Recent forest cover developments and determinants
Decreasing forest cover has been documented all over Madagascar by various authors (e.g., Harper et al.
2007; Vieilledent et al.
2018) and our study area is no exception to this development. Although the mean annual deforestation rate did not deviate from country-wide estimates (e.g., Vieilledent et al.
2018), more localized studies already reported significantly higher rates from the study region (e.g., Schüßler et al.
2018b; Llopis et al.
2019). This discrepancy is probably due to geographic restrictions towards the forest frontier compared to the broader scope of larger-scale studies and inter-annual variations that can be triggered by political instability, the establishment of protected areas and/or price fluctuations of major cash crops (Llopis et al.
2019).
Shifting cultivation was already identified as the main driver of forest cover loss in northeastern Madagascar (Zähringer et al.
2015). Our results illustrate that despite deforestation happening on small and local scales, it is scattered all over the region and has major effects on regional forest cover. Major determinants of forest conversion are the status of forest protection and proximity to the forest edge. Protected areas effectively prevented deforestation inside their boundaries, although some illegal selective logging is nevertheless reported from national parks such as Masoala NP (Burivalova et al.
2015) and Mananara-Nord NP (pers. comm. Madagascar NP staff member). In contrast, deforestation is omnipresent outside of the protected areas and increasingly along their boundaries as it can be seen around Zahamena NP and along the eastern border of Makira NatP (Fig.
2). Areas within the proximity of the forest edge are primary subjects to deforestation with smaller fragments being particularly vulnerable. Their ongoing loss from the landscape mosaic leads to landscape homogenization and the loss of habitat complexity and connectivity (Urech et al.
2015; Zähringer et al.
2015).
We detected two major interruptions of the former continuous rain forest belt of eastern Madagascar in which only few smaller forest fragments remained in 2018 (red arrows in Fig.
2). We want to highlight that these landscapes are distinctively different from those in other ecoregions of the island. In northern and northwestern Madagascar, for example, at least some lemur species (used here as indicators, Muldoon & Goodman
2015) appear to be able to maintain population connectivity among forest fragments that are embedded in a rather open Savannah matrix (e.g., Quéméré et al.
2010; Salmona et al.
2015; Aleixo-Pais et al.
2018; Steffens and Lehman
2018). Although less studied in the humid ecoregion, it is much more difficult for lemurs and other forest-dwelling species to migrate through the degraded areas of secondary vegetation (Nunziata et al.
2016; Schüßler et al.
2018b; Galán-Acedo et al.
2019). These areas are usually characterized by a mosaic of dense stands of the perennial
Aframomum spp. intermixed with smaller tree species like
Harungana madagascariensis and
Psiadia altissima or, when severely degraded, dense and almost impassable fern- and grasslands with
Sticherus flagellatis and the exotic
Psidium cattleianum as typical species (Styger et al.
2007; Burivalova et al.
2015; Miller et al.
2018; Schüßler et al.
2018b). Homogenization of landscapes and the disruption of previously continuous forest tracts have major implications for landscape connectivity (Haddad et al.
2015). Increasing fragmentation lowers the resilience of species communities to adapt to environmental changes (Brown et al.
2015; Brown and Yoder
2015). Most alarmingly, landscape connectivity and thus the potential for range shifts along the rainforest belt was already largely restricted and definitely precluded in at least two areas of forest interruption in 2018.
The remaining explanatory variables, despite having lower relative importance values, nevertheless contributed to explaining the observed pattern of deforestation. For example, the relevance of the proximity to villages and to major pathways indicates that deforestation is mainly the result of expanding the size of arable lands within daily walking distance from villages (Kremen et al.
1999; Allnutt et al.
2013). These distances are usually between 1.17 h and 2.5 h of one-way walking time from the village (Fedele et al.
2011; Allnutt et al.
2013) but can also reach almost 3 h corresponding to 7–8 km (Schüßler unpubl. data). Indeed, the degree of forest disturbance and degradation was shown to be inversely correlated with distance to the nearest village (Schüßler et al.
2018b).
Lowest relative importance values were returned for inclination and the distance to the next major river. Shifting cultivation in this area is mainly focused on hill rice, beans or cassava, for which even steep slopes can be suitable for cultivation if located within daily walking distance from the village. Major rivers can be a significant way of transportation (e.g., Kremen et al.
1999), but appear to have had less pronounced effects on deforestation in the region within the last three decades.
We documented a relatively low natural forest regrowth. This is in accordance with other studies (Styger et al.
2007; Manjaribe et al.
2013; Zwartendijk et al.
2017). Natural succession is assumed to be particularly slow when soils are severely degraded (Klanderud et al.
2010; Randriamalala et al.
2012). Limited forest regrowth further increases the pressure on the remaining forests concerning the extraction of wood for construction or cooking as well as for non-timber-forest products (NTFP) like medical plants or fibers. Although fallow lands can provide some of these products (Pfund et al.
2011), forest resources play a key role in local livelihoods (Ratsimbazafy et al.
