Detecting ecological thresholds for biodiversity in tropical forests: Knowledge gaps and future directions

Protecting tropical forests and their biological diversity is a global priority. Understanding if thresholds of forest cover exist beyond which biodiversity displays non-linear declines key to developing appropriate conservation strategies policies, but uncertainty remains around the identification characteristics these thresholds. We performed systematic review studies using gradient identify an ecological threshold across ecosystems. Our finds 68 reported in 33 peer-reviewed publications. Three main conclusions emerged: first, we show clear geographical gaps studies, with 72% found South America, over half Brazil; second, see follow taxonomic biases line wider research; third, there lack homogeneity comparability metrics sampling designs used threshold. This shows interest continues grow, further evidence needed understand application management. knowledge provide guidance focus research efforts on six aspects better potential as policy-making tool for conservation. Abstract Spanish available online material. La protección de los bosques tropicales y su diversidad biológica es una prioridad global. Mejorar el conocimiento científico sobre umbrales ecológicos cobertura forestal, tras cuales la biodiversidad muestra un declive no-lineal, clave para desarrollo estrategias políticas conservación. Pero existe falta métodos detección, las características, existencia estos en diferentes regiones tropicales. En este estudio realizamos revisión sistemática artículos científicos ecológicas forestal ecosistemas reveló encontrados científicos. Presentamos tres conclusiones principales: existen brechas grupos taxonómicos; identificados provienen América del Sur, más mitad Brasil; homogeneidad comparabilidad entre utilizados identificar forestal. Esta que interés comunidad científica usar continúa aumento, pero se requiere investigación entender sus características uso gestión Identificamos principales existentes recomendamos enfocar seis áreas potencial como herramienta política conservación Tropical play major role conserving providing ecosystem services (Barlow et al., 2018; Ferreira 2018). Covering 10% Earth's surface, they contain world's known species including 90% terrestrial bird Gibson 2011) are central climate change mitigation (Gibbs 2007). Estimates suggest tropics lost 12 million ha 2018, average rate loss intact tripling last 10 years 4.3 per year, area size Belgium (Schulte 2019). Deforestation be driver (Maxwell 2016; Morris, 2010a) effects loss, habitat fragmentation, degradation well understood (Giam, 2017; Haddad 2015; 2010b). As implementation period Convention Biological Diversity's (CBD) Strategic Plan Biodiversity 2011−2020 comes end (Convention Diversity, 2019), scientists policy-makers discussing new targets form basis national international plans protect next decades (Purvis, 2020; Rounsevell 2020). While success policy processes depends issues governance, socioeconomics, politics, also questions addressed creation (Green 2019; Noss Purvis, Svancara 2005). In creating protection forests, such through protected areas, or restoration, responses must understood. How much enough maintain biodiversity? Ecological have emerged past 20 help design effective action based preventing “break-point,” after drastic observed (Francesco Ficetola & Denoel, 2009). context ecosystems, increasingly applied method identifying minimum amount native vegetation avoid changes species–habitat relationships, leading (Huggett, 2005; Kelly 2015) (Figure 1). They been management decisions Australia, Brazil, Canada, USA (van der Hoek define limits deforestation set restoration targets. Despite support important link (“tipping points”) functioning at risk (Arroyo-Rodríguez Banks-Leite 2014; Oliveira Roque Johnson, 2013; Lindenmayer Luck, I. Melo 2018), strong scientific debate exists existence use planning (Brook Foley van Muradian, 2001). Concerns surrounding concept include criticisms its “rule thumb” value globally, while local regional (Banks-Leite 2021), publication bias meaning negative results less likely published. addition, some investigating response may not specific tipping point change, abundance often happen non-linearly biome 2013). points might identifiable, nor useful 2013), landscape scale continue investigated (Estavillo Newbold 2018) way changes, anthropogenic drivers, loss. Given continued management, compilation from ecosystems world can inform decision-making (Adams Sandbrook, 2013) practices. To identification, characteristics, use, conducted approaches findings landscapes. present overview discuss uncertainties regions variation methods them. had following objectives: (a) compile existing cover; (b) summarize data; (c) methodological inconsistencies identification; (d) highlight particular concern lacking information additional empirical work most urgently needed. literature SciVerse's Scopus bibliographical Google Scholar “Guidelines Evidence Synthesis Environmental Management” established by Collaboration (CEE) (CEE, two searches, first between April June 2018 second search January 2020. Due volume available, were restricted English. Studies selected fit criteria: (i) study (as defined Olson al. 2001); (ii) analyzed metric change; (iii) recorded gradient; (iv) analysis aimed numerical threshold, expressed percentage cover. Considering selection criteria, designed string sub-strings divided into categories (see Supporting Information). The strategy resulted 3593 initial (Table S1). included all 1970 An screening was reviewing title abstract each determine whether subject scope article criteria. produced 289 articles, reviewed abstract, methods, quantitative (percent cover) criteria outlined above. Keywords terms during process: “forest cover”, “cover,” “%,” “percentage,” “threshold.” ensure representation possible reporting bias, targeted searches carried out data: SE Asia Central Africa. Following this method, one