Ecology

Sarwar, N. et al. Phytoremediation strategies for soils contaminated with heavy metals: Modifications and future perspectives. Chemosphere 171, 710–721 (2017).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Lin, H. M. et al. Cadmium-stress mitigation through gene expression of rice and silicon addition. Plant Growth Regul.: Int. J. Nat. Synthetic Regul. 81(1), 91–101 (2017).Article 
CAS 

Google Scholar 
Pan, F. S. et al. Enhanced Cd extraction of oilseed rape (Brassica napus) by plant growth-promoting bacteria isolated from Cd hyperaccumulator Sedum alfredii Hance. Int. J. Phytorem. 19(1/6), 281–289 (2017).Article 
CAS 

Google Scholar 
Puangprasert, S. & Prueksasit, T. Health risk assessment of airborne Cd, Cu, Ni and Pb for electronic waste dismantling workers in Buriram Province, Thailand. J. Environ. Manag. 252, 109601 (2019).Article 
CAS 

Google Scholar 
Tipu, M. I. et al. Growth and physiology of maize (Zea mays L.) in a nickel-contaminated soil and phytoremediation efficiency using EDTA. J. Plant Growth Regul. 40(2), 774–786 (2021).Article 
CAS 

Google Scholar 
Chaturvedi, N., Dhal, N. K. & Patra, H. K. EDTA and citric acid-mediated phytoextraction of heavy metals from iron ore tailings using Andrographis paniculata: A comparative study. Int. J. Min. Reclam. Environ. 29(1), 33–46 (2015).Article 
CAS 

Google Scholar 
Wang, G. Y. et al. Heavy metal removal by GLDA washing: Optimization, redistribution, recycling, and changes in soil fertility. Sci. Total Environ. 569–570, 557–568 (2016).Article 
ADS 
PubMed 

Google Scholar 
Kołodyńska, D. Cu(II), Zn(II), Co(II) and Pb(II) removal in the presence of the complexing agent of a new generation. Desalination 267(2–3), 175–183 (2011).Article 

Google Scholar 
Guo, X. F. et al. Mixed chelators of EDTA, GLDA, and citric acid as washing agent effectively remove Cd, Zn, Pb, and Cu from soils. J. Soils Sediments 18(2), 835–844 (2017).
Google Scholar 
Wang, X. et al. Subcellular distribution and chemical forms of cadmiun in Bechmeria nivea L. Gaud. Environ. Exp. Bot. 62(3), 389–395 (2008).Article 
CAS 

Google Scholar 
Gallego, S. M. et al. Unravelling cadmium toxicity and tolerance in plants: Insight into regulatory mechanisms. Environ. Exp. Bot. 83, 33–46 (2012).Article 
CAS 

Google Scholar 
Clemens, S., Aarts, M. G. M., Thomine, S. & Verbruggen, N. Plant science: The key to preventing slow cadmium poisoning. Trends Plant Sci. 18(2), 92–99 (2013).Article 
CAS 
PubMed 

Google Scholar 
Zhou, J. T. et al. Integration of cadmium accumulation, subcellular distribution, and physiological responses to understand cadmium tolerance in apple rootstocks. Front. Plant Sci. 8, 966 (2017).Article 
PubMed 
PubMed Central 

Google Scholar 
Yang, L. P., Zhu, J., Wang, P., Lyu, D. G. & Li, H. F. Effect of Cd on growth, physiological response, Cd subcellular distribution and chemical forms of Koelreuteria paniculata. Ecotoxicol. Environ. Saf. 160, 10–18 (2018).Article 
CAS 
PubMed 

Google Scholar 
Wang, W. J., Zhang, M. Z. & Liu, J. N. Subcellular distribution and chemical forms of Cd in Bougainvillea spectabilis Willd. as an ornamental phytostabilizer: An integrated consideration. Int. J. Phytorem. 20(11), 1087–1095 (2017).Article 

Google Scholar 
Weigel, H. J. & Jäger, H. J. Subcellular distribution and chemical form of cadmium in bean plants. Plant Physiol. 65(3), 480–482 (1980).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Khanna, K., Kohli, S. K., Ohri, P., Bhardwaj, R. & Ahmad, P. Agroecotoxicological aspect of Cd in soil–plant system: Uptake, translocation and amelioration strategies. Environ. Sci. Pollut. Res. 29, 30908–30934 (2022).Article 
CAS 

Google Scholar 
Wei, Z. B., Chen, X. H., Wu, Q. T. & Tan, M. Biodegradable chelator GLDA induced remediation of heavy metal contaminated soil in Southeast Jingtian. Environ. Sci. 36(5), 1864–1869 (2015).CAS 

Google Scholar 
Wang, K., Liu, Y. H., Song, Z. G., Wang, D. & Qiu, W. W. Chelator complexes enhanced Amaranthus hypochondriacus L. phytoremediation efficiency in Cd-contaminated soils. Chemosphere 237, 124480 (2019).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Meng, N., Wang, M., Chen, L., Zheng, H. & Chen, S. B. Remediation effects of different herbaceous plants intercropping on Cd-contaminated soil. China Environ. Sci. 38(7), 2618–2624 (2018).CAS 

Google Scholar 
Jones, D. & Willett, V. Experimental evaluation of methods to quantify dissolved organic nitrogen (don) and dissolved organic carbon (doc) in soil. Soil Biol. Biochem. 38(5), 991–999 (2006).Article 
CAS 

Google Scholar 
Su, F. L. et al. The distribution and enrichment characteristics of copper in soil and Phragmites australis of Liao River estuary wetland. Environ. Monit. Assess.: Int. J. 190(6), 1–9 (2018).Article 
CAS 

Google Scholar 
Shahid, M., Dumat, C. & Khalid, S. Reviews of Environmental Contamination and Toxicology Vol. 241, 3–137 (Springer, 2016).
Google Scholar 
Yuliya, V. et al. Comparison of soil-to-root transfer and translocation coefficients of trace elements in vines of Chardonnay and Muscat white grown in the same vineyard. Sci. Hortic. 192, 89–96 (2015).Article 

Google Scholar 
Liu, Q. Q., Chen, Y. H., Shen, Z. G. & Zheng, L. Q. Roles of cell wall in plant heavy metal tolerance. Plant Physiol. J. 50(5), 605–611 (2014).
Google Scholar 
Zhen, S. et al. Foliar application of Zn reduces Cd accumulation in grains of late rice by regulating the antioxidant system, enhancing Cd chelation onto cell wall of leaves, and inhibiting Cd translocation in rice. Sci. Total Environ. 770, 145302 (2021).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Shi, Y. X. et al. Simulation of the absorption, migration and accumulation process of heavy metal elements in soil-crop system. Environ. Sci. 37(10), 3996–4003 (2016).
Google Scholar 
Yan, X. X. et al. Effect of foliar application of different manganese fertilizers on cadmium accumulation and subcellular distribution in pak choi. J. Agro Environ. Sci. 38(8), 1872–1881 (2019).
Google Scholar 
He, S., Wu, Q. & He, Z. Effect of DA-6 and EDTA alone or in combination on uptake, subcellular distribution and chemical form of Pb in Lolium perenne. Chemosphere 93(11), 2782–2788 (2013).Article 
ADS 
CAS 
PubMed 

