More stories

  • in

    Substantial oxygen consumption by aerobic nitrite oxidation in oceanic oxygen minimum zones

    Nitrite oxidation rates in the ETNPWe sampled six stations in the ETNP OMZ with DO concentrations 1 µM at 100 m (Fig. 2A). Chlorophyll concentrations were also high in the upper water column (up to 5 mg m−3 at 20 m), with an SCM spanning 70–125 m (Fig. 2B). Nitrite oxidation displayed a local maximum at the base of the EZ at Station 1 (20–30 m), and then increased to higher levels ( >100 nmol L−1 day−1; Fig. 2C). This increase at 100 and 125 m corresponded with the overlap between the bottom of the SCM and the top of the SNM. Nitrite oxidation rates then reached higher values at 150 m within the SNM at Station 1. Stations 2 and 3 displayed similar nitrite oxidation rate profiles to each other, including elevated rates in the SCM (Fig. 2G, K). Nitrite oxidation rates were similar in magnitude, and peak values at the base of the EZ and in the OMZ were also similar (69–96 nmol L−1 day−1). Depth patterns tracked oceanographic differences across the three AMZ stations, as the depth of all features increased moving offshore from Stations 1 to 2 to 3. For example, the SCM extended from 105 to 155 m at Station 2, while nitrite concentrations began to increase below 100 m; nitrite oxidation rates were elevated at 140 m and declined slightly with increasing depth (Fig. 2E–G). At Station 3, the SCM (120–180 m) and SNM ( >140 m) depths were deeper, and nitrite oxidation rates increased from 100 to 200 m (Fig. 2I–K).Fig. 2: Biogeochemical depth profiles.Profiles of A, E, I dissolved oxygen (solid lines) and nitrite (data points connected by dashed lines), B, F, J chlorophyll a, C, G, K nitrite oxidation rates, and D, H, L oxygen consumption rates (OCR; data presented as mean values of five independent replicates ±1 SD) show consistent variation across A–D Station 1, E–H Station 2, and I–L Station 3 (denoted by different colors). Black horizontal lines denote the depth of the oxygen minimum zone (OMZ), and shaded areas show the secondary chlorophyll maximum (SCM) at each station. Rates measured below the SCM should be considered potential rates (see main text). Maximum chlorophyll values at Station 1 plot off-axis.Full size imageIn contrast to these three AMZ stations (Stations 1–3), rate profiles at Stations 4–6 showed peaks at the base of the EZ followed by decreases with depth and lacked a pronounced rate increase within the OMZ (Supplementary Fig. 1). Parallel measurements of ammonia oxidation rates also showed this type of pattern at all stations (Supplementary Fig. 1). Subsurface maxima in ammonia oxidation tracked variations in the EZ across all six stations, but rates were not elevated in OMZ/AMZ waters—again contrasting with nitrite oxidation rate profiles at the AMZ stations. These data accord with earlier work in OMZs showing contrasting ammonia and nitrite oxidation rate profiles, and particularly high rates of nitrite oxidation in OMZ waters6,7,8,29,30,31.Initial DO concentrations for these measurements closely matched in situ values above the SCM (where DO concentrations are higher), and starting DO ranged from 260–1500 nM for measurements in and below the SCM. These DO concentrations are generally lower than those used for previous nitrite oxidation rate measurements in OMZs6,9, but similar to work examining the oxygen affinity of nitrite oxidation22 and overall oxygen consumption16,19. Elevated nitrite oxidation in the limited number of samples (n = 5) collected below the SCM ( >125 m at Station 1, >155 m at Station 2, and >180 m at Station 3)—where little to no DO is typically available—should be considered potential rates and could have a number of possible explanations discussed below. Within the SCM, our data support the idea that nitrite oxidation contributes to ‘cryptic’ oxygen cycling15—i.e., that DO produced via oxygenic photosynthesis is rapidly consumed.Oxygen consumption via nitrite oxidationWe determined the contribution of nitrite oxidation to overall oxygen consumption via parallel measurements of OCRs using in situ optical sensor spots—which are noninvasive, provide nearly identical results as other low-level measurement approaches32, are the only effective means of achieving substantial replication, and for which sensitivity increases as DO decreases32,33. Decreases in DO were measured in both nitrite and ammonia oxidation rate sample bottles, as well as in three additional replicates, to leverage statistical power for increased sensitivity to low-level DO consumption (see “Methods”). Water column OCR profiles at all stations showed exponential declines with depth and decreasing DO concentrations (Fig. 2D, H, L and Supplementary Fig. 1). Rates were highest in the upper water column and declined sharply within the upper portion of the OMZ above the SCM. The majority of measurements within the SCM—where DO may be produced via photosynthesis—were 100 s of nmol L−1 day−1, with an overall range of 160–1380 nmol L−1 day−1. Below the SCM, DO would be available more rarely (e.g., ref. 16), and OCR measurements represent potential rates should oxygen be supplied; OCR ranged from 120 to 390 nmol L−1 day−1. OCR also tracked variations in DO across stations, with progressively steeper declines in OCR with depth from Station 6 to Station 1.These OCR results are similar to the limited previous measurements that have been conducted in OMZ regions, with some key differences. In particular, they are consistent with previous measurements of rapid DO consumption in the SCM, with OCR rates ranging from 482 to 1520 nmol-O2 L−1 day−1 in the ETSP, and from 55 to 418 nmol-O2  L−1 day−1 in the ETNP15. Earlier OCR measurements conducted in the ETNP near Stations 1 and 3 (across a wide range of DO values) likewise ranged from 420 to 828 nmol L−1 day−1 in the SCM near Station 1, and from 101 to 269 nmol L−1 day−1 in the SCM near Station 3 (ref. 16). Above the SCM, previous OCR measurements in the ETNP spanned 2260 to 662 nmol L−1 day−1 from the EZ to the edge of the OMZ; these values are lower than our measurements at 44 and 67 m depth at Station 2, but in line with our remaining measurements above the SCM. OCR reached 1610 nmol L−1 day−1 in the SCM in Namibian shelf waters and 200–400 nmol L−1 day−1 in the SCM off Peru18. Kalvelage et al.18 furthermore observed sharp decreases with depth in the ETSP, with rates declining from >1000 nmol L−1 day−1 above the SCM.This pattern of declining OCR with increasing depth and decreasing DO was also evident in our dataset and contrasted with that of nitrite oxidation rates, which were notably elevated in the SCM at the AMZ stations (Fig. 2). We directly compared nitrite oxidation rates with OCR, assuming that each mole of nitrite is oxidized using ½ mole of O2 (ref. 5). We found that nitrite oxidation systematically increased as a proportion of overall OCR at lower DO levels (Fig. 3A, B). Nitrite oxidation was responsible for up to 69% of OCR at Station 1, although most values were closer to 10–40% at Stations 2 and 3 (Fig. 3A, B). In contrast, ammonia oxidation contributed 100 s of nM represent potential rates. For OMZ edge samples, OCR values in the µM range were higher than observed in profiles—most likely due to the effects of bubbling19, which could physically break down the organic matter present in higher concentrations at these depths (Table 1). Throughout all experiments, rate magnitudes in the 100 s of nM DO concentration range (11–820 nmol L−1 day−1) were similar to profile measurements (Fig. 2), as well as to previous measurements in OMZs15,16,18,19 (see above).DO concentrations were also continuously monitored in a subset of experimental bottles, and DO consumption was consistently linear (see “Methods”). The few exceptions occurred in several experiments conducted at DO concentrations More

  • in

    Temperature-dependent modelling and spatial prediction reveal suitable geographical areas for deployment of two Metarhizium anisopliae isolates for Tuta absoluta management

    1.Sibomana, M. S., Workneh, T. S. & Audain, K. A review of postharvest handling and losses in the fresh tomato supply chain: A focus on Sub-Saharan Africa. Food Secur. 8, 389–404 (2016).
    Google Scholar 
    2.Ochilo, W. N. et al. Characteristics and production constraints of smallholder tomato production in Kenya. Sci. Afr. 2, e00014 (2019).
    Google Scholar 
    3.Pratt, C. F., Constantine, K. L. & Murphy, S. T. Economic impacts of invasive alien species on African smallholder livelihoods. Glob. Food Sec. 14, 31–37 (2017).
    Google Scholar 
    4.Aigbedion-Atalor, P. O. et al. The South America tomato leafminer, Tuta absoluta (Lepidoptera: Gelechiidae), spreads its wings in Eastern Africa: distribution and socioeconomic impacts. J. Econ. Entomol. 112, 2797–2807 (2019).PubMed 

    Google Scholar 
    5.Brévault, T., Sylla, S., Diatte, M., Bernadas, G. & Diarra, K. Tuta absoluta Meyrick (Lepidoptera: Gelechiidae): A new threat to tomato production in sub-Saharan Africa. African Entomol. 22, 441–444 (2014).
    Google Scholar 
    6.Guedes, R. N. C. & Picanço, M. C. The tomato borer Tuta absoluta in South America: Pest status, management and insecticide resistance. EPPO Bull. 42, 211–216 (2012).
    Google Scholar 
    7.Desneux, N. et al. Biological invasion of European tomato crops by Tuta absoluta: ecology, geographic expansion and prospects for biological control. J. Pest Sci. 83, 197–215 (2010).
    Google Scholar 
    8.Abbes, K., Harbi, A. & Chermiti, B. The tomato leafminer Tuta absoluta (Meyrick) in Tunisia: Current status and management strategies. EPPO Bull. 42, 226–233 (2012).
    Google Scholar 
    9.Biondi, A., Guedes, R. N. C., Wan, F.-H. & Desneux, N. Ecology, worldwide spread, and management of the invasive South American tomato pinworm, Tuta absoluta: past, present, and future. Annu. Rev. Entomol. 63, 239–258 (2018).PubMed 
    CAS 

