1.1 Background

Climate change is an increasingly urgent global environmental and humanitarian problem, threatening disruption of ecological processes, alterations of land-based and aquatic food production systems, increasing risks to human health, challenges to maintaining and improving human wellbeing, risks to biodiversity and species survival, undermining of economic growth and resilience, and increasing conflict and violence (IPCC 2014a). Tropical deforestation and forest degradation are one key part of the problem, with 12% of total GHG emissions in the period 2000–2009 coming from forests and other land uses (IPCC 2014b, 16) and therefore potentially an important part of the solution (Goodman and Herold 2014).

When REDD+1 was launched at the Bali COP in 2007, it involved several innovations aimed at overcoming decades of failure in attempts to reduce tropical deforestation. REDD+ would raise billions of US dollars to pay the opportunity costs of forest conservation. It would create a performance-based, conditional system for delivering rewards to those stakeholders who avoid elimination or degradation of forests. Remote sensing technology would be applied toward verifying avoided deforestation against a pre-established reference level. Policies and measures would be implemented at the national and subnational levels to lay the institutional groundwork for REDD+ and exert leverage toward assuring its success.

Seven years after the Bali COP, there is good and bad news to report. On the positive side, the idea of REDD+ has taken hold and grown, reflected in the large number of countries and initiatives participating in the REDD+ experiment, the many publications written about it and the large commitment of donor funds (Angelsen and McNeill 2012, 32–33). There has been explosive growth in the number of REDD+ subnational initiatives and there are now more than 300 (Simonet et al. 2014). REDD+ has been given prominent attention year by year in UN technical SBSTA meetings and COP negotiations because it has been viewed as a leading option for addressing climate change early and affordably. Although negotiations at the COPs from 2008 to 2012 led to incremental progress, the breakthrough finally came at the 2013 COP in Warsaw, with progress made on six key issues involving the coordination of: financial arrangements, transparency and safeguards, national forest monitoring systems, verification at the international level, institutional arrangements, and drivers of deforestation (Stolle and Alisjahbana 2013). Brazil experienced a 79% reduction in its rate of deforestation between 2004 and 2013 (INPE 2014a) and it has made the largest contribution to reduced GHG emissions of any country to date (Springer and Wolosin 2014) (see Box D on Brazil). It is doubtful this achievement can be counted as a REDD+ success because the impetus predates REDD+, and the policy leverage applied is not directly related to REDD+. Nevertheless it demonstrates conclusively the high potential for massive GHG emissions reduction in the forest sector. There are other, more minor, ‘deforestation success stories’ that are attributed to REDD+ (Boucher et al. 2014).

The bad news is that for all the investment in REDD+-readiness in the last seven years, implementation of REDD+ has fallen far short of what was hoped. The large amount of funding meant to drive REDD+ has not yet materialized. To date, less than USD 10 billion has been mobilized (Voluntary REDD+ Database 2014; Norman and Nakhood 2014), whereas it is estimated that USD 5.0–12.5 billion is needed annually (Stern 2006, 217; Angelsen 2013, 13). At current prices, the supply of forest carbon credits is 13–39 times larger than demand and there will be a USD 15–48 billion funding gap in the coming years (IFF 2014, 8). Public sector funding of REDD+ was meant to be temporary, but it continues to fill the gap because of failure of the market for forest carbon credits (both voluntary and compliance) to develop. Many REDD+ subnational initiatives are ‘treading water,’ waiting for conditions to be more propitious. Some are drifting away from the concept of REDD+ (Sunderlin et al. 2014a). Formulation of REDD+ policy at the national level has in some cases met stiff resistance by forces aligned against it (Brockhaus et al. 2013). Systems for MRV meant to assure the efficient functioning of REDD+ are substandard in some countries (Romijn et al. 2012; Joseph et al. 2013). Perhaps most importantly, although a binding global climate change agreement would likely propel the regulatory environment necessary for funding REDD+, the prospect of an agreement at the Paris COP in 2015 is said to be dim by some sources (CICERO 2014; Davenport 2014).

It is unclear where REDD+ is going, and there are equal measures of hope and discouragement about prospects for fulfilling its lofty goals. At this juncture, scientific evidence on REDD+ implementation is critically important to provide insights on what is going right, what is going wrong and to propose course corrections. This book provides preliminary information on ‘REDD+ on the ground’ in the form of REDD+ subnational initiatives. These initiatives seek to move beyond readiness to actually reduce forest carbon emissions and are thus critical empirical reference points on the successes and failures of REDD+ at delivering both reduced emissions and co-benefits for local livelihoods and environmental services. Ultimately, implementation of REDD+ will depend on decisions made at the subnational and local levels, as with all climate mitigation strategies (UNDP 2014). Independent scientific research on subnational initiatives is required to assess whether they provide ‘proof of concept’ or reason for concern about all of REDD+, as well as to extract lessons for other forest conservation efforts.

Drawing on research in Brazil, Peru, Cameroon, Tanzania, Indonesia and Vietnam, this book describes 23 subnational REDD+ initiatives, including their institutional, socioeconomic and biophysical context, and their structure, strategy and implementation. We discuss the concerns and challenges facing these initiatives, as well as the lessons they offer for REDD+. For each country, the national context is briefly described, and the final chapter draws some synthesis conclusions on the nature of the challenges they face in fulfilling their goals.

This introductory chapter is structured as follows. In Section 1.2, we describe the research project that has produced the case study information in this book (namely Module 2 on subnational initiatives of CIFOR-GCS), with particular attention to how we defined our sample and data collection methods. In the following section (Section 1.3), we describe the locations of the initiatives and discuss implications for the generalizability of the research findings. We then describe the structure of the book and the contents of the 22 case report chapters, and explain the aim of the concluding synthesis chapter (Section 1.4). We close with thoughts on the urgent need for effective forest-based climate change mitigation, and how this book can contribute toward that goal (Section 1.5).