Google Scholar 
Li, C. C. et al. Integration of metal chemical forms and subcellular partitioning to understand metal toxicity in two lettuce (Lactuca sativa L.) cultivars. Plant Soil 384(1/2), 201–212 (2014).Article 
CAS 

Google Scholar 
Li, D., He, T., Saleem, M. & He, G. Metalloprotein-specific or critical amino acid residues: Perspectives on plant-precise detoxification and recognition mechanisms under cadmium stress. Int. J. Mol. Sci. 23(3), 1734 (2022).Article 
CAS 
PubMed 
PubMed Central 

Google Scholar 
Perriguey, J., Sterckeman, T. & Morel, J. L. Effect of rhizosphere and plantrelated factors on the cadmium uptake by maize(Zea mays L.). Environ. Exp. Bot. 63(1/3), 333–341 (2008).Article 
CAS 

Google Scholar 
Dai, S. et al. Effects of biochar amendments on speciation and bioavailability of heavy metals in coal-mine-contaminated soil. Hum. Ecol. Risk Assess. Int. J. 24(7), 1887–1900 (2018).Article 
CAS 

Google Scholar 
Hou, S., Zheng, N., Tang, L., Ji, X. F. & Li, Y. Y. Effect of soil pH and organic matter content on heavy metals availability in maize (Zea mays L.) rhizospheric soil of non-ferrous metals smelting area. Environ. Monit. Assess. 191(10), 634 (2019).Article 
PubMed 

Google Scholar 
Wu, H. J. et al. Effects of Astragalus smicuson cadmium effectiveness in paddy soil and cadmium accumulation in rice plant. Chin. Agric. Sci. Bull. 33(16), 105–111 (2017).ADS 

Google Scholar 
Jin, P. K., Liu, K. J. & Wang, X. B. Conversion and utilization of slowly biodegradable organic matter. Chin. J. Environ. Eng. 10(5), 2168–2174 (2016).CAS 

Google Scholar 
Kopáček, J. et al. Factors affecting the leaching of dissolved organic carbon after tree dieback in an unmanaged European mountain forest. Environ. Sci. Technol. 52(11), 6291–6299 (2018).Article 
ADS 
PubMed 

Google Scholar 
Anwar, S. et al. Impact of chelator-induced phytoextraction of cadmium on yield and ionic uptake of maize. Int. J. Phytorem. 19(6), 505–513 (2017).Article 
CAS 

Google Scholar 
Wu, J. M., Xi, M. & Kong, F. L. Review of researches on the factors influencing the dynamics of dissolved organic carbon in soils. Geol. Rev. 59(5), 953–961 (2013).CAS 

Google Scholar 
AkzoNobel. Dissolvine GL® Technichal Brochure 1–5 (AkzoNobel Amsterdam, 2010).
Google Scholar 
Beygi, M. & Jalali, M. Assessment of trace elements (Cd, Cu, Ni, Zn) fractionation and bioavailability in vineyard soils from the Hamedan, Iran. Geoderma 337, 1009–1020 (2019).Article 
ADS 
CAS 

Google Scholar 
Gul, I. et al. Comparative effectiveness of organic and inorganic amendments on cadmium bioavailability and uptake by Pelargonium hortorum. J. Soils Sediments 19(5), 2346–2356 (2019).Article 
CAS 

Google Scholar 
Wang, H., Sun, L. N., Li, H. B. & Sun, T. Y. Effect of different chelators application on Cd accumulation in metal polluted soils by Beta vulgaris var. cicla L. Ecol. Environ. 17(6), 2249–2252 (2008).
Google Scholar 
Zhang, G. X. et al. Effects of biochars on the availability of heavy metals to ryegrass in an alkaline contaminated soil. Environ. Pollut. 218, 513–522 (2016).Article 
CAS 
PubMed 

Google Scholar 
Gu, M. H. et al. Effects of manganese application on the formation of manganese oxides and cadmium fixation in soil. Ecol. Environ. Sci. 229(2), 360–368 (2020).
Google Scholar 
Bradl, H. B. Adsorption of heavy metal ions on soils and soils constituents. J. Colloid Interface Sci. 277(1), 1–18 (2004).Article 
ADS 
CAS 
PubMed 

Google Scholar  More

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.This is a summary of: Marin, F. R. et al. Protecting the Amazon forest and reducing global warming via agricultural intensification. Nat. Sustain. https://doi.org/10.1038/s41893-022-00968-8 (2022). More

Longline fishing boats such as these at Honolulu’s harbour in Hawaii must respect a large no-fishing zone off the western side of the archipelago.Credit: Sarah Medoff

