Heterogeneous impacts of community forestry on forest conservation and poverty alleviation: Evidence from Indonesia

1. Community forestry is a participatory approach aiming to achieve sustainable forest management while also reducing poverty among rural communities. Yet, evidence of the impacts of community forestry programmes on both forest con‐ servation and poverty alleviation is scarce, and there is limited understanding of impacts across different social and biophysical contexts. 2. We applied a matching method to assess the extent to which deforestation has decreased and village well‐being has improved as a result of Indonesia's commu‐ nity forestry scheme, Hutan Desa (Village Forest). We assessed five dimensions of well‐being: basic (living conditions), physical (access to health and education), financial (income support), social (security and equity) and environmental (natural hazard prevention). 3. We found that Hutan Desa was associated with reduced deforestation and pov‐ erty. ‘Win‐win’ outcomes were found in 51% of cases, comprising (a) positive out‐ comes for both forests and poverty, (b) a positive outcome for one aspect and a negligible outcome for the other, or (c) a positive outcome for poverty in areas where natural forest had already been lacking prior to Hutan Desa tenure. Benefits to forests and people systematically differed depending on land‐use zones, re‐ flecting subtle interactions between anthropogenic pressures and community livelihood characteristics. 4. In Watershed Protection Zones, which are dominated by subsistence‐based for‐ est livelihoods, community forestry provided mild conservation benefits, but re‐ sulted in the greatest improvements in well‐being through improved land tenure. In Limited Production Zones, community forestry provided modest benefits for on outcomes. Crucially, our study provides vital objective information for future policy develop‐ in and other community


| INTRODUC TI ON
The developing world is experiencing unprecedented degradation of the natural environment (Hoekstra & Wiedmann, 2014). While economic growth has lifted millions of people out of poverty, rural deprivation remains prevalent (Akram-Lodhi, Borras, & Kay, 2007).
In recognition of the dual objectives of reducing poverty and improving ecosystem conservation, developing country governments and international donors are promoting policies that involve communities in environmental management, such as community forestry (CF; Bowler, Buyung-Ali, Healey, Jones, & Knight, 2012;Sayer & Margules, 2017). The Indonesian government adopted CF in 2014, setting an ambitious target of 10% of the country's forests (12.7 Mha)  while also avoiding deforestation. This shift in land tenure is unprecedented in Indonesia, and is one of the largest CF policies of any tropical country . Around onethird of the proposed CF area is in the tropical forests of Kalimantan, Indonesian Borneo (MEF, 2018a). With much more land set to be allocated to CF, it is timely to evaluate whether Indonesia's policy as currently interpreted is fulfilling its main conservation and social objectives.
Empirical evaluations of CF have often focused on a single perspective; either the conservation outcome (e.g. Luintel, Bluffstone, & Scheller, 2018;Rasolofoson, Ferraro, Jenkins, & Jones, 2015;Santika et al., 2017), or well-being (e.g. Rahut, Ali, & Behera, 2015;Rasolofoson et al., 2017). Evaluations that combine socioeconomic development with environmental conservation outcomes are imperative to understand potential trade-offs or synergies between the two potential impacts, and how these vary across different social and landscape contexts (Agrawal & Chhatre, 2006). Nonetheless, many CF evaluations have focused on well-being aspects, and have typically done so via localized case studies using a limited subset of well-being measures (Bowler et al., 2012;Burivalova, Hua, Koh, Garcia, & Putz, 2017;Mongabay, 2017). For example, although there have been numerous studies examining CF programmes in Indonesia, these have focused mainly on financial outcomes and conflict (Table   S1). Assessments that integrate both conservation and well-being perspectives are rare (Gilmour, 2016;Newton et al., 2015).
Despite the rich literature detailing institutional arrangements for CF, there are few studies that examine the role of socioeconomic, market and biophysical factors in shaping actual outcomes (Hajjar & Oldekop, 2018;Hajjar et al., 2016;Newton et al., 2015). There is therefore limited objective information with which to guide the development of policies, monitor performance, and scale up implementation. Deeper understanding of what influences performance would also allow CF policies to be better tailored for communities to enhance both social and environmental outcomes (Ostrom & Nagendra, 2006). At present, there is a risk that CF policies that fulfil environmental objectives could result in negative impacts on poverty, or vice versa; yet comprehensive baseline assessments that enable reliable monitoring and evaluation are often lacking.
For CF policy evaluation to be robust, any attribution of changes in forest cover or poverty due to the policy should be compared to the counterfactual condition of no CF or other confounding factors (Baylis et al., 2016;Lan & Yin, 2017). Policy instruments usually target certain populations with specific underlying characteristics, both conservation and well-being because restrictions on timber harvest due to Hutan Desa designation reduced the financial well-being of logging communities.
The greatest conservation benefits were experienced in Permanent or Convertible Production Zones, but well-being improvements were minimal. Here, living conditions and environmental well-being were reduced due to pressure to intensify agricultural production under increased land scarcity in these predominantly cash crop-oriented communities. 5. Our results highlight the spatial and contextual variation in impacts of community forestry policies on poverty alleviation and forest conservation outcomes.
Crucially, our study provides vital objective information for future policy development in Indonesia and other tropical countries implementing community forestry schemes.

