Tree species that ‘live slow, die older’ enhance tropical peat swamp restoration: Evidence from a systematic review

The restoration of degraded forested lands is a global priority incentivised by international commitments to counteract decades rapid deforestation (United Nations Framework Convention on Climate Change, 2019). For some types less forests, passive (including assisted natural regeneration) without planting trees can be resilient and cost-effective form reforestation (Chazdon & Guariguata, 2016; Crouzeilles et al., 2017; Molin 2018). However, many forests may require active reforestation, specifically the trees. Ensuring that planted survive not simple task, entails multiple ecological, economic social considerations (Di Sacco 2021; Meli 2014). Selecting which tree species plant represents central consideration depends upon goal project, while being constrained logistical issues such as seed availability (Chechina Hamann, 2015; Trial error through pilot trials, local knowledge accumulated past experience helps guide selection Graham 2017). Syntheses using seedling monitoring data from projects remain rare, particularly in tropics (Dimson Gillespie, 2020; Suding, 2011). This dearth syntheses due perception performance too site specific, physical linguistic difficulties accessing ‘grey literature’ (e.g. government reports, conference proceedings, working papers), legitimate lack dropped during following budget cuts (Corlett, 2011; Holl, Nevertheless, evidence-based are valuable tools aid future improve outcomes. Tropical peat swamp (TPSF) wetland forest ecosystems globally valued for carbon storage, centres endemism, floral faunal diversity, regional water cycling, livelihoods communities, public health cultural landscapes (Harrison, Ottay, Harrison, Wijedasa, Page Posa Globally tropical peatlands cover 185 470 million km2, equivalent ~3% land surface area, with large expanses TPSF found across Southeast Asia, South America equatorial Africa (Dargie Gumbricht Xu have experienced degradation recent timber logging, conversion agricultural purposes, associated drainage fire, widespread loss Asia Miettinen 2009). Between 1990 2015, area ‘intact’ declined 76% 29% (11–4.6 Mha) its original 25 ha (Miettinen 2016). Intact poor drainage, permanent waterlogging anaerobic conditions slow decomposition lead formation peat, is, partially decomposed organic matter. Conversion cause oxidation elevated CO2 emissions, subsidence (lowering surface) flooding, increased risk fires (particularly drier El Niño conditions), vegetation primarily comprised dense thicket sedges ferns competes naturally recolonising seedlings (Blackham 2014; Mishra highest levels impede regeneration over several decades, thus necessitating (Giesen Sari, 2018; Page, While substantive efforts been made reforest peatlands, ecological needed reliably inform decision-making remains limited. There literature reviews detailing their survival rates (see (Dohong Giesen van der Meer, 2009; Graham, Taylor Although useful, strength systematic review quantitative meta-analysis provides more robust approach (Andivia 2019; Romanelli 2021), instance, accounting variations numbers duration assessing variability survival. Harsh environmental TPSF, stemming draining, burning, periodic flooding canopy known threaten establishing Lampela Eijk In an attempt growth plethora treatments applied. Several these generic restoration, example fertilisation, shading weeding (Graham, 2013; 2019) inoculating mycorrhizal fungi (Graham Turjaman Yuwati 2008). Other largely specific rewetting preventing further reducing fire Wösten 2006), control mounding prevent submergence reduce mortality (Giesen, 2004; Rotinsulu most, if all, native tolerant severe above height increase Flooding exacerbated occurring micro-topography intact TPSF; highs—‘hummocks’—and lows—‘hollows’—caused fall, buttresses, root pneumatophores deposits matter (Freund Constructing artificial mounds raise heights flood line designed mimic hummocks. Yet, has shown both positive (Nuyim, 2000; Santosa Wibisono 2005) neutral effects (Lampela 2018) Knowledge about functional traits increasingly applied predict how respond site-specific barriers (Wainwright Broadly, biochemical, morphological, phenological