What is out there? a typology of land restoration projects in Latin America and the Caribbean

Restoring degraded lands is high on the international agenda and the number of restoration projects in Latin America and the Caribbean (LAC) has increased considerably over the past decades. However, the variety of approaches used have not been systematically documented and analyzed. We aimed to develop a restoration typology as a function of the geographical and socio-economical setting, planning, timeframe, finances, implementation, monitoring and potential impact, which will help to discern broad patterns and identify gaps in project implementation in LAC. We categorized 97 restoration projects using Multiple Correspondence Analysis and a Hierarchical Clustering on Principal Components. Three main restoration types emerged from the clustering, with the main defining variables being: (1) project area under restoration, (2) amount of funding received, (3) source of funding and (4) monitoring efforts. The first type are large-scale projects, which receive high amounts of funding provided by international donors, and with a well-established monitoring plan; the second type are projects financed with private money, typically lacking a sound monitoring program; and the third type represents small projects with low amounts of funding, financed with public funds from national governments, often with a rudimentary monitoring plan. The typology enables a comparative analysis of the status and trends of restoration activities across Latin America. We conclude that, despite growing awareness and recognition that integrated approaches are needed to revert complex and interconnected socio-economic and environmental issues like land degradation, the socio-economic dimension remains underexposed in the majority of restoration projects, whereas monitoring is still regarded as an extra cost instead of a necessary investment.


Introduction
Land degradation is a growing global problem resulting in annual losses between 6.2 and 10.3 trillion US dollars (ELD 2015) and affecting at least 1.6 billion people worldwide (Bai et al 2008). Restoring degraded lands is urgently needed to protect biodiversity and ecosystem services, and to halt and reverse land degradation (IPBES 2018), and is high on the international agenda, e.g. Aichi Biodiversity Target 15 (CBD 2011), Bonn Challenge (IUCN 2011), the New York Declaration on Forests (UN 2014) and Sustainable Development Goal 15 (UN 2015).
At the ecosystem level the Society of Ecological Restoration (SER) distinguishes three groups of restorative activities (McDonald et al 2016): Ecological restoration aims at the complete recovery of the ecological functions and processes, and the biotic community structure of the original ecosystem. Rehabilitation strives to repair key ecosystem functions, processes and services not necessarily leading to a complete recovery of the original ecosystem and reclamation is limited to repairing only minimal functions of the land. When restoration is applied at the landscape scale it is referred to as forest landscape restoration (FLR) which is a holistic approach that aims to regain ecological functionality and enhance human well-being across deforested or degraded landscapes (Sabogal et al 2015, GPFLR 2018. FLR may include all three types of restorative activities abovementioned as they are applied to different ecosystems within a landscape. Even though the focus of FLR is human wellbeing, a sustainable FLR project requires a balance between natural and productive systems, because productive systems and humans depend on the natural ecosystems as well. Therefore, in a FLR project, one could be restoring natural ecosystems in some areas, and ensuring they have a full complement of native species, rehabilitating land for agricultural productivity in other areas, and reclaiming areas where e.g. only mineral soil remains. In essence, FLR is a deliberate intervention where a suite of different land uses, varying from natural forest cover to commercial plantations, natural and assisted regeneration, and agroforestry and silvo-pastoral systems, coexist within a 'multifunctional landscape' (Laestadius et al 2015, Aronson et al 2017. In this paper, we will use the term restoration in a broad sense, referring to both ecological restoration and FLR.
The stiff competition between land used for agriculture and demands for conservation of biodiversity and provision of ecosystem services is evident in Latin American and the Caribbean (LAC) (Harvey et al 2008, Estrada-Carmona et al 2014. The LAC region not only contains seven of the 25 biodiversity hotspots worldwide (Myers et al 2000) but also one of the four main global AFOLU (Agriculture, Forestry and Other Land Uses) emission hotspots (de Sy et al 2015, Roman-Cuesta et al 2016. Restoration is likely to increase biodiversity (Rey Benayas et al 2009) and has great potential for carbon sequestration . Furthermore, restoring degraded agricultural lands to increase agricultural production can contribute to meeting growing global food demands, diminishing pressure on undisturbed ecosystems by agricultural expansion, and improving rural livelihoods (Lamb et al 2005, Sayer et al 2013. The multiple benefits of restoration and the growing awareness that sectoral approaches are not suitable for complex and interconnected socio-economic and environmental issues (Reed et al 2016) have led to a significant increase in both the design and implementation of projects and initiatives worldwide. As a response to the Bonn Challenge, seventeen national governments, three subnational government and three non-governmental organizations in LAC have committed, to date, to restore about 53 million hectares of land under Initiative 20×20 while other programs that target climate change mitigation such as the Forest Investment Program (FIP) or the Global Environmental Facility (GEF) have similarly adopted restoration as an entry point to sustainably manage natural resources (GEF 2018).