1.2 Sample and methods

The aim of CIFOR-GCS, begun in 2009, is to identify the challenges and enabling conditions for REDD+ to achieve outcomes that are effective, efficient and equitable, and fulfill co-benefits such as protection and improvement of livelihoods, tenure and gender rights, and biodiversity. (These outcomes are abbreviated ‘3Es and co-benefits’ or ‘3E+.’) The results are meant to guide policy makers, governments and initiative proponents in designing and implementing REDD+. CIFOR-GCS is being conducted in 14 countries2 and is composed of four research modules and one information sharing module.3

This book summarizes field data gathered in 2009–2013 by CIFOR-GCS’s Module 2 on subnational initiatives. Module 2 is working in the CIFOR-GCS core countries, that is: Brazil, Peru, Cameroon, Tanzania, Indonesia and Vietnam. It is carrying out its research in collaboration with 23 subnational initiatives (Figure 1.1, Table 1.1 and Appendices 1–3 and 5–6). The proponent organizations developing these initiatives are predominantly private nonprofit organizations (13), but also include private for-profit (4), private nonprofit/government (2), government (2), government to government partnership (1), and public bilateral (1) organizations. Seventeen of the initiatives operate at the project scale (smaller than and not developed as part of a government administrative unit), whereas six are jurisdictional (encompassing a government jurisdictional unit). Funding to date comes largely from bilateral public sector funding. Other sources of funding include philanthropic organizations, private companies, international organizations and subnational governments (Appendix 1). The initiatives range widely in area from Acre’s 157,490 km2 (approximately the area of Bangladesh) to SE Cameroon’s 28 km2, encompassing just two villages. Seventeen of the initiatives are in the tropical wet climate region, whereas all six of the Tanzania initiatives are in the tropical dry region (Appendix 2).


Figure 1.1 CIFOR-GCS abbreviated and formal names and locations of subnational initiatives.

Table 1.1 Forest area, forest loss and carbon emissions in the CIFOR-GCS initiative sites.




Forest area 2000b

Forest loss






(Gg C)d









Bolsa Florestae



































Madre de Dios














Mt Cameroon







SE Cameroon






































































Rimba Rayai







TNC within BFCP







Ulu Masen







Cat Tien






a Here, areas (in ha) are calculated using the cells of the Hansen raster data, assuming equal cell sizes of 30 x 30 m. Hence, these sizes may differ from size calculations based on projected GIS data.

b Forests are defined as areas with >10% tree cover.

c Forest loss is defined as a change from >10% tree cover to ~0% tree cover.

d Gg C = gigagrams of carbon (1 Gg = 1000 tons).

e Here, the total Bolsa Floresta Program area is included, although only one part of the total program (i.e. Juma) is considered an official REDD+ initiative.

f Only the individual plots in the total initiative area are included.

g Only the protected area within the total initiative area is included.

h Only the CIFOR-GCS sample villages within the total initiative area are included.

i Only the ‘carbon accounting area’ was included, which corresponds to the area known as ‘Proposed project area in 2010.’

Box A
Estimating forest loss and carbon emissions

Astrid B Bos and Valerio Avitabile

To compare forest area, forest loss and carbon emissions associated with that loss in the various REDD+ intervention areas of CIFOR-GCS, there is a need for one consistent method. Here, we follow FAO (2000) and define forests as areas with at least 10% tree cover. We make use of existing global datasets that cover the entire tropics and thus all of the initiatives in the CIFOR-GCS sample. Potentially higher quality, local data are not always available or directly comparable due to different methods and assumptions.

The intervention areas vary significantly in location, size and composition (i.e. ranging from complete provinces to a group of relatively small scattered plots). These differences may have a considerable influence on the forest loss and emission values.

Forest area and forest loss

Forest area and forest loss in all of the CIFOR-GCS sites were estimated on the basis of the dataset provided by Hansen et al. (2013), which contains the results of a time series analysis of Landsat 7 ETM+ images on global forest extent and forest change between 2000 and 2012 at 30 m spatial resolution. There has been significant debate over the extent to which this dataset actually represents forest and deforestation in the tropics. We use these data mindful of the fact that there is a controversy, yet believing these are currently the best globally comparable data for describing forest cover change at the 23 sites. We define deforestation as tree cover loss (according to the Hansen et al. (2013) data on global forest cover loss) that occurred in areas with at least 10% tree cover in 2000, consistent with the FAO definition of forests (FAO 2000).

Carbon emissions

Carbon emissions from aboveground biomass due to forest loss at the CIFOR-GCS sites were estimated by multiplying the area of forest loss (activity data) obtained from the Hansen et al. (2013) dataset with the respective forest carbon stock density values (emission factors). Carbon stocks at the CIFOR-GCS sites were derived from the IPCC Tier 1 default values (IPCC 2006). The activity data were stratified by continent and ecological zones according to the FAO Global Ecological Zone map (FAO 2001), and the respective average IPCC aboveground biomass density value for forest was applied to each deforestation unit, using a conversion factor of biomass to carbon of 0.5. The estimates refer only to the carbon emissions from aboveground biomass and do not include other carbon pools (belowground biomass, litter, dead wood, soil carbon) or emissions from forest degradation. Emissions from aboveground biomass represent the major source of carbon to the atmosphere in most forest types, but in specific contexts the total carbon emissions can be substantially higher, such as in peat forests where large amounts of carbon are stored in the soil.

Module 2 has adopted “before–after/control–intervention” (BACI) as its core method. Through this counterfactual approach, Module 2 aims to provide robust evidence on the performance of subnational initiatives in making REDD+ effective, efficient, equitable and able to provide co-benefits such as poverty reduction, improved livelihoods, secure tenure and biodiversity protection (the so-called 3E+ goals of REDD+). The approach involves comparing changes before and after the introduction of REDD+ in forests, villages and households that are inside the intervention area for REDD+ with matched areas that are outside (controls) (Jagger et al. 2009, 2010). At most of the 23 sites, there are other conservation and development initiatives ongoing in parallel to REDD+. This raises challenges for the attribution of outcomes to the REDD+ initiative, both in our own impact evaluation and in terms of the MRV of carbon additionality and co-benefits. In this book, we are describing baseline conditions in the intervention area and are thus presenting only the ‘before’ data from intervention villages and households. In the course of 2015, Module 2 will be analyzing the full panel data in order to assess the impacts of the REDD+ interventions. Case reports are based primarily on CIFOR-GCS research, including information obtained through interviews with villagers, proponents, government officials, other stakeholders and critics of REDD+. All unattributed information in the case reports is based on CIFOR-GCS research. Where we draw information directly from secondary literature, we cite those sources.