    Google Scholar 
    10.Niassy, S., Ekesi, S., Migiro, L. & Otieno, W. Sustainable management of invasive pests in Africa. (Springer International Publishing, 2020). https://doi.org/10.1007/978-3-030-41083-4.11.Mansour, R. et al. Occurrence, biology, natural enemies and management of Tuta absoluta in Africa. Entomol. Gen. 38, 83–112 (2018).
    Google Scholar 
    12.Lietti, M. M. M., Botto, E. & Alzogaray, R. A. Insecticide resistance in Argentine populations of Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae). Neotrop. Entomol. 34, 113–119 (2005).
    Google Scholar 
    13.Guedes, R. N. C. C. et al. Insecticide resistance in the tomato pinworm Tuta absoluta: patterns, spread, mechanisms, management and outlook. J. Pest Sci. 92, 1329–1342 (2019).
    Google Scholar 
    14.APRD. Arthropod Pesticide Resistance Database, Michigan State University. https://www.pesticideresistance.org/display.php?pa.15.Lichtenberg, E. & Zimmerman, R. Adverse health experiences, environmental attitudes, and pesticide usage behavior of farm operators. Risk Anal. 19, 283–294 (1999).PubMed 
    CAS 

    Google Scholar 
    16.Soares, M. A. et al. Botanical insecticide and natural enemies: a potential combination for pest management against Tuta absoluta. J. Pest Sci. 92, 1433–1443 (2019).
    Google Scholar 
    17.Aigbedion-Atalor, P. O. et al. Host stage preference and performance of Dolichogenidea gelechiidivoris (Hymenoptera: Braconidae), a candidate for classical biological control of Tuta absoluta in Africa. Biol. Control 144, 1–8 (2020).
    Google Scholar 
    18.Akutse, K. S., Subramanian, S., Khamis, F. M., Ekesi, S. & Mohamed, S. A. Entomopathogenic fungus isolates for adult Tuta absoluta (Lepidoptera: Gelechiidae) management and their compatibility with Tuta pheromone. J. Appl. Entomol. 1–11 (2020). https://doi.org/10.1111/jen.12812.19.Agbessenou, A. et al. Endophytic fungi protect tomato and nightshade plants against Tuta absoluta (Lepidoptera: Gelechiidae) through a hidden friendship and cryptic battle. Sci. Rep. 10, 22195 (2020).ADS 
    PubMed 
    PubMed Central 
    CAS 

    Google Scholar 
    20.Lacey, L. A. et al. Insect pathogens as biological control agents: Back to the future. J. Invertebr. Pathol. 132, 1–41 (2015).PubMed 
    CAS 

    Google Scholar 
    21.Chandler, D. et al. The development, regulation and use of biopesticides for integrated pest management. Philos. Trans. R. Soc. B Biol. Sci. 366, 1987–1998 (2011).22.Shang, Y., Feng, P. & Wang, C. Fungi that infect insects: Altering host behavior and beyond. PLoS Pathog. 11, 1–6 (2015).
    Google Scholar 
    23.Bayissa, W. et al. Selection of fungal isolates for virulence against three aphid pest species of crucifers and okra. J. Pest Sci. 90, 355–368 (2017).
    Google Scholar 
    24.Jackson, M. A., Dunlap, C. A. & Jaronski, S. T. Ecological considerations in producing and formulating fungal entomopathogens for use in insect biocontrol. Biocontrol 55, 129–145 (2010).
    Google Scholar 
    25.Fang, W., Azimzadeh, P. & St. Leger, R. J. Strain improvement of fungal insecticides for controlling insect pests and vector-borne diseases. Curr. Opin. Microbiol. 15, 232–238 (2012).26.Tumuhaise, V. et al. Temperature-dependent growth and virulence, and mass production potential of two candidate isolates of Metarhizium anisopliae (Metschnikoff) Sorokin for managing Maruca vitrata Fabricius (Lepidoptera: Crambidae) on cowpea. African Entomol. 26, 73–83 (2018).
    Google Scholar 
    27.Onsongo, S. K., Gichimu, B. M., Akutse, K. S., Dubois, T. & Mohamed, S. A. Performance of three isolates of Metarhizium anisopliae and their virulence against Zeugodacus cucurbitae under different temperature regimes, with global extrapolation of their efficiency. Insects 10, 1–13 (2019).
    Google Scholar 
    28.Dimbi, S., Maniania, N. K., Lux, S. A. & Mueke, J. M. Effect of constant temperatures on germination, radial growth and virulence of Metarhizium anisopliae to three species of African tephritid fruit flies. Biocontrol 49, 83–94 (2004).
    Google Scholar 
    29.Ekesi, S., Maniania, N. K. & Lux, S. A. Effect of soil temperature and moisture on survival and infectivity of Metarhizium anisopliae to four tephritid fruit fly puparia. J. Invertebr. Pathol. 83, 157–167 (2003).PubMed 
    CAS 

    Google Scholar 
    30.Jaronski, S. T. Ecological factors in the inundative use of fungal entomopathogens. Biocontrol 55, 159–185 (2010).
    Google Scholar 
    31.Klass, J. I., Blanford, S. & Thomas, M. B. Development of a model for evaluating the effects of environmental temperature and thermal behaviour on biological control of locusts and grasshoppers using pathogens. Agric. For. Entomol. 9, 189–199 (2007).
    Google Scholar 
    32.Klass, J. I., Blanford, S. & Thomas, M. B. Use of a geographic information system to explore spatial variation in pathogen virulence and the implications for biological control of locusts and grasshoppers. Agric. For. Entomol. 9, 201–208 (2007).
    Google Scholar 
    33.Allen, C. & Mehler, D. M. A. Open science challenges, benefits and tips in early career and beyond. Plos Biol. 17, 1–14 (2019).
    Google Scholar 
    34.McCammon, S. A. & Rath, A. C. Separation of Metarhizium anisopliae strains by temperature dependent germination rates. Mycol. Res. 98, 1253–1257 (1994).
    Google Scholar 
    35.Ekesi, S., Maniania, N. K. & Ampong-Nyarko, K. Effect of temperature on germination, radial growth and virulence of Metarhizium anisopliae and Beauveria bassiana on Megalurothrips sjostedti. Biocontrol Sci. Technol. 9, 177–185 (1999).
    Google Scholar 
    36.De Croos, J. N. A. & Bidochka, M. J. Effects of low temperature on growth parameters in the entomopathogenic fungus Metarhizium anisopliae. Can. J. Microbiol. 45, 1055–1061 (1999).
    Google Scholar 
    37.Dahlberg, K. R. & Etten, J. L. V. Physiology and biochemistry of fungal sporulation. Annu. Rev. Phytopathol. 20, 281–301 (1982).CAS 

    Google Scholar 
    38.Hywel-Jones, N. L. & Gillespie, A. T. Effect of temperature on spore germination in Metarhizium anisopliae and Beauveria bassiana. Mycol. Res. 94, 389–392 (1990).
    Google Scholar 
    39.Acheampong, M. A., Coombes, C. A., Moore, S. D. & Hill, M. P. Temperature tolerance and humidity requirements of select entomopathogenic fungal isolates for future use in citrus IPM programmes. J. Invertebr. Pathol. 174, 107436 (2020).40.Allen, P. J. Metabolic aspects of spores germination in fungi. Annu. Rev. Phytopathol. 3, 313–342 (1965).CAS 

    Google Scholar 
    41.de Campos, M. R. et al. Thermal biology of Tuta absoluta: demographic parameters and facultative diapause. J. Pest Sci. (2004). (2020) doi:https://doi.org/10.1007/s10340-020-01286-8.42.Vidal, C., Fargues, J. & Lacey, L. A. Intraspecific variability of Paecilomyces fumosoroseus: Effect of temperature on vegetative growth. J. Invertebr. Pathol. 70, 18–26 (1997).
    Google Scholar 
    43.Smits, N., Brière, J.-F. & Fargues, J. Comparison of non-linear temperature-dependent development rate models applied to in vitro growth of entomopathogenic fungi. Mycol. Res. 107, 1476–1484 (2003).PubMed 

    Google Scholar 
    44.Cabanillas, H. E. & Jones, W. A. Effects of temperature and culture media on vegetative growth of an entomopathogenic fungus Isaria sp. (Hypocreales: Clavicipitaceae) naturally affecting the whitefly, Bemisia tabaci in Texas. Mycopathologia 167, 263–271 (2009).45.Guimapi, R. A. et al. Decision support system for fitting and mapping nonlinear functions with application to insect pest management in the biological control context. Algorithms 13, 1–21 (2020).
    Google Scholar 
    46.Smith, J. D. et al. Host range of the invasive tomato pest Tuta absoluta Meyrick (Lepidoptera: Gelechiidae) on solanaceous crops and Weeds in Tanzania. Florida Entomol. 101, 573–579 (2018).
    Google Scholar 
    47.Tumuhaise, V., Khamis, F. M., Agona, A., Sseruwu, G. & Mohamed, S. A. First record of Tuta absoluta (Lepidoptera: Gelechiidae) in Uganda. Int. J. Trop. Insect Sci. 36, 135–139 (2016).
    Google Scholar 
    48.Kassa, A., Brownbridge, M., Parker, B. L. & Skinner, M. Whey for mass production of Beauveria bassiana and Metarhizium anisopliae. Mycol. Res. 112, 583–591 (2008).PubMed 

    Google Scholar 
    49.Jenkins, N. E., Heviefo, G., Langewald, J., Cherry, A. J. & Lomer, C. J. Development of mass production technology for aerial conidia for use as mycopesticides. Biocontrol News Inf. 19, 21–32 (1998).
    Google Scholar 
    50.Barra, P., Barros, G., Etcheverry, M. & Nesci, A. Mass production studies in solid substrates with the entomopathogenic fungus, Purpureocillium lilacinum. Int. J. Adv. Agric. Res. 6, 78–84 (2018).
    Google Scholar 
    51.Jackson, M. A. Optimizing nutritional conditions for the liquid culture production of effective fungal biological control agents. J. Ind. Microbiol. Biotechnol. 19, 180–187 (1997).CAS 