Large no-fishing areas can drive the recovery of commercially valuable fish species, a study suggests. Ten years’ worth of fisheries data have shown that catch rates of two important types of tuna increased drastically in the vicinity of a marine protected area surrounding the northwestern Hawaiian islands.“It’s a win–win for fish and fishermen,” says Jennifer Raynor, an economist at the University of Wisconsin–Madison and a co-author of the study, which was published on 20 October in Science1.The results highlight the value of large-scale marine protected areas — a type of environmental management that has emerged in the past two decades, mostly in the Pacific Ocean, says Kekuewa Kikiloi, who studies Hawaiian culture at the University of Hawaii at Mānoa. Countries around the world have committed to protecting 30% of their land and oceans by 2030.Previous research showed that marine protected areas can help to restore populations of creatures that don’t move around much or at all, such as corals2 and lobsters3. Raynor and her colleagues wanted to test whether the areas could also drive the recovery of migratory species and provide spillover benefits for fisheries. The researchers looked at one of the largest such areas in the world, the 1.5-million-square-kilometre Papahānaumokuākea Marine National Monument, which was created in 2006 and expanded in 2016 to protect biological and cultural resources.The team focused on the Hawaiian ‘deep-set’ longline fishery, which mainly targets yellowfin tuna (Thunnus albacares) and bigeye tuna (Thunnus obesus).The researchers analysed catch data collected on fishing vessels between 2010 and late 2019. Then, they compared catch rates at various distances up to 600 nautical miles (1,111 kilometres) from the protected area, before and after its expansion in 2016. (The protected area itself currently extends for 200 nautical miles from the northwestern part of the Hawaiian archipelago.) They found that after the expansion, catch rates — defined as the number of fish caught for every 1,000 hooks deployed — went up, and that the increases were greater the closer the boats were to the no-fishing zone. At distances of up to 100 nautical miles, the catch rate for yellowfin tuna increased by 54%, and that for bigeye tuna by 12%. Some other types of catch rate also increased, but not by equally significant margins.The size of the Papahānaumokuākea Marine National Monument — more than three times the surface area of California — probably played a part in the positive effects, as did its shape. It spans about 2,000 kilometres from west to east, protecting large swathes of ocean waters at tropical latitudes. This means that tropical fish such as yellowfin and bigeye tuna — which tend to move along an east–west axis to stay in their preferred temperature range — can travel a long way and still stay in the no-fishing zone.What’s more, says Raynor, Papahānaumokuākea is a spawning ground for yellowfin tuna. Because the animals don’t travel far from their birthplace, the no-take zone provides refuge from fishing, helping tuna to aggregate and reproduce.“It is exciting to see that there are benefits to the fishing industry from this marine protected area,” says David Kroodsma, director of research and innovation at Global Fishing Watch in Oakland, California, a US non-governmental organization that monitors fishing activity worldwide. However, he adds, it’s unclear whether the results can be generalized to other areas of the world.Regardless, the findings could help others to design marine protected areas so that benefits trickle down to fisheries, says Steve Gaines, a marine ecologist at the University of California, Santa Barbara. The study, he says, “provides a platform to definitively evaluate what is working and what isn’t”.Co-managed by Indigenous populations, the state of Hawaii and the US government, Papahānaumokuākea is an example of a collaborative management strategy that bridges Indigenous knowledge and modern science, Kikiloi says. The approach, he adds, “can work successfully in other places too, if given a chance”. More

Field samplingLake Masoko fish were chased into fixed gill nets and SCUBA by a team of professional divers at different target depths determined by diver depth gauge (12× male benthic, 12× male littoral). Riverine fish (11× Mbaka River and 1× Itupi river) were collected by local fishermen. On collection, all fish were euthanized using clove oil. Collection of wild fish was done in accordance with local regulations and permits in 2015, 2016, 2018 and 2019. On collection, fish were immediately photographed with color and metric scales, and tissues were dissected and stored in RNAlater (Sigma-Aldrich); some samples were first stored in ethanol. Only male specimens (showing bright nuptial coloration) were used in this study for the practical reason of avoiding any misassignment of individuals to the wrong population (only male individuals show clear differences in phenotypes and could therefore be reliably assigned to a population). Furthermore, we assumed that any epigenetic divergence relevant to speciation should be contributing to between-population differences in traits possessed by both sexes (habitat occupancy, diet). To investigate the role of epigenetics in phenotypic diversification and adaptation to different diets, homogenized liver tissue – a largely homogenous and key organ involved in dietary metabolism, hormone production and hematopoiesis – was used for all RNA-seq and WGBS experiments.Common-garden experimentCommon-garden fish were bred from wild-caught fish specimens, collected and imported at the same time by a team of professional aquarium fish collectors according to approved veterinary regulations of the University of Bangor, UK. Wild-caught fish were acclimatized to laboratory tanks and reared to produce first-generation (G1) common-garden fish, which were reared under the same controlled laboratory conditions in separate tanks (light–dark cycles, diet: algae flakes daily, 2–3 times weekly frozen diet) for approximately 6 months (post hatching). G1 adult males showing bright nuptial colors were culled at the same biological stages (6 months post hatching) using MS222 in accordance with the veterinary regulations of the University of Bangor, UK. Immediately on culling, fish were photographed and tissues collected and snap-frozen in tubes.Stable isotopesTo assess dietary/nutritional profiles in the three ecomorph populations, carbon (δ13C) and nitrogen (δ15N) isotope analysis of muscle samples (for the same individuals as RRBS; 12, 12 and 9 samples for benthic, littoral and riverine populations, respectively) was undertaken by elemental analyzer isotope ratio mass spectrometry by Iso-Analytical Limited. It is important to note that stable isotope analysis does not depend on the use of the same tissue as the ones used for the RRBS/WGBS samples45. Normality tests (Shapiro–Wilk, using the R package rstatix v.0.7.0), robust for small sample sizes, were performed to assess sample deviation from a Gaussian distribution. Levene’s test for homogeneity of variance was then performed (R package carData v.3.0-5) to test for homogeneity of variance across groups. Finally, Welch’s ANOVA was performed followed by Games–Howell all-pairs comparison tests with adjusted P value using Tukey’s method (rstatix v.0.7.0). Mean differences in isotope measurements and 95% CI mean differences were calculated using Dabestr v.0.3.0 with 5,000 bootstrapped resampling.Throughout this manuscript, all box plots are defined as follows: centre line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers.RNA-seqNext-generation sequencing library preparationTotal RNA from liver tissues stored in RNAlater was extracted using a phenol/chloroform approach (TRIzol reagent; Sigma-Aldrich). Of note, when tissues for bisulphite sequencing samples were not available, additional wild-caught samples were used (Supplementary Table 3). The quality and quantity of RNA extraction were assessed using TapeStation (Agilent Technologies), Qubit and NanoDrop (Thermo Fisher Scientific). Next-generation sequencing (NGS) libraries were prepared using poly(A) tail-isolated RNA fraction and sequenced on a NovaSeq system (S4; paired-end 100/150 bp; Supplementary Table 3), yielding on average 32.9 ± 3.9 Mio reads.Read alignment and differential gene expression analysisAdaptor sequence in reads, low-quality bases (Phred score  More

Wolf, S. et al. Nat. Ecol. Evol. https://doi.org/10.1038/s41559-022-01904-x (2022).Article 

Google Scholar 
Kattge, J. et al. Glob. Change Biol. 26, 119–188 (2020).Article 

Google Scholar 
Sabatini, F. M. et al. Glob. Ecol. Biogeogr. 30, 1740–1764 (2021).Article 

Google Scholar 
Łopucki, R., Kiersztyn, A., Pitucha, G. & Kitowski, I. Ecol. Modell. 468, 109964 (2022).Article 

Google Scholar 
Kosmala, M., Wiggins, A., Swanson, A. & Simmons, B. Front. Ecol. Environ. 14, 551–560 (2016).Article 

Google Scholar 
White, C. R. et al. Funct. Ecol. 35, 1572–1578 (2021).Article 

Google Scholar 
Xirocostas, Z. A., Debono, S. A., Slavich, E. & Moles, A. T. Methods Ecol. Evol. 13, 596–602 (2022).Article 