Given the multiple objectives of restoration to satisfy environmental and socioeconomic goals and the millions of hectares of land committed to undergo restoration so far, a typology of restoration projects may help to discern broad patterns and identify gaps in project implementation in LAC. Although others have carried out assessments of ecological restoration projects at the national level (Murcia and Guariguata 2014, Cerrón et al 2017, Méndez-Toribio et al 2018 and of integrated landscape initiatives at the regional level (Estrada-Carmona et al 2014), a multidimensional assessment is lacking. That said, we developed a typology of restoration projects in LAC as a function of socio-environmental settings, planning, timeframe, finances, implementation, monitoring and potential impact, using multivariate techniques. The typology enables a comparative analysis to provide insights on what is happening on the ground.
Restoration projects are complex. Their successful implementation requires careful project planning to minimize conflicts of interests and strike a balance among the multiple functions of a landscape and all relevant stakeholders (Sayer et al 2013, Meli et al 2016, Murcia et al 2016. Also, clarity about land tenure and rights (Duchelle et al 2014, Ceccon et al 2015, and understanding and addressing the drivers of degradation are essential for the long-term sustainability of interventions (Crouzeilles et al 2016. Finally, an objective-based monitoring plan grounded on a baseline assessment is required to provide evidence on the socio-economic and environmental benefits (Hobbs and Norton 1996, Chazdon 2008, Evans et al 2018. Monitoring also allows for improving best practices Aide 2011, Suding 2011), adaptive management (IUCN and WRI 2014), long-term evaluation of different approaches Guariguata 2016, Meli et al 2017) and measuring the socio-economic and environmental impact of the project (Adams et al 2016, Aronson et al 2010. In this paper we focus on recently finished, ongoing and pledged restoration projects in LAC and we aim to provide a coherent overview of how these projects are, in practice, contributing to satisfying multiple goals. This knowledge is urgently needed to guide and improve efforts to up-and outscale restoration practices (Estrada-Carmona et al 2014, Murcia et al 2016). Accordingly, our first objective is to define a typology of restoration activities based on the variables related to socio-environmental settings, planning, timeframe, finances, implementation, monitoring and potential impact. The second objective is to evaluate gaps in each restoration type through comparative analysis.

Selection of projects and variables
We selected 97 restoration projects that started in 2000 or later, with the majority starting after 2008, and covered areas of at least 10 ha from a previously assembled database of projects carried out across LAC by multilateral donors, national and sub-national governments, and NGOs (Romijn and Coppus 2019, Open Access). For more information about the compilation of the database, see Supporting Information. The multilateral projects were financed by the Global Environment Facility (GEF), that supports countries to develop and implement biodiversity, and climate change adaptation and mitigation strategies (GEF 2018), or the Forestry Investment Program (FIP), which addresses the drivers of deforestation and forest degradation to achieve REDD+objectives. For private sector/NGO projects associated with Initiative 20×20, the World Resources Institute (WRI) provided descriptions of projects. We selected projects from two restoration databases for Colombia (Murcia and Guariguata 2014) and Mexico (Méndez-Toribio et al 2018) compiled by the Centre for International Forestry Research (CIFOR). We also used projects from a database compiled by Bioversity International, the World Agroforestry Centre (ICRAF) and Servicio Forestal Nacional y de Fauna Silvestre (SERFOR) for Peru (Cerrón et al 2017).