    Google Scholar 
    52.Goettel, M. S. & Inglis, D. G. Fungi: Hyphomycetes. Manual of Techniques in Insect Pathology https://doi.org/10.1016/B978-012432555-5/50013-0 (1997).Article 

    Google Scholar 
    53.Fargues, J., Maniania, N., Delmas, J. & Smits, N. Influence de la température sur la croissance in vitro d’hyphomycètes entomopathogènes. Agronomie 12, 557–564 (1992).
    Google Scholar 
    54.Santana, P. A., Kumar, L., Da Silva, R. S. & Picanço, M. C. Global geographic distribution of Tuta absoluta as affected by climate change. J. Pest Sci. 92, 1373–1385 (2018).
    Google Scholar 
    55.Migiro, L. N., Maniania, N. K., Chabi-Olaye, A. & Vandenberg, J. Pathogenicity of entomopathogenic fungi Metarhizium anisopliae and Beauveria bassiana (Hypocreales: Clavicipitaceae) isolates to the adult pea leafminer (Diptera: Agromyzidae) and prospects of an autoinoculation device for infection in the field. Environ. Entomol. 39, 468–475 (2010).PubMed 
    CAS 

    Google Scholar 
    56.Campbell, A., Frazer, B. D., Gilbert, N., Gutierrez, A. P. & Mackauer, M. Temperature requirements of some aphids and their parasites. J. Appl. Ecol. 11, 431–438 (1974).
    Google Scholar 
    57.Brière, J.-F., Pracros, P., Le Roux, A.-Y. & Pierre, J.-S. A novel rate model of temperature-dependent development for arthropods. Environ. Entomol. 28, 22–29 (1999).
    Google Scholar 
    58.Archontoulis, S. V. & Miguez, F. E. Nonlinear regression models and applications in agricultural research. Agron. J. 107, 786–798 (2015).
    Google Scholar 
    59.Logan, J. A., Wollkind, D. J., Hoyt, S. C. & Tanigoshi, L. K. An analytic model for description of temperature dependent rate phenomena in arthropods. Environ. Entomol. 5, 1133–1140 (1976).
    Google Scholar 
    60.Steiniger, S. & Hunter, A. J. S. Free and open source GIS software for building a spatial data infrastructure. in Geospatial Free and Open Source Software in the 21st Century 247–261 (2012).61.Abbott, W. S. A method of computing the effectiveness of an insecticide. J. Econ. Entomol. 18, 265–267 (1925).CAS 

    Google Scholar 
    62.R Core Team. R: A language and environment for statistical computing R Foundation for Statistical Computing, Vienna, Austria, https://www.R-project.org/. (2019). More

  • in

    Modelling ocean acidification effects with life stage-specific responses alters spatiotemporal patterns of catch and revenues of American lobster, Homarus americanus

    Model speciesThe American lobster Homarus americanus, is a crustacean found in temperate regions across the Northwest Atlantic Ocean. It is a highly valuable fishery species, caught off the coast of Eastern Canada and Northeast USA. For the past decade, they have been the most valuable single-species fishery in all of Canada and the US30,48. In Canada, catch was estimated at 104,000 tonnes and 14% of all of Canada’s total marine fisheries catch in 2019. However, their landed value was almost $1.6 billion and 44% of the Canada’s total marine fisheries landed value, and over 49% of the landed value from Canada’s Atlantic coast fisheries30,49. Canada’s Atlantic marine fisheries provide employment to over 40,000 people in primary harvesting and processing sectors, and supports many rural populations30.Dynamic bioclimate envelope modelWe used a dynamic bioclimate envelope model (DBEM)31,50 to assess spatial and temporal changes in the abundance and fisheries catch under different scenarios of climate change and fishing pressure. The model infers the environmental preference of the modelled species51, and simulates future changes in biomass and maximum catch potential. Notably, the DBEM integrates growth models52, ecophysiological models53, advection–diffusion models54, and surplus production population dynamics models55 to determine ocean change effects on species distribution, abundance, and catch. Since H. americanus is primarily a benthic invertebrate, we used values at sea floor for environmental variables in our model. However, they also have a pelagic larval phase and we used surface environmental variables to model this life stage.Initial distribution and abundanceThe DBEM uses an initial species distribution determined using a rule-based algorithm56,57,58. This algorithm determines species’ distribution range base on a series of geographic constraints, including latitudinal range, depth range, occurring ocean basins, and published or expert provided ‘bounding box’. It assumes that the relative abundance of the species distributes along gradients within these geographic limits with the centroid of the range having the highest relative abundance (Palomares et al. 2016). Species distributions are mapped on a global grid with a resolution of 0.5° longitude by 0.5° latitude cells with values representing relative abundance. Historical reconstructed catch data (http://www.seaaroundus.org)59 was used to estimate the global abundance and distributed accordingly with relative abundance values60. These catch data are based on various government and non-government reports, primary and grey literature, and also mapped on a 0.5° longitude by 0.5° latitude grid.The initial species distribution is then overlaid on climatologies of historical environmental conditions (e.g. temperature, salinity, oxygen, pH) simulated from outputs of Earth system models (e.g. Bopp et al. 2013; Dunne et al. 2013). The DBEM assumes that species distribution is in equilibrium with the average historical environmental condition (1971–2000 average) and abundance in cells are assumed to be at carrying capacity.Modelling individual growthGrowth is modelled using a derived von Bertalanffy growth model to incorporate how environmental stressors affects body size of lobsters31,32,61. We model the following important life history parameters as a function of relative changes in temperature, oxygen content [O2], and pH [H+]. Growth rate (dB/dt) is dependent on weight-specific anabolic and catabolic rates:$$frac{dB}{{dt}} = H_{i,t} W^{d} – k_{i,t} W$$
    (1)
    where H and k represent the coefficients for oxygen supply (anabolism) and oxygen demand for maintenance metabolism (catabolism), respectively, for cell i at time t. Body weight is scaled to anabolism with the exponent d  40 ppt), mixoeuhaline ( > 29 ppt), polyhaline ( > 18 ppt), mesophaline ( > 5 ppt), oligohaline ( > 0 ppt)—and SAssoc is the association of the species with each salinity class; and Icei is the sea ice % area coverage in a cell and IceP is the ice-dependency of the species. For H. americanus, they are not specifically associated with any habitat and thus only restricted by depth parameters. However, they are limited to mixoeuhaline and polyhaline salinities62.The TPP was estimated using the initial predicted relative abundance (described above) overlaid with the inputs of earth system models of initial environmental conditions. The relative weight for each temperature class z of the temperature preference profile was calculated as (TPP_{z} = R_{z} /sum R_{z}), where Rz is the relative abundance in each temperature class.A fuzzy logic model was used to model the movement between neighbouring cells based on differences in habitat suitability50. Emigration into a cell is favoured if habitat suitability is higher than surrounding cells, and immigration out of a cell is favoured if habitat suitability is lower than surrounding cells.We estimated larval production as 30% of spawning population biomass for each cell i, while larval mortality was 0.85 day−1 and settlement rate was 0.15 day−1—these values were chosen based on the sensitivity testing of these parameters50.Larval dispersal is modelled using an advection–diffusion54 and a larval duration model based on temperature66, such that abundance Ai,t in each cell is numerically solved for using the equation:$$frac{{partial A_{i,t} }}{partial t} = frac{partial }{partial x}left( {D_{i,t} frac{{partial A_{i,t} }}{partial x}} right) + frac{partial }{partial y}left( {D_{i,t} frac{{partial A_{i,t} }}{partial y}} right) – frac{partial }{partial y}left( {u cdot A_{i,t} } right) – frac{partial }{partial y}left( {v cdot A_{i,t} } right) – lambda cdot A_{i,t}$$
    (16)