1.2.1 Selecting REDD+ countries

Module 2’s target countries were chosen as part of a CIFOR-GCS-wide exercise to identify the optimum countries for all modules. Among the criteria were the following: (i) key tropical forest countries, in particular, those that are pioneering REDD+ (e.g. Brazil and Indonesia); (ii) balance in the number of countries in each of the three main continental regions (Latin America, Africa and Asia); (iii) diversity of stages on the forest transition, for example, a relatively high rate of deforestation and degradation (e.g. Indonesia) and relatively stable forest cover (e.g. Vietnam); (iv) countries with sufficiently stable governance and security conditions to enable productive research; and (v) strong donor interest (Sunderlin et al. 2010, 19). In addition to these criteria, Module 2 favored the selection of countries where the presence of a CIFOR office could support field research activities (initially Bolivia, Brazil, Cameroon, Indonesia and Vietnam) and where REDD+-readiness was sufficiently far along to justify site-level research (which is why we did not choose DRC). Module 2 began to work in Bolivia, but when the government turned against REDD+ in 2010, we switched to Peru.

1.2.2 Selecting subnational initiatives

Candidate subnational initiatives were chosen on the basis of the following five criteria: (i) they conform to our operational definition of the term ‘REDD+’;4 (ii) they intend to monitor, report, and/or transact reductions in carbon emissions or increases in carbon stock (in a quantified manner); (iii) they define site boundaries and intervention villages before the beginning of our field research; (iv) REDD+ incentives were not planned to begin until after May 2010, assuring us a risk-free period in which to collect the ‘before’ data; and (v) the REDD+ incentives had a reasonable chance of being implemented and maintained in the subsequent 1.5 years (July 2010 – December 2011) (Sunderlin et al. 2010, 19–20). The combination of criteria (3) and (4) meant that only initiatives in a particular stage of development were eligible for our sample, and we essentially selected all initiatives in our study countries that we could verify were in that stage.

1.2.3 Selecting villages and households

Our core method was to collect data at the household and village levels in both intervention and control villages. (The 17 initiatives where we collected both village and household data we called our ‘intensive’ sites.) However, in order to expand our sample within our available research budget, we gathered data only at the village level and only in intervention villages at five additional sites. (These we called our ‘extensive’ sites.) Finally, we also took advantage of an opportunity to collect data at a site where REDD+ was already well advanced. One among the 23 initiatives (Bolsa Floresta in Brazil) is not part of the BACI analysis. Bolsa Floresta had already introduced REDD+ conditional incentives when the research began, meaning it was not possible to collect ‘before’ data at this site. We chose to do research at Bolsa Floresta because it was the first REDD+ initiative in Brazil to make direct conditional payments to households. Appendix 3 shows the distribution of the 23 initiatives across countries, type of study design (intensive, extensive, non-BACI), and the number of villages and households researched.

The selection of villages and households at intensive sites followed five steps. First, at each site, the field research teams identified a pool of 15 candidate REDD+ villages. In initiatives that cover a large region, they identified the set of villages where direct interventions are planned and where recent deforestation rates are average or higher than average for the initiative or region, because proponents had identified this as a key criterion for site selection (Lin 2012).

Second, for this set of candidate REDD+ villages, field teams used their knowledge and understanding of site characteristics to identify a pool of potential control villages that were not targeted for the REDD+ intervention but were similar in terms of market accessibility, deforestation pressures and socioeconomic factors. However, potential controls had to be far enough away that they would not be affected by direct spillovers or leakages (i.e. activity shifting) from the intervention area.


Small-scale farmers in Mato Grosso being interviewed by the Brazilian news station, Rede Globo, about their agroforestry production. (Icaro Cooke Vieira/CIFOR)

Third, the field teams collected data on 22 key characteristics considered likely to influence both initiative placement, and land use and welfare outcomes. These data were collected from secondary data sources, key informant interviews and by using other rapid rural appraisal techniques. Examples of variables in this data collection include population, village area, ethnicity, distance to roads and markets, forest dependence, forest cover, experience with a forest conservation NGO and major deforestation threats.

Fourth, we compiled these rapid rural appraisal data on the 30 villages per site (15 intervention and 15 control) for each country, in order to statistically match REDD+ villages to similar control villages, based on a set of key characteristics related to the probability of being selected for REDD+ interventions. (The rapid rural appraisal data for a few sites arrived later, and in those cases, villages were matched within the site.) The characteristics that ended up being most influential in matching were: (i) deforestation pressures; (ii) presence of NGO; (iii) forest tenure; (iv) number of village organizations; (v) population; (vi) village forest cover; (vii) forest dependence; and (viii) distance to main road.

Finally, in each site, we selected the four intervention and four control villages that were most closely matched, i.e., that were closest in terms of the Mahalanobis metric estimated at the country level. (One of the four intervention villages was included on the basis of having the highest potential for reducing deforestation, in the view of the proponent.) This statistical matching procedure was conducted by a central analytical team, not engaged in field data collection, to ensure procedural consistency across all countries and to minimize the influence of researcher preferences for particular field sites.

At all intensive sites, a minimum of 30 households were selected through simple random sampling in each of the eight villages (four intervention and four control) for a total of 240 households in the sample (Sunderlin et al. 2010, 27–29). If, due to local circumstances or matching, the number of villages fell below eight, the number of sample households was increased in each village to attain the minimum of 240 households at the level of the site.5 At three sites, the REDD+ initiative involved specific households, and thus, we stratified our sample based on household participation in the initiative.6 In this book, we focus exclusively on results from the intervention villages.

