of forest cover changes for the Congo Basin are largely influenced by DRC component due to the large extent of DRC forests (62%). Also in terms of area deforested in 2000-2012 derived from the GFC product, DRC presents the highest value of forest loss. DRC is followed by Cameroon, RoC and Gabon (Table 2). The forest areas and forest area losses in Equatorial Guinea and CAR are much smaller than the four other countries of the region.

In our study, the rural complex class is representing 10% in term of area of the combined forest and rural complex classes over the Congo Basin. Presented for 2000-2012 in Table 1b, the rates of loss for the rural complex class are much higher than the rates in the forest class. Loss rates are about four to six times higher in the rural complex class than in the forest class for Cameroon, the RoC, DRC, Eq. Guinea. Extreme difference is found in Gabon with a rate of loss of 0.04% in the forest class and a rate of 0.36% in the rural complex class.

In Table 2, the loss identified on the GFC product and occurring in the area mapped as rural complex is presented in comparison to the loss in the area identified as forest in this study. From Table 2, the total area of loss in the rural complex class is of the same order of magnitude as the area of forest loss (2622 [10.sup.3] ha for the forest class and 2097 [10.sup.3] ha for the rural complex class). Compared to countries with a similar forest class area (RoC and Gabon), Cameroon has a higher proportion of rural complex class and area of loss within that class.

In the JRC dataset, a decrease of the forest loss rates is observed between the first decade (1990-2000) and the second decade (2000-2010) for all six countries. The annual gross deforestation rates for the entire Congo Basin remain relatively low over the two decades (0.19% and 0.14%). For Cameroon specifically, a decrease is observed in the annual rates of forest losses from 0.13% during the 1990s to 0.08% during the 2000s.

While the JRC estimates are only available as 10 year averages, the GFC dataset allows us to decompose the estimates for the period 2000-2012 into yearly estimates. For 2000-2012, while a decrease in the annual rates is observed in the GFC data from 2000 to 2004 when the minimal value is observed, a strong increase is seen from 2005 to 2010 (Figure 2). The year 2010 presents a particularly high rate of gross forest cover loss. The rates fall in subsequent years (2011 and 2012). The general decrease in annual deforestation rates during the early 2000s and upturn towards the end of the 2000-2010 decade observed in Cameroon is also visible for the different Congo Basin countries individually and when taking the whole Congo Basin into account.

Land cover classes transitions in Cameroon

Information about the type of land cover transitions occurring between 1990s and 2000s can be retrieved from the JRC dataset. Concerning the loss of Tree Cover (conversion to other land cover classes), the most important conversion observed is from 'Tree Cover' to 'Tree Cover Mosaics' for both 1990s and 2000s (representing respectively 74% and 60% of the total Tree Cover loss). A significant reconversion from 'Tree Cover mosaics' to 'Tree Cover' (representing 41% and 46% of the total Tree Cover gain for 1990s and 2000s respectively) is also observed highlighting the dynamics of shifting cultivation or abandoned land.

Spatial distribution of forest cover changes in Cameroon

An analysis of the spatial distribution of the forest losses is carried out using the GFC product. From the ten administrative regions of Cameroon provided by the WRI, five contain significant amounts of forest cover: Centre, East, Littoral, South and South West. While forest losses are identified in all of these five regions, annual deforestation rates vary from 0.17% in the Littoral region to 0.04% in the East region (Figure 3a). The Littoral region presents a rate of loss that exceeds the other four regions and the GFC average rate of the country (0.07%). This region, situated near the ocean, includes the city of Douala and has extensive agro-industrial plantations. Oil palm and rubber trees plantations are also found in the South West and South regions. The East region is more affected by logging and has a very low annual rate of tree cover loss. The dynamic is slightly different in the rural complex class. The rates of loss are higher than in the forest class with rates varying between 0.45% (Littoral region) and 0.26% (East region).

The main dynamics of forest cover changes can be identified in Cameroon by visual interpretation of the GFC product ranging from small scale disturbances to large agro-industrial installation (Figure 4). Major disturbances are observed in the vicinity of Douala, the largest city of the country (Figure 5a). Some of this loss is due to agro-industrial plantations and this loss is quite clearly detected in the north west and the south east of the city. There are also large areas of disturbances observed in the north of Douala and in the north west that are not easily attributed to a specific land use or deforestation process. Forest loss linked to the expansion in oil palm and rubber plantations is observed in the South West of the country in the coastal zone near Kribi.