    while adult dispersal is similarly modelled,$$frac{{partial A_{i,t} }}{partial t} = frac{partial }{partial x}left( {D_{i,t} frac{{partial A_{i,t} }}{partial x}} right) + frac{partial }{partial y}left( {D_{i,t} frac{{partial A_{i,t} }}{partial y}} right)$$
    (17)
    Advection was modelled for larval dispersal using parameters u and v for horizontal (east–west) and vertical (north–south) directions for surface current velocity (m2 s−1), respectively, between neighbouring cells x and y in the east–west and north–south direction, respectively. Instantaneous rate of larval mortality, ML, and settlement, SL was integrated into Eq. (16), where (lambda = 1 – e^{{ – left( {M_{L} + S_{L} } right)}}). The coefficient Di,t is the diffusion parameter:$$D_{i,t} = frac{{D_{i,0} cdot m}}{{1 + e^{{(tau cdot P_{i,t} cdot rho_{i,t} )}} }}$$
    (18)
    and$$rho_{i,t} = 1 – frac{{emptyset_{i,t} }}{{left( {C_{i,t} /overline{W}_{i,t} } right)}}$$
    (19)
    where Di,0 is the initial diffusion coefficient and a function of the spatial grid size (GR): (D_{i,0} = left( {1.1 times 10^{4} } right) cdot GR cdot 1.33). Parameters m and (tau)—both set at 2 in the model—determine the curvature of the functional relationship between D, P, and (rho)50. Parameter (rho_{i,t}) represents density-dependent factors and a function of population density (number of individuals), (emptyset_{i,t}), carrying capacity ((C_{i,t})), and mean body weight ((overline{W}_{i,t})) in each cell i.Models of ocean acidification effectsThe DBEM operates to model larval dispersal using advection–diffusion models. Survival is calculated at each time step (biweekly) based on a static annual survival rate. We recently tested a simple linear relationship between survival rate and pH, represented by percent changes in the survival rate given a change in pH32. We used parameters derived from previous experimental studies, where they observed a ~ 15% increase in mortality in larval and juvenile stages37 from a doubling of hydrogen ion concentration.We explore the OA effects by modelling variations in life stage-specific sensitivities to OA. Larvae, juveniles, and adults are modelled based on size classes, and the weight at maturity determines the size at which juveniles transition into adults65. Therefore, impacts of OA on survival can be modelled for various size classes. We model the effects of OA on the three major life history stages—larval, juvenile, and adult—and use a correlative approach to link changes in ocean acidity with changes in survival.The length transition matrix (Eqs. (8) and (9)) is split up into 40 length size classes, divided evenly from 0 to (l_{infty ,i,t}). We assume larvae transition from the pelagic phase to the growth transition matrix, and enter as the ‘larval’ stage for only the first size class. Next, juvenile size classes comprise all size classes below the length at maturity, lmat, as determined in Eq. (12), and lobster in any size classes greater are considered adults. While our models do not incorporate lobster-specific life cycle traits (i.e. transitioning between larval stages then to juvenile stages), we use more general models that can be broadly applied to many species.Modelling effects on survivalOA effects can be modelled as relative changes in survival rate for all life stages in Table 2. In other words, percent changes in acidity (i.e. hydrogen ion concentration) from baseline initial conditions results in a percent change in baseline survival rate. We use a model structure similar to that of previous work we have done32:$$Surv_{t} = Surv_{init} *left[ {1 + left( {p*left( {frac{{left[ {H^{ + } } right]_{t} }}{{left[ {H^{ + } } right]_{init} }} – 1} right)^{w} } right)} right]$$
    (20)
    Table 2 Scenario settings explored with model projections.Full size tableSurv is the survival rate per year and used here as an example but can be applied to other life histories affected by OA (e.g. growth, reproduction). Survival rate in year t is derived from the initial (init) survival rate and the relative change in [H+] between year t and initial [H+] conditions. Note that in our previous model, p represents the point value of the percent change effect size with a doubling of [H+]. This model utilizes single point effect size estimates that have no underlying assumed relationship between acidity and survival. In our model, we used an exponent value, w, equal to 1, which assumes a linear relationship32.Fishing pressureFishing mortality was assumed to be at maximum sustainable yield (MSY), which is the theoretical maximum biomass that can be sustainably removed from the population indefinitely. MSY is calculated using a Gordon Schaefer population growth model67:$$MSY = frac{{B_{infty } cdot r}}{4}$$where B∞ is the population carrying capacity and r is the intrinsic population growth rate. We use this measure of MSY as a proxy for the maximum catch potential (MCP) into the future, thus we assume that fisheries management are optimized and operate at MSY.Furthermore, the fishing mortality rate, Fi,t—i.e. the annual proportion at which biomass is taken from the current population biomass—in each cell i at time t at MSY can be calculated as:$$F_{MSY,i,t} = frac{r}{2}$$The fishing mortality rate was adjusted to explore scenarios of reduced fishing pressure and interactions with climate change and OA on the population dynamics of lobster. Any reductions in fishing pressure began in 2010 to represent how changes in fishing implemented now could change the state of lobster populations with the added stressors of climate change.Fishing size limitsFishing size limits were set to represent management scenarios and to observe their effects on lobster populations and size distribution of the population. We set four scenarios of minimum body size restrictions of lobster catch: no limit, canner small ( > 0.5 lb, 220 g), canner large ( > 0.75 lb, 320), and market ( > 1 lb, 430 g). Canner lobsters are smaller lobsters that are often sold at a cheaper price. They range from 0.5 to 1 lb, and are largely caught in Northumberland Strait where size limits are currently set lower due to warmer waters and smaller size at maturity44. For these scenarios of fishing size limits, we continue to assume catch is at maximum sustainable yield. Therefore, the same catch biomass (calculated using fishing mortality rate, F) will be the same for the various size limit restrictions, and more biomass will be taken from upper size classes where size limits are implemented. The no size limit results in fishing mortality to all size classes (including undersized lobsters).Climate change scenariosWe use outputs from three Earth system models for projections of various future climate change outcomes: NOAA’s Geophysical Fluid Dynamics Laboratory (GFDL-ESM), Institute Pierre Simon Laplace Climate Modelling Centre (IPSL-ESM), and Max Planck Institute for Meteorology (MPI-ESM)68. These models are included in the Coupled Model Intercomparisons Projection Phase 5 (CMIP5). We used two Representative Concentration Pathways (RCPs)69 which are greenhouse gas (GHG) concentration trajectories derived to reflect possible combinations of various socioeconomic assumptions, RCP 2.6 and RCP 8.5. The number associated with RCPs represent the radiative forcing values by the year 2100 based on greenhouse gas concentrations. RCP 2.6 corresponds with a low climate change scenario and assumes immediate mitigation of GHG emissions where annual emissions peak by mid-decade (year 2025) but is reduced considerably. This scenario is more in line with the 2015 Paris Agreement to limit warming to + 1.5 °C relative to other emissions scenarios applied in most CMIP5 Earth system models. Conversely, RCP 8.5 corresponds to a high climate change scenario and our current trajectory where we continue to use fossil fuels, have little to no change to switch to renewable energy sources, and GHG emissions continue to increase with no implementation of any mitigation action. We chose these three Earth system models as they provide sea surface and bottom layers and the full range of environmental variables required by the DBEM (i.e. sea temperature, dissolved oxygen, primary production, pH, current advection, salinity, sea ice extent) for both RCP scenarios2.All statistical analyses and figures were generated using the programming software R v4.0.370. More