Google Scholar 
Callaghan, C. T. et al. Bioscience 71, 55–63 (2020).
Google Scholar  More

Data collectionIn phase 1 (2017), we first circulated a comprehensive multi-language questionnaire and associated guidelines on protected area personnel numbers to major national protected area agencies, focusing on the 50 countries listed in the WDPA as having the most protected areas. The questionnaire requested information on personnel numbers, type of employers and management levels (from executive to skilled practical workers). Protected area personnel were defined as those spending at least 50% of their work time on protected area-related tasks. The questionnaire also requested information about job titles used for personnel equivalent to rangers. This phase produced usable data for 28 countries/territories.In phase 2 (2018 onwards), we conducted online searches for published data on protected area personnel numbers in the countries/territories not included in the questionnaire survey or where questionnaire responses were incomplete or unclear. The resulting information came from official organizational reports (10 countries/territories), published external studies, project documents and journal papers (35 countries/territories) and websites of protected area organizations or individual sites (9 countries/territories).In phase 3 (2018–2021), we directly requested personal contacts to locate or supply information from official sources both for the remaining countries/territories and to improve or verify data from phases 1 and 2. The minimum data requested were the overall number of protected area personnel, the number of those personnel that could be categorized as rangers, the terrestrial area of protected areas managed by the listed personnel and the source of the information. This phase contributed usable data for 68 countries and territories. Data for a further 17 countries/territories were assembled from multiple sources.The final dataset covered 176 countries/territories: 167 surveyed countries/territories and a further 9 countries/territories that have no WDPA-listed protected areas (Supplementary Table 1), with contributions from more than 150 individuals.Initial data processingTo assess and, where necessary, improve the reliability of data obtained in a wide range of formats and levels of detail and from multiple sources, we scored the data for each country/territory from 0 to 5 for each of four criteria—detail, accuracy, source and age of the data—with a maximum score of 20 (Supplementary Table 1 and Supplementary Fig. 1). For all low-scoring records (a score of less than 15), we sought more-reliable sources in later phases of the study, rejecting any final scores of less than 10.On reviewing the data, we excluded from the analysis protected areas identified in the WDPA as predominantly or entirely marine, Antarctica and countries/territories categorized in the WDPA as polar (Greenland, French Southern Territories, Bouvet Island, Heard Island and McDonald Islands, South Georgia and the South Sandwich Islands). These large, remote and/or largely uninhabited areas are likely to have quite different management models and scales of staffing from terrestrial protected areas (although marine protected areas are also widely understaffed11). For example, in 2012 the 972,000 km2 of Northeast Greenland Protected Area (categorized by the WDPA as polar) was only periodically visited by six two-person teams of naval personnel47, and the 2008 management plan of the 1.51 million km2 Papahānaumokuākea Marine National Monument (Hawai’i, USA) specifies just nine personnel, working in conjunction with several other agencies48. Data for one country were supplied by officials on the agreement that the country was not specifically identified in publications (the country is given the three-letter code ZZZ in relevant tables and figures).Because the format, completeness and level of detail of the data varied widely, from comprehensive personnel lists to single figures, we restricted our raw dataset to six variables that could be consistently extracted from data obtained for each country/territory:

1.

Total number of non-ranger personnel (if known)

2.

Total number of rangers (if known)

3.

Total number of protected area personnel (either the sum of 1 and 2 or provided as an undifferentiated total)

4.

Terrestrial area of protected areas covered by surveyed personnel (km2)

5.

Total terrestrial area of protected areas of the country/territory (km2)

6.

Year of the data

We used the WDPA, official publications and websites to determine (or verify) the area of terrestrial protected areas covered by the personnel listed for each country/territory, using WDPA data if there were discrepancies. Total national terrestrial protected area coverage was taken from the WDPA, with the exception of Turkey, where the area officially reported to the WDPA is significantly less than the nationally published area.The raw data from the survey are shown in Supplementary Table 1.Candidate predictorsTo predict the number of rangers and non-rangers in countries and territories for which we had no data (Statistical analysis), we collected information on the following set of variables, hereafter referred to as candidate predictors:Location dataThe WGS84 latitude and longitude of the centroid of the largest land mass associated with each country/ territory (to obtain the polygons defining the land masses, we used the R package rnaturalearth version 0.1.0; https://github.com/ropensci/rnaturalearth)2020 data from the World Bank (https://data.worldbank.org/indicator)

Area of the country/territory

Population density: the mid-year population divided by land area

Gross domestic product (GDP) in US dollars

GDP per capita in US dollars (GDP divided by mid-year population)

Growth rate of GDP

The proportion of rural inhabitants

The proportion of unemployed inhabitants

The forested proportion of the country/territory

2020 data for each country/territory from the WDPA (https://www.protectedplanet.net/)

The total terrestrial area of WDPA-listed protected areas

The proportion of the terrestrial area of all IUCN-categorized protected areas (Categories I–VI) that falls within protected areas in Category I or II

The proportion of the terrestrial area of all IUCN-categorized protected areas (Categories I–VI) that falls within protected areas in Categories I–IV

2020 data from the Yale Center for Environmental Law and Policy Environmental Performance Index (https://epi.yale.edu/)

Environmental Performance Index (EPI): a composite index using 32 performance indicators across 11 categories

Ecosystem Vitality Index (EVI): an indicator of how well countries preserve, protect and enhance ecosystems and the services they provide

Species Protection Index (SPI): an indicator of the species-level ecological representativeness of each country’s/territory’s protected area network