Timber extraction is a driver of forest cover disturbances in Cameroon. Logging infrastructure (i.e. logging roads) is observed in the East region (Figure 5b). Commercial logging is mainly happening in the Forest Managements Units (FMUs). Created under the 1994 Forest code, FMUs are allocated by a competitive bidding process for a 15-year period and require a management plan (Mertens et al. 2007). The observed forest cover losses inside FMUs (forest gaps and logging roads) varies a lot between FMUs (Figure 3b). Higher annual rate of loss is found in units localised in the South West and the East regions. Most of FMUs present an annual rate of forest cover openings lower than the country annual rate of forest cover loss (0.06%) and many present an annual rate lower than 0.02%.

Agricultural expansion within the forests is visible at the south of lake Bankim along the Mbam river basin. It is also the cause of the forest loss observed around Mbitom nearby the National park of Mbam and Djerem (Figure 4c). Extensive small scale shifting cultivation occurs along the road network across the country (Figure 4d). This loss is often linked with the presence of small villages, probably due to rural agriculture and fuel wood collection. Besides forest loss observed around villages and smaller cities, urban expansion is visible around Douala and Yaounde.

Inside the study area, Cameroon has 24 protected areas (national park, wildlife reserve or wildlife sanctuaries) with a "classified" or "in classification" status (Mertens et al. 2012). According to the spatial distribution of forest loss mapped in the GFC product, some encroachment is observed in these protected areas. Among national parks, the highest forest loss rate (0.03%) is observed in the national park of Deng-Deng, situated in the north of the East region and entirely covered by dense moist forest. Encroachments of agriculture fields are visible at the border of the park (Figure 4e). As this park has only been created (classified) in 2010, the changes detected may have occurred before the classification.

Agricultural fields are visible in the vicinity of the future national park of Ndongere. The park, situated in the South West region, near the coast, is only in the process of being classified. The other national parks have annual forest loss rates below 0.01%.

Some wildlife reserves present significant forest loss rates, e.g. the wildlife reserve of Douala Edea with an annual forest loss rate of 0.08%. This rate is lower than the 0.14% of the Littoral region where it is situated but higher than the GFC country rate of 0.06%. Some encroachments are visible in the north east part of the park. According to the shapefile of the WRI the limits of an oil palm concessions actually superimpose that of the wildlife reserve.

The law in Cameroon defines two types of forest, i.e. permanent and non-permanent, and their respective ownership. The permanent forest domain is owned by the State and is designated to remain forested in the long term. It includes the production forests. The non-permanent forest domain is composed of agro-forestry areas and other forests (Mertens et al. 2001). By combining the Interactive Forestry Atlas of Cameroon of the WRI (Mertens et al. 2012) with the GFC forest loss mapped for the 2000-2012 period, we can observe the following distribution of the loss areas: 44% is happening in the Permanent Forest Estate (Table 3), 14% in the Nonpermanent forest domain with 5% of the loss occurring in Agro-industrial zone (oil palm and rubber trees). The impact of mining concessions could not be assessed as most mining concession in Cameroon are currently in an explorative status. The Forestry Atlas of Cameroon allows us to identify around 58% of the 2000-2012 losses happening in the forest by specific forestry land uses. This seems to indicate that one third of the tree cover losses detected in the GFC product in the forest are happening outside identified forestry use categories.


Currently, the JRC and GFC are the only remote sensing based products providing information about the forest cover change at both the national scale of Cameroon and in the context of the Congo Basin regional scale. The differences noticed for the rate of change per year in the two products for the overlapping period of 2000-2010 have to be analysed in the light of the respective specificities of each product. They diverge in term of methodology, period of time analysed, definition of forest cover change processes and imagery used. Their main divergence is that the JRC product is based on a systematic sampling while the GFC product is a wall to wall product. Accordingly, the GFC product is flexible for further analysis whereas the JRC estimations rely on statistical extrapolations which are more complicated to replicate by the user.