  • in

    Global patterns of potential future plant diversity hidden in soil seed banks

    Our global database was derived from studies measuring soil seed bank diversity and density of natural plant communities across all continents, albeit with a strong data availability bias towards North America, Europe, eastern Asia and Oceania as compared to elsewhere (Fig. 1). The database contains 15,698 records for soil seed banks worldwide, including 6,480 for diversity (represented here by species richness) and 9,218 for density (number of seeds per soil surface area). The database represents more than a century of research with the oldest publication dating back to 191815. This most exhaustive and comprehensive set of research data on soil seed bank to date allowed us to identify the determinants and patterns of soil seed bank at the global scale.Fig. 1: Locations of the soil seed bank studies included in our database.a Diversity; b Density.Full size imageTo make data among studies comparable, we standardized them using a three-step process. First, we identified soil seed banks that showed seasonal patterns in both diversity and density, all of which peaked slightly in winter (Supplementary Fig. 1a, b). Thus, we standardized all data (from non-(sub-)tropical regions) for other seasons to winter. Second, sampling area for soil seed bank diversity varied among studies, with 0.01 m2 being the most commonly reported (Supplementary Fig. 2), to which we standardized all data using a species-area curve (Supplementary Table 1). Third, sampling depth also varied among studies, with 0–5 cm being the most frequently reported soil depth (Supplementary Fig. 3). Therefore, 0–5 cm was chosen as the soil depth for standardization of data in various soil depths. Such standardization is needed to find the relationships of seed bank data between different soil depths. We used the upper and lower limits of soil depths (e.g., for 0–5 cm, the upper limit was 0 cm and the lower 5 cm). The log-scale regressions showed that both soil seed bank diversity and density decreased significantly with lowering upper boundaries of soil depths but increased with lower ones (Supplementary Table 2), and thus we standardized all data to 0–5 cm depth using these relationships. To account for possible variation among biomes, the second and third standardization procedures were conducted for each biome separately. The analyses during standardization confirmed the need to standardize empirical findings when comparing seed bank patterns across studies, as previously stressed in a study on grassland soil seed banks10. Our standardization procedures made all data comparable in terms of season, sampling area and soil depth.Non-parametric Kruskal–Wallis tests showed that soil seed banks differed significantly among ecosystem types. Mangroves, tundra and tropical & subtropical dry broadleaf forests had a lower diversity of soil seed banks, whereas Mediterranean forests, woodlands & scrub, tropical & subtropical moist broadleaf forests and tropical & subtropical coniferous forests had a higher diversity (Supplementary Fig. 4a). For density, mangroves and flooded grasslands & savanna had the lowest value, while temperate broadleaf & mixed forests and temperate conifer forests had the highest value (Supplementary Fig. 4b).Prior to spatial analyses, we computed semivariograms to determine whether spatial autocorrelation could affect our models. We found that there was no obvious spatial autocorrelation in the data of soil seed bank diversity or density (Supplementary Fig. 5), indicating no spatial dependence in our data. We then used the random-forest algorithm (see Methods for details) to determine the importance (as increase in node purity) of the influence of 31 variables related to climate, soil, human disturbance and spatial coordinates (Supplementary Table 3) on diversity and density of soil seed banks. These variables previously were reported to affect plant performance at the global scale16,17,18, and thus they could affect soil seed banks via their effects on seed production. Moreover, we expected that potentially these variables could affect seed longevity in the soil. Full models using all 31 predictors showed that climate and soil were important in predicting soil seed banks (Fig. 2a and Supplementary Fig. 6). Moreover, spatial coordinates (absolute latitude) were the most important predictor for diversity, i.e., diversity of soil seed banks exhibit clear spatial patterns at the global scale. Net primary productivity (NPP) and soil characteristics were important in predicting the density of soil seed banks (Fig. 2b and Supplementary Fig. 6).Fig. 2: Variable importance (increase in node purity) of random forests run with all 31 predictors.a Soil seed bank diversity; b Density. abs.latit, absolute latitude; AMT, annual mean temperature; AP, annual precipitation; ATR, annual temperature range; AVWAC, available water capacity (%); BULK, bulk density; CEC, cation exchange capacity; CLAYC, clay (mass %); diversity, plant diversity; HFP, human footprint; Isoth, isothermality; npp, plant productivity (net primary production); ORGNC, organic carbon content; PCQ, precipitation of coldest quarter; PDM, precipitation of driest month; PDQ, precipitation of driest quarter; pH, pH measured in water; Pseason, precipitation seasonality (coefficient of variation); PWeQ, precipitation of wettest quarter; PWM, precipitation of wettest month; PWQ, precipitation of warmest quarter; SANDC, sand (mass %); SILTC, silt (mass %); TCM, min temperature of coldest month; TCQ, mean temperature of coldest quarter; TDQ, mean temperature of driest quarter; TDR, mean diurnal range (mean of monthly (max temp–min temp)); Tseason, temperature seasonality (standard deviation *100); TWeQ, mean temperature of wettest quarter; TWM, max temperature of warmest month; TWQ, mean temperature of warmest quarter.Full size imageWe then built final random-forest models using the most important predictors of seed banks selected from full models: nine variables for diversity and five for density (Supplementary Fig. 7). Final models explained more of the total variance than did full models (Supplementary Table 4), and they were robust to K-fold cross-validation (Supplementary Fig. 8), indicating that a small number of variables predicted soil seed bank diversity and density. Absolute latitude (abs.latit) was the most important predictor for diversity, which varied between 0–55° and then decreased beyond this range (Fig. 3a). Five climatic variables were important for diversity. Diversity peaked at intermediate annual temperature ranges (ATR), while it was the lowest at intermediate mean temperature of driest quarter of the year (TDQ), precipitation of the coldest quarter (PCQ) and precipitation of the driest quarter (PDQ). Diversity increased with increasing annual precipitation (AP). In addition, three soil variables were important for diversity. Diversity showed a humped relationship with soil pH, with pH 6–7 having the highest diversity. Diversity increased with soil cation exchange capacity (CEC) and soil silt content (SILT). These results indicate that diversity exhibits strong spatial patterns at the global scale. However, our spatial patterns differ from those found for a specific ecosystem worldwide (e.g., grasslands), where there were only weak latitudinal gradients in seed bank diversity10. In addition, climate emerged as an important predictor for seed bank diversity, which is consistent with the report that climate acts as environmental filters affecting soil seed bank of grasslands around the world13. Our results agree with a continental study in Europe, where ATR was more important than mean annual temperature for determining seed bank richness and warmer temperatures were associated with lower seed bank richness11. Possible mechanisms by which temperature affects soil seed banks are that it (1) influences seed bank inputs via its effects on seed production; (2) cues dormancy-breaking and germination1, thus determining germinable seed output from seed banks; and (3) affects seed metabolic activity and soil fungal activity19, thereby determining seed viability and persistence in the soil. Finally, our findings of a significant effect of soil pH are supported by some regional and local studies. For instance, seed bank composition is significantly associated with soil pH at high elevations on the Tibetan Plateau20. A negative effect of low pH also has been reported in a large-scale study of acidic and calcareous grasslands in England21. Two possible mechanisms for the effects of soil pH are that (1) low pH may cause loss of seed viability due to the toxicity from aluminum or other metals that become more readily available in soils with low pH22; and (2) high pH may accelerate decomposition and promote growth of pathogens that negatively affect seed persistence23. In our study, the two mechanism may operate synchronously, thereby resulting in the highest diversity of soil seed banks at intermediate pH at the global scale. Further, our results show that soil CEC and SILT affect seed bank diversity, which agrees with a study on the Tibetan Plateau20. The physical and chemical properties of soils can affect seed bank directly by affecting seed germination and aging via regulating soil water-holding capacity24, or indirectly by affecting seed viability via controlling the activity of soil pathogens21,22,25.Fig. 3: Partial feature contributions (the marginal effect of a variable on response) of the most important variables for soil seed banks.a diversity; b density. Variable importance (inc. node) is the decrease in the residual sum of squares that results from splitting regression trees using the variable. The percentage increase in mean squared error (% inc. MSE) is the increase in model error as a result of randomly shuffling the order of values in the vector. abs.latit, absolute latitude; AP, annual precipitation; ATR, annual temperature range; BULK, bulk density; CEC, cation exchange capacity; npp, plant productivity (net primary production); PCQ, precipitation of coldest quarter; PDM, precipitation of driest month; PDQ, precipitation of driest quarter; pH, pH measured in water; SILTC, silt (mass %); TDQ, mean temperature of driest quarter; TWM, max temperature of warmest month.Full size imageFor soil seed bank density, soil bulk density (BULK) was the most important predictor; density increased below 750 g/cm3 BULK but remained stable when BULK was higher than 800 g/cm3 (Fig. 3b). Density peaked when temperature of the warmest month (TWM) was 34 °C. Density showed similar variation with NPP, precipitation of the driest quarter of the year (PDQ) and of the driest month (PDM), i.e., it peaked at intermediate values of these variables. Precipitation influences the success of sexual reproduction of plants and the size of the seed bank through seed input26, and it also affects soil pathogenic fungi, which cause seed mortality27. Therefore, precipitation has a strong effect on seed bank density, as reported for 27 alpine meadows on the Tibetan Plateau28. Our results further illustrate that PDQ and PDM are the key factors determining seed bank density worldwide, suggesting that moisture fluctuation in soils triggered by precipitation of the driest time of the year can affect seed bank density. If soil moisture fluctuations are high, seed germination will be primed by increasing moisture24.At the global scale, we mapped soil seed bank diversity and density using the final random-forest models. Mapping soil seed bank values onto global maps revealed considerable geospatial variation, the pattern of which varied between diversity and density (Fig. 4). For diversity, western North America, central South America, central Africa, central Europe, southern and eastern Asia and eastern Oceania had high values. In contrast, eastern and central North America, northern Africa and central Asia had low values (Fig. 4a). For density, northern North America, northern Europe and northern Asia had higher values than elsewhere (Fig. 4a). Our results are consistent with the reports that larger seed banks are more common in cooler temperate climates19,29. The latitudinal pattern of higher density in colder regions in the Northern Hemisphere may be driven by lower seed mortality in colder soils6, resulting in stable seed bank densities of long-lived seeds that counteract low seed production in some years at cold northern latitudes, as shown in a study of temperate forests along a 1900 km latitudinal gradient in northwestern Europe29. The latitudinal pattern highlights that particularly species rich low-latitude biomes such as tropical rainforests generally have very low seed bank densities, while their seed bank diversity does not exceed that in higher latitudes biomes. However, our global assessment should be interpreted with caution since some studies in azonal vegetation or in rare habitats in our database did not fully reflect soil seed banks in that region, and thus these data shortcomings may have induced bias in our global predictions. Moreover, data gaps in our database are also likely to have had an effect on the global predictions, i.e., fewer data available from some continents (e.g., northern Asia and Africa) could lead to less confidence for prediction in these regions. For example, Russia has very few soil seed bank data, which may have led to an inaccurate prediction for this country. Nevertheless, based on our global patterns of soil seed bank diversity and density, the latitudinal pattern strongly suggests that the biodiversity of (sub-)tropical forests is particularly vulnerable to large-scale climatic or land-use disturbances. However, in-depth investigation is needed to quantify the extent to which temporal integration of seed bank effects for long-lived trees and seed masting events may buffer the effects of low seed bank diversity and density at any given time of sampling. In contrast, the higher-latitude plant diversity, while currently low compared to that in tropical rainforest, may rely on high soil seed bank densities to boost its resilience to large-scale climate- or land-use induced disturbances. Further, our analyses suggest that the least vulnerable ecosystems in terms of hidden diversity should be those that combine high seed-bank diversity with high density; and therefore the relationships between the two variables across the global map certainly would be an interesting topic worthy of further study.Fig. 4: Extrapolated global maps of soil seed banks.a diversity in terms of number of species per 0.01 m2; b density as number of seeds per m2. In b, values are log10-transformed to facilitate viewing. The spatial resolution of grid cells is 5 arcmin-by-5 arcmin.Full size imageOur global assessment reveals that both diversity and density exhibit clear spatial patterns of soil seed banks but differ in their environmental determinants. These findings alone do not necessarily mean that this biodiversity reservoir has strong buffering capacity under climate change, because both climate and soil conditions influence seed bank diversity and density. Based on a large number and long history of studies globally, we provide quantitative evidence of how environmental conditions shape soil seed bank distributions and spatially explicit maps of this biodiversity reservoir in plant communities worldwide. Our quantification of environmental determinants and global mapping can be readily applied to dynamic global vegetation and plant diversity models to enable a more complete and accurate prediction of the impact of ongoing environmental changes on plant diversity (both above- and belowground) at the global scale. The next research challenge will be to plot current (visible) aboveground plant diversity (ideally using the available data in the studies themselves) against soil seed bank diversity under global change scenarios in order to pinpoint even more accurately which plant communities, ecosystems and biomes (and their turn-over) are most at risk of losing their diversity due to global changes. More

  • in

    Heat stress reduces the contribution of diazotrophs to coral holobiont nitrogen cycling

    1.Thompson JR, Rivera HE, Closek CJ, Medina M. Microbes in the coral holobiont: partners through evolution, development, and ecological interactions. Front Cell Infect Microbiol. 2014;4:176.PubMed 

    Google Scholar 
    2.Pogoreutz C, Voolstra CR, Rädecker N, Weis V, Cardenas A, Raina J-B. The coral holobiont highlights the dependence of cnidarian animal hosts on their associated microbes. In: Bosch TCG, Hadfield MG, editors. Cellular dialogues in the holobiont. CRC Press; 2020. p. 91–118.3.Stanley GD, van de Schootbrugge B. The evolution of the coral–algal symbiosis. In: van Oppen MJH, Lough JM, editors. Coral bleaching: patterns, processes, causes and consequences. Berlin, Heidelberg: Springer; 2009. p. 7–19.4.LaJeunesse TC, Parkinson JE, Gabrielson PW, Jeong HJ, Reimer JD, Voolstra CR, et al. Systematic revision of Symbiodiniaceae highlights the antiquity and diversity of coral endosymbionts. Curr Biol. 2018;28:2570–80.e6.CAS 
    PubMed 