At the five extensive sites, villages were selected by identifying a pool of 15 candidate REDD+ villages (or the total number of villages at the site if there were fewer than 15). Four of those villages were then selected for the sample. One village was chosen purposively for having the highest potential for success in reducing deforestation, according to the proponent. The remaining three were chosen randomly.

1.2.4 Survey instruments

Most of the quantitative primary data reported in this volume are from a household questionnaire. The main aim of the household questionnaire (in terms of time consumed in the interview) was to record all household livelihood activities from all sources and the value of household income in the 12-month period prior to the date of the interview. The household income survey records environmental income, distinguishing between forest and non-forest environmental income. For definitions of these and other variables used in the study, see Appendix 4. Other purposes of the questionnaire were to record the type and value of all household assets including land and household goods, tenure of household lands, access to utilities (water, toilet, electricity, cooking technology and fuel), household activities on forest lands in the two-year period prior to the interview, subjective perception of well-being change in the two-year period prior to the interview, and knowledge of and involvement in REDD+. A village questionnaire and a women’s questionnaire collected the same kinds of data, although at a much lower level of specificity for household income.7 Also used were a “Proponent Appraisal Form” to get preliminary data on the initiative, and a “Survey of Project Implementation” for gaining insights into the background, history, institutional dynamics and politics of initiative development. The Technical Guidelines, survey instruments and code book used during phase 1 (2009–2013) for this research can be found on the CIFOR-GCS website.8

1.2.5 Methods limitations

Several issues are flagged here to help the reader understand the strengths and weaknesses of our approach. First, we have a sample of just four intervention villages in what are, in many cases, large and diverse sites. Second, in a small subset of ‘intervention’ villages, the proponent interventions have not begun, either because the initiative is on hold (e.g. Ulu Masen) or because the proponent shifted attention to other villages.

Third, the external validity of our results would have been maximized if we randomly selected tropical forest countries, initiatives, villages and households. However, given budget constraints on our sample size, that procedure would have limited our statistical power to detect impacts of a variety of REDD+ initiatives. Thus, countries and initiatives were selected purposively, villages were selected based mostly on statistical matching9 and households were selected randomly. Given our constraints, purposive selection of all countries and initiatives was a practical requirement to assure conformity with our data needs. Our purposive selection of initiatives was mostly driven by their timing and stage of development and not biased toward particular types of proponents, initiatives or regions. We selected a sample of villages that we could match to ‘control’ villages outside the intervention area in terms of basic socioeconomic and geographic characteristics, from the set of villages where proponents were definitely planning interventions and where deforestation and degradation rates were at least average for the region. On the whole, we believe our sample to be a source of valuable data, with information on a wide variety of initiative types, including in the two main REDD+ countries (Brazil and Indonesia).

Fourth, the six country teams had somewhat different approaches in gathering the household income data. The Brazil and Peru teams used highly elaborated, pretested and written call-out lists for inquiring about all sources of household income, whereas the other country teams used a mix of written and mental (unwritten) call-out lists of likely sources of local income. They developed these mental call-out lists during the interviewing process, so the interview of the last household is likely to have been conducted differently than the interview of the first household. While this variation in technique is unlikely to have introduced any systematic bias across intervention and control villages, it does mean that there may have been more complete and consistent reporting of income in Latin America, and this is one of many caveats on drawing comparisons across continents.

1.3 Location of initiatives and implications for generalizability

The sample of 23 initiatives is a broad cross section of sites with a wide diversity of types of proponents, scope, sources of funding (Appendix 1); biophysical characteristics including area, elevation, forest type, ecozone, climate region (Appendix 2); sources of pressure on forests (Appendix 5); and rates of deforestation (Table 1.1). In the period 2000–2012, forest cover loss ranged from a low of 0.3% (Madre de Dios, Peru) to a high of 19.2% (Transamazon, Brazil).

To what extent is it possible to generalize research findings from the CIFOR-GCS sample to the broader universe of REDD+ initiatives? The CIFOR-GCS sample suffers the disadvantage of being only 23 initiatives out of 329 worldwide, and being drawn from only 6 of 47 REDD+ countries. Nevertheless, comparison of the CIFOR-GCS sample to the global database of Simonet et al. (2014) demonstrates that the CIFOR-GCS sample is a reasonable if imperfect representation of the wider universe of subnational REDD+ (see Table 1.2).10

Table 1.2 Comparison of key characteristics of the CIFOR-GCS sample and REDD+ worldwide.


CIFOR-GCS sample

REDD+ worldwide
(Simonet et al. 2014)

Countries with REDD+ initiatives

Brazil, Indonesia and Peru are in the CIFOR-GCS sample, but DRC is not

Countries with the largest number of initiatives: Brazil (47), Indonesia (42), Peru (23), DRC (19)

Area of initiatives – range

2,664–15,749,099 ha

<1,000 to >1 million ha

Area of initiatives – mean

1,605,808 ha

300,000 ha

Area of initiatives – total

36,933,575 ha

73,500,000 ha

Ecological zone/forest type

Humid, 17 of 23 = 74%

No data, 0 of 23 = 0%

Dry, 3 of 23 = 13%

Dry and humid, 3 of 23 = 13%

Humid 47%

No data 35%

Dry 10%

Dry and humid 8%

Proportion private initiatives

17 of 23 = 74%


Of private initiatives, proportion for profit

4 of 17 = 24%


Of private initiatives, proportion not for profit

13 of 17 = 76%


Proportion selling forest carbon credits

4 of 23 = 17%


The CIFOR-GCS sample contains the three countries (Brazil, Indonesia, Peru) with the most initiatives. The range in the land area of initiatives in the CIFOR-GCS sample is roughly comparable to the world range, but the average area is substantially larger in the CIFOR-GCS sample, which includes two exceptionally large initiatives (Acre and SFX). Because of the inclusion of these two large sites, the area of the CIFOR-GCS sample is half the size of all initiatives worldwide. In the CIFOR-GCS sample, 74% of the sites have humid forests whereas in the global data set the share is 47%. The comparison is complicated, however, by the fact that there are no data for one-third of the cases in the global dataset. The CIFOR-GCS proportion of private initiatives (74%) is similar to the world average (82%), but the inclusion of for-profit initiatives is somewhat smaller and of nonprofit initiatives somewhat higher. The CIFOR-GCS and global figures for the sale of forest carbon credits are similar (17% and 20%).