A disadvantage of a systematic sampling approach is that it is more likely to underestimate or overestimate those forest change processes which occur more rarely. The observations will depend on the localisation of the sampling. This could explain the absence of reforestation estimates in 2000-2010 in the RoC and DRC. The design of the sampling grid for the JRC study (every half degree) is lacking robustness for providing country statistics over a limited forest cover extent. Therefore, the estimates for Eq.Guinea and CAR have to be analysed with caution. On the other hand, the added value of the JRC product is that after the processing, every sample has been visually validated by an expert for the three time periods. The JRC product is part of the FAO/FRA initiative and findings can be compared to other regions of the world. The fact that the JRC product is giving estimates of forest changes happening in the period 1990-2000 is also unique. Experts can also detect changes such as oil palm plantation where an automatic approach is likely to consider it as reforestation.

Regarding the GFC product, its wall-to-wall approach makes it quite flexible for further analysis in land cover, land use or administrative stratification. However, the impact of the important cloud cover is not assessed while it clearly has an impact on the estimation of the tree cover percentage over the Atlantic Coast. In this study, an external land cover map was use to refine the definition of the area considered for the forest loss and gain estimations. A final observation about the GFC product is that being a pixel by pixel product, it presents an important level of noise in the Central Africa area that is not specifically assessed.

The difficulty to harmonize the results from forest cover changes products due to differences in methods, types of imagery used and definitions of deforestation is well explained by Ickowitz et al. (2015) in the context of DRC. In that case, the different products were however agreeing on the importance of agriculture as a driver of deforestation. Here too, some commons trends can be found in the two products. The two datasets show that the annual gross deforestation rates for the entire Congo Basin is relatively low over the two decades (0.19% for JRC 1990-2000, 0.14% for 2000-2010 and 0.12% for 2000-2012) compared to the two other tropical forest basin (Achard et al. 2014). The results from both the JRC and the GFC contradict some findings of the analysis of Hosonuma et al. (2012), classifying countries of the Congo Basin as pre-transition phase countries with a high forest cover and low deforestation rates except for Cameroon, classified in an early transition phase, where forest cover is already being converted at a rapid rate. DRC is shown to be presenting higher rates of forest loss while Cameroon is presenting rate of loss similar to other countries with a similar proportion of forest cover. But it confirms that in the Congo Basin, the dynamics of forest cover changes are dominated by deforestation processes with all countries presenting a higher rate of forest loss than of forest gain.

Thanks to the flexibility of the GFC products delivered in layers, the dynamic of change happening in the rural complex was evaluated. It is important to note that the rural complex is coming from a land cover map at a coarser spatial resolution than the 30 meter of the GFC product but it is capturing well areas composed of a mixture of crops and secondary forest. The dynamic in the rural complex was shown to be much more important that in the forest area in term of loss and gain. However, even there, the gain observed in 12 years remains much lower than the loss.

The combination of the JRC and GFC products gives an overview of the temporal evolution of forest cover losses over more than 20 years. First, an important decrease of gross deforestation rates is observed between the 1990s and 2000s. This is also observed at the regional scale for the Congo basin. For Cameroon, the high deforestation rate in the 90s can be associated to the economic crisis of 1986. As studied by Mertens et al. (2000) the economic crisis of was associated in time with a strong increase of deforestation rate related to population growth, increased marketing of food crops, modification of farming systems, and colonization of new agricultural areas in remote forest zones. In 1994, a new forest law setting out regulations on the use and management of forest and wildlife resources with the allocation of logging concessions has been implemented in Cameroon (Cerutti and Tacconi 2006; Mertens et al. 2012; Cerutti et al. 2013; Dkamela et al. 2014). According to Mertens et al. (2012), those changes in policies are translated in total observed area logged. However, additional studies about volumes and surfaces harvested would be required to link it to the remote sensing observations.

The two main activities to be considered under REDD+ are (i) deforestation (conversion of forest land to other land uses) and (ii) forest degradation (defined in the REDD+ context as a long term decrease of carbon stocks within forest lands). While the satellite imagery at circa 30 m resolution used in the JRC and the GFC products generally succeed in capturing forest disturbances associated to deforestation, measuring fine-scale local disturbances, forest degradation and forest regrowth are still considered as a challenging task with methodologies and imagery currently used (GOFCGOLD 2014).