physiological properties individuals enable them survive, grow, thereby higher fitness within given environment (Violle 2007). ideal would fast growing quickly, high enough rate maintain other recruit under them, inexpensive rear meet targets. To maximise photosynthesis growth, faster generally larger thinner leaves, nutrient contents lower wood densities, but at cost vulnerability disturbance damage enemies (Philipson Russo Wright 2010). context successfully identify growth-mortality spectrum (Charles, Charles Martínez-Garza 2013). example, densities probability when restore wet Despite research leaf pristine (Tuah 2003; Wedeux, Yanbuaban 2007), our knowledge, there no formal reforestation. this study, we undertake Asia. focus namely facilitate agroforestry systems, fruit aim generate comprehensive assessment factors influencing ultimately help Using species-specific compiled review, carried out address questions: (1) Which grow best peatlands? (2) treatments, (notably drained vs. rewetted) strongest impact growth? (3) Can used rates? A search was conducted between May October 2020. Bibliometric searches followed guidelines outlined Collaboration Environmental Evidence (Collaboration Evidence, 2018), targeted scholarly grey literature, latter including non-government guidelines, theses unpublished datasets (Haddaway Haddaway Bayliss, 2015). total 15 engines, 94 journal archives, 32 full proceedings (conferences spanning 1995–2018) 28 institute forestry department repositories were searched (Tables S1–S5, Appendix 1 Supporting Information). Online engines explored combination 64 keywords optimised text-mining keyword co-occurrence networks R package ‘LitSearchR’ (Grames (Table S6, All English, Indonesian Malay languages. Potentially relevant articles English titles or abstracts main text Japanese German also included. Collectively, resulted 605 potentially screened eligibility (Figure S1, Results same sometimes reported studies publication formats journals, reports). Therefore, duplicate information consolidated into single study-site analysis. defined identifiable location (i.e. georeferenced name) period. Studies split study-sites based different locations habitat open enrichment) timing planting, instance replanting (near-) complete planting. Experimental units sites, blocks, transects plots, considered separate study-sites. Conversely, presenting results period grouped study-site. texts fulfilling criteria read in-depth gather study location, planted, condition history hydrological condition, duration, data. Almost all lacked necessary none provided raw We contacted 46 authors request missing additional one 16 partial studies. reports four since published S7, bibliometric author datasets, contained S7; Figure collected S8, Information), supplemented those three online databases: PROSEA (https://www.prota4u.org/prosea/search.aspx), ICRAF (http://db.worldagroforestry.org//) DRYAD Global Wood Density (https://datadryad.org/stash/dataset/doi:10.5061/dryad.234). total, 27 literature-matched identified review. subset seven included analyses sufficient corresponding S9, Information) below). Species taxonomic names verified against records assigned categories: Asian occur non-peat (mineral soil) habitats only, exotic species. excluded they had unknown name, believed mis-identified, only region distant site, genus only. Taxonomic checked database vascular (The Plant List, 2013) synonyms unified most up-to-date records. Our dataset mainly comprises species, includes tree-like palms Areca Metroxylon species). refer ‘seedlings’ simplicity. variety studies, impossible fully incorporate analyses. therefore broad categories S10, mycorrhiza fungal (AM ECM) ‘mycorrhizal-inoculated’. mounding, fertilisation) kept treatment combinations rather than treatments. Treatments intensities 50% shade 100% shade) ‘shading’ category Across little consistent reporting conditions. Given importance recovery impacts survival, concentrated determining whether rewetted. categorical variable table measurements, descriptions management interventions blocked canals) articles, where author's responses questions. tried determine post-fire before then monitored post-fire) later excluded, because undertook measures