The list of variables and categories used for this study was adopted from the national inventories in Colombia, Peru and Mexico (Murcia and Guariguata 2014, Cerrón et al 2017, Méndez-Toribio et al 2018 and adjusted to fit the regional scope of this study (table 1). The variables were grouped into classes that reflect the socio-economic, biophysical, organizational, technical and financial aspects of restoration projects. The variables were divided into categories. In general, the categories were not mutually exclusive, which means that a combination of categories can represent a variable. All categories were represented by two or more levels. The levels of the non mutually exclusive categories were 0 if absent, 1 if present or NA when missing, whereas the levels of the mutually exclusive categories corresponded with ordinal integer values. For example, the variable land use prior to restoration could consist of the categories agriculture, grazing and forestry whereas the category levels of the variable project area varied from 1 (<50 ha) to 8 (>100,000 ha) (table 1).
To minimize the amount of missing data affecting the multivariate analyses, we combined categories to the extent possible. Both projects with missing values for more than four categories, and categories with more than 50% missing values were excluded from the multivariate analyses. For more information on the variables and categories included, see Supporting Information.

Multivariate analysis
To define a typology of restoration activities, we used Multiple Correspondence Analysis (MCA) followed by a Hierarchical Clustering on Principal Components (HCPC). MCA is an exploratory technique to analyze correspondence in a multi-way frequency table (Bartholomew et al 2008) and illustrates the most important relationships among the variable´s response categories in a graphical way (Sourial et al 2010). The MCA was run with the FactoMineR package (Lê et al 2008) using R software (R Core Team 2016). In the resulting biplot inferences can be made of the relative associations between projects and levels. Only levels with a contribution to the definition of the particular dimension higher than 1% and a cos 2 (its quality of representation) higher than 0.2 were used to define restoration types (see supporting information Multivariate analyses and table S1 is available online at stacks.iop.org/ERC/1/041004/mmedia).
The first 20 dimensions of the MCA, which together explained 61% of the variance in the dataset, were used in a Hierarchical Clustering on Principal Components (HCPC) to create groups of restoration projects that share similar characteristics, i.e. restoration types. The HCPC was run with the FactoMineR package (Lê et al 2008) which uses the Ward criterion as it is based on the multidimensional variance calculated in the MCA. The final partitioning was obtained with K-means clustering (Lê et al 2008). In the output of the HCPC the values of Mod/Cla and Cla/Mod provide information about the content of the clusters, where Mod refers to modality or category level and Cla is the class or cluster (Husson et al 2010). The column Mod/Cla shows the percentage of restoration projects in the cluster that meet the category level and can be interpreted as a measure of robustness of a cluster for the specific category level. The column Cla/Mod shows the percentage of all projects that meet the category level and belong to the specific cluster and can be interpreted as a measure of uniqueness of a specific category level for the cluster. For example, when the category improving livelihoods has a value of 100%

Characterization of restoration projects
The projects included in the analysis were distributed over 15 countries in LAC and 11 biomes according to Olson et al (2001; figure 1). Half of the restoration projects covered areas smaller than 1,000 ha and 20% were larger than 100,000 ha ( figure 2(a)). Pasture (67%) and agriculture (79%) were the most frequent types of land use reported at the onset of the restoration intervention. Secondary forest appeared in 47% of the projects and in 29% patches of the original ecosystem were still present ( figure 2(b)). Objectives related to nature conservation guided nearly all projects (97%). Economic objectives (improvement of local employment and eco-tourism) and objectives related to alternative agricultural production systems (agro-forestry and silvo-pastoral systems) were mentioned for 60% and 63% of projects, respectively (figure 2(c)). Half of the projects were financed by national governments (52%) followed by international donors (46%) and impact investors (27%; figure 2(d)). In 31% of  the cases the amount of funding received were less than 500,000 US$ while 30% of the projects received funding of more than 10,000,000 US$ (figure 2(e)). In 16% of the projects the funding amounts were not mentioned. Natural regeneration of vegetation (53%) and assisted regeneration (36%) were the two most commonly used approaches (figure 2(f)). Twenty seven percent of the projects used mixed tree plantations.

