Coupled forest zoning and agricultural intervention yields conflicting outcomes for tropical forest conservation in the Democratic Republic of the Congo (DRC)

In the Democratic Republic of the Congo (DRC), which contains approximately 60% of Central Africa's lowland forests, poverty rates are among the highest in the world [1]. Agricultural activity and fuelwood collection are increasing as a result of growing human populations and comprise the majority of deforestation [2, 3]. This is predicted to rise as pressures to increase food production force smallholders to expand their agricultural areas into primary forests [4].

Africa's agricultural productivity on a per capita basis has fallen behind other parts of the developing world [5, 6] for a variety of reasons, including poor soil fertility and related nutrient deficiencies [7] and lack of access to modern agricultural technologies [8, 9]. Research has shown that to feed increasing populations and support sustainable rural livelihoods, agricultural productivity in Africa must be sustainably intensified, i.e. increase agricultural output per unit of area while minimizing environmental impact [6, 10, 11], especially in the light of Africa's projected fourfold population expansion expected by 2100 [12]. Promoting sustainable agricultural intensification strategies can help optimize the allocation of land for both conservation and food production [13–15]. However, intensification can spur increased deforestation caused by raising the profitability of agricultural production—known as the rebound effect [16, 17]—and can bring about other perverse market outcomes [18]. These negative outcomes can be lessened by implementing complementary land-use planning, governance and policy frameworks [19–21].

Achieving appropriate land use governance is a challenge where rural dwellers lack access to land tenure systems and are often marginalized in favor of private-sector natural resource extraction [22, 23]. Participatory mapping and planning processes can help achieve governance through community-based adoption of rules and enforcement of land use designations [24]. The DRC lacks a system of land tenure; legally, the country's forests and natural resources belong to the state. In 2002, however, the DRC adopted a Forest Code that allowed forest-dependent communities to formally request the right to manage their own forests [25]. The Ministerial Decree 14/018 (signed in 2014), which was then followed by the Ministerial Order 0/25 (adopted in 2016), completed the legal framework for communities to govern their own forest concessions up to a maximum of 50 000 hectares [26, 27]. DRC's rural villages, however, generally have very little capacity to initiate the mapping and legal recognition process required to obtain these management rights.

From 2009 until 2018, a consortium of institutions in partnership with the Government of DRC, has built capacity for community-based participatory forest zoning coupled with agricultural intensification interventions in Tshuapa Province, located in northern central DRC, an important habitat for the endangered bonobo (Pan paniscus) great ape. The consortium engaged 90 village communities (figure 1 and table S1 available online at stacks.iop.org/ERL/17/064002/mmedia) in voluntary participatory zoning and a subset received agricultural interventions to boost cassava production around a quid-pro-quo agreement [28, 29]. In this agreement, the subset of communities who received the agricultural interventions agreed to abide by commonly agreed land use designations that limited agricultural expansion to designated zones, and thus limited forest conversion outside these areas. A primary goal, alongside enhancing food production and protecting forests and their ecosystem services, was building local stakeholder capacity, empowerment, and ability to obtain formal government recognition of land management rights. Members of all 90 villages collaboratively defined, via a participatory mapping process, the geographic boundaries for the following land use zones: (a) agricultural micro-zones (AMZs), which contained the villages' agricultural activities, and (b) permanent forest micro-zones (PFZs), which were protected from agricultural expansion (figure 1).

Zoom In Zoom Out Reset image size

For this study, we assessed household perceptions of the zoning process, assessed farmers' decisions related to field and fallow management, and estimated the uptake rates of improved cassava varieties and assessed yields of improved versus local cassava varieties. Subsequently, we analyzed and compared changes in annual rates of observed tree cover loss in the forest zones between project intervention and pre-intervention time periods. For each, we compared results between households that received the interventions with those that did not. We also conducted an accuracy assessment of our tree cover loss detections. Our research elucidates the importance of tropical forest monitoring, improves methods for monitoring human impacts on natural forests in areas where development initiatives are occurring, and helps inform policy debates centered around achieving forest conservation while increasing food productivity.