Simonet et al. (2014, 20) found that the main drivers of deforestation, in terms of the number of initiatives where types of drivers are experienced, are, in rank order: local livelihoods, industrial agriculture, slash and burn agriculture, illegal logging, fire, energy wood, industrial wood exploitation, and oil. It is not possible to make a direct comparison to the drivers identified in the CIFOR-GCS sample because of differences in terminology and clustering. Nevertheless, there are notable similarities to the patterns of pressure experienced at CIFOR-GCS sites (see Appendix 5). Similar to the worldwide pattern observed by Simonet et al. (2014), the most common pressures experienced at the CIFOR-GCS sites are from small-scale agriculture. Unlike the worldwide pattern, pressures from industrial agriculture appear to be less frequent in the CIFOR-GCS sample.

There is strong commonality in the types of forest protection interventions introduced at the sample sites of CIFOR-GCS initiatives (Appendix 6). Almost all initiatives have applied restrictions on forest access and conversion, and non-conditional livelihood enhancements. These are the tandem (combined negative and positive incentives) found in all ICDPs, and most initiatives in the sample are actually continuations of pre-existing ICDPs. The ICDP lineage in REDD+ has been well documented in the literature (Blom et al. 2010; Sunderlin and Sills 2012; Minang and van Noordwijk 2013). Note, importantly, that Appendix 5 lists only the first year of (re)implementation of an intervention in the REDD+ period. In many cases, interventions were also applied prior to the REDD+ period. All but two of the sites have carried out environmental education, which is a key part of the FPIC process in REDD+. Likewise, tenure clarification has been carried out at most sites since it is an important preparatory step for REDD+. Based on follow-up research at the sites, we have found that by 2014, conditional livelihood incentives – which is to say performance-based REDD+ incentives – have been carried out at 14 of 23 initiatives, but in almost all cases only on a pilot basis. Only four of the case initiatives are currently selling forest carbon credits. In a similar vein, Simonet et al. (2014, 19) found that REDD+ initiatives tend to be structured along the lines of the ICDP model, and performance-based incentives have not been applied as extensively as expected.

1.4 Structure and content of the book

This book uses CIFOR-GCS Module 2’s baseline data to provide ex ante insights on REDD+ development on the ground. This introductory chapter gives a thumbnail sketch of the ‘big picture’ of REDD+ development and why research on it is necessary. It explains the sample and methods of the study, presents contextual information on all 23 cases, and it describes in broad terms some of the content of the case reports (with a focus on the strategies of the initiatives) and of the closing synthesis chapter.


Each of the 22 case report chapters11 follows a pre-set template on information to be presented:

  • basic factual information on the initiative (geography, stakeholders and funding, motivation, timeline)
  • the strategy pursued by the proponent organization, including interventions deployed
  • characterization of the villages and smallholders at the site, including information on livelihoods, forest dependence and deforestation
  • the key challenges facing the initiative in meeting its goals
  • aspects of the initiative that are unique and/or provide lessons about REDD+.

This template approach enables comparative analysis among the cases, and synthesis of key insights in the culminating chapter.


The case chapters in this book help disseminate the experience of these pilot initiatives and provide a means to widen and deepen our understanding of the contextual conditions shaping REDD+ in the early stages. Specifically, these chapters enrich our understanding of: the institutional make-up of proponent organizations; the political, economic, social and biophysical conditions that motivate their actions and circumscribe their opportunities and choices; and the socioeconomic characteristics of local communities that proponents collaborate with to carry out their programs and interventions. The cases give first-hand insights into the reasons why initiatives have evolved in the ways they have, and perhaps most importantly, on the range of challenges they have faced in attempting to fulfill their goals and objectives.

One challenge stands out above all the rest. This is still a world where interests favoring the conversion of forests to non-forest uses in tropical countries often have the upper hand in land-use decisions. This legacy has deep roots in the political economies of all tropical developing countries, and the persistence of the past is nourished by various factors – among them: the need of states to increase employment opportunities and foreign exchange earnings through agricultural development; the imperative to feed and provide land for an ever growing population; and infrastructural development in increasingly remote places. Forest conservation need not always be at odds with agricultural development and economic growth, as the case of Brazil mentioned earlier illustrates well. The world is at a crossroads where REDD+ has opened up an opportunity for forest conservation, leveraged by world concern about climate change. But as this book shows, REDD+ faces a steep uphill climb in reaching its objectives.

The proponents are well aware of this fundamental challenge. At the most basic level, all proponents share one common strategy to meet this challenge: reduce incentives to deforest. As explained earlier, most initiatives have an ICDP approach involving restricting forest access to local stakeholders, and compensating that with livelihood enhancements. Livelihood enhancements both compensate for restricting forest access to local stakeholders and secure the support of local stakeholders in efforts to exclude large commercial development. But beyond these common points of departure, there is a wide array of strategies among proponents. Some tend to be common to a country, whereas others are found across countries. Strategies evolve over time in response to changing international, national, and local opportunities and constraints.

Here we summarize some of the most common strategies reported in the case chapters. They are clustered under five headings: obtaining financing for REDD+ and thereby paying the opportunity costs of forest conservation; addressing the tenure status of custodians of and claimants on REDD+ forests; collaborating with government at various scales; creating a system for MRV; and assuring that local stakeholders and biodiversity are not placed at risk by REDD+ (social and environmental safeguards). In the paragraphs that follow, we briefly summarize the nature and scope of these key strategies as they play out in some of the case chapters.