individual datapoint unique Half-life RGR required time-series minimum measurements line-fitting. Asymptotic model fits could negative estimates, presumably exceptionally low RGR, predictions omitted. Study-sites analysis least 14 sampling variances (three times average) 14. half-life calculations, averaged first 2 months, replaced dead few months plantings took month initial monitoring. patchiness relation conditions, models inclusion addressing questions Information; Hector Survival, each analysed separately five models, totalling models. tested differences among following: (4) (drained rewetted), (5) experimental design site-level rewetting; instead rewetted sites Furthermore, replicated trait analyses, represented paired below) Survival final proportion surviving individuals. investigate multilevel linear mixed covariates (moderators) ‘rma.mv’ function metafor (Viechtbauer, Model tests confidence intervals computed Hartung–Knapp–Sidik–Jonkman method (Pappalardo 2020). covariate. Seedling covariate, density contrasted Random inner random factor outer factor, scaled identity matrix. structure assumes independent correlated. second did account durations. model, per selected coverage (leaf calcium, nitrogen, magnesium, phosphorus, potassium, RGR), level component phylogenetic correlation constructed ‘phylo.maker’ v.phylomaker (Jin Qian, S2, generalised ‘glmmTMB’ glmmtmb (Brooks log transformed square homoscedasticity requirements. fitted assuming Gaussian distribution fixed treatment, respective (the above). Study-site effect except (outlined above) part ‘glmmTMBphylo’ phyloglmm (https://github.com/wzmli/phyloglmm). generated contrasts releveling running comparator two commonly (Shorea balangeran Dyera polyphylla), temporal trends above, Niño-Southern Oscillation (ENSO) (El Niño, La Niña neutral) climatic dominating 6 project moderator term (National Oceanic Atmospheric Administration, 2021). performed statistical software v4.0.2 (R Core Team, diagnostics residual vs values, bias funnel plots sensitivity outliers ‘metafor’ Residuals dharma (Hartig, Of majority located Indonesia (83%), principally Central Kalimantan province (50%) 1a). year spanned 1988 2019 number increasing substantially after 1997/1998 1b). 30 ± 34 (mean standard deviation, median 18 range 2.5–180 months) 1c). diversity low, one-third average 4 maximum 23 1d). ranged almost 56,000 1e). Overall, 141 palm comprising: 113 12 2). Shorea (39 study-sites) polyphylla (34), Gonystylus bancanus (17), Melaleuca cajuputi (11) Alstonia pneumatophora (10), 62 modelling 43 half-life, 24 S1; Table S11, eight TPSF. overall mean 62% (95% CI: 118%) 439 study-site, Average 33 ranging 0.6 174 months. differed significantly likely Validating measure, time until 45%–55% positively correlated (Spearman's correlation; rs = 0.97, df 82, p <0.001; S3, longer compared Lophopetalum cf. rigidum, Cratoxylum arborescens, balangeran, spatulata Madhuca motleyana 1; contrast, shorter included: Calophyllum hosei, Aglaia rubiginosa, Horsfieldia crassifolia, Licania splendens Xylopia fusca Additional S. D. polyphylla, showed ENSO influenced S12, Compared years (wetter) (drier) Meanwhile, S12; S4, Mycorrhizal inoculation sufficiently 10 subsets Seedlings inoculated small statistically greater 78% 86% 3a–c; S13, significant mycorrhizal-inoculated Six arbuscular (AM), (S. balangeran) ectomycorrhizal (ECM). Due imbalance, type analysis, although it 3. 11 effects, 3d–f; S14, corroborated RGR; occurred sooner shaded controls 3; Fertilisation reduced difference On average, 87% fertilised 5% higher. <3%) observed metric specifically: versus averaging 69% 59%, respectively affect Only exhibited ranges (included analyses; 1.1 24.6 mg/g calcium; 7.5 18.5 nitrogen; 6.6 magnesium; 0.1 1.8 1.6 10.6 potassium. 0.28 0.9 g/cm3 predicted traits, magnesium potassium S15, negatively related 4). words, cohorts slower last reaching mortality. When expanded include 66 relationship similar trend, 45 synthesis lasted 2.5 long lived (Russo 2010), finds give promise assist sense realism remaining challenges. One major challenge seemingly marginal inconsistent growth. Another chronic underutilisation flora reforestation: one-tenth 1173 2011) planted. select tremendous variation half-l