In Cameroon specifically, forest cover changes are currently related to small-scale agriculture in the form of shifting cultivation, cash (coffee and cocoa) and food (plantain, cassava, corn, yam,...) crops (Sunderlin et al. 2000; Bellassen and Gitz 2008), cattle ranching, large scale conversion of land to agricultural use (mainly for palm oil, rubber, rice, and maize production), fuelwood extraction, timber extraction, urbanisation, increase of oil and mineral extraction concessions and large infrastructure and energy projects (Hosonuma et al. 2012, Mertens et al. 2012, Ernst et al. 2013, Oyono et al. 2013, Dkamela et al. 2014, Carodenuto et al. 2015). The relatively high population density of Cameroon is an underlying cause of forest cover changes, so is the poor governance, weak political stability and weak economy in the region (Ernst et al. 2013).

Selective logging in Forest Management Units presents a specific pattern easily identified from satellite imagery. Development of logging road networks appears clearly in Landsat images time series (e.g. from 1990 to 2010). In the GFC product, some recent logging roads are well detected, e.g. in the East region. But when the logging roads are a bit older, e.g. in the South East of Yokadouma or in the north of Mindourou or in other regions, they are not so well detected in the GFC product. The network is not as visible as the more recent network of logging roads of the neighbouring Republic of Congo. It is not clear to what extent this is due to the lack of imagery combined with important cloud cover in the area. Concerning selective logging, to capture the actual canopy gaps created by the wood extraction rather than just the logging roads, satellite imagery with finer spatial and higher temporal resolutions would be needed. Besides what is occurring in FMUs, it is known that illegal logging plays an important role in deforestation in Cameroon (Cerutti and Tacconi 2006). An advantage of a remote sensing approach is that forest cover losses can be captured independently of any administrative or land use boundaries. But complementary GIS data are needed to attribute the forest loss detected to specific land uses.

In Cameroon, the majority of agro-industrial lands are planted with rubber trees or oil palm (Mertens et al. 2012). Industrial plantations are well visible on Landsat images due their large extents and regular patterns. However, only the most recent plantations are clearly detected in the GFC product as forest loss. In fact, the clearing which occurs just before the establishment of plantation of palm or rubber trees is actually detected. This is also highlighted by Tropek et al. (2014) in their review of the GFC product. Moreover, oil palm plantations are sometimes considered as "forest gain" in the GFC product whereas it is not considered as forest domain. However, the GFC product shows interesting perspective to monitor losses linked to new agro-industrial exploitation.

Beside industrial agriculture, smaller scale agricultural expansion and shifting cultivation is an important driver of forest loss. In the north of Cameroon important agricultural expansion in the forest is well detected. But when the forest cover changes represent small areas of shifting cultivation around roads and villages, the detection is more random. The small crops and household plantations visible on fine resolution imagery available through Google Earth platform are not well captured on Landsat imagery. Finer spatial resolution is needed to better identified loss linked with this type of change.

Regarding urbanisation, forest losses due to roads or villages are visible in Landsat images and detected in the GFC product but it is not clear what proportion of the losses is depicted. Extension of larger cities (around Douala or Yaounde) is however clearly visible in image time series.

Despite the fact that protected areas play a strong role in the mitigation of deforestation (Ernst et al. 2013), some forest losses are observed in those areas. While explanations can be that some of the protected areas were created/classified only in the late 2000s, it demonstrates that a remote sensing approach can help highlighting encroachment of forest loss in protected areas.

While information reported by the Forestry Atlas of Cameroon for year 2012 might need some updates, the Atlas clearly highlights that about 40% of the forest losses detected by the GFC product are happening outside allocated forestry uses. In this case, a remote sensing approach proves its utility by putting into evidence the losses linked to infrastructure development, illegal logging, rural population pressure or urban expansion.

It is interesting to recall that Central Africa is historically characterized by lower Landsat satellite data acquisition (Wulder et al. 2015) making this kind of estimations exercise of forest cover change more difficult than in other regions of the world.


To monitor forest changes in tropical areas, remote sensing products have become more available since the last 10 years. Unfortunately, studies using the same methodology allowing for a regional analysis of the different countries of the Congo Basin are still scarce. Two recent remote sensing derived products providing assessment of forest cover change in the Congo Basin for the last two decades were compared and combined (JRC and GFC products). Due to the differences in thematic legends and time periods, the combination of the two products was not straightforward. Nevertheless, as these two products present many similarities, we exploited the advantages of each product to provide an overview of national trends and differences in deforestation dynamics over the Congo Basin during the last two decades. Going back to 1990, the JRC product give a unique overview of the forest loss dynamic in time, highlighting a decrease of the annual rate of forest cover losses between 1990s and 2000s in the humid domain of Cameroun and the rest of the Congo Basin. The GFC product, with yearly estimates from 2000 to 2012, brings more accurate description of the processes, highlighting a new trend of increase from 2004 to 2010.The analysis of the GFC spatial distribution of the forest loss in Cameroon reveals that an important part of the loss is happening outside the permanent and the non-permanent domain. Remote sensing approaches, both from a sample or a wall to wall scheme, provide a global view of the forest losses and are not limited to some specific land uses.