2.1. Study area

2.2. Participatory forest zoning and agricultural interventions

2.3. Household surveys: micro-zoning perceptions and field and fallow management

2.4. Field study: in-situ cassava yields

2.5. Tree cover loss analysis

3.1. Forest zoning perceptions

Although all household survey respondents belonged to villages that had undergone the forest zoning process, household awareness varied between domains. For households that received the agricultural interventions (hereby called those in the 'intervention domain'), 80 ± 3% (hereby defined as the mean ± standard error of mean) of the respondents had heard of the zoning process, while for those in the non-intervention domain, only 53 ± 5% had heard of it (figure 2(a)). Additionally, household participation in the zoning process varied between domains: 59 ± 3% of the households participated in the intervention domain, while only 38 ± 6% participated in the non-intervention domain (figure 2(b)). Furthermore, a higher proportion of respondents in the intervention domain (55 ± 3%) knew the location of the AMZ boundaries compared to the non-intervention domain (37 ± 4%) (figure 2(c)). Finally, about half of the households in the intervention villages (50 ± 3%) thought that the AMZ limits were respected in their village, whereas only 24 ± 4% thought so in the non-intervention villages (figure 2(d)).

Zoom In Zoom Out Reset image size

3.2. Cassava improved varietal uptake and yield

Household uptake of improved cassava varieties differed between domains. In the intervention domain, 72.4 ± 2.7% (1796 ± 66 households) reported using improved varieties, while in the non-intervention domain only 13.9 ± 3.0% (416 ± 90 households) reported using them (figure 3).

Zoom In Zoom Out Reset image size

On average, improved varieties yielded 62% more than the local varieties (24.5 ± 1.5 Mg ha⁻¹ and 15.1 ± 1.1 Mg ha⁻¹, respectively, significant to p < 0.001) (figure 4). Assessment of yields for the eight improved cassava varieties showed that varieties Obama and Ilona had on average both the highest total yield (33.4 and 33.9 Mg ha⁻¹, respectively) and highest marketable yield (27.5 and 24.6 Mg ha⁻¹, respectively), although they were not frequently encountered (table S2). From the local varieties, Mbandaka and Isakakala were the most productive, with total yields of 19.0 and 21.5 Mg ha⁻¹ respectively, and with marketable yields of 13.8 and 13.4 Mg ha⁻¹ respectively. The share of marketable tubers in the total yield was strongly variety-dependent and ranged between 52% (Vuvu) and 87% (Mayombe). There was limited evidence for a difference in the share of marketable yield between local and improved varieties (69% for local compared to 73% for improved varieties, p < 0.05, two-sample t-test).

Zoom In Zoom Out Reset image size

3.3. Field and fallow management

3.4. Tree cover loss

3.4.1. Primary forest analysis (2001)

The percentage of land covered by primary forest in all three zones in year 2001 was similar between the intervention and non-intervention domains, which proved consistency in their forest land cover prior to the study (table S3).

3.4.2. Tree cover loss mapping by forest zone

Approximately 70% of the tree cover loss that took place in the PFZs from 2001 to 2018 occurred within the 3 km buffer zone, which suggests that the loss is driven by agricultural expansion. In the buffer zone, average annual tree cover loss between pre- and post-intervention periods increased by 0.17% yr⁻¹ (from 0.08% yr⁻¹ to 0.25% yr⁻¹) in the intervention domain compared to 0.11% yr⁻¹ (from 0.29% yr⁻¹ to 0.41% yr⁻¹) in the non-intervention domain (figure 5(a)). Inside the PFZs, average annual tree cover loss between time periods was not statistically different for either domain but increased by 0.01% yr⁻¹ (from 0.024% yr⁻¹ to 0.031% yr⁻¹) in the intervention domain and remained unchanged in the non-intervention domain (figure 5(b)). Annual tree cover loss across the 18 year period was higher overall in the non-intervention domain for both the buffer zones and the PFZs (figures S5(a) and (b)). However, between 2016 and 2018, tree cover loss inside the buffer zone in the intervention domain escalated to match and even surpass that of the non-intervention domain (figure S5(a)). Meanwhile, annual tree cover loss inside the PFZs remained relatively parallel between domains, with more yearly variation appearing in the non-intervention domain (figure S5(b)). Tree cover loss decreased for both domains in each zone during the intervention period between 2010 and 2013, and more steeply decreased for the non-intervention domain in both zones (figures S5(a) and (b)).