Typology of restoration projects
On the basis of the MCA (figure 3), three clusters representing the restoration types, emerged from the HCPC. Type 1 (33 projects, 34%) is defined by funding by international donors, large project areas (>100,000 ha) and large amount of funding (>10,000,000 US$; figure 3 and table S1). The objectives of these projects included recovery of biodiversity, improvement of ecological processes, generation of local employment and the capture and storage of carbon. Areas with secondary forest and patches of the original ecosystem occurred within the project boundaries, and natural and assisted regeneration of vegetation were part of the implementation approaches (tables S1 and 2). All projects determined the causes and the degree of degradation, and undertook actions to address these drivers. The monitoring plans included a base line study and the variables monitored related to the objectives. Implementation and monitoring was mainly carried out by public institutions (figure 3). Projects were largely aimed at improving livelihoods, capacitation of local communities, regulating hydrologic processes and ensuring water availability, and addressing climate change by increasing C stocks in biomass and/or soils. Financing by impact investors and timber products as economic revenue were the main determinants of type 2 (22 projects, 23%, figure 3 and table S1) and 78% of all projects financed by companies or private land owners were grouped in this cluster (table 2). The degree of degradation in the project areas was often not determined and many projects did not have a monitoring plan and did not carry out a baseline assessment. In general, the restoration approaches did not explicitly consider natural forest regeneration. Fifty-five percent of type 2 projects focused on mixed tree plantations (table 2).

Comparative analysis of restoration types
Addressing multiple objectives was a common feature in all restoration types and reflects the multifunctional aspect of restoration (table 3). The projects of type 1 had sound planning that addressed the causes of degradation. In contrast, projects grouped as type 2, despite having identified the main causes of degradation, often did not take actions against these causes and the degree of degradation was poorly understood.
Natural and assisted regeneration was the preferred implementation approach in projects funded by international donors (type 1) whereas impact investor projects (type 2) typically promoted mixed tree plantations (table 3). All types preferred the use of native species although exotic species were used in 41% of the projects financed with private money, which are mostly found in type 2.
The projects of type 1 had a coherent monitoring plan, which included a baseline study and in which variables to be monitored were related to the objectives of the restoration effort, contrary to the projects of type 2, where monitoring was often lacking (table 3).
The economic incentives and revenues of projects with funding from impact investors focused on timber and non-timber products, and to some extent carbon sequestration (41%). Payment for ecosystem services (PES) schemes were to a certain degree associated with funding from international donors (30% of the projects of type 3) but almost absent in the other types.
Apart from improving food security (58%), the environmental and socio-economic impact was of great importance to the international donor projects. Type 3 prioritized environmental benefits such as improving biodiversity and ensuring water availability, though to a lesser extent than type 1 Table 3. Comparative analysis of the restoration types. The relative frequency distribution of categories per restoration type is shown with low referring to <50%, intermediate to 50%-75% and high to >75%.

Discussion and conclusions
The 97 projects analyzed for LAC fit into three general restoration types, which can be discerned based on four main variables: (1) project area under restoration, (2) amount of funding received, (3) source of funding and (4) monitoring efforts (table 4).
Our results show that projects of restoration type 1 follow country-level agendas and are aligned with international agreements such as the Convention on Biodiversity (CBD) and the Sustainable Development Goals. Projects allocated in type 2 operate at the local level and focus on the primary interests of impact investors and companies, i.e. making profit, with nature conservation as secondary objective. Restoration activities of type 3 focus on improving local environmental conditions and do not appear to be strongly linked with national policies to meet international commitments.
The results suggest that many restoration projects take a top-down approach, not necessarily considering the interests of communities. Apart from the international donors funded projects, few aimed at generating local employment and improving livelihoods. In addition, communities were often not consulted in the planning phase and their participation was mostly limited to the implementation phase. Rarely were local communities involved in monitoring or acting as main responsible. This is in line with the national restoration inventories of Colombia (Murcia and Guariguata 2014), Peru (Cerrón et al 2017) and Mexico (Méndez-Toribio et al 2018), and a regional assessment of integrated landscape initiatives in LAC (Estrada-Carmona et al 2014), which reported low active community involvement and showed that poverty alleviation efforts are not a priority in many projects.