reference added December (Kupsch addition publications captured selecting those thresholds, scanned lists relevant papers Information details). identified containing Although interested links structure responses, our concerning direct exclusively fragmentation included. effect independent populations different ways (Fahrig, 2003; so, clarity interpretation results, focused modeling assessing landscape-scale only. Prior analysis, pilot same approach answer question Atlantic Forest Brazil (Laurance, 2009), allowing refinement best aims smaller geographic more complete data pool. tested against list articles stemming expert discussions search. then three extended maximum inclusion considering practicalities related number screened test accuracy process, co-authors subset discrepancies process extraction, randomly sharing extracted interpretation. A database created final articles. article: measure (e.g., richness, phylogenetic diversity, abundance). authors grouped 9 general S2 Table recorded, definitions 2015). Each conditions met: unique level specialization) sample location site, that is, classed and/or dependent variable, location, size. Threshold values authors, cases zone-type (when band given rather than single number), mid-point analysis. response, measured within region. GEO/IPBES1 classification (Brooks 2016) sub-regions conducted. Species classified authors. amphibians, birds, insects, mammals, plants, “mixed” when multiple classes (amphibians N = 1, birds 23, insects 3, mammals 22, plants 15, mixed 4). specialization specialist generalist. “sample area” site (km2) publication. Typically either areas varying levels compared, area, response. Size (if reported) region located, represents calculated. contains sites where measured, emerges differences sites. detect individual Jamaican fruit bat, groups (families) phyllostomid bats). First, summarized regionally. Shapiro–Wilk normality (Ghasemi Zahediasl, 2012) assess distribution data. calculated 95% confidence interval bootstrap non-normal (Jung Wang, 2001), R “boot” package (Canty Ripley, Second, descriptive statistics patterns metrics, sizes gradients used. Third, nested linear models investigate influence parameters “lme4” (Bates 2015): class, specialization, size, gradient. nature dataset collection random nesting account fact (A) (B) paper. “Region” “Paper ID” (representing paper extracted) effects. full dataset, sub-analyses look five representative groups: (N 23) 22). “ANOVA” function “CAR” compare model (Fox Weisberg, All statistical analyses Studio v3.6.2 (R Core Team ArcMap v10.6 (ESRI create map studies. covering 2) find both “point-type” “zone-type” F. P. L. Zone-type zones bands instead range 30‒40%, 35%) numeric break could identified. Of Martensen 2012; Pardo 41.9% (95% CIs 38.08−45.82, 1.97) S1) geographically biased 2); 82% published date Neotropics, 49 7 Mesoamerica, 2 Caribbean S5). Africa severely under-represented, only eight identified, respectively. vary 35% 1) America 49) 62% 8). There wide analyses, ranging large datasets 500 species. gap invertebrates. Twenty-three 22 15 amphibians (figures do total due 4 being combined mammal data). indicate variance affected 2). Model comparison reveals has strongest (p 0.0007 ***) sub-group 22; p 0.0006***) 3). Within 23), none fixed significantly region, mostly 44; cover)). Adding variables higher mean (49.8% compared without (41.9% cover). measuring finding decline variety used, mainly richness (46%; 31) (25%; 17) S2). break-point vary, applying techniques generalized models, piecewise regression (Muggeo, Toms Lesperance, 2003), logistic 2019) S3). Empirical visually S3) (Balkenhol Benchimol Bergman 2006; Boesing Estavillo Püttker (average 36.5% varies widely. small plots larger 56.06 km2 (SE 18.81) S4); 13 100 km2. generally high, 68% landscapes 80 100% (Mean 81.62%, 2.34). 26% 45% 60%. promising easy communicate concept, receive attention community, regions, gaining traction (Dunning, calls kept above 30% 40% Fahrig, Rompré 2010). become widely referred ecology since Andrén (1994) below exerts increased pressures populations. message maintaining least put forward (2010), who ranges (using boreal temperate forests) 40%, 33.6% latitudes 7) (I. (2014) Brazil's Forest, adopted government Thresholds canopy (a similar measure, although (Asrat 2018)) industry “Biodiversity-friendly” coffee certification shown positive (Caudill Rice, 2016). presence 4), highlights congruence indeed definition itself. review, analytical biodiversity, consistent defining (percentage) “drastic” measured. 42% cover, conceptual framework does allow robustly assessed (Martin However, it enough? highly 2020), considered (Rezende grow draw several meaningful forests. unevenly distributed among regions. Data remain toward birds. hotspots (Myers 2000), showing follows (Di Marco Trimble Aarde, 2012). pollinating (jaguars (Zemanova 2017)). concentrated many biome, (Ávila-Gómez Kupsch non-independence observation means comparisons inferences validity scales locations cannot made, landscapes, (Roque replications improve understanding taxa Ávila-Gómez Conservation interventions policies apply entire countries biomes (Paloniemi 2012), few quantifying relationship km2, 54% Some representing considered, reach up 1.3 (in case Brazilian Forest) (e.g. (sample areas) remaining scalability dynamics, nearly Information), estimates home mammalian fauna ~788.4 (Bogoni thousand kilometers (Pe’er Purvis 2000). Analyzing implications range-sizes dispersal needs, apex predators, crucial (Holland 2004), currently under-represented Arroyo-Rodríguez (2020), guidelines applicable “most species”, requirements lower 1,000 “smaller landscapes” (of 3000 km2). insufficient mismatch made challenge development successful poorly (Gonzalez Thus, dictate keep portions should approached care examples long-term land-use biodiversity. Time lags (Norris, 2016), prevents impacts These delayed species-specific differ scales. Existing time-series bridge Dornelas Hansen vast majority