Financing REDD+ interventions. Most initiatives in our sample initially planned to access the forest carbon market to fund livelihood enhancement at the local level. Some intend to apply the rewards at the level of the individual or household (e.g. Transamazon, Kilosa, Lindi), and others at the level of the village or community (e.g. SE Cameroon). Some initiatives are treating potential revenues from REDD+ as just one stream of funding among other possible sources, including, for example, sales of FSC-certified timber (e.g. Acre, Jari/Amapá, Madre de Dios, Ucayali, Mpingo, TNC within BFCP). But as of now, only four have obtained funding through the forest carbon market (Bolsa Floresta, Jari/Amapá, Madre de Dios, Rimba Raya). Most of the initiatives are in standby mode, waiting for the forest carbon market to become sufficiently reliable and predictable to serve as a long-term foundation for REDD+’s reward mechanism. Because this has not yet happened, most continue to rely on short- to medium-term funding sources that include the proponent organization itself, an affiliated implementer, government, a donor organization or some combination of these. Some initiatives have decided not to seek funding through the forest carbon market and are steering a different course, including engagement in low-carbon development at the landscape level (e.g. SFX).

Addressing tenure. Initiative forests face threats from actors large and small, and from both inside and outside the boundaries of the initiative. This state of affairs underscores the need to provide local stakeholders with the legal means to exclude claimants, as well as the right to benefit from the stream of REDD+ rewards as compensation for keeping forests standing. It also highlights the need to review and revise tenure arrangements that have historically provided privileged access to forest resources to large actors. Some approaches to addressing tenure are found at many sites, including ascertaining village and forest boundaries, participatory mapping, and assessing, clarifying and strengthening the forest tenure rights of those local stakeholders the proponent wants to become the custodians of forests. Above and beyond various conventional approaches, some proponents have also relied on various legal mechanisms to link tenure rights to environmental outcomes. Examples are CAR in Brazil (São Félix do Xingu, Cotriguaçu, Transamazon), the village forest (hutan desa) tenure classification for local stakeholders in Indonesia (KCCP, KFCP, TNC within BFCP) and the ERC for private REDD+ proponents in Indonesia (Katingan, Rimba Raya).

Collaborating with government at various scales. Proponent collaboration with governments has potential benefits, and there may be further advantages for subnational initiatives that operate at the jurisdictional scale instead of the project scale. Government officials and agencies might, for example, exert leverage for getting licenses, enforce environmental laws, facilitate the provision of services to local stakeholders or support helpful tenure reforms. While collaboration with government at various scales may provide advantages such as these, promote local buy-in and lend legitimacy to initiatives, implementing REDD+ across subnational jurisdictions does not guarantee these benefits. In our sample, there is considerable variation in how subnational initiatives work with governments. Six of the initiatives are subnational jurisdictional programs. The other initiatives are nonjurisdictional projects but still have varying levels of coordination with governments at different scales, ranging from informal agreements with districts to jointly provide technical support to communities, to legal agreements between projects and national governments. These differences have implications for how initiatives are implemented, and these are explored in later chapters.

Establishing an MRV system. For REDD+ to function properly, the proponent organization must put in place a sophisticated, durable and credible MRV system for documenting its performance in terms of carbon emissions and removals. This is an indispensable necessity for being able to sell carbon credits, among other purposes. For example, a well-functioning MRV system is necessary for assuring the efficient and equitable distribution of REDD+ benefits among stakeholders (Skutsch et al. 2014), and is therefore closely tied to the issue of social safeguards (Duchelle et al. in press). The experiences of the 23 initiatives show that their success in developing an adequate MRV system is closely linked to the forest monitoring capacity of the country where the initiative is located. Such success is thus linked to whether the proponent collaborates with a developed country institution, whether the proponent organization is international and whether the level of in-house expertise available to the proponent organization is adequate. For example, the initiatives in Brazil and Peru have access to high quality data and are equipped with sophisticated monitoring technologies, primarily as a result of the high monitoring capacity of their governments. Organizations of international origin such as TNC (SFX in Brazil, TNC within BFCP in Indonesia) and Flora and Fauna International (KCCP in Indonesia) are also making significant progress in MRV in collaboration with their local partners. Official development assistance in the Asia–Pacific region enables the proponents in Indonesia to acquire key components in their MRV systems. It is also important to note that highly motivated in-house experts alone can make a significant difference, as is evident in the case of the MRV accomplishment at the Kilosa and Lindi sites.

Fulfillment of social and environmental safeguards. The six study countries are in the process of elaborating formal, national-level REDD+ social and environmental safeguards. In terms of social safeguards, all of the initiatives either have or plan to carry out FPIC consultations with the local population. In most (but not all) cases, attention to social safeguards is built into the logic of subnational initiatives for instrumental (means-ends) reasons. After all, proponents, at least in principle, have a stake in fulfilling the needs and rights of their chief collaborators – the local stakeholders vested with responsibility to keep forests standing. Attention to the needs of local stakeholders is reinforced by the expectations of donors, the corporate social responsibility mandate of private organizations, the ethical mandate that NGO proponents have brought into REDD+ and the certification process. Most proponents in our sample either have, or are striving for, third-party certification through the CCBA or REDD+ Social & Environmental Standards Initiative, which motivates them to pay attention to how their initiatives affect local stakeholders and the environment (Jagger et al. 2014). However, fulfillment of this social mandate is not automatic given (among other reasons), that REDD+ initiatives almost always by design restrict access to forests that local people depend on for part of their livelihood. Six of the case initiatives (Ucayali, Kilosa, Lindi, KCCP, Rimba Raya and Cat Tien) are likely to give dedicated attention to biodiversity protection, given the proponent’s high ranking of this goal.

In the synthesis chapter (Chapter 24), these five categories are analyzed in greater detail, with a focus on the challenges that they present for proponents. These challenges are framed in the context of secondary literature, and then examined in light of the evidence as seen in the case chapters. Comparisons and contrasts are made among the cases with an eye to revealing the inner workings of these obstacles to climate change mitigation in the forest sector. The synthesis closes with recommendations on possible pathways to surmount the challenges encountered.

1.5 Closing thoughts

With scientific forecasts of the consequences of anthropogenic climate change becoming ever more ominous, the need for effective, fast and global-scale approaches to mitigation is ever more urgent. For seven years, REDD+ has been considered a frontline strategy for achieving near-term reductions of GHG emissions in the forest sector. Subnational initiatives are the laboratory in which the REDD+ experiment is being conducted, and in which the human and biophysical consequences can be measured. This volume provides useful insights into progress made, setbacks encountered and possible pathways toward improvement.