Many projects financed by project investors were associated with Initiative 20×20 but the relation between national restoration projects and the initiative was less clear. Although the 20×20 projects seem to be disconnected from the national restoration agendas, it is very likely they will be used to demonstrate the countries efforts to meet their national restoration pledges. Projects that are dominantly financed with either private (type 2) or public funds (type 3) are complementary in the sense that they focus on economic and environmental benefits, respectively, and both operate at the local level. Combining both types of projects could create synergies if they are embedded in public-private-civic partnerships that adopt a landscape approach. Such partnerships are increasingly being promoted to meet commitments like the Bonn Challenge, the Aichi biodiversity targets, the New York Declaration on Forests, Initiative 20×20 and the Sustainable Development Goals (Pistorius andFreiberg 2014, Scherr et al 2017).
National and local governments should play a more active role in convening multi-stakeholder partnerships needed for out-and upscaling. However, large scale restoration programs cannot solely rely on top-down approaches as they often focus on short-time results and do not stimulate better practices, resulting in low involvement of the participating stakeholders, especially in regions with poor governance and weak legal enforcement Despite repeated calls that decision makers and land managers need evidence-based assessments to evaluate the success of restoration efforts (e.g. Palmer andFiloso 2009, Menz et al 2013), in many projects monitoring is still being regarded as an extra cost rather than a necessary investment. Monitoring as an integral part of the restoration activity was only common in projects funded by international donors. Monitoring in projects financed with public funds often only included a limited number of biological indicators related to plant survival, growth and health (Murcia and Guariguata 2014, Cerrón et al 2017, Méndez-Toribio et al 2018. To improve monitoring, capacity strengthening of communities, which was a recurring component of most restoration projects, should be used for organizing participatory workshops to discuss with local stakeholders which indicators to monitor in a cost-effective way. Participatory monitoring would greatly increase active community involvement and strengthen collective ownership of the restoration activity (Danielsen et al 2011, Evens et al 2018. Monospecific plantations make up only a small proportion of the applied restoration approaches in all types. Barlow et al (2007) and Lamb (2014) pointed out that monospecific plantations are not very likely to deliver multiple benefits such as sustainable livelihoods and improvement of ecosystem services, and it seems that this message has been internalized by international donors, impact investors and national governments. Moreover, in all restoration types preference was given to native species although in privately financed projects exotic species were used as well.
PES schemes were not frequently incorporated in restoration activities in LAC. This is probably due to uncertainties in the long-term sustainability of projects (Bullock et al 2011) and the limited effectiveness of PES in promoting forest restoration (Pirard et al 2014). Also, PES schemes tend to be more efficient when a single, clearly defined ecosystem service is targeted (Wunder et al 2008) which is often not the case, given the multifunctional character of most projects. Many projects claimed to have identified the drivers of degradation and actions to address them were proposed in their designs. However, most focused on the biophysical aspects of unsustainable land use only, whereas the socio-economic aspect remained underexposed in the majority of restoration projects. If the underlying causes like inequitable access to resources and political economy factors are to be understood, economic and institutional conditions must be studied (Pacheco and Poccard-Chapuis 2012). To tackle the underlying drivers, national governments should start developing intersectoral legal frameworks, including economic development, forestry and environmental policies, and remove perverse incentives that encourage degradation of forests and deforestation. These contextual factors need to be addressed as we move from largescale restoration commitments to successful large-scale implementation.
This typology does not pretend to be complete. Although our database covered projects from twelve biomes across LAC using a wide range of restoration approaches, the level of detail captured in the database was a compromise between completeness of project data and the easiness of access to information sources. The typology could probably be further improved through the inclusion of additional variables. For example, variables related to economic and financial factors such as return on investment could serve as a useful indicator for financial gains, particularly for impact investors, but was only mentioned in a few project descriptions. Costbenefit analyses are also needed to allocate the available financial resources efficiently and to facilitate prioritization (Rey Benayas et al 2009, Menz et al 2013 but the economic assessment of natural capital is still in its infancy (Bullock et al 2011).
Moreover, it is well known that uncertain or insecure land tenure negatively affect motivation and capacities to invest capital or labor to improve land, and thus are a major limitation to the sustainability of restoration projects (Lamb et al 2005, Duchelle et al 2014. Unfortunately, explicit information on land tenure issues was not provided in many projects. However, and in spite of the information limitations, our analysis outlines a general picture of how restoration projects are being implemented in LAC.