K E Y W O R D S
avoided deforestation, human well-being, impact evaluation, multidimensional poverty, rural development, sustainable development, tropics thereby introducing selection bias when measuring performance and masking true outcomes. Matching techniques select controls with similar observed characteristics as the populations receiving the intervention, and thus can overcome such bias by providing a fair and reliable way of comparing sites with different intervention exposure (Dehejia & Wahba, 2002). Matching methods have been used widely in assessing the impact of protected areas (e.g. Ferraro & Hanauer, 2011) and certification of logging operations (e.g. Miteva, Loucks, & Pattanayak, 2015), but their use in CF evaluation is uncommon (e.g. Rasolofoson et al., 2015;Rasolofoson et al., 2017).
Here, we determine the extent to which CF in Indonesia has resulted in both reduced deforestation and improved well-being.
Well-being is multidimensional, and incorporates economic, social and environmental perspectives (Alkire & Santos, 2014). The aspects of well-being examined include: (a) basic (living conditions), (b) physical (access to health and education), (c) financial (income support), (d) social (security and equity) and (e) environmental (natural hazard prevention). We focus our socio-ecological analysis on Hutan Desa or 'Village Forest', the main CF scheme being applied in Kalimantan and elsewhere in Indonesia. Hutan Desa aims to reduce poverty and improve the social welfare and forest use rights of marginalized communities by allowing forests to be protected and managed communally through the authority of a village head (Myers & Ardiansyah, 2014). Hutan Desa licences are granted by the Ministry of Environment and Forestry through a rigorous selection process, where the approval of the licence is based on the provision of a management plan with goals towards sustainable development and conservation, strong participation from the local community, and collaborative relationships with external partners, such as nongovernmental organizations (Siscawati et al., 2017). We assessed: (a) the association between Hutan Desa, deforestation rates and the change in different aspects of village well-being; and (b) variation of these associations in different areas in Kalimantan. From our results, we provide recommendations to improve CF policy that reflect how performance varies in different contexts. Our assessment included 41 Hutan Desa management areas (total 1,300 km 2 ) that had been approved and facilitated by external organizations between 2009 and 2014, the latest period corresponding to sufficient forest cover and well-being data. The spatial unit of the analysis was 1 × 1 km 2 grid-cell resolution for the deforestation outcome, and village boundaries for well-being. We assessed the performance of Hutan Desa for each land-use zone, because regulation on the amount of timber extraction permitted within Hutan Desa boundaries differs by zone. Timber extraction is prohibited in Hutan Desa granted in Watershed Protection Zone, but in Production Zones  Figure   S1). In the Permanent or Convertible Production Zones, plantations, both outside and within oil-palm concessions (livelihood categories 3 and 4), dominate (Figure 1c), and the proportion of non-indigenous people, including migrants from Java and Sumatra, is also relatively high ( Figure S1).