Zoom In Zoom Out Reset image size

A sample-based validation confirmed a high level of accuracy for the forest cover loss map. The overall accuracies of both 2001–2010 and 2013–2018 forest cover loss maps were 99.9%. Other accuracy metrics were likewise comparable between the time intervals (table S4).

With growing populations dependent on farming as a major source of employment and livelihoods, Africa is faced with a critical socio-ecological challenge: boost food security while conserving the rich biological integrity of its forests and ecosystems [14]. Although environmental trade-offs will be necessary to raise global food production [35], research suggests that negative environmental impacts can be lessened by achieving higher yields on existing croplands in under-yielding countries [15, 36]. However, there is conflicting and sparse evidence about the extent to which agricultural intensification can conserve natural lands in Central and Sub-Saharan Africa. Importantly, there is now greater recognition that intensification of commodity crops (as opposed to staple caloric crops consumed by food insecure households that is the focus of this study) is more likely to result in the rebound effect that can exacerbate deforestation [35]. Other research stresses the importance of complementary participatory land use zoning to reduce deforestation [37, 38].

Here, we investigated whether a quid-pro-quo approach resulted in positive socio-ecological outcomes. Results revealed potential shifts in social norms and perceptions that signaled positive support for forest zoning. Households belonging to the intervention domain had enhanced awareness of, and more positive attitudes about, the land sparing and forest zoning concepts, which indicated that the intervention households would be more likely to adhere to zoning rules governing agricultural expansion. There is therefore potential in this region to engage in participatory forest zoning processes that encourage more sustainable management of land and resources through direct participation of local communities through these types of agreements. The interventions produced clear gains in cassava yields; furthermore, around 75% of households in the intervention domain and 14% of households in the non-intervention domain cultivated improved varieties of cassava, which indicates a potential spill-over effect from the interventions. This demonstrates significant potential to increase food productivity through uptake of improved cassava varieties in this region. However, sustaining increased yields and income in the longer term will require additional interventions for greater market development, which could in turn negatively affect forest conservation [39].

Although the household surveys showed support for positive conservation outcomes from the agricultural interventions, the results of the satellite-based analysis did not. Tree cover loss continued to increase during the post-intervention period for both domains in both the buffer zones and the PFZs, and no real effect of the interventions on forest conservation was detected. In fact, the escalation of tree cover loss inside the buffer zone of the intervention domain during the post-intervention period signals risk for a rebound effect. The disconnection between household perceptions of zoning adherence and actual behavior warrants discussion. First, there is a risk of bias known as social desirability bias when conducting surveys with beneficiaries of interventions, since these beneficiaries could respond in ways that seek to minimize self-stigmatization and generate a more positive image [40], especially when interviews are conducted by project staff. Second, although natural resource management interventions can result in increased positive conservation outcomes [41, 42], these types of interventions can be less effective when presented to landholders who are dependent on clearing land to maximize livelihood gains [43]. Interventions that aim to generate income via alternative livelihoods that do not require deforestation such as beekeeping, soap-making, tailoring and food processing might therefore have greater potential of conserving forest in the study area.

Our results emphasize the importance of strengthening the formalization of land rights and governance structures for community-based monitoring and enforcement. Coupled with agricultural interventions, participatory adoption of rules and enforcement of land use designations as part of a formal zoning process is crucial to ensure that the PFZ areas are maintained long-term [20]. Formalization of land tenure and strengthened land management rights can lead to positive outcomes for human socio-economic well-being, women's empowerment, food security, and forest conservation [44–47]. However, because communities may be managing lands in order to maximize livelihood gains, it is important that arrangements like the quid-pro-quo agreements used in this study provide clear incentives for communities to actively conserve their forested land [45].

Because 70% of tree cover loss in the PFZs occurred inside the 3 km buffer zone, pressure on the PFZs is most likely exerted by direct agricultural expansion from the AMZs. Households in the non-intervention domain reported walking longer durations to access both their furthest fields and their nearest fallows, which indicates that they were farming geographically farther from their homesteads and extending potentially into the buffer zones and/or PFZs. This presents important social implications beyond deforestation, especially for women, who have important roles in farming and harvesting crops in the DRC and who face increased safety risks when required to travel farther from their homes to cultivate land [48]. A previous study in the same area [28] found that villages that supported higher human populations had generally larger AMZ surface areas; therefore, human population size was likely not the main driver of a village's decision to expand its agricultural activities into the PFZ. There is therefore theoretical support that agricultural interventions aimed at boosting production have potential to decrease agricultural expansion into the PFZs.