The chapters in this book document how subnational REDD+ initiatives set out to contribute to climate change mitigation while also fulfilling a range of other goals – both social and environmental. In response to various challenges, they have adapted and innovated to keep forging a path toward their destination. A few of the 23 initiatives have begun to demonstrate that it is possible to implement REDD+ on the ground more or less in the way originally envisioned, and others are hoping to eventually follow suit. Some, due in part to unresolvable challenges, are either falling back on previous conservation strategies or steering a new course beyond what was initially envisioned for REDD+. Unfortunately, some proponents have had to yield to the reality that sustained efforts to reduce forest emissions require enabling conditions that are not yet in place, and have therefore brought their initiatives to a close. It is abundantly clear that if forest-based climate change mitigation is to rise to the expectations set for it seven years ago, there will have to be an unprecedented groundswell of collective concern and action to move this path-breaking idea towards realization. We hope this book provides both helpful knowledge, and also inspiration, in that direction.

Box B
Challenges to measuring emission sources and sinks in REDD+ subnational initiatives

Shijo Joseph

A functional MRV system, capable of estimating net emissions and removals against an REL, is a major step in REDD+ readiness. The IPCC (2006) put forth guidelines focused on activity data and emission factors, which are central to estimating emission reductions. Activity data reflect the area in which carbon changes occur over time, while emission factors reflect changes in carbon stock densities per unit area. Activity data are generally produced by repeated measurements from a remote-sensing platform, while emission factors are estimated by ground measurements of carbon and noncarbon GHGs.

An assessment of REDD+ proponents’ MRV capacity and readiness in 20 of the 23 subnational REDD+ initiatives described in this book found that about half had capacity deficiencies for generating activity data and emission factors (Joseph et al. 2013). This study employed 19 performance criteria and 76 indicators under three categories of capacity and readiness: (i) remote sensing and GIS; (ii) carbon pool measurements; and (iii) REL and monitoring. Capacity and readiness tended to be highest at the Latin American sites and somewhat lower in Africa and Southeast Asia. Landsat was the primary source of activity data, and about half of the proponents had access to high-resolution (>10 m) satellite imagery. The majority of organizations (70%) showed in-house expertise and used advanced classification and change detection techniques for generating activity data. With respect to emission factors, only a few initiatives monitored all five carbon pools, none inventoried nitrous oxide or methane, and about half had site-specific allometric equations. For REL and monitoring, there were limitations in all the initiatives. A few had well-defined strategies to slow and halt proximate causes of deforestation, but all were limited in addressing underlying causes that originated outside the intervention area. A few initiatives showed reasonable monitoring plans, while the rest had loosely defined monitoring plans or no plan at all.

Technical challenges are the main reason for delays in attaining MRV readiness. REDD+ requires monitoring of three trajectories: deforestation, degradation and regrowth. Remote sensing is the most important tool available to monitor these trajectories, with hundreds of images available for a given area. However, these images are affected by atmospheric scattering, clouds, cloud shadows, geometric errors and other sensor-related factors, and require careful filtering before meaningful interpretations of the data can be made. The filtering can be either at the pixel or image level. In the pixel-based approach, individual pixels are evaluated and those that meet the quality criteria are selected. In the image-based approach, each image is taken as a whole, and those that do not pass the quality criteria are filtered out. In most cases, very few images meet the quality criteria, which makes remote-sensing-based monitoring fragmented and less real time. Although deforestation can be detected using these multi-temporal data, the monitoring of degradation and regrowth requires denser time series (hyper-temporal) data. Several research efforts are progressing toward that end (see Verbesselt et al. 2010, 2012). Emission factors are also equally important when converting activity data on degradation and regrowth to estimates of emission sources and sinks, and recent methodological advice from GFOI (2013) should help REDD+ proponents in mainstreaming their efforts.

Remote sensing is rapidly progressing with new sensors, analytical algorithms and high data-handling capabilities (Joseph et al. 2011). The constellations of mini- and nanosatellites and the dawn of civilian drones/unmanned aerial vehicles open up a new age where it is possible to gather ultra-dense time series data of hundreds of daily observations (The Economist 2014). The entry of technological giants such as Google into the space- and location-based business (Samuels 2013; Oremus 2014) can bolster forest monitoring. Quite soon, detection, monitoring and quantification of even minute resource extractions from the forest (e.g. felling, skidding and transport of logs) will be possible. Algorithms for analyzing data are also being developed so as to synthesize terabyte data and deliver meaningful results. Questions remain on how to institutionalize such technological innovations for the benefits of society, which is key in REDD+ readiness and in determining the fate of REDD+.

Box C
REDD+ in-depth costing

Eduardo Marinho, Noah Greenberg, Demetrius L Kweka and Erin O Sills


The Stern Review (Stern 2006) identified the reduction of emissions from deforestation as a highly cost-effective approach to climate change mitigation. This provided strong support for the implementation of REDD+ initiatives throughout the world. Now, the time has come to assess the actual costs of REDD+. Here, we compare and contrast two very different REDD+ initiatives in Brazil and Indonesia – two countries that together store more than one-third of the world’s forest carbon (Achard 2002; Gibbs 2006; IPCC 2006; Gibbs et al. 2007) and account for more than half of the world’s forest carbon emissions. To gain insight into how much REDD+ really costs, we quantify and analyze the implementation costs as budgeted during the design phase of two subnational initiatives operating in different institutional frameworks, facing different drivers of deforestation and choosing different strategies to combat them.


Figure C.1 Budgeted annual costs of the REDD+ initiatives over time.