Not all this information was available for all countries/territories. Most of the missing data were for small territories that account for only a very small proportion of the total area of protected areas worldwide (Supplementary Table 2c).Statistical analysisOur primary objective was to estimate the total number of all personnel engaged in managing all the world’s WDPA-listed terrestrial protected areas and the number categorized as rangers. Our raw data collection yielded full, partial or no information on total personnel and ranger numbers for each country/territory (Supplementary Table 1 shows the completeness of all the data collected). Our first task, therefore, was (1) to impute the information for unsurveyed protected areas on the basis of information from surveyed protected areas within the same countries/territories and (2) to predict those numbers for countries/territories where no information was available on overall personnel numbers and/or ranger numbers on the basis of relationships we could establish between available information and candidate predictors in other countries/territories (Supplementary Table 7). A brief description of these two approaches follows, and full details on the analysis are provided in Supplementary Information.Data imputationFor countries/territories where we had obtained information about numbers of personnel and/or rangers for only some protected areas, our strategy was to populate the unsurveyed protected areas in proportion to the densities of personnel or rangers from the surveyed protected areas of the same countries/territories. For example, for Spain we obtained evidence that there are 619 rangers responsible for protected areas covering 44,328 km2, out of a national total protected area system covering 142,573 km2. To impute the number of rangers for the remaining 98,245 km2, we used the density of rangers in the surveyed area (one ranger per 44,328/619 = 71.6 km2) and applied that to the unsurveyed area, giving a total of 1,991 rangers (619 + (98,245/71.6)). This imputation assumes that unsurveyed areas are staffed at the same density as surveyed areas, whereas in reality the relative densities are likely to vary in unknown ways within different countries/territories. To study the sensitivity of our results to the assumed proportion, we repeated our analysis using the following proportions of the observed densities: 0, 0.25, 0.50, 0.75 and 1.00. This provided a range of personnel numbers from a minimum (based on a proportion of 0) to a presumed maximum (based on a proportion of 1.00). From the data obtained, it was not possible to calculate the actual proportions, but based on the experience of the practitioners in the author team, the unsurveyed areas are highly unlikely to be staffed at higher densities than surveyed areas and, on average, are very likely to be staffed at lower densities. After all, most survey respondents were national or subnational agencies responsible for protected areas subject to stronger formal requirements for protection and management and therefore likely to have larger workforces. Unsurveyed protected areas are more likely to be managed by local entities, with fewer resources, less-stringent management obligations and therefore fewer personnel. The range of proportions we considered to populate unsurveyed areas should therefore yield predictions encompassing the actual (unknown) numbers of rangers and non-rangers with a conservative margin of error. In the main text, we have reported the results of imputation assuming a proportion of 1, which is probably the most optimistic assessment of the current workforce in protected areas within the proportions of the observed densities considered. Results using lower proportions are shown in Extended Data Fig. 2 and Supplementary Tables 4 and 5.Data predictionOur imputation approach was not possible for countries/territories where (1) zero ranger or personnel data had been obtained and (2) specific data had not been obtained that allowed imputation either for rangers or for total personnel (where only total personnel numbers or only ranger numbers had been obtained). To predict the missing information, we used two different statistical approaches: linear mixed models (LMMs)49 and a general implementation of random forests, which we term RF/ETs because it encompasses both random forests sensu stricto (RFs)50 and a variant called extremely randomized trees (ETs)51. LMMs and RFs have been extensively discussed and reviewed in the literature49,52,53. We adopted these approaches because both have proved successful in producing accurate predictions for a wide range of applications and because both are well suited to our data since they both produce predictions from a set of predictors and allow for the consideration of spatial effects54,55. Furthermore, comparing predictions generated through very different methods informs us about the robustness of our results with respect to key statistical assumptions. LMMs come from the ‘data modelling culture’56 and belong to parametric statistics; RF/ETs come from the ‘algorithmic modelling culture’ and belong to non-parametric statistics.We followed the same workflow for both statistical approaches, comprising eight steps: (1) general data preparation; (2) preparation of initial training datasets; (3) selection of predictor variables and of the method used for handling spatial autocorrelation; (4) preparation of final training datasets; (5) fine tuning; (6) final training; (7) preparation of datasets for predictions and simulations; and (8) predictions and simulations (see Supplementary Information for details).Both approaches yielded very similar results with our data. We chose to present the LMM results in the main text, but we provide and compare the results obtained by both approaches in Supplementary Information.SoftwareWe performed all the data analyses using the free open-source statistical software R version 4.157. We used the R package spaMM version 3.9.13 to implement LMMs58 and the R package ranger version 0.13.1 to implement RF/ETs59. To reformat and plot the data, we used the Tidyverse suite of packages60. Details are provided in an R package we specifically developed so that findings presented in this paper can readily be reproduced (see Code availability). Using a workstation with an AMD Ryzen Threadripper 3990 × 64-core processor and 256 GB of RAM, our complete workflow ran in ~3,000 CPU hours.Estimation of required numbers and densities of personnelTo estimate the numbers of personnel and rangers required for effective management of existing protected areas, we referred to ref. 25. This estimates that the minimum budget needed to adequately manage the existing protected area system is US$67.6 billion per year and that current annual expenditure is US$24.3 billion. From these figures, we can calculate that resources invested in the current global system of protected areas are approximately 36% of what is required. We consulted data from https://ourworldindata.org to determine that the proportion of global public expenditure on employee compensation has remained between 21.01% and 23.33% in the years from 2006 to 2019. We obtained these figures from the ‘Government Spending’ section of the site, consulting the chart ‘Share of employee compensation in public spending, 2002 to 2019’ and selecting data for ‘World’. On the basis of this broadly constant proportion and the assumption that total employee compensation is an indicator of total employee numbers, we inferred that current numbers of protected area employees are also around 36% of what is required. We therefore multiplied our estimations of personnel and ranger numbers by 1/0.36 and recalculated the densities on this basis (current requirement = 1/0.36 × current estimate).To estimate staffing requirements for 30% global coverage of protected areas—the global target intended to be reached by 2030—we used the mean personnel and ranger densities calculated as being required at present to ‘populate’ a global area of terrestrial protected areas if increased from the percentage at the time of our study (15.7%) to 30% (current requirement × (0.300/0.157)).Economic calculationsWe based our calculations on published data from 202025, which estimate that expanding the protected areas to 30% would generate higher overall output (revenues) than non-expansion (an extra US$64–454 billion per year by 2050). This figure is only an indicative, partial estimate, generated for the purposes of comparison and to illustrate the substantial return on investment that protected area staff investments imply. Using these figures and our estimates of personnel requirements to ensure effective management of 30% coverage, we calculated the range of sums that each additional protected area staff member has the potential to generate (Supplementary Table 8). For clarity, we rounded these figures to the nearest hundred US dollars in the main text.Our estimates of the gross value added per worker in forestry and agriculture (sectors responsible for similar proportions of the world as protected areas) are included to provide a point of comparison for the figures showing the economic benefit generated per protected area personnel member (see the preceding). The data for the gross annual value of world agricultural production (US$3,550,231,736,000) and the number of workers employed in agriculture (343,527,711) come from the Food and Agriculture Organization of the United Nations30, providing an average gross value of annual agricultural production per worker of US$10,335. We adjusted these 2018 data to 2020 price levels using a deflator based on the US consumer price index (CPI) from the World Economic Outlook database61 (Supplementary Table 9). This ensures that all the economic value data we present are directly comparable for protected area, agricultural and forestry workers. We calculated the gross value of forest production per worker on the basis of direct contribution of forestry of more than US$539 billion to world GDP in 201162 and total forest-sector employment of 11.881 million full-time-equivalent jobs in 201032. These were the most up-to-date global estimates we could locate from credible sources that presented comparable estimates of forest-sector employment and contribution to GDP. This gives an average gross value of forest production per worker of US$45,367 per year. We used the same method as for agriculture to bring these figures to 2020 price levels (Supplementary Table 9). These figures are rounded to the nearest hundred US dollars in the main text. More