| Data on well-being
Indicators of well-being were derived from Potensi Desa (PODES; 'Village Potential') data from the Indonesian government. PODES is a publicly available village-level socioeconomic dataset collected every 2-4 years by the Bureau of Statistics Indonesia (BPS Indonesia, 2017). The data represent the overall socioeconomic conditions in a village, and thus do not capture the variation and disparity in socioeconomic indicators among different sub-villages or households. Rather, the data provide a useful way to compare village administrative units over large spatial extents and over time.
PODES data are collected from village head offices. The reliability of data, therefore, may vary across different villages, resulting in potential for bias. However, should this bias propagate sufficiently to affect the outcome of analysis, then we would expect it to override the signal of CF (i.e. no impact of Hutan Desa would be observed), as opposed to overstating (or understating) the impact. PODES is the best socioeconomic dataset available at sufficient spatial resolution in Indonesia. The data have been used extensively in rural development studies (Table S2) and have proven useful for monitoring the various socioeconomic impacts of land-use policy interventions (e.g. Barron, Kaiser, & Pradhan, 2009;Jagger & Rana, 2017;Miteva et al., 2015).
We used 16 indicators derived from PODES 2008 and 2014 collections as proxies for the change in five aspects of well-being (Table 1). The choice of indicators and directionality of the effects on well-being was informed by existing methodologies used to assess poverty, such as the Multidimensional Poverty Index (Alkire & Santos, 2014), the Sustainable Livelihood Approach (Scoones, 1998), and the Nested Spheres of Poverty tool (Gönner, Haug, Cahyat, Wollenberg, F I G U R E 1 The distribution of Hutan Desa and land-use zones across Kalimantan by February 2015 (a); village primary livelihoods sectors (b); and the break down of livelihoods according to zone (c). Black lines in the maps indicate provincial and national boundaries TA B L E 1 Potensi Desa (PODES) variables used as proxies for five aspects of well-being: basic, physical, financial, social and environmental. w k denotes the directional effect of the change in indicator k that defines improvement in well-being. If w k = 1, then positive change (i.e. an increase) in indicator k represents improvement in well-being. If w k = −1, then negative change (or reduction) in indicator k represents improvement in well-being

Aspect of well-being PODES variable (k)
Variable response

Water pollution in the last year
Categorical (1 = none, 2 = mild, 3 = severe) −1 Well-being improves when there is a change to a lower water pollution category, e.g. a change from category 2 (mild) to 1 (none)

Air pollution in the last year
Categorical (1 = none, 2 = mild, 3 = severe) −1 Well-being improves when there is change to lower air pollution category, e.g. a change from category 2 (mild) to 1 (none) a Per 1,000 people.  Figure S2). Further details on indicators and justification for their directionality are provided in Supplementary Materials and Table S3. We recognize that more subjective, non-material, indicators exist to measure poverty and human well-being (Biedenweg et al., 2014;Breslow et al., 2016;Chan, Satterfield, & Goldstein, 2012). However, these are difficult to aggregate at the village level and are not available within the PODES dataset. Our analyses should therefore be interpreted as documenting changes in objective, material aspects of well-being.

| Data on deforestation
We used deforestation rate as an indicator of forest conservation.
We assessed the impact of Hutan Desa on deforestation rates based on data between 2010 and 2014 to roughly match the time period covered by the PODES data. Deforestation rates were derived from the forest loss variable in the Global Forest Change (GFC) dataset (Hansen et al., 2013). The GFC dataset does not distinguish between the loss of natural forest and that of tree plantations. Therefore, to restrict our analysis to natural forest loss, we used the extent of natural forest in 2010 derived from Margono, Potapov, Turubanova, Stolle, and Hansen (2014). Natural forest is defined as a mature forest that has not been completely cleared in the last 30 years. The GFC dataset was then restricted to the extent of natural forest in 2010. The GFC and natural forest extent data both have spatial resolution 30 × 30 m 2 , and we analysed forest cover change annually in hectares at a spatial resolution of 1 × 1 km 2 . We focused on deforestation of intact natural forest, that is, 1 × 1 km 2 grid-cells with >80 hectares of natural forest in the beginning of Hutan Desa tenure. Among the 41 Hutan Desa areas we examined, 32 were mostly (>70%) covered by natural forest in 2010, and were included in our deforestation assessment. Because interannual climate variation between 2010 and 2014 was within the normal range (mostly categorized as non-dry years, Figure S3), we assumed that deforestation was largely driven by anthropogenic activities during this period, rather than drought-induced fire.