The budgets for these two initiatives vary in duration, application, total costs and per-unit costs (see Figure C.1). The budget for the Transamazon initiative (Chapter 7) covers a period of five years, as specified by the donor. Its interventions are expected to change local conservation practices and livelihoods. In contrast, the Katingan Project (Chapter 18) aims to restore degraded peatlands and prevent future large-scale forest conversion through a 60-year ERC license, which was granted in 2013. The Transamazon initiative’s focus on promoting sustainable practices among smallholders (local deforestation agents) translates into higher per-household costs. The Katingan Project has relatively higher costs per hectare (see Figure 1-2) due to its ecosystem restoration and patrolling activities, as well as higher costs for MRV and marketing because it plans to sell carbon credits.


Figure C.2 Cost per household and per hectare of the Transamazon initiative and the Katingan Project.

The Transamazon initiative

Country: Brazil

Proponent: IPAM

Duration: Five years

Funding: Public (Amazon Fund)

Households: 2926

Area: 245,485 ha

Yearly cost: USD 2,250,245

Although deforestation rates have significantly decreased in Brazil, the success of the national policy is less apparent among smallholders living in Amazonian settlements. The Transamazon initiative seeks to address this problem by increasing settlers’ income with no need for additional forest clearing. Community development (agricultural inputs and technical guidance, tree nurseries and PES) accounts for 54% of the costs of the initiative, while expenditures in protection and enforcement are negligible (at least in part because this is a government function). Notably, finance and administration is the second major source of costs (19%) because of the complex legal requirements and relatively high salaries in Brazil. High salaries also explain why more than half of the initiative budget is spent on personnel. Other reasons for this high contribution of personnel are high costs of taxes and benefits and a relatively small number of activities run by consultants. The direct cost of PES accounts for only 9% of the budget, although this number would be substantially higher if all households, and not only 12% of them, received payments.


Figure C.3 Costs of the Transamazon initiative.

The Katingan Project

Country: Indonesia

Duration: Undefined

Proponent: PT. Rimba Makmur Utama

Households: 12,880

Funding: Private (carbon markets)

Area: 200,000 ha

Yearly cost: USD 5,149,629

Indonesia has demonstrated the largest increase in forest loss over the past decade and hence offers significant opportunities to reduce emissions. There are a number of legal pathways toward REDD+ initiative development in Indonesia, each with different implications for initiative structure and cost profile. The recent advent of ERCs has provided a relatively clear regulatory pathway for private sector REDD+ initiatives. Our detailed budgeting exercise for the Katingan Project reveals that the ERC pathway has high upfront costs driven by concession fees (IDR 39.4 billion, or USD 3.2 million as of this writing). The greatest overall cost driver is ecosystem restoration activities (36%), required under the ERC model to “restore biological equilibrium.” Significant human resources are required to implement ecosystem restoration and protection (14%) activities leading to high expenditures, much funneled to community and other local partners, categorized here as contracted services (40%). Overall, however, a comparatively lower level of economic development in Indonesia appears to keep personnel costs relatively low in the Katingan case (26%) compared to what we find in the Transamazon initiative in Brazil.


Figure C.4 Costs of the Katingan Project.


As might be expected given the wide variation in strategies employed by REDD+ initiatives documented throughout this book, their costs also vary substantially. This is partly due to national context. In Brazil, the strong command and control policy at the federal level relieves initiatives of protection and enforcement costs while imposing higher needs for fostering sustainable livelihoods through community development. In Indonesia, the ERC license fees result in front-loaded costs. In our research on the costs of various REDD+ initiatives, we have found that personnel costs are always a significant budget item, albeit in the Brazilian context they are overwhelmingly the largest cost driver.

This is a first attempt to understand REDD+ initiatives through analysis of their costs. Although our focus on only two initiatives prevents any generalizations, our analysis suggests that the source of emissions being targeted, the strategies used by initiatives, and the institutional and legal context, all affect the implementation costs of REDD+ on the ground.

1 We define REDD+ both broadly and narrowly, as follows: “A broad definition, based on the official COP13 terminology, holds that REDD+ comprises local, subnational, national and global actions whose primary aim is to reduce emissions from deforestation and forest degradation and enhance carbon stocks (increase removals) in developing countries. A narrower definition is that REDD+ also includes results-based or conditional payments, which was a core idea when REDD+ was launched” (Angelsen et al. 2012, 381).

2 CIFOR-GCS as a whole is conducting research in Bolivia, Peru, Brazil, Democratic Republic of Congo, Cameroon, Tanzania, Ethiopia, Burkina Faso, Mozambique, Nepal, Indonesia, Papua New Guinea, Vietnam and Laos.

3 The research modules involve international and national policies and processes (Module 1); subnational initiatives (Module 2); MRV (Module 3); and multiscale governance (Module 4).

4 Specifically, by our definition, they aim to reduce net carbon emissions primarily by: (a) reducing deforestation/degradation; or (b) implementing forest conservation/restoration/management. That is, they do not derive most of their carbon benefits from afforestation/reforestation outside of existing forest.

5 For example, at the SE Cameroon site (Cameroon), there are only two villages within the boundaries of the site. In each of these intervention villages, 60 households were selected, and four control villages were selected with 30 households each.

6 At the Acre (Brazil) site, the sample was stratified based on participants and nonparticipants in the Certification of Smallholder Properties Program. At the Transamazon (Brazil) site, the sample was stratified based on participants and nonparticipants in the Proambiente program. At the Shinyanga (Tanzania) site, the sample was stratified based on households that have their own forest enclosure (ngitili) and those that do not. In each of these cases, approximately 15 participants and 15 nonparticipants were selected in each community.

7 Both the village and women’s surveys were conducted through focus group discussions with 10 to 15 adult (16 years or older) respondents selected through collaboration with the village leadership.

9 Three of four villages at extensive sites were chosen randomly.

10 Whereas the CIFOR-GCS sample is limited to initiatives that aim to reduce net carbon emissions primarily by reducing deforestation/degradation or implementing forest conservation/restoration/management, the Simonet et al. (2014) database includes these and also initiatives aiming primarily at afforestation and reforestation.

11 Although there are 23 initiative cases, there are 22 case reports because the initiative called “Making REDD Work for Communities and Forest Conservation in Tanzania” encompasses both the Kilosa and the Lindi cases.