Imagery from Landsat satellites at 30 m spatial resolution easily picks up change caused by clearing of forest prior to the plantation of industrial oil palm or rubber trees, i.e. large agro-industrial exploitation, but also the conversion of forests to other type of croplands when the disturbed area is large enough. Also forest loss around villages and along roads is partially identified. Selective logging is indirectly detected when the pattern of logging roads is recent. A monitoring of the integrity of protected areas is also possible. It however fails to monitor forest degradation processes, small scale deforestation (e.g. for fuel wood collection, small crops or village concessions), plantations already implemented before the acquisition date of the imagery and logs in selective logging concession.

From a satellite imagery point of view, finer spatial resolution images (<10m) combined with more frequent acquisitions would be useful to monitor small scale change and degradation. Currently fine resolution data are not available at large area coverage and remains relatively expensive but the situation might change in the coming years. The recent improvements in the availability of Landsat-8 images, due to better programing (Belward and Skoien 2015) and the recently launched Sentinel-2 satellite, promise to provide more frequent acquisitions of medium to fine spatial resolution (circa 30-10m resolution) data.

Until recently, only the decadal remote sensing assessments provided by the JRC together with the FAO (FAO and JRC 2012) have been feasible, as a result of lack of imagery, data acquisition and processing costs. These have a number of land cover classes, and are quality controlled by regional experts, providing efficient statistical estimate for period reporting on forest cover changes and land cover trajectories. However, a wall-to-wall coverage providing spatially explicit information may be desirable for national coordination of land management planning. While currently the class definition remains basic (tree cover percentage) and false detection are not corrected by the automatic classification procedures, it is only the beginning of this kind of extensive and repeatable monitoring.

To support REDD+ policies in the best way, some improvement taking advantage of the higher resolution of the forthcoming sensors and the increased computing capacity can be foreseen. The needs for a future monitoring system could be express as following: a wall to wall monitoring system, providing consistent reference forest maps and forest loss area estimations on a yearly basis. Monitoring smaller disturbances in the forest need to be addressed. Also it is important in the future to develop a capacity to quantify not only the disappearance of forest (deforestation) but also forest degradation by means of higher spatial and temporal imageries. Finally, an increased consistency between remote sensing products in terms of forest class definitions as well as validation should become a greater priority in the future development of products.


ACHARD, F., BEUCHLE, R., MAYAUX, P., STIBIG, H.-J., BODART, C., BRINK, A., CARBONI, S., DESCLEE, B., DONNAY, F., EVA, H.D., LUPI, A., RASI, R., SELIGER, R. and SIMONETTI, D. 2014. Determination of tropical deforestation rates and related carbon losses from 1990 to 2010. Global change biology 2010: 1-15.

BELLASSEN, V. and GITZ, V. 2008. Reducing Emissions from Deforestation and Degradation in Cameroon--Assessing costs and benefits. Ecological Economics 68: 336-344.

BELWARD, A.S. and SK0IEN, J.O. 2015. Who launched what, when and why; trends in global land-cover observation capacity from civilian earth observation satellites. ISPRS Journal of Photogrammetry and Remote Sensing 103: 115-128.

CARODENUTO, S., MERGER, E., ESSOMBA, E., PANEV, M., PISTORIUS, T. and AMOUGOU, J. 2015. A Methodological Framework for Assessing Agents, Proximate Drivers and Underlying Causes of Deforestation: Field Test Results from Southern Cameroon. Forests 6: 203-224.

CERUTTI, P.O. and TACCONI, L. 2006. Forests, illegality, and livelihoods in Cameroon. Working Paper 35. Bogor, Indonesia, Center for International Forestry Research

CERUTTI, P.O., TACCONI, L., LESCUYER, G. and NASI, R. 2013. Cameroon's Hidden Harvest: Commercial Chainsaw Logging, Corruption, and Livelihoods. Society & Natural Resources 26: 539-553.