    Google Scholar 
    5.Muscatine L. The role of symbiotic algae in carbon and energy flux in reef corals. Coral Reefs. 1990;25:75–87.
    Google Scholar 
    6.Stolarski J, Kitahara MV, Miller DJ, Cairns SD, Mazur M, Meibom A. The ancient evolutionary origins of Scleractinia revealed by azooxanthellate corals. BMC Evol Biol. 2011;11:316.PubMed 
    PubMed Central 

    Google Scholar 
    7.Frankowiak K, Wang XT, Sigman DM, Gothmann AM, Kitahara MV, Mazur M, et al. Photosymbiosis and the expansion of shallow-water corals. Sci Adv. 2016;2:e1601122.PubMed 
    PubMed Central 

    Google Scholar 
    8.Wild C, Hoegh-Guldberg O, Naumann MS, Colombo-Pallotta MF, Ateweberhan M, Fitt WK, et al. Climate change impedes scleractinian corals as primary reef ecosystem engineers. Mar Freshw Res. 2011;62:205–15.CAS 

    Google Scholar 
    9.Hughes TP, Barnes ML, Bellwood DR, Cinner JE, Cumming GS, Jackson JBC. et al. Coral reefs in the Anthropocene. Nature. 2017;546:82–90.CAS 
    PubMed 

    Google Scholar 
    10.Kleypas J, Allemand D, Anthony K, Baker AC, Beck MW, Hale LZ, et al. Designing a blueprint for coral reef survival. Biol Conserv. 2021;257:109107.
    Google Scholar 
    11.Anthony KRN, Hoogenboom MO, Maynard JA, Grottoli AG, Middlebrook R. Energetics approach to predicting mortality risk from environmental stress: a case study of coral bleaching. Funct Ecol. 2009;23:539–50.
    Google Scholar 
    12.Hughes TP, Anderson KD, Connolly SR, Heron SF, Kerry JT, Lough JM. et al. Spatial and temporal patterns of mass bleaching of corals in the Anthropocene. Science. 2018;359:80–83.CAS 
    PubMed 

    Google Scholar 
    13.Voolstra CR, Suggett DJ, Peixoto RS, John E, Parkinson KM, Quigley CB, Silveira M, et al. Extending the natural adaptive capacity of coral holobionts. Nat Rev Earth Environ. 2021;2:747–62; https://doi.org/10.1038/s43017-021-00214-3.Article 

    Google Scholar 
    14.Suggett DJ, Smith DJ. Coral bleaching patterns are the outcome of complex biological and environmental networking. Glob Chang Biol. 2020;26:68–79.PubMed 

    Google Scholar 
    15.Rädecker N, Pogoreutz C, Gegner HM, Cárdenas A, Roth F, Bougoure J, et al. Heat stress destabilizes symbiotic nutrient cycling in corals. Proc Natl Acad Sci USA. 2021;118:e2022653118.PubMed 
    PubMed Central 

    Google Scholar 
    16.Morris LA, Voolstra CR, Quigley KM, Bourne DG, Bay LK. Nutrient availability and metabolism affect the stability of coral–Symbiodiniaceae symbioses. Trends Microbiol. 2019;27:678–89.CAS 
    PubMed 

    Google Scholar 
    17.Muscatine L, Porter JW. Reef Corals: mutualistic symbioses adapted to nutrient-poor environments. Bioscience. 1977;27:454–60.
    Google Scholar 
    18.Falkowski PG, Dubinsky Z, Muscatine L, Porter JW. Light and the bioenergetics of a symbiotic coral. Bioscience. 1984;34:705–9.CAS 

    Google Scholar 
    19.Falkowski PG, Dubinsky Z, Muscatine L, McCloskey L. Population control in symbiotic corals. Bioscience. 1993;43:606–11.
    Google Scholar 
    20.Peng S-E, Chen C-S, Song Y-F, Huang H-T, Jiang P-L, Chen W-NU, et al. Assessment of metabolic modulation in free-living versus endosymbiotic Symbiodinium using synchrotron radiation-based infrared microspectroscopy. Biol Lett. 2012;8:434–7.PubMed 

    Google Scholar 
    21.Krueger T, Horwitz N, Bodin J, Giovani M-E, Escrig S, Fine M, et al. Intracellular competition for nitrogen controls dinoflagellate population density in corals. Proc R Soc B. 2020;287:20200049.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    22.Roberty S, Béraud E, Grover R, Ferrier-Pagès C. Coral croductivity is co-limited by bicarbonate and ammonium availability. Microorganisms. 2020;8:640CAS 
    PubMed Central 

    Google Scholar 
    23.Baker DM, Freeman CJ, Wong JCY, Fogel ML, Knowlton N. Climate change promotes parasitism in a coral symbiosis. ISME J. 2018;12:921–30.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    24.Cunning R, Muller EB, Gates RD, Nisbet RM. A dynamic bioenergetic model for coral- Symbiodinium symbioses and coral bleaching as an alternate stable state. J Theor Biol. 2017;431:49–62.CAS 
    PubMed 

    Google Scholar 
    25.Rädecker N, Pogoreutz C, Voolstra CR, Wiedenmann J, Wild C. Nitrogen cycling in corals: the key to understanding holobiont functioning? Trends Microbiol. 2015;23:490–7.PubMed 

    Google Scholar 
    26.Wiedenmann J, D’Angelo C, Smith EG, Hunt AN, Legiret F-E, Postle AD, et al. Nutrient enrichment can increase the susceptibility of reef corals to bleaching. Nat Clim Chang. 2012;3:160–4.
    Google Scholar 
    27.Houlbrèque F, Ferrier-Pagès C. Heterotrophy in tropical scleractinian corals. Biol Rev Camb Philos Soc. 2009;84:1–17.PubMed 

    Google Scholar 
    28.Fiore CL, Jarett JK, Olson ND, Lesser MP. Nitrogen fixation and nitrogen transformations in marine symbioses. Trends Microbiol. 2010;18:455–63.CAS 
    PubMed 

    Google Scholar 
    29.Grover R, Maguer J-F, Allemand D, Ferrier-Pagès C. Uptake of dissolved free amino acids by the scleractinian coral Stylophora pistillata. J Exp Biol. 2008;211:860–5.CAS 
    PubMed 

    Google Scholar 
    30.Lema KA, Willis BL, Bourne DG. Corals form characteristic associations with symbiotic nitrogen-fixing bacteria. Appl Environ Microbiol. 2012;78:3136–44.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    31.Pogoreutz C, Rädecker N, Cárdenas A, Gärdes A, Wild C, Voolstra CR. Nitrogen fixation aligns with nifH abundance and expression in two coral trophic functional groups. Front Microbiol. 2017;8:1187.PubMed 
    PubMed Central 

    Google Scholar 
    32.Olson ND, Lesser MP. Diazotrophic diversity in the Caribbean coral, Montastraea cavernosa. Arch Microbiol. 2013;195:853–9.CAS 
    PubMed 

    Google Scholar 
    33.Lesser MP, Morrow KM, Pankey SM, Noonan SHC. Diazotroph diversity and nitrogen fixation in the coral Stylophora pistillata from the Great Barrier Reef. ISME J. 2018;12:813–24.CAS 
    PubMed 

    Google Scholar 
    34.Moynihan MA, Goodkin NF, Morgan KM, Kho PYY, dos Santos AL, Lauro FM, et al. Coral-associated nitrogen fixation rates and diazotrophic diversity on a nutrient-replete equatorial reef. ISME J. 2021; https://doi.org/10.1038/s41396-021-01054-1.35.Tilstra A, Pogoreutz C, Rädecker N, Ziegler M, Wild C, Voolstra CR. Relative diazotroph abundance in symbiotic Red Sea corals decreases with water depth. Front Mar Sci. 2019;6:372.
    Google Scholar 
    36.Bednarz VN, Cardini U, van Hoytema N, Al-Rshaidat MMD, Wild C. Seasonal variation in dinitrogen fixation and oxygen fluxes associated with two dominant zooxanthellate soft corals from the northern Red Sea. Mar Ecol Prog Ser. 2015;519:141–52.
    Google Scholar 
    37.Rädecker N, Meyer FW, Bednarz VN, Cardini U, Wild C. Ocean acidification rapidly reduces dinitrogen fixation associated with the hermatypic coral Seriatopora hystrix. Mar Ecol Prog Ser. 2014;511:297–302.
    Google Scholar 
    38.Cardini U, Bednarz VN, Naumann MS, van Hoytema N, Rix L, Foster RA, et al. Functional significance of dinitrogen fixation in sustaining coral productivity under oligotrophic conditions. Proc R Soc B. 2015;282:20152257.PubMed 
    PubMed Central 

    Google Scholar 
    39.Bednarz VN, van de Water JAJM, Grover R, Maguer J-F, Fine M, Ferrier-Pagès C. Unravelling the importance of diazotrophy in corals – combined assessment of nitrogen assimilation, diazotrophic community and natural stable isotope signatures. Front Microbiol. 2021;12:1638.
    Google Scholar 
    40.Santos HF, Carmo FL, Duarte G, Dini-Andreote F, Castro CB, Rosado AS, et al. Climate change affects key nitrogen-fixing bacterial populations on coral reefs. ISME J. 2014;8:2272–9.PubMed 
    PubMed Central 