Study areaThe present study was conducted in Uttarkashi district, Uttarakhand, located between 38° 28′ to 31°28′ N latitude and 77°49′ to 79°25′ E longitude with an area of about 8016 km2, covering primarily hilly terrain with an altitudinal range of 715–6717 m (Fig. 3). The terrain is mountainous, consisting of undulating hill ranges and narrow valleys with temperate climatic conditions. The district lies in the upper catchment of two major rivers of India, viz., the Ganges (Bhagirathi towards upstream) and the Yamuna. The major vegetation types of the study area are Himalayan moist temperate forest, sub-alpine forest and alpine scrub59. The Uttarkashi district forests are managed under three Forest Divisions viz., (i) Uttarkashi Forest Division (ii) Upper Yamuna Badkot Forest Division and (iii) Tons Forest Division) with two Protected Areas (PAs) (i) Gangotri National Park and (ii) Govind Pashu Vihar National Park. The forested habitats of the study landscape are home to top conservation priority species, including Asiatic Black bear (Ursus thibetanus), Musk deer (Moschus spp.), Common leopard (Panthera pardus), Himalayan brown bear (Ursus arctos isabellinus) and Western Tragopan (Tragopan melanocephalus), Himalayan monal (Lophophorus impejanus). The study was conducted after a study permit issued by the Chief Wildlife Warden, Forest Department, Uttarakhand government, vide letter no. 848/5-6 dated 31/08/2019, we have not handled the species for doing research. Instead, remote camera traps have been used for collecting the data with the permission of the Chief Wildlife Warden, Government of Uttarakhand. Further, informed consent was taken before interviewing the local communities. The data was collected according to the institutional guidelines and approved by the Research Advisory and Monitoring Committee of the Zoological Survey of India.Figure 3Map of the study area Uttarkashi, Uttarakhand. ArcGIS 10.6 (ESRI, Redlands, CA) was used to create the map. (Map created using ArcGIS 10.6; http://www.esri.com).Full size imageSampling protocolThe basic sampling protocol and assumptions for multi-species occupancy modelling are identical to the single-species case7. Briefly, a set of 62 intensive sites, were randomly selected, and each site i was surveyed j times. During each survey, detection/non-detection of S focal species was recorded. Additionally, direct or indirect evidences of species presence from the different areas were also recorded.Data collectionThe complete study area was divided into 10 × 10 km grids, consisting of n = 60 grids. Based on the reconnaissance survey, out of these 60 grids, we selected 25 girds that were accessible to conduct the survey and have the species presence. Further, these grids were divided into 2 × 2 km grids to maximize our effort so that all logistically accessible grids could be covered, and we conducted intensive sampling in N = 62 grids after excluding the grids with human settlements. T The field surveys were conducted during 2018–2019, and a team of researchers systematically visited selected grids to collect data on the detection/non-detection of these ungulates. A total of 62 camera traps were deployed in selected grids, and 650 km were traversed, accounting for N = 54 trails in these sampled grids. These camera traps were visited once in every fifteen days for replacing the batteries as well as documenting the presence of the species through the sign surveys. The ultra-compact SPYPOINT FORCE-11D trail camera (SPYPOINT, GG Telecom, Canada, QC) and Browning trail camera (Defender 850, 20 MP, Prometheus Group, LLC Birmingham, Alabama, https://browningtrailcameras.com) camera traps were used to detect the presence/absence of ungulate species. The cameras were mounted 40–60 cm above ground on natural trails without lures.Data explorationWhile deploying camera traps, we also noted habitat variables through on-site observation such as distance to the village and human disturbance. We tested site covariates for collinearity and discarded one of a pair if the Pearson’s correlation was greater than 0.760. Hence, we assumed each of the site covariates could influence the occupancy and detectability of these ungulates.CovariatesWe hypothesized that habitat variables may influence these ungulates’ occupancy and detection probability. A total of 21 variables were extracted either from the field or using the ArcGIS v. 10.6 software (ESRI, Redlands, CA), and only 14 were retained after collinearity testing60 (Table 3). These covariates were classified into the following categories (Topographic variables, Habitat variables and anthropogenic variables). The topographic variables (elevation, slope and aspect) were generated using 30× resolution SRTM (Shuttle Radar Topography Mission) image downloaded from EarthExplorer (https://earthexplorer.usgs.gov/). The habitat/ land cover classification was carried out using Landsat 8 satellite imagery (Spatial resolution = 30 m) downloaded from Global Land Cover Facility by following the methodology suggested by61 using the ArcGIS v. 10.6 software (ESRI, Redlands, CA). The study area was classified into nine Land use/land cover (LULC) classes viz., West Himalayan Sub-alpine birch/fir Forest (FT 188), West Himalayan upper oak/fir forest (FT 162), West Himalayan Dry juniper forest (FT 180), Ban oak forest (FT 152), Moist Deodar Forest (FT 155), Western mixed coniferous forest (FT 156), Moist temperate Deciduous Forest (FT 157) which were used for further analysis considering their importance to species ecology and behavior60. The values for all the covariates were extracted at 30 m resolution, and a single value per site was obtained by averaging all the pixel values within each sampling site (camera trap locations).Table 3 Habitat variables used for multi species occupancy analysis of three ungulate species in Uttarkashi, Uttarakhand.Full size tableOccupancy modelling frameworkWe used multi-species occupancy modelling62 of barking deer, goral and sambar to estimate the probability of the species (s) occurred within the area (i) sampled during our survey period (j), for accounting the imperfect detection of the species8. Distinguishing the true presence/absence of a species from detection/non-detection (i.e., species present and captured or species present but not captured) requires spatially or temporally replicated data. We used camera stations to record the presence/absence of species along with sign survey in all the studied grids. The camera traps were placed along the trail/transects in the studied grids hence each grid needs to be visited once in every fifteen days to check the camera traps as well as to document the presence of the studied species. Therefore, we treated 15 trap nights as one sampling occasion at a particular camera station resulting in ~ 7 sampling occasions per camera station.Our aim was to record the presence/ absence of the species at a particular gird hence we incorporated sign survey data if the species was not detected in camera station but recorded through sign survey. We pooled the presence/absence data in a single sheet of each species following6 and fitted occupancy and detectability models using programme Mark63,64. We model the species (s) presence (ysij = 1) and absence (ysij = 0) at site i during survey j, and the sampling protocol was identical to single species case65, where the Bernoulli random variable was conditional on the presence of species s (Zs = 1) following6$${text{y}}_{sij} sim {text{ Bernoulli}}left( {{text{p}}_{sij} {text{z}}_{si} } right),$$