DESCLEE, B., MAYAUX, P., HANSEN, M., AMANI, P.L., SANNIER, C., GOND, V., RAHM, M., LUBAMBA, J.-P.K., TURUBANOVA, S., ALTSTATT, A., FICHET, L.-V., RAMMINGER, G., CORNU, G., BOURBIER, L., JUNGERS, Q., DEFOURNY, P., LETOAN, T., HIRSCHMUGL, M., JAFFRAIN, G., PINET, C., KEMAVO, A., DORELON, P., PEDRAZZANI, D., SEIFERTGRANZIN, J., MANE, L., BANAK, L.N., VRIELING, A. and MERMOZ, S. 2014. Evolution of forest cover at a national and regional scale and drivers of change. The Forests of the Congo Basin--State of the Forest 2013. C. de Wasseige, J. Flynn, D. Louppe, F. Hiol Hiol and P. Mayaux, Weyrich. Belgium: 21-46.

DKAMELA, G.P., BROCKHAUS, M., KENGOUM DJIEGNI, F., SCHURE, J. and ASSEMBE MVONDO, S. 2014. Lessons for REDD+ from Cameroon's past forestry law reform: a political economy analysis. Ecology and Society 19: art30.

ERNST, C., MAYAUX, P., VERHEGGHEN, A., BODART, C., CHRISTOPHE, M. and DEFOURNY, P. 2013. National forest cover change in Congo Basin: deforestation, reforestation, degradation and regeneration for the years 1990, 2000 and 2005. Global Change Biology 19: 1173-1187.

FAO and JRC. 2012. Global forest land-use change 19902005. FAO Forestry Paper. Rome, FAO, Food and Agriculture Organization of the United Nations and European Commission Joint Research Centre. 169: 40 p.

GOFC-GOLD. 2014. A sourcebook of methods and procedures for monitoring and reporting anthropogenic greenhouse gas emissions and removals associated with deforestation, gains and losses of carbon stocks in forests remaining forests, and forestation. GOFC-GOLD Report version COP20-1. The Netherlands, GOFC-GOLD Land Cover Project Office, Wageningen University.

HANSEN, M.C. and DEFRIES, R.S. 2004. Detecting long-term global forest change using continuous fields of tree-cover maps from 8-km advanced very high resolution radiometer (AVHRR) data for the years 1982-99. Ecosystems 7: 695-716.

HANSEN, M.C., POTAPOV, P.V., MOORE, R., HANCHER, M., TURUBANOVA, S.A., TYUKAVINA, A., THAU, D., STEHMAN, S.V., GOETZ, S.J., LOVELAND, T.R., KOMMAREDDY, A., EGOROV, A., CHINI, L., JUSTICE, C.O. and TOWNSHEND, J.R.G. 2013. High-resolution global maps of 21st-century forest cover change. Science (New York, N.Y.) 342: 850-3.

HIRSCHMUGL, M., STEINEGGER, M., GALLAUN, H. and SCHARDT, M. 2014. Mapping Forest Degradation due to Selective Logging by Means of Time Series Analysis: Case Studies in Central Africa. Remote Sensing 6: 756-775.

HOSONUMA, N., HEROLD, M., DE SY, V., DE FRIES, R.S., BROCKHAUS, M., VERCHOT, L., ANGELSEN, A. and ROMIJN, E. 2012. An assessment of deforestation and forest degradation drivers in developing countries. Environmental Research Letters 7: 044009.

ICKOWITZ, A., SLAYBACK, D., ASANZI, P. and NASI, R. 2015. Agriculture and deforestation in the Democratic Republic of the Congo: A synthesis of the current state of knowledge. Occasional Paper. CIFOR. Bogor, Indonesia, 18 pp.

MAYAUX, P., PEKEL, J.-F., DESCLEE, B., DONNAY, F., LUPI, A., ACHARD, F., CLERICI, M., BODART, C., BRINK, A., NASI, R. and BELWARD, A. 2013. State and evolution of the African rainforests between 1990 and 2010. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 368: 20120300.

MERTENS, B., FORNI, E. and LAMBIN, E.F. 2001. Prediction of the impact of logging activities on forest cover: A case-study in the East province of Cameroon. Journal of Environmental Management 62: 21-36.