2012). To counteract the progressive isolation of forest fragments and the loss of crucial forest products for local communities, active reforestation measures are urgently needed. Land users in the region are particularly experienced in raising and planting tree saplings, for example to plant clove, coffee and cacao trees. Cash crops are obviously beneficial to invest in, since income opportunities from crop sales are likely. Plantation of non-cash crop trees in terms of reforestation, however, will require incentives to facilitate the involvement of local communities. Frameworks still need to be developed, evaluated and communicated scientifically, but payments for certain numbers of planted trees with controlling for their survival after some time may be conceivable. Only few restoration projects have been scientifically evaluated to date, but promising approaches are based on raising fast growing early successional native tree species in tree nurseries in order to restore soil conditions and to promote natural succession (see Manjaribe et al.
2013; Rajaonarimamy et al.
2017; Zwartendijk et al.
2017). In turn, we highly discourage the use of non-native tree species, since unpredictable side-effects can have massive negative effects on Madagascar’s fragile endemic biodiversity.
Settlement development and determinants
Many low-income countries are characterized by strong population growth (UNDP
2016). Madagascar’s population rose from about 7.9 Mio. in 1975 (UNDP
2005) to 24.2 Mio. in 2015 and may further increase (3.0% annual growth rate) to 36.0 Mio. people by 2030 (UNDP
2016). It is difficult to predict how human population density will develop in our study area. Apart from birth rates, domestic immigration triggered by recent price booms in major cash crops (vanilla and cloves) may accelerate population growth in the region (Hänke et al.
2018; Zhu
2018). So far, Jones et al. (
2018) documented only rather local migratory trends in the southern part of the study area (< 50 km movements reported by about 90% of interviewed households from the Ankeniheny-Zahamena Corridor). These short distance movements were mainly driven by the search for accessible lands (Jones et al.
2018). At least one fifth of all villages in our study area were newly established since 1990 in areas that were still forested at that time. We identified inclination and proximity to the nearest neighboring village as the major predictors for places in which new villages were built. Additionally, the altitude of the area and the proximity to the nearest major pathway also contributed to explaining the observed settlement expansion. This resembles a housing sprawl along the forest edge, which is in accordance with short distance movements to access and acquire new lands for agricultural use (Nawrotzki et al.
2012; Jones et al.
2018). Although there are first tendencies towards land use intensification and the establishment of permanent land use systems (Zähringer et al.
2015; Llopis et al.
2019), this documented housing sprawl until 2018 is alarming, since it concerns almost all forests outside protected areas and steep mountains, irrespective of their potential importance as landscape corridors and biodiversity reservoirs.
Although we could not quantify this development in areas that were already deforested prior to 1990, it occurred to some degree. Most villages in these places, the southeastern part of the study area, grew in size in accordance with the documented population growth (UNDP
2016). Houses are traditionally built from natural material extracted from the forests around the villages. Interestingly, the proportion of houses with natural leave roofs significantly declined and most houses were equipped with iron roofs in 2018. This development can be explained in two ways: first, we investigated an area that has only restricted access to forest resources and intact forest tracts are far away from the villages. Under these circumstances, the availability of leaves to tile roofs may be low, and people may be forced to use other materials. On the other hand, this observation implies that households may be shifting away from pure subsistence farming towards the production and sale of products in regional and even global markets. In fact, recent price booms of vanilla and cloves have resulted in significant income increases for smallholder farmers and in remarkable improvements of the socio-economic situation in some parts of northeastern Madagascar (Hänke et al.
2018; Zhu
2018). Furthermore, our study shows that most of the larger villages lie in areas close to the coastline that were already almost entirely deforested in 1990. These places are still well suited for land use intensification (Zähringer et al.
2016) and may therefore allow higher population densities than the more rugged terrain along the remaining forest frontier.
Future development in forest cover
Forest cover loss in northeastern Madagascar is mainly driven by smallholder farmers practicing shifting cultivation (Zähringer et al.
2015). This is similar to most of the world’s tropics, though, the extent of shifting cultivation is expected to decline within the next few decades (Heinimann et al.
2017). First indications for this development were found in our study area with land use intensification and the establishment of permanent cultivation systems (Zähringer et al.
2015; Llopis et al.
2019). However, the transition from traditional shifting cultivation towards permanent land use practices and thus a significant decrease in deforestation is unlikely to happen within a short time interval in those remote areas that are at the frontier of deforestation activities. We therefore aimed at identifying areas that are particularly threatened by further deforestation under a “business-as-usual” scenario.
Although our models weighted input variables differently, the results of the predictive modeling are mostly consistent across modeling approaches. Protected areas and highland forests face the lowest deforestation probabilities due to their inaccessibility and the relative efficiency of forest protection inside protected areas compared to those outside their boundaries. However, our predictions also illustrate that the pressure on at least some parts of the protected areas may also be increasing in the future.
The risks of landscape homogenization, i.e., the decrease of habitat complexity towards more uniform land use types, were already noted by other authors (Urech et al.
2015; Zähringer et al.
2015) as well as by our own analysis that predicts further loss of smaller forest fragments from the landscape mosaic. Despite presumed negative effects on biodiversity due to decreasing habitat size and diversity (Haddad et al.