| Confounding variables
We controlled for potentially confounding variables in the assessment of Hutan Desa performance in terms of which locations were selected for Hutan Desa and the outcome being measured (Table 2). as proxies for sociopolitical factors. Decentralization of government functions to provincial levels has been identified as a key driver of agriculture expansion (Moeliono & Limberg, 2012;Resosudarmo, 2004). Economic growth can also vary across different provinces (Suryahadi, Suryadarma, & Sumarto, 2009). NGO partnerships are pre-requisite to applying for a Hutan Desa license (Siscawati et al., 2017), and recognized as an important factor in improving the performance of community forestry (Akiefnawati et al., 2010). Therefore, the net impact of Hutan Desa tenure should account for the NGO influence in the deforestation and well-being outcome within Hutan Desa area. The indicator of well-being prior to Hutan Desa designation provides a baseline to control for initial conditions that may bias impact estimates.
We used elevation (variable ELEV), slope (variable SLOPE), proximity to large cities or arterial roads (variable CITY) and human population density (variable POP) as proxies for market value.
Communities living in areas closer to roads, at lower elevation and flat terrain, and in areas of higher human population density tend to have better socioeconomic welfare than those living in remote areas without exposure to the market economy (Resosudarmo & Jotzo, 2009;Sunderlin, Dewi, & Puntodewo, 2007).
We used long-term seasonal rainfall patterns (variables DRY and is also an important factor driving agricultural conversion (Carlson et al., 2013). The decline in forest area in Kalimantan had been partly attributed to an increase in agricultural area, much of which is linked to transmigration sites (Dennis & Colfer, 2006). Extreme climate, such as prolonged dry months and heavy rains, can decrease agriculture productivity (Iizumi & Ramankutty, 2015;Oettli, Behera, & Yamagata, 2018) and increase natural disasters such as wildfire and flood (Field, Werf, & Shen, 2009). Such extreme events can lead to reduced economic growth and adversely affect community social welfare (Herawati & Santoso, 2011).

| Matching method
We employed propensity score matching (Rosenbaum & Rubin, 1983) to select a set of control grid-cells outside Hutan Desa boundaries that exhibited the same baseline characteristics as grid-cells with Hutan Desa. We used a nonparametric generalized boosted regression model implemented in the R-package gbm (Ridgeway & Southworth, 2015) to generate the propensity scores using the variables described in Table 2   locations with and without Hutan Desa had a higher degree of overlap after matching than before matching ( Figure S4 and Table S4).
After  Desa is considered to be effective at improving a single indicator of well-being k if the difference between the change in the value of indicator in the treated village (C j,k ) and the control village To assess whether or not our estimate based on matching was robust to the possible presence of an unobserved confounder, a sensitivity analysis was applied based on the principle of randomization inference (Keele, 2014). The results indicated that if an omitted confounding variable does exist, it has to increase the likelihood of the non Hutan Desa village to receive intervention by a factor greater than at least 1.81 (Table S5). This magnitude suggests that our matching method is robust to hidden bias.
Detailed explanations of the matching method are provided in the Supplementary Materials.

| Association between Hutan Desa and avoided deforestation
Hutan Desa was associated with reduced deforestation rates overall compared to the counterfactual (Figure 2a).

| D ISCUSS I ON
Our study provides the first broad jurisdictional assessment of the impacts of CF on forest conservation and rural well-being seen through a multidimensional lens. We found that Indonesia's main CF scheme, Hutan Desa, was associated with both reduced deforestation and improved indicators of well-being for Kalimantan's rural communities (Figure 4). Positive outcomes for both conservation and well-being, or a positive outcome for one aspect and negligible for the other, were detected in 51% of cases (sum of cells A in Figure 4), suggesting that, under some circumstance, forest conservation can go hand-in-hand with poverty alleviation. Negligible outcomes for both conservation and well-being were detected in 14% of cases (sum of cells B in Figure 4). Conversely, trade-offs between conservation and well-being were detected in 17% of cases (sum of cells C in Figure 4); with 13% of cases poverty alleviation occurred at the expense of forest, but in only 4% of cases avoided deforestation was achieved at the expense of poverty outcomes. Negative outcomes Hutan Desa on deforestation imply that deforestation rates within Hutan Desa are considerably lower, higher than or similar to those outside Hutan Desa areas with similar baseline biophysical characteristics. Positive, negative or negligible effects of Hutan Desa on well-being imply that improved well-being in villages with Hutan Desa occurs considerably faster, slower or at a similar pace as that in villages without Hutan Desa with similar baseline biophysical and socioeconomic characteristics. A total of 41 Hutan Desa management areas were assessed for well-being outcome and 32 Hutan Desa areas were assessed for deforestation (as intact natural forest was lacking in nine Hutan Desa areas before tenure). Positive outcomes for both conservation and well-being, or a positive outcome for one aspect and negligible for the other are labelled "A" (total 51% of cases. Negligible outcomes for both conservation and well-being are labelled "B" (14%). Trade-offs between conservation and well-being are labelled as "C" (17%), while negative outcomes for both conservation and well-being or a negative outcome for one aspect and negligible for the other are labelled "D" (18%) Production Zone. Our findings are comparable to those of Ferraro and Hanauer (2011) on protected areas in Costa Rica (mainly composed of national parks and community-based protected areas, IUCN categories II and VI, respectively), which highlighted protected areas with the most avoided deforestation to be associated with the least poverty alleviation, and protected areas where conservation effectiveness was limited were associated with the most improved community well-being.