MERTENS, B., SHU, G.N., STEIL, M., TESSA, B., MINNEMEYER, S., DOUARD, P., JEAN-DANIEL MENDOMO BLANG, A.L., MBOUNA, D., MBOUA, P., FUEZING, A., METHOT, P. and NGLILAMBI, H. 2012. Interactive Forest Atlas of Cameroon--Atlas Forestier Interactif du Cameroun (Version 3.0).

MERTENS, B.T., SUNDERLIN, W.D., NDOYE, O. and LAMBIN, E.F. 2000. Impact of Macroeconomic Change on Deforestation in South Cameroon: Integration of Household Survey and Remotely-Sensed Data. World Development 28: 983-999.

OYONO, P.R., MORELLI, T.L., SAYER, J., MAKON, S., DJEUKAM, R., HATCHER, J., ASSEMBE, S., STEIL, M., DOUARD, P., LIMA, R., MAKAK, J.S., TESSA, B., MBOUNA, D. and NDIKUMAGENGE, C. 2013. Allocation and use of forest land: current trends, issues and perspectives. State of the Forests 2013: 215-240.

SUNDERLIN, W.D., NDOYE, O., BIKIE, H., LAPORTE, N., MERTENS, B. and POKAM, J. 2000. Economic crisis, small-scale agriculture, and forest cover change in Southern Cameroon. Environmental Conservation 27: 284-290.

TROPEK, R., SEDLA EK, O., BECK, J., KEIL, P., MUSILOVA, Z., IMOVA, I. and STORCH, D. 2014. Comment on "High-resolution global maps of 21st-century forest cover change". Science 344: 981-981.

UNFCCC. 2014. Key decisions relevant for reducing emissions from deforestation and forest degradation in developing countries (REDD+). Decision booklet REDD+. Bonn, UNFCCC secretariat.

VERHEGGHEN, A., MAYAUX, P., DE WASSEIGE, C. and DEFOURNY, P. 2012. Mapping Congo Basin vegetation types from 300 m and 1 km multi-sensor time series for carbon stocks and forest areas estimation. Biogeosciences 9: 5061-5079.

WHITE, F. 1983. The vegetation of Africa. A descriptive memoir to accompany the Unesco/AEFTFAT/UNSO vegetation map of Africa.

WULDER, M.A., WHITE, J.C., LOVELAND, T.R., WOODCOCK, C.E., BELWARD, A.S., COHEN, W.B., FOSNIGHT, E.A., SHAW, J., MASEK, J.G. and ROY, D.P. 2015. The global Landsat archive: Status, consolidation, and direction. Remote Sensing of Environment. In Press.


European Commission, Joint Research Centre (JRC), Institute for Environment and Sustainability (IES), Forest Resources and Climate Unit, Via E. Fermi, 2749, TP 261, Ispra, VA 21027, Italy

Email: and

TABLE 1 Forest cover changes in the Congo Basin Humid domain. a) JRC
estimates for the 1990s and 2000s and b) GFC estimates for the period
2000-2012 for the forest and the rural complex classes determined as
such: the forest class correspond to the humid forest classes in
Verhegghen et al. (2012) and the intersection of remaining forest
classes with a stratum of 70-100% Tree Cover (Hansen et al., 2013),
the rural complex class is the intersection of the rural complex class
(Verhegghen et al., 2012) with a stratum of50-100% Tree Cover (Hansen
et al., 2013).


                           1990-2000 rates (%/year)

country        n             gross                  gross
                         deforestation          reforestation

Cameroon        45    0.13 [+ or -] (0.04)   0.04 [+ or -] (0.01)
Congo           65    0.09 [+ or -] (0.02)   0.03 [+ or -] (0.01)
Gabon           63    0.07 [+ or -] (0.02)   0.02 [+ or -] 0
Eq. Guinea       7    0.13 [+ or -] (0.08)   0.10 [+ or -] (0.06)
CAR             26    0.11 [+ or -] (0.03)   0.03 [+ or -] (0.01)
DRC            114    0.24 [+ or -] (0.05)   0.03 [+ or -] (0.01)
Congo Basin    171    0.19 [+ or -] (0.03)   0.03 [+ or -] (0.01)

                 1990-2000 rates        2000-2010 rates
                     (%/year)               (%/year)

country                net                   gross
                  deforestation          deforestation