2015), the complete loss of forests in some areas will have adverse effects on local climates and water availability. Land users in northeastern Madagascar already noticed problems in rice paddy irrigation due to decreasing forest cover and water availability around their fields in two study sites (Llopis et al.
2019). Our survey walks revealed that this problem is rather widespread across the region. Diminishing forest resources will therefore most likely decrease the productivity of intensive land use systems that produce the most important staple crop in Madagascar. Unfortunately, awareness about the scarcity of forest resources and their ecosystem services seems to be highest in areas with limited or absent forest cover, whereas people in currently forest-rich areas appear to be less aware about decreasing forest coverage (Urech et al
2015). This mismatch in perceptions needs to be targeted in forest frontier areas to illustrate the finiteness of forests and the crucial services they provide. Promising approaches using environmental education are described elsewhere (e.g., Schüßler et al.
2019).
Almost all of the already restricted lowland forests of the study area are predicted to face high deforestation probabilities (except Mananara-Nord NP). These findings highlight that lowland forests require increased conservation attention to protect locally endemic lowland forest species, but also that effective conservation management is crucial for Mananara-Nord NP as the only officially protected tract of lowland rain forest in the region. Why do comparably small forests matter as long as there are larger forested areas with low deforestation probabilities available in the highlands of the southern part of the region and in Makira NatP in the northern part? The study area can be subdivided into 13 inter-river-systems (IRSs) that may differ in their species inventories. Even single IRSs can be of particular importance for micro-endemic species like mouse lemurs (Radespiel et al.
2008,
2012), amphibians (Gehring et al
2010; Vences et al.
2010) or reptiles (Gehring et al.
2010). Furthermore, the lowland rain forests of eastern Madagascar have been stated the key for lemur conservation in the long term (Campera et al.
2019). At least two of these IRSs are predicted to have almost no areas of low deforestation probability (fields C7 and D8 as well as field D6 in Fig.
5), but some forests are already protected by the local community (see Miller et al.
2018; Schüßler et al.
2018b). These forests still harbor a high lemur diversity and undisturbed habitats despite being surrounded by a deforested matrix. Although these community-managed areas are so far effectively protected against deforestation (Miller et al.
2018; Schüßler et al.
2018b), it appears that shrinking resources around these forests will most likely increase the pressure on the currently protected forest patches (A. Miller pers. com.; Schüßler et al.
2018b).
Four other IRSs contain rather limited areas of low deforestation probability or areas predominantly restricted to higher elevations (fields C9, C10, D5 in Fig.
5) without any current forest protection scheme. Deforestation of whole IRS will inevitably increase extinction risk of micro-endemic species in an area that is still not well covered with biodiversity surveys and even new species to be found. Certain areas should be therefore highlighted in the following to be considered for the expansion of the protected area network of the region. These forests could act as stepping stones offering habitats for dispersing species or refuges for species retreating from human-dominated landscapes. We focus on areas that are predicted to have low deforestation probabilities as they may have lower value for communities in terms of expanding shifting cultivation or agroforestry systems due to their remoteness and/or inaccessibility to minimize land use conflicts. However, our village map illustrates that all forests are surrounded by villages. Protection schemes should therefore be promoted that encourage communities to set aside a certain portion of their forest area for biodiversity conservation. As described above, community-managed forests are already established in some places away from the major protected areas as well as in the buffer zones around Makira NatP (Brimont et al.
2015). This scheme of community-managed forest should also be promoted in other areas by focusing particularly on land users most vulnerable to environmental or other external changes (Harvey et al.
2014; Brimont et al.
2015). Recent evaluations of community-based approaches have illustrated that the danger of marginalization of poorer households, non-targeting of remote but most affected households and elite capture are severe issues to tackle (Brimont et al.
2015; Cullman
2015; Ward et al.
2017). The growing body of studies illustrating the pitfalls of previous approaches should be considered carefully when establishing new community-managed forests.
One of these potential stepping stone forests is located in the highlands west of Mananara-Nord NP (field C6 in Fig.
5) spanning two IRS in the source region of the Anove river. As of 2018, there is still a large tract of forest left in this area and a zonation scheme integrating different land use types may be beneficial to achieve protection of a core forest zone. Further forested areas that should receive more attention are the lowland forests of Manompana (see above) that are currently protected by the local community, as well as the lowland and mid-altitude forest south of Makira NatP (fields C5 and D5 in Fig.
5). The latter area represents the only forests with low deforestation probabilities for this IRS and could therefore become an important stepping stone forest. Another important stepping stone forest is located in the highlands of the southern part of the region (fields A8 and B8 in Fig.
5). Encouraging forest management based on a zonation system should conserve again important habitats spanning two IRS. We want to emphasize that we are not promoting to set apart extensive forest areas to establish new protected areas. The forests we highlighted here rather represent the minimal recommendation to protect most of the regional biodiversity at least in a shorter term. Community-based forest management approaches appear to be valuable in several parts of the study area and should receive further attention in protecting key forest areas.