In addition, the satellite analysis of tree cover loss presented two notable caveats that illustrate technical complexities of effective monitoring of the interventions based on satellite analysis alone. One caveat was related to the timing of our monitoring efforts, which was limited to a 5 year post-intervention period. Although we do not have evidence from the literature that a 5 year post-intervention period is insufficient, it is likely a minimal timeframe to comprehensively monitor change [44]. The timing of the study's interventions also presented some challenges for accurate monitoring. Non-intervention villages began their sensitization and formal participatory zoning activities in early 2014, which marked the second year of the post-implementation period for the intervention domain. Therefore, some land use decisions made by the non-intervention domain beginning in 2014 might reflect the community sensitization and forest zoning activities that co-occurred around that time.

A second caveat was related to satellite data quality and availability. The Landsat data archive grew enormously in 2013 with the launch of Landsat 8. A higher Landsat observation frequency is reflected by the difference between the user's and producer's accuracies of the forest loss class (table S4). Specifically, the higher producer's accuracy of 97.5 for the 2013–2018 period compared to 95.6 for the 2001–2010 period indicates slightly higher sensitivity of the change detection model to forest loss after the year 2013. Nevertheless, the differences in map accuracies were very small and were not reflected in the overall accuracies, which suggests that the forest loss results for both time intervals can be directly compared.

We did not find a reason for bias contributed by the non-random distribution of intervention and non-intervention village domains. Although we were only able to test for similarity in the proportion of primary forest land cover between domains for each zone type prior to the study, we believe that other characteristics of the study area likely enabled a somewhat consistent comparison between domains. For example, the villages located in both domains were similar in their remoteness and farther proximity to the region's larger market hubs of Kisangani and Boende; the only roads connecting them to these areas are unpaved, dirt tracks that connect the villages. In fact, the clustered distribution of villages in the two domains likely enabled a more spatially consistent interpretation of deforestation; we believe that it might have potentially been more difficult to attribute deforestation to a single village in one domain if the surrounding villages belonged to the opposite domain, where the potential for spillover effects could be quite high.

However, we note other limitations to our study. While conducting the household surveys, we were unable to interview as many households in the non-intervention domain as in the intervention domain due to non-responsiveness. In addition, a variety of other factors that were not accounted for in our study could play a significant role in the motivation and ability of farmers to adopt new techniques and/or self-limit their agricultural expansion. These factors might include measures of market access, off-farm income activities like the collection and sale of non-timber forest products, labor costs, land availability, and the quantity of agricultural products consumed versus sold [14]. Since we did not account for these factors, we cannot assume that our findings are transferrable to other geographical or socio-economic contexts outside our study site.

As increasing human populations exert pressure on Congo Basin forests, developing strategies that strengthen forest governance will be critical. Hence, voluntary agreements like the quid-pro-quo agreements used in this activity should strive to ensure that agricultural interventions are complemented by land management rights and rules that constrain the expansion of agricultural activity into nearby forests. Developing a more formal process for communities to manage their forest lands will encourage them to take ownership and responsibility for preserving the community forests that will extend beyond just a project-based approach [49]. These types of longer-term successes will be achieved best through education, active collaboration with local authorities and local and national institutions, and community empowerment through the direct participation of women, men and youth.

This research was made possible by the United States Agency for International Development (USAID) Central Africa Regional Program for the Environment (CARPE) Program (AID6660-A-13-00009 and NNX14AR46G). We would like to thank the Bill & Melinda Gates Foundation and USAID for cassava variety development. For assistance with the participatory micro-zoning delimitation, we would like to thank Zozime Esombo and Etienne Ingoli-Ingoli. For collection of household survey data, we would like to thank IITA's local facilitators and students from the local college, Institut Supérieur du Developpement Rural (ISDR-Djolu). The paper's lead author, Dr Janet Nackoney, was not working at USAID when the research for the current paper was conducted. The views and opinions expressed in this paper are those of the authors and not necessarily the views and opinions of the United States Agency for International Development.

The data that support the findings of this study are available upon reasonable request from the authors.