    Google Scholar 
    41.Cardini U, van Hoytema N, Bednarz VN, Rix L, Foster RA, Al-Rshaidat MMD, et al. Microbial dinitrogen fixation in coral holobionts exposed to thermal stress and bleaching. Environ Microbiol. 2016;18:2620–33.CAS 
    PubMed 

    Google Scholar 
    42.Bednarz VN, van de Water JAJM, Rabouille S, Maguer J-F, Grover R, Ferrier-Pagès C. Diazotrophic community and associated dinitrogen fixation within the temperate coral Oculina patagonica. Environ Microbiol. 2019;21:480–95.CAS 
    PubMed 

    Google Scholar 
    43.Pogoreutz C, Rädecker N, Cárdenas A, Gärdes A, Voolstra CR, Wild C. Sugar enrichment provides evidence for a role of nitrogen fixation in coral bleaching. Glob Chang Biol. 2017;23:3838–48.PubMed 

    Google Scholar 
    44.Bednarz VN, Grover R, Maguer J-F, Fine M, Ferrier-Pagès C. The assimilation of diazotroph-derived nitrogen by scleractinian corals depends on their metabolic status. mBio. 2017;8:e02058–16.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    45.Petrou K, Nunn BL, Padula MP, Miller DJ, Nielsen DA. Broad scale proteomic analysis of heat-destabilised symbiosis in the hard coral Acropora millepora. Sci Rep. 2021;11:19061.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    46.Roth F, Rädecker N, Carvalho S, Duarte CM, Saderne V, Anton A, et al. High summer temperatures amplify functional differences between coral‐ and algae‐dominated reef communities. Ecology. 2021;102:e03226.PubMed 

    Google Scholar 
    47.Andersson AF, Lindberg M, Jakobsson H, Bäckhed F, Nyrén P, Engstrand L. Comparative analysis of human gut microbiota by barcoded pyrosequencing. PLoS ONE. 2008;3:e2836.PubMed 
    PubMed Central 

    Google Scholar 
    48.Bayer T, Neave MJ, Alsheikh-Hussain A, Aranda M, Yum LK, Mincer T, et al. The microbiome of the Red Sea coral Stylophora pistillata is dominated by tissue-associated Endozoicomonas bacteria. Appl Environ Microbiol. 2013;79:4759–62.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    49.Gaby JC, Buckley DH. A comprehensive evaluation of PCR primers to amplify the nifH gene of nitrogenase. PLoS ONE. 2012;7:e42149.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    50.Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods. 2016;13:581–3.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    51.R Core Team. R: a language and environment for statistical computing computer program. Vienna, Austria: R Foundation for Statistical Computing; 2021.52.Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res. 2013;41:D590–6.CAS 
    PubMed 

    Google Scholar 
    53.McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS ONE. 2013;8:e61217.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    54.Oksanen J, Kindt R, Legendre P, O’Hara B, Stevens MHH, Oksanen MJ, et al. The vegan package. Community Ecol Package. 2007;10:631–7.
    Google Scholar 
    55.Roesch LFW, Dobbler PT, Pylro VS, Kolaczkowski B, Drew JC, Triplett EW. pime: A package for discovery of novel differences among microbial communities. Mol Ecol Resour. 2020;20:415–28.CAS 
    PubMed 

    Google Scholar 
    56.Guo K. microbial: do 16s data analysis and generate figures. R package version 1.14.4. 2021.57.Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet J. 2011;17:10–12.
    Google Scholar 
    58.Wang Q, Quensen JF, 3rd, Fish JA, Lee TK, Sun Y, Tiedje JM. et al. Ecological patterns of nifH genes in four terrestrial climatic zones explored with targeted metagenomics using FrameBot, a new informatics tool. mBio. 2013;4:e00592–13.PubMed 
    PubMed Central 

    Google Scholar 
    59.Meunier V, Geissler L, Bonnet S, Rädecker N, Perna G, Grosso O, et al. Microbes support enhanced nitrogen requirements of coral holobionts in a high CO2 environment. Mol Ecol. 2021;30:5888–99 .60.Angel R, Nepel M, Panhölzl C, Schmidt H, Herbold CW, Eichorst SA, et al. Evaluation of primers targeting the diazotroph functional gene and development of NifMAP – a bioinformatics pipeline for analyzing nifH amplicon data. Front Microbiol. 2018;9:703.PubMed 
    PubMed Central 

    Google Scholar 
    61.Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol. 2013;30:772–80.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    62.Frank IE, Turk-Kubo KA, Zehr JP. Rapid annotation of nifH gene sequences using classification and regression trees facilitates environmental functional gene analysis. Environ Microbiol Rep. 2016;8:905–16.PubMed 

    Google Scholar 
    63.Berger SA, Krompass D, Stamatakis A. Performance, accuracy, and web server for evolutionary placement of short sequence reads under maximum likelihood. Syst Biol. 2011;60:291–302.PubMed 
    PubMed Central 

    Google Scholar 
    64.Hardy RW, Holsten RD, Jackson EK, Burns RC. The acetylene-ethylene assay for N2 fixation: laboratory and field evaluation. Plant Physiol. 1968;43:1185–207.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    65.Breitbarth E, Mills MM, Friedrichs G, LaRoche J. The Bunsen gas solubility coefficient of ethylene as a function of temperature and salinity and its importance for nitrogen fixation assays: Bunsen coefficient and N2fixation. Limnol Oceanogr Methods 2004;2:282–8.
    Google Scholar 
    66.Zehr JP, Montoya JP. Measuring N2 fixation in the field. In: Bothe H, Ferguson SJ, Newton WE, editors. Biology of the nitrogen cycle. Elsevier; 2007. p. 193–205.67.Soper FM, Simon C, Jauss V. Measuring nitrogen fixation by the acetylene reduction assay (ARA): is 3 the magic ratio?. Biogeochemistry. 2021;152:345–51.CAS 

    Google Scholar 
    68.Lavy A, Eyal G, Neal B, Keren R, Loya Y, Ilan M. A quick, easy and non‐intrusive method for underwater volume and surface area evaluation of benthic organisms by 3D computer modelling. Methods Ecol Evol. 2015;6:521–31.
    Google Scholar 
    69.Wilson ST, Böttjer D, Church MJ, Karl DM. Comparative assessment of nitrogen fixation methodologies, conducted in the oligotrophic North Pacific Ocean. Appl Environ Microbiol. 2012;78:6516–23.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    70.Hoppe P, Cohen S, Meibom A. NanoSIMS: Technical aspects and applications in cosmochemistry and biological geochemistry. Geostand Geoanal Res. 2013;37:111–54.CAS 

    Google Scholar 
    71.Neave MJ, Rachmawati R, Xun L, Michell CT, Bourne DG, Apprill A, et al. Differential specificity between closely related corals and abundant Endozoicomonas endosymbionts across global scales. ISME J. 2017;11:186–200.PubMed 

    Google Scholar 
    72.Savary R, Barshis DJ, Voolstra CR, Cárdenas A, Evensen NR, Banc-Prandi G, et al. Fast and pervasive transcriptomic resilience and acclimation of extremely heat tolerant coral holobionts from the northern Red Sea. Proc Natl Acad Sci USA. 2021;118:e2023298118.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    73.Lesser MP, Mazel CH, Gorbunov MY, Falkowski PG. Discovery of symbiotic nitrogen-fixing cyanobacteria in corals. Science. 2004;305:997–1000.CAS 
    PubMed 

    Google Scholar 
    74.Lesser MP, Falcón LI, Rodríguez-Román A, Enríquez S, Hoegh-Guldberg O, Iglesias-Prieto R. Nitrogen fixation by symbiotic cyanobacteria provides a source of nitrogen for the scleractinian coral Montastraea cavernosa. Mar Ecol Prog Ser. 2007;346:143–52.CAS 

    Google Scholar 
    75.Tilstra A, El-Khaled YC, Roth F, Rädecker N, Pogoreutz C, Voolstra CR, et al. Denitrification aligns with N2 fixation in Red Sea corals. Sci Rep. 2019;9:19460.PubMed 
    PubMed Central 

    Google Scholar 
    76.Rädecker N, Pogoreutz C, Ziegler M, Ashok A, Barreto MM, Chaidez V, et al. Assessing the effects of iron enrichment across holobiont compartments reveals reduced microbial nitrogen fixation in the Red Sea coral Pocillopora verrucosa. Ecol Evol. 2017;7:6614–21.PubMed 
    PubMed Central 

    Google Scholar 
    77.Ainsworth TD, Fine M, Blackall LL, Hoegh-Guldberg O. Fluorescence in situ hybridization and spectral imaging of coral-associated bacterial communities. Appl Environ Microbiol. 2006;72:3016–20.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    78.van de Water JAJM, Ainsworth TD, Leggat W, Bourne DG, Willis BL, van Oppen MJH. The coral immune response facilitates protection against microbes during tissue regeneration. Mol Ecol. 2015;24:3390–404.PubMed 

    Google Scholar 
    79.Wada N, Ishimochi M, Matsui T, Pollock FJ, Tang S-L, Ainsworth TD, et al. Characterization of coral-associated microbial aggregates (CAMAs) within tissues of the coral Acropora hyacinthus. Sci Rep. 2019;9:14662.PubMed 
    PubMed Central 

    Google Scholar 
    80.Udvardi M, Poole PS. Transport and metabolism in legume-rhizobia symbioses. Annu Rev Plant Biol. 2013;64:781–805.CAS 
    PubMed 

    Google Scholar 
    81.Pernice M, Meibom A, Van Den Heuvel A, Kopp C, Domart-Coulon I, Hoegh-Guldberg O, et al. A single-cell view of ammonium assimilation in coral–dinoflagellate symbiosis. ISME J. 2012;6:1314–24.CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    82.Shashar N, Cohen Y, Loya Y, Sar N. Nitrogen fixation (acetylene reduction) in stony corals: evidence for coral-bacteria interactions. Mar Ecol Prog Ser. 1994;111:259–64.CAS 