where Psij represents the probability of detecting species S during replicate survey j at site i and Zsi = presence or absence of species s at site i.Furthermore, we model the latent occupancy state of species s at site i as a multivariate Bernoulli random variable:$${text{Z}}_{i} sim {text{MVB}}left( {uppsi _{i} } right)$$
where Zi = {Z1i, Z2i….., ZSi} is an S-dimensional vector of 1’s and 0’s denoting the latent occupancy state of all S species and (ψi) is a 2S-dimensional vector denoting the probability of all possible sequences of 1’s and 0’s Zi can attain such that ∑ ψi = 1 with corresponding probability mass function (PMF) adopted from6,64.$$fleft( {{text{Z}}_{i} } right) = {text{ exp}}left( {left( {{text{Z}}_{i} {text{log}}(uppsi_{{text{i}}} {1}/uppsi_{{text{i}}} 0} right) , + {text{ log}}left( {uppsi_{{text{i}}} 0} right)} right).$$The quantity f = log (ψi1/ψi0), is the log odds species S occupies a site often referred to as a ‘natural parameter’.Since we are modeling three ungulate species (S = 3), 2S = 23 the possible encounter histories included in the dataset were eight, if neither of the two species were detected the value of ‘00’ was assigned; similarly ‘01’ indicates detection of species 1; ‘02’ indicates detection of species 2; ‘03’ indicates detection of both the species; ‘04’ indicates detection of species 3; ‘05’ indicates detection of species 1 and species 3; ‘06’ indicates detection of species 2 and species 3 and ‘07’ indicates detection of all the three species. We modelled constant occupancy and detection probability for each of the three species. Hence, we specified 6 f and p parameters, an intercept (β) for each of one-way f parameter and detection parameter p following64.$$f_{{1}}=upbeta_{{{1},}} ;;{text{p}}=upbeta_{{4}}$$$$f_{{2}} = upbeta_{{{2},}} ;{text{p }} = , upbeta 5$$$$f_{{3}} = , upbeta_{{{3},}}; {text{p }} = , upbeta_{{6}}$$We fit a set of models including the detection probability as a constant, p(.), and variable function to occupancy ψ(covariate) for site-specific covariates and models include occupancy as constant ψ(.) and variable function of the detection p(covariates) for the respective site covariates.As we have assumed the independence among all three species, the model shows marginal occupancy probabilities of species 1, species 2 and species 3 varies as a function of environmental variables. We incorporated site-level characteristics affecting species-specific occurrence (f1: occupancy of species 1, f2: occupancy of species 2, & f3: occupancy of species 3) and detection probabilities using a generalized linear modelling approach42. This requires 9 parameters: an intercept (β1, β3, β5) and slope (β2, β4, β6) coefficient for each 1-way f parameter f1, f2, f3 and an intercept parameter for each detection parameter (β7, β8, β9). Below mentioned is the model for 1-way f parameters.$$f_{{1}} = , upbeta_{{{1 } + }} upbeta_{{2}} left( {{text{Covariate}}} right),;;{text{ p }} = , upbeta_{{7}}$$$$f_{{2}} = , upbeta_{{{3 } + }} upbeta_{{4}} left( {{text{Covariate}}} right),;;{text{ p}} = , upbeta_{{8}}$$$$f_{{3}} = , upbeta_{{5}} + , upbeta_{{6}} left( {{text{Covariate}}} right),;;{text{ p }} = , upbeta_{{9}} .$$All covariates were standardized before model fitting. We fitted the most complex model to each species and considered all possible combinations of covariates using the logit link function. Our rationale for including these variables in the occupancy and detectability component of the model was that we expected these variables to influence the occupancy and detectability of the study species.Since multi-species occupancy simultaneously model environmental variables, & interspecific interactions. Further it also allows to understand the influence of environmental variables on one species occupancy, in the presence or absence of other sympatric species64. Hence, we also modeled two species occur together as a function of covariates. We examined how the variables of each camera site influenced the pair-wise interaction of the three ungulate species. This model assumes that the conditional probability of one species varies in the presence or absence of other species. We assumed f123: co-occurrence of species 1, species 2 & species 3 = 0, hence we did not include higher-order interactions in any of our models, we assumed the conditional probability of 3 species occurred together was purely a function of species-specific (f1, f2, f3) and pair-wise interaction (f12: co-occurrence of species1 & species 2, f13: co-occurrence of species 1 & species 3, f23: co-occurrence of species 2 & species 3) parameters. We modeled pair-wise interaction of species varies as a function of environmental variables keeping detection probability constant. Hence, we specified 15 f and p parameters, an intercept and slope coefficient for each of the one-way (f1, f2, f3) and the two-way f parameters (f12, f13, and f23); as well as an intercept parameter for each of the detection models. The model equation below implies for 2-way f parameters:$$f_{{{12}}} = , upbeta_{{{7 } + }} upbeta_{{8}} left( {{text{Covariate}}} right),;;{text{ p }} = , upbeta_{{{13}}}$$$$f_{{{13}}} = , upbeta_{{{9 } + }} upbeta_{{{1}0}} left( {{text{Covariate}}} right),;;{text{ p }} = , upbeta_{{{14}}}$$$$f_{{{23}}} = , upbeta_{{{11 } + }} upbeta_{{{12}}} left( {{text{Covariate}}} right),;;{text{ p }} = , upbeta_{{{15}}} .$$We also fitted models including co-occurrence and detection probability of a species varies as a function of environmental variables. Hence, we specified 18 f and p parameters, an intercept and slope coefficient for each of one-way (f1, f2, f3) and two-way f parameters (f12, f13, f23); and an intercept as well as the slope parameters for each of the detection models. The model equation below implies for 2-way f parameters:$$f_{{{12}}} = , upbeta_{{{7 } + }} upbeta_{{8}} left( {{text{Covariate}}} right),{text{ p }} = , upbeta_{{{13 } + }} upbeta_{{{14}}} left( {{text{covariate}}} right)$$$$f_{{{13}}} = , upbeta_{{{9 } + }} upbeta_{{{1}0}} left( {{text{Covariate}}} right),{text{ p }} = , upbeta_{{{15}}} + , upbeta_{{{16}}} left( {{text{covariate}}} right)$$$$f_{{{23}}} = , upbeta_{{{11 } + }} upbeta_{{{12}}} left( {{text{Covariate}}} right),{text{ p }} = , upbeta_{{{17}}} + , upbeta_{{{18}}} left( {{text{covariate}}} right)$$A total of 38 models were run to test the influence of environmental variables on occupancy and detection probability of species-specific (f1, f2, f3) and pair-wise interaction of the three ungulate species. The best-supported model was identified by selecting the model with the lowest AICc value and highest model weights66, where higher model weights indicate a better fit of the model to the data. Second-Order Information Criterion (AICc)67 values were used to rank the occupancy models, and all the models whose ΔAICc  More