| Why do benefits vary across land-use zones?
The benefits of Hutan Desa were moderated by baseline conditions and pressures in the different land-use zones. Hutan Desa areas in the Watershed Protection and Limited Production Zones tend to be located in areas of high forest cover and away from major cities and roads ( Figure S7). In these areas, anthropogenic pressure is generally low to moderate, and forest encroachment mainly arises from illegal logging and shifting cultivation by local farmers (Purwanto, 2016;Resosudarmo, 2004). Because anthropogenic pressure is mild, any reduced deforestation rates are also expected to be mild with the introduction of Hutan Desa tenure.
Hence, community forestry schemes are well placed to maintain forest cover simply because the pressures on forests are inherently low. Communities here often lack basic facilities, such as health clinics and schools, have limited access to electricity, poor housing conditions and are dependent on wood fuel ( Figure S8).
Despite these conditions, malnutrition among infants is rare ( Figure S8), most likely due to high food self-sufficiency and large variety of micronutrient-rich food sources, as is often typical of forest-dependent communities (Harper, 2002;Ickowitz, Powell, Salim, & Sunderland, 2014;Ickowitz, Rowland, Powell, Salim, & Sunderland, 2016;West, 2006). These communities often rely on subsistence livelihoods (farming, fishing and gathering of forest products; Figure 1c). Hutan Desa facilitation by external organizations in the Watershed Protection and Limited Production Zones has likely led to improved financial well-being according to our indicators and improved infrastructure compared to the counterfactual. Hutan Desa licenses have also provided tenure clarity and reduced illegal logging and forest encroachment by people from outside the village, which could explain why social conflict has declined and the environment reported to be better preserved than in areas without Hutan Desa ( Figure S6d). This pattern of improved income and infrastructure provision and reduced social conflicts has also been observed in other studies of forest-dependent communities with CF (e.g. Rahut et al., 2015;Rasolofoson et al., 2017).
Areas designated as Permanent or Convertible Production Zone are typically located in places of low forest cover, near to markets, cities and major roads, where oil palm plantations dominate ( Figure   S7) and infrastructure is nearby ( Figure S8). Competition for land is high (Sahide, Supratman, Maryudi, Kim, & Giessen, 2016) and typically involves a complex network of actors and stakeholders (Santoso, 2016). Because anthropogenic pressure is strong, deforestation is typically high, and there is potentially much to be gained for conservation with the introduction of effective Hutan Desa tenure. A large proportion of communities in this zone depend on cash crops, mainly oil palm (Figure 1c), giving them livelihood options outside the forestry sector. However, employment opportunities are often distributed unequally among community members (Obidzinski, Andriani, Komarudin, & Andrianto, 2012); inequities that may reduce overall well-being in intensively managed landscapes (Rasmussen et al., 2018). These factors provide an explanation as to why improvement to financial well-being is comparatively lower in this zone due to Hutan Desa designation compared to locations in the Watershed Protection Zone (Figure 3). A reliance on cash crops, which encourages people to purchase processed foods with limited nutritional value and results in poor environmental conditions, may also explain higher infant malnutrition in these areas ( Figure S8a; Gómez et al., 2013;Ickowitz et al., 2016). Forest protection can induce a spillover effect of agricultural expansion and intensification to sub-optimal areas that often require high fertilizer inputs (Didham et al., 2015;Duncan, Dorrough, White, & Moxham, 2008). Excessive fertilizer usage has negative environmental effects particularly on water quality (Obidzinski et al., 2012), which explains why water pollution levels had reportedly increased in Permanent or Convertible Production Zone with Hutan Desa than those without ( Figure S6e), particularly on peatland. The analysis of PODES data also indicates that the number of agricultural labourers was markedly reduced in villages with Hutan Desa in Permanent or Convertible Production Zone compared to villages without ( Figure S6d). Agricultural intensification often requires higher labour input to increase production per hectare (Rasmussen et al., 2018). Given shortage in agricultural labourers, and combined with a decrease in farmland due to land scarcity, this could lead to a decrease in household production and income (Angelsen & Kaimowitz, 2001). These impacts are reflected also in the reduction in the basic well-being indicator in these production zones with Hutan Desa relative to those without (Figure 3).