Cameroon       0.09 [+ or -] (0.04)   0.08 [+ or -] (0.03)
Congo          0.05 [+ or -] (0.02)   0.07 [+ or -] (0.02)
Gabon          0.05 [+ or -] (0.02)   0.03 [+ or -] (0.01)
Eq. Guinea     0.03 [+ or -] (0.07)   0.04 [+ or -] (0.03)
CAR            0.09 [+ or -] (0.03)   0.06 [+ or -] (0.02)
DRC            0.22 [+ or -] (0.04)   0.20 [+ or -] (0.04)
Congo Basin    0.16 [+ or -] (0.03)   0.14 [+ or -] (0.03)

                        2000-2010 rates (%/year)

country               gross                   net
                  reforestation          deforestation

Cameroon       0.02 [+ or -] (0.01)    0.06 [+ or -] (0.04)
Congo                   0              0.07 [+ or -] (0.02)
Gabon          0.01 [+ or -] 0         0.01 [+ or -] (0.01)
Eq. Guinea     0.05 [+ or -] (0.03)   -0.01 [+ or -] (0.02)
CAR            0.01 [+ or -] 0         0.05 [+ or -] (0.02)
DRC                     0              0.19 [+ or -] (0.04)
Congo Basin             0              0.14 [+ or -] (0.03)


2000-2012 rates (%/year)

                    Dense forests

               Loss   Gain   Loss minus Gain

Cameroon       0.07   0.01        0.06
Congo          0.05   0.01        0.05
Gabon          0.04   0.01        0.03
Eq. Guinea     0.09   0.01        0.08
CAR            0.13   0.02        0.11
Democratic     0.15   0.03        0.12
  of Congo
Congo Basin    0.12   0.02        0.10

                      Rural Complex

               Loss   Gain   Loss minus Gain

Cameroon       0.28   0.03        0.25
Congo          0.30   0.05        0.25
Gabon          0.36   0.06        0.30
Eq. Guinea     0.32   0.03        0.28
CAR            0.28   0.08        0.20
Democratic     1.00   0.43        0.58
  of Congo
Congo Basin    0.77   0.30        0.47

TABLE 2 Estimates of Forest and Rural Complex class and area of loss
derived from the GFC product in those classes for the period 2000-
2012 (in thousand hectares)

                                Dense forests

                                Total area        Loss area 2000-2012
                                ([10.sup.3] ha)   ([10.sup.3] ha)

Cameroon                                19,640                   170
Congo                                   21,367                   136
Gabon                                   22,913                   112
Eq. Guinea                               2,016                    21
CAR                                      7,217                   112
Democratic Republic of Congo           113,999                  2071
Congo Basin                            187,152                  2622

                                Rural Complex

                                Total area        Loss area 2000-2012
                                ([10.sup.3] ha)   ([10.sup.3] ha)

Cameroon                                 3,290                   109
Congo                                    1,886                    68
Gabon                                    1,240                    54
Eq. Guinea                                 457                    17
CAR                                        472                    16
Democratic Republic of Congo            15,218                 1,833
Congo Basin                             22,563                 2,097

TABLE 3 Repartition of the forest loss in the National Forest Domain
of Cameroon. The area are derived from the GFC product for the period
2000-2012 (in thousand hectares) and the administrative limits of the
WRI Atlas (Mertens et al., 2012)

                                             Loss area      Proportion
                                             2000-2012         (%)
                                          ([10.sup.3] ha)

Loss in the Permanent forest domain               75,116           44
  Logging concession (FMU)                        18,120           11
  Council forest                                   2,595            2
  Forest Reserve                                   3,404            2
  Protected Areas                                  3,544            2
  Licenses--expired by 2000                       47,451           28
  (remove everything else from it)
Loss in the Non-permanent forest domain           24,136           14
  agro industrial zone                            13,298            5
  community forest                                 5,604            6
  SSV (Sales of Standing Volume)                   2,169            3
Total Loss in Cameroon                           170,151          100
Outside of the National Forest Domain             70,898           42
COPYRIGHT 2016 Commonwealth Forestry Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Verhegghen, A.; Eva, H.; Desclee, B.; Achard, F.
Publication:International Forestry Review
Article Type:Report
Geographic Code:6CAME
Date:Dec 1, 2016
Previous Article:Cameroon's forest policy within the overall national land use framework: from sectorial approaches to global coherence?/Les politiques forestieres et...
Next Article:The technical and political challenges of the industrial forest sector in Cameroon/Les defis techniques et politiques du secteur forestier industriel...

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters