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

Mistakes in a landmark paper that reported the first orangutan genomes might have implications for breeding programmes.Credit: Fiona Rogers/Nature Picture Library

Susie the Sumatran orangutan was a genetic pioneer — the first of her species to have her genome fully sequenced. Her genetic library, and that of ten other orangutans, appeared in a landmark paper in Nature in 20111 that has underpinned hundreds of subsequent studies.But in August, researchers revealed that eight of the sequences in this paper had mistakenly been assigned to the wrong orangutans2. Nature issued a correction from the authors of the original paper3.The scale of the errors sparked ire on social media, and some scientists have warned that the mistakes could have repercussions for orangutan breeding programmes. “Well that’s a bit of a f&£k up orang-utan genome researchers — only mildly embarrassing guys and girls”, tweeted Michael Sweet, a molecular ecologist at the University of Derby, UK.
Africa: sequence 100,000 species to safeguard biodiversity
It’s not clear how these swapped identities have affected orangutan research. But researchers involved in the new analysis believe the discovery might highlight how issues in the scientific community — including the pressure to publish and a reliance on peer review to catch mistakes — could allow such errors to slip into the scientific record.“I think there are errors like this in many, many published papers,” says Graham Banes, an evolutionary biologist formerly at the University of Wisconsin–Madison who led the reanalysis of the 2011 paper. “In some ways, we’re lucky that this was just orangutans. What if this was a biomedical paper and people were developing therapies based on published data?”“It’s fairly easy for these things to occur,” adds Robert Fulton, a genomic scientist at Washington University School of Medicine in St Louis, Missouri, who was part of the team behind the original paper and is a co-author on the reanalysis. “What’s important is that that the data are now correct.” Devin Locke, who led the preparation of the 2011 paper and was formerly a colleague of Fulton’s at Washington University, did not respond to questions about the work.Hybrid headacheDetailed ‘reference’ genomes, such as those published in the 2011 Nature paper, are a key tool for biologists. In 2017, Banes and his team were using the genomes to study what happens when different species of orangutan interbreed, a process called hybridization.They noticed that the names given to some of the samples didn’t match the animals’ reported sex. For example, the 2011 paper reported that an orangutan named Dolly was male. But according to the orangutan studbook — a record of orangutans living in zoos — Dolly was female. Even stranger, Banes found that some of the genomes marked as male lacked a Y chromosome. “There was just this series of things that didn’t make sense,” he recalls.
Major wildlife report struggles to tally humanity’s exploitation of species
Banes and his colleagues eventually found that the 2011 paper had misidentified all but two of the orangutan genomes. Some mistakes seem to be the result of typos. In one case, a sample from a male orangutan was given an ID number that actually corresponded to a sample from an African pig in a tissue repository. Other samples seem to have had their identities swapped during laboratory work. The 2011 study helped to pin down when Bornean and Sumatran orangutans split into separate species, and compared their genomes with those of other primates. These conclusions are largely uncompromised by the mix-up. But Banes says that the errors could have implications for other research, including his own.Banes uses genetic data to provide zoos with recommendations about their captive breeding programmes. Zoos try to avoid crossbreeding orangutan species, partly to mimic wild populations and also because hybrids can suffer high rates of miscarriage and birth defects, says Banes. While re-examining the samples from the 2011 paper, the team realized that one of the sequences thought to be Sumatran (Pongo abelii) was actually Tapanuli (Pongo tapanuliensis), a third species of orangutan that was only described in 20174.Unfortunately, the 2011 paper had wrongly assigned the Tapanuli genome to Baldy, a male orangutan, rather than its actual owner, a female orangutan named Bubbles (both are now dead). Banes says that his team came “perilously close” to announcing in a paper that Baldy was Tapanuli.Although Baldy has no living descendants, Bubbles has several offspring at zoos around the world, all of which are Sumatran–Tapanuli hybrids. Zookeepers will now have to decide whether to stop breeding Bubbles’ descendants to avoid further hybridization, says Vincent Nijman, an anthropologist at Oxford Brookes University, UK.‘Bigger concerns’However, Nijman also argues that the errors will have little effect on orangutan conservation as a whole. Zoos often bill their animals as a back-up for endangered species, but conservationists are much more focused on the thousands of orangutans in the wild that are threatened by deforestation. “I think we have bigger concerns than some mixed-up samples,” says Erik Meijaard, a conservation scientist at Borneo Futures, a conservation consultancy company based in Bandar Seri Begawan, Brunei.
Scientists warn deal to save biodiversity is in jeopardy
Michael Krützen, an evolutionary geneticist at the University of Zurich in Switzerland, agrees that although the errors are “annoying”, their impact on downstream research is probably minimal. However, he says that the problems might be an example of how academia’s publish-or-perish environment could lead to “sloppy” work, as researchers race to publish their work in high-tier journals.Banes agrees that this kind of pressure — along with an over-reliance on a peer-review system that does not offer its volunteer reviewers tangible financial or professional benefits — could lead to errors slipping into published manuscripts.A spokesperson for Nature declined to comment on why the errors in the 2011 paper were not caught by peer review, citing concerns about confidentiality. (Nature’s news team is editorially independent of its academic publishing operation). “However, we would like to stress that we take our responsibility to maintain the accuracy of the scientific record very seriously,” they wrote in an e-mail. “If issues are raised about any paper we have published, we will look into them carefully and update the literature where appropriate.”Banes says that it’s important not to blame individual scientists for such errors, not least because it could discourage efforts to correct mistakes in future. “I think any scientist could have made these mistakes,” he says. “But if we all jump out and say, ‘oh my god, how could they have been so stupid?’, no one is ever going to correct anything. That shame is detrimental to science.” More

Sequencing and assembly of the H. schachtii genomeWe measured (Supplemental Fig. 1), sequenced (BioProject PRJNA722882), and assembled the genome of H. schachtii (population Bonn) using a combination of flow cytometry, Pacific Biosciences sequencing, and Illumina sequencing. H. schachtii has the largest genome (160–170 Mb) of any cyst nematode measured/sequenced to date (Supplementary Table 1). It was sequenced to 192-fold coverage using Pacific Biosciences sequencing (fragment n50 of 16 kb), and 144-fold coverage using Illumina sequencing (150 bp Paired-end reads). The final, polished, contamination-free (Supplemental Fig. 2), assembly (v1.2) included ~179 Mbp contained within 395 scaffolds: 90% of the sequence is contained on scaffolds longer than 281,463 bp (n = 154). The assembly is a largely complete haploid representation of the diploid genome, as evidenced by core eukaryotic genes being largely present, complete and single copy (CEGMA 93.15% complete with an average of 1.12 copies each, and BUSCO (Eukaryota odb9) 79% complete with 8.2% duplicated—Supplementary Table 2). Over three million variants were phased into haplotypes (2029 blocks, N50 239.5 kb, covering 94.7% of the reference) which can be used to predict true protein variants (Supplementary data 1), and 601 larger structural variants were identified (Supplementary data 2).The trans-kingdom, lifestage-specific, transcriptomes of H. schachtii and A. thaliana provide a holistic view of parasitismWe devised a sampling procedure to cover all major life stages/transitions of the parasitic life cycle to generate a simultaneous, chronological, and comprehensive picture of nematode gene expression, and infection-site-specific plant gene expression patterns. We sampled cysts and pre-infective second-stage juveniles (J2s), as well as infected segments of A. thaliana root and uninfected adjacent control segments of root at 10 hours post infection (hpi – migratory J2s, pre-establishment of the feeding site), 48 hpi (post establishment of the feeding site), 12 days post infection females (dpi – virgin), 12 dpi males (differentiated, pre-emergence, most if not all stopped feeding), and 24 dpi females (post mating), each in biological triplicate (Fig. 1A). We generated approximately nine billion pairs of 150 bp strand-specific RNAseq reads (Supplementary data 3) covering each stage in biological triplicate (for the parasite and the host): in the early stages of infection we generated over 400 million reads per replicate, to provide sufficient coverage of each kingdom.Fig. 1: Trans-kingdom, lifestage-specific, transcriptome of H. schachtii and A. thaliana.A Schematic representation of the life cycle of H. schachtii infecting A. thaliana, highlighting the 7 stages sampled in this study. For each stage, the average number of trimmed RNAseq read pairs per replicate is shown, with the proportion of reads mapping to either parasite or host in parentheses. B Principle components 1 and 2 for H. schachtii and A. thaliana expression data are plotted. Arrows indicate progression through the life cycle/real-time. Hours post infection (hpi), days post infection (dpi).Full size imageStrand-specific RNAseq reads originating from host and parasite were deconvoluted by mapping to their respective genome assemblies (H. schachtii v.1.2 and TAIR10). For the parasite, ~500 million Illumina RNAseq read pairs uniquely mapping to the H. schachtii genome were used to generate a set of 26,739 gene annotations (32,624 transcripts – detailed further in the next section), ~77% of which have good evidence of transcription in at least one lifestage (≥10 reads in at least one rep). Similarly for the host, ~2.8 billion Illumina RNAseq read pairs uniquely mapping to the A. thaliana genome show that ~77% of the 32,548 gene models have good evidence of transcription in at least one stage (≥10 reads in at least one rep, even though we only sampled roots). A principal component analysis of the host and parasite gene expression data offers several insights into the parasitic process. Principle component 1 (60% of the variance) and 2 (19% of the variance) of the parasite recapitulate the life cycle in PCA space (Fig. 1B). The 12 dpi female transcriptome is more similar to the 24 dpi female transcriptome than to the 12 dpi male transcriptome. Principle components 1 (75% of the variance) and 2 (10% of the variance) of the host show that the greatest difference between infected and uninfected plant tissue is at the early time points (10 hpi), and that the transcriptomes of infected and uninfected plant material converge over time, possibly due to systemic effects of infection. A 12 dpi male syncytium transcriptome is roughly intermediate between a control root transcriptome and a 12 dpi female syncytium transcriptome. Given that at this stage most if not all of the males will have ceased feeding, this could be due to inadequate formation of the feeding site, or regression of the tissue. In any case, by comparing both principal component analyses, we can see that what is a relatively small difference in the transcriptomes of the feeding sites of males and females is amplified to a relatively large difference in the transcriptomes of the males and females themselves (Fig. 1B).The consequences, and possible causes, of large-scale segmental duplication in the Heterodera lineageTo understand the evolutionary origin(s) of the relatively large number of genes in H. schachtii in particular, and Heterodera spp. in general, we analysed the abundance and categories of gene duplication in the predicted exome. Compared to a related cyst nematode, Globodera pallida (derived using comparable methodology and of comparable contiguity) the exomes of H. schachtii and H. glycines are characterised by a relatively smaller proportion of single-copy genes (as classified by MCSanX toolkit17, and a relatively greater proportion of segmental duplications (at least five co-linear genes with no >25 genes between them), with relatively similar proportions of dispersed duplications (two similar genes with >20 other genes between them), proximal duplications (two similar genes with  +0.5 or  More