| Implications for CF investments
The Indonesian government has pledged to allocate extensive land to CF by the end of 2019, which presents great challenges in terms of capital requirements and distribution. Our study provides objective information to guide the focus and priorities of the Hutan Desa programme and investments as it develops (as summarized in Figure 5, and translated in Bahasa Indonesia in Figure S9), as well as lessons that are broadly applicable to schemes in other national contexts. For communities living within the boundaries of Watershed Protection Zones, investment in Hutan Desa or payments for ecosystem services could be directed towards improving basic living conditions (e.g. sanitation, education and health programmes; Sunderlin et al., 2005). As the proportion of indigenous communities is higher in Watershed Protection Zones and Limited Production Zones ( Figure   S3), Hutan Desa can additionally provide a platform for enhancing recognition of indigenous wisdom and knowledge of forest and nature (Boedhihartono, 2017 Figure 5 and Figure S9). However, the extent of deforestation that is avoided due to Hutan Desa in this zone is also greatest (Figure 2b). This implies that safeguarding forests  (Field et al., 2016;Herawati & Santoso, 2011). During the severe El Niño drought year in 2015 ( Figure S3), Hutan Desa in Permanent or Convertible Production Zone, particularly on peatland, performed poorly in preventing deforestation due to wildfire (Gaveau et al., 2018;Santika et al., 2017). Therefore, the avoided deforestation we detected reflected anthropogenic factors, and might not hold during periods of extreme climatic events.
Some of the well-being indicators we used focus on the prevalence of socioeconomic programmes, such as cooperative schemes, credits and small businesses. While information about the rates of community participation could provide a better proxy for well-being than merely the prevalence of these programs, such data are unfortunately not available in the PODES dataset over the spatial and temporal scale of our study. Moreover, we were restricted to objective and material well-being indicators. There is growing recognition of the need to include subjective non-material indicators in poverty assessment, such as intrinsic values related to culture, ethnicity and social embedding or spiritual attachment to places (Biedenweg et al., 2014;Breslow et al., 2016;Russell et al., 2013). Unfortunately, in many developing countries, including Indonesia, such data are typically unavailable at broad geographical scales. Where such data do exist, we encourage researchers to include them within CF assessments, and explore ways to scale up beyond local-level appraisals in order to provide comprehensive assessment of both objective and subjective well-being.

| Conclusions
Our study highlights that the successful implementation of CF, where forest conservation is implemented, will require investment in different activities in different land-use zones to support transition of livelihoods and to prevent exacerbating environmental degradation, poverty, and socioeconomic disparity. We provide a robust framework for monitoring and evaluating CF, and an appraisal of performance over the first five years of Indonesia's leading CF scheme, which can serve as a crucial baseline for long-term monitoring. This monitoring and evaluation framework has broad applicability for other countries implementing community forestry.

CO N FLI C T O F I NTE R E S T
Nothing to declare. We thank editors and reviewers for providing valuable inputs on our manuscript. We also thank the many stakeholders in government, NGOs and local communities in Kalimantan that have shared their insights in community forestry and provided useful feedback on our analyses.

DATA ACCE SS I B I LIT Y
The data used for analyses are publicly available under license via Indonesia's Bureau of Statistics (https ://mikro data.bps.go.id/mikro data/index.php/catal og/PODES ), and University of Maryland (http://earth engin epart ners.appsp ot.com/scien ce-2013-globalfores t/downl oad_v1.6.html). Data linking specific community forest areas with poverty and/or deforestation outcomes will not be made