We hypothesized that functionally distinct forest types can be mapped at moderate spatial resolutions, using a combination of canopy foliar traits and canopy structure information. Our analysis of LiDAR and imaging spectroscopy data at spatial resolutions ranging from 4 to 200 m (16 m2–40,000 m2), with an emphasis on the 30 m (900 m2) spaceborne hyperspectral spatial resolution, reveals that few remotely sensed canopy properties are needed to successfully identify ecologically distinct forest types at two diverse tropical forest sites in Malaysian Borneo. In testing our second hypothesis that mapped forest types exhibit distinct ecosystem function, we found that forest types identified using remotely sensed leaf P, LMA, Max H, and canopy cover at 20 m height (Cover20) closely align with forest types defined from field-based floristic surveys29,30,31,32,33 and inventory plot-based measurements of growth and mortality rates (Fig. 4b). Our approach, however, enables mapping of their entire spatial extent (Fig. 1) and reveals important structural and functional variation within areas characterized as a single forest type in previous studies (Fig. 3). Current and forthcoming satellite hyperspectral platforms, including PRISMA (30 m), CHIME (20–30 m), and SBG (30 m), have or will have comparable spectral resolution, higher temporal revisits, and much greater geographic coverage. The ability to conduct this type of analysis using remote sensing measurements at 30 m resolution suggests that our method can be applied to these emerging spaceborne imaging spectroscopy data to reveal important differences in structure and function across the world’s tropical forests.Nested functional forest types revealedTo test our first hypothesis, rather than making an a priori decision about the number of k-means clusters (k), we explored the capacity of remotely sensed data to reveal ecologically relevant variation in forest types. Baldeck and Asner took a similar unsupervised approach to estimating beta diversity in South Africa34. Because the choice of k directly influences analysis outcomes, careful selection of k is required. Different approaches for identifying the number of clusters, using the Gapk and Wk elbow metrics35, yielded varying optimal numbers of clusters for the Sepilok and Danum landscapes (Fig. 1, Supplementary Figs. 4 and 5). However, at both sites, a comparison of results based on different values of k revealed ecologically meaningful structural and functional differences and graduated transitions between forest types (Fig. 2, Supplementary Figs. 7 and 8), indicating that the exploration of traits that aggregate or separate forest types as k changes is a valuable exercise. Overlap between the remotely sensed forest type boundaries and inventory plots within distinct forest types indicate that the series of clustered forests align closely with forest types defined based on in situ data on species composition and ecosystem structure. In part, this type of analysis requires careful selection of the number of clusters. Additionally, however, we gained valuable insights via the exploration of varying numbers of clusters as it relates to biologically meaningful categorization of forest types. Extending this method to other parts of the tropics will require similar decision-making, which will either require user input, or the development of robust automated algorithms for selecting k.Forest types capture differences in ecosystem dynamicsWe further evaluated the canopy traits and structural attributes that were most critical for mapping distinct forest types, hypothesizing that mapped forest types exhibit distinct ecosystem function. Forest types revealed by the cluster analyses were distributed along the leaf economic spectrum, where the leaf economic spectrum characterizes a tradeoff in plant growth strategies36. LMA, which can covary strongly with leaf N and P, is a key indicator of plant growth strategies along the spectrum37. At the slow-return end of the leaf economics spectrum, plants in nutrient-poor conditions with low leaf nutrient concentrations invest in leaf structure and defense, expressed as high LMA, strategizing longer-lived, tougher leaves with slower decomposition rates. This strategy comes at the cost of slower growth. At the quick-return end of the spectrum, plants in nutrient-rich environments with higher leaf nutrient concentrations invest less in structure and defense, enabling faster growth and more rapid leaf turnover, i.e., shorter leaf lifespans. This quick-return growth strategy supports higher photosynthetic rates and more rapid carbon gain36.In this study, the principal components and clustering results yielded forest types that are indicative of community level differences associated with leaf economic spectrum differences. The nutrient rich sites (Danum1 and Danum2, Supplementary Fig. 8) show high canopy N and P and low LMA compared to the nutrient poor and acidic sites (Sandstone and Kerangas), which contributes to lower leaf photosynthetic capacity (Vcmax) and growth (Fig. 4b). Foliar N:P also increased with site fertility, confirming that tropical forests are primarily limited by phosphorus, and not nitrogen38,39, with large implications for carbon sequestration in these forests. Orthogonal differences in canopy structure and architecture between Danum forest types and Sepilok Sandstone and Alluvial forests could be indicative of ecosystem scale differences in the sensitivity of these forests to endogenous disturbance processes40.The significant differences in aboveground carbon stocks and growth and mortality rates between forest types further suggests strong differences in ecosystem dynamics. In general, growth rates varied inversely to aboveground carbon, and higher aboveground carbon corresponded to lower mortality rates. As an example, the Sepilok sandstone forests, which are largely comprised of slow-growing dipterocarp species29,33, had the highest median aboveground carbon (236 Mg C ha−1), with higher canopy P and N, and lower LMA. The taller canopy and low canopy leaf nutrient concentrations are consistent with the low growth and mortality rates found in the sandstone forest, indicating a slow-growth strategy yielding larger trees and higher aboveground carbon stocks. In contrast, alluvial forests exhibit high turnover with mortality and growth rates higher relative to Sandstone forests corresponding to lower aboveground carbon on average. Kerangas forests exhibited low aboveground carbon despite an intermediate plot-level growth rate, and mortality rates that were significantly lower than the Danum or alluvial forest types. Kerangas forests, which were characterized by the highest LMA, lowest foliar P and N (Fig. 2a), and the lowest plot-level aboveground carbon density (186 Mg C ha−1; Fig. 4a), are known to have higher stem densities, lower canopy heights, and long-lived leaves5,32,41, suggesting well-developed strategies for nutrient retention42. Interestingly, despite significantly different aboveground carbon and demography, the kerangas and sandstone forests did not differ in LAI or canopy architecture (P:H); although maximum height, Cover20, and Hpeak LAI were significantly higher in the sandstone forest, highlighting the need to account for differences beyond LAI when scaling processes from leaves to ecosystems.In addition, when three forest types were distinguished at Sepilok, the alluvial inventory plot had significantly higher aboveground carbon than the remote sensing-derived alluvial forest extent (Fig. 4a, p  More