    Google Scholar 
    83.Inomura K, Bragg J, Riemann L, Follows MJ. A quantitative model of nitrogen fixation in the presence of ammonium. PLoS ONE. 2018;13:e0208282.PubMed 
    PubMed Central 

    Google Scholar 
    84.Agawin NSR, Rabouille S, Veldhuis MJW, Servatius L, Hol S, van Overzee HMJ, et al. Competition and facilitation between unicellular nitrogen-fixing cyanobacteria and non-nitrogen-fixing phytoplankton species. Limnol Oceanogr. 2007;52:2233–48.CAS 

    Google Scholar 
    85.El-Khaled YC, Roth F, Tilstra A, Rädecker N, Karcher DB, Kürten B, et al. In situ eutrophication stimulates dinitrogen fixation, denitrification, and productivity in Red Sea coral reefs. Mar Ecol Prog Ser. 2020;645:55–66.CAS 

    Google Scholar 
    86.Fine M, Loya Y. Endolithic algae: an alternative source of photoassimilates during coral bleaching. Proc R Soc B. 2002;269:1205–10.PubMed 
    PubMed Central 

    Google Scholar 
    87.Sangsawang L, Casareto BE, Ohba H, Vu HM, Meekaew A, Suzuki T, et al. 13C and 15N assimilation and organic matter translocation by the endolithic community in the massive coral Porites lutea. R Soc Open Sci. 2017;4:171201.PubMed 
    PubMed Central 

    Google Scholar 
    88.Fine M, Roff G, Ainsworth TD, Hoegh-Guldberg O. Phototrophic microendoliths bloom during coral “white syndrome. Coral Reefs. 2006;25:577–81.
    Google Scholar 
    89.Fine M, Steindler L, Loya Y. Endolithic algae photoacclimate to increased irradiance during coral bleaching. Mar Freshw Res. 2004;55:115–21.CAS 

    Google Scholar 
    90.Meunier V, Bonnet S, Benavides M, Ravache A, Grosso O, Lambert C, et al. Diazotroph-derived nitrogen assimilation strategies differ by scleractinian coral species. Front Mar Sci. 2021;8:1018.
    Google Scholar  More

  • in

    Counting using deep learning regression gives value to ecological surveys

    DatasetsIn this study, datasets from two fundamentally different real-world ecological use cases were employed. The objects of interest in these images were manually counted in previous studies2,8,36,37, without the aim of DL applications.Microscopic images of otolith ringsThe first dataset consists of 3585 microscopic images of otoliths (i.e., hearing stones) of plaice (Pleuronectes platessa). Newly settled juvenile plaice of various length classes were collected at stations along the North Sea and Wadden Sea coast during 23 sampling campaigns conducted over 6 years. Each individual fish was measured, the sagittal otoliths were removed and microscopic images of two zoom levels ((10times 20) and (10times 10), depending on fish length) were made. Post-settlement daily growth rings outside the accessory growth centre were then counted by eye6,7. In this dataset, images of otoliths with less than 16 and more than 45 rings were scarce (Fig. 6). Therefore, a stratified random design was used to select 120 images to evaluate the model performance over the full range of ring counts: all 3585 images were grouped in eight bins according to their label (Fig. 6) and from each bin 15 images were randomly selected for the test set. Out of the remaining 3465 images, 80% of the images were randomly selected for training and 20% were used as a validation set, which is used to estimate the model performance and optimise hyperparameters during training.Figure 6Distribution of the labels (i.e., number of post-settlement rings) of all images in the otolith dataset ((n=3585)).Full size imageAerial images of sealsThe second dataset consists of 11,087 aerial images (named ‘main dataset’ from now onwards) of hauled out grey seals (Halichoerus grypus) and harbour seals (Phoca vitulina), collected between 2005 and 2019 in the Dutch part of the Wadden Sea2,36. Surveys for both species were performed multiple times each year: approximately three times during pupping season and twice during the moult8. During these periods, seals haul out on land in larger numbers. Images were taken manually through the airplane window whenever seals were sighted, while flying at a fixed height of approximately 150m, using different focal lengths (80-400mm). Due to variations in survey conditions (e.g., weather, lighting) and image composition (e.g., angle of view, distance towards seals), this main dataset is highly variable. Noisy labels further complicated the use of this dataset: seals present in multiple (partially) overlapping images were counted only once, and were therefore not included in the count label of each image. Recounting the seals on all images in this dataset to deal with these noisy labels would be a tedious task, compromising one of the main aims of this study of reducing annotation efforts. Instead, only a selection of the main dataset was recounted and used for training and testing. First, 100 images were randomly selected (and recounted) for the test set. In the main dataset, images with a high number of seals were scarce, while images with a low number of seals were abundant (Fig. 7, panel A). Therefore, as with the otoliths, all 11,087 images were grouped into 20 bins according to their label (Fig. 7, panel A), after which five images were randomly selected from each bin for the test set. Second, images of sufficient quality and containing easily identifiable were selected from the main dataset (and recounted) for training and validation, until 787 images were retained (named ‘seal subset 1’). In order to create images with zero seals (i.e., just containing the background) and to remove seals that are only partly photographed along the image borders, some of these images were cropped. The dimensions of those cropped images were preserved and, if required, the image-level annotation was modified accordingly. The resulting ‘seal subset 1’ only contains images with zero to 99 seals (Fig. 7, panel B). These 787 images were then randomly split in a training (80%) and validation set (20%). In order to still take advantage of the remaining 10,200 images from the main dataset, a two-step label refinement was performed (see the section “Dealing with noisy labels: two-step label refinement” below).Figure 7Distribution of the labels (i.e., number of seals) in (A) the seal main dataset ((n=11{,}087)), (B) ‘seal subset 1’ ((n=787)) and (C) ‘seal subset 2’ ((n=100)).Full size imageConvolutional neural networksCNNs are a particular type of artificial neural network. Similar to a biological neural network, where many neurons are connected by synapses, these models consist of a series of connected artificial neurons (i.e., nodes), grouped into layers that are applied one by one. In a CNN, each layer receives an input and produces an output by performing a convolution between the neurons (now organised into a rectangular filter) and each spatial input location and its surroundings. This convolution operator computes a dot product at each location in the input (image or previous layer’s output), encoding the correlation between the local input values and the learnable filter weights (i.e., neurons). After this convolution, an activation function is applied so that the final output of the network can represent more than just a linear combination of the inputs. Each layer performs calculations on the inputs it receives from the previous layer, before sending it to the next layer. Regular layers that ingest all previous outputs rather than a local neighbourhood are sometimes also employed at the end; these are called “fully-connected” layers. The number of layers determines the depth of the network. More layers introduce a larger number of free (learnable) parameters, as does a higher number of convolutional filters per layer or larger filter sizes. A final layer usually projects the intermediate, high-dimensional outputs into a vector of size C (the number of categories) in the case of classification, into a single number in the case of regression (ours), or into a custom number of outputs representing arbitrarily complex parameters, such as the class label and coordinates of a bounding box in the case of object detection. During training, the model is fed with many labelled examples to learn the task at hand: the parameters of the neurons are updated to minimise a loss (provided by an error function measuring the discrepancy between predictions and labels; in our case this is the Huber loss as described below). To do so, the gradient and its derivative with respect to each neuron in the last layer is computed; modifying neurons by following their gradients downwards allows reducing the loss (and thereby improving model prediction) for the current image accordingly. Since the series of layers in a CNN can be seen as a set of nested, differentiable functions, the chain rule can be applied to also compute gradients for the intermediate, hidden layers and modify neurons therein backwards until the first layer. This process is known as backpropagation38. With the recent increase of computational power and labelled dataset sizes, these models are now of increasing complexity (i.e., they have higher numbers of learnable parameters in the convolutional filters and layers).CNNs come in many layer configurations, or architectures. One of the most widely used CNN architecture is the ResNet20, which introduced the concept of residual blocks: in ResNets, the input to a residual block (i.e., a group of convolutional layers with nonlinear activations) is added to its output in an element-wise manner. This allows the block to focus on learning residual patterns on top of its inputs. Also, it enables learning signals to by-pass entire blocks, which stabilises training by avoiding the problem of vanishing gradients39. As a consequence, ResNets were the first models that could be trained even with many layers in series and provided a significant increase in accuracy.Model selection and trainingFor the otolith dataset, we employed ResNet20 architectures of various depths (i.e., ResNet18, ResNet34, ResNet50, ResNet101 and ResNet152, where the number corresponds to the number of hidden layers in the model, see Supplementary S1). These ResNet models were pretrained on ImageNet40, which is a large benchmark dataset containing millions of natural images annotated with thousands of categories. Pre-training on ImageNet is a commonly employed methodology to train a CNN efficiently, as it will already have learned how to recognise common recurring features, such as edges and basic geometrical patterns, which would have to be learned from zero otherwise. Therefore, pre-training reduces the required amount of training data significantly.Figure 8Schematic representation of the CNN used in this study. The classification output layer of the pretrained ResNet18 is replaced by two fully-connected layers. The model is trained with a Huber loss.Full size imageWe modified the ResNet architecture to perform a regression task. To do so, we replaced the classification output layer with two fully-connected layers that map to 512 neurons after the first layer and to a single continuous variable after the second layer23 (Fig. 8). Since the final task to be performed is regression, the loss function is a loss function that is tailored for regression. In our experiments we tested both a Mean Squared Error and a Smooth L1 (i.e., Huber) loss21 (see Supplementary S1). The Huber loss is more robust against outliers and is defined as follows:$$begin{aligned} {mathscr {L}}(y,{hat{y}})=frac{1}{n}sum _i^{n} z_i end{aligned}$$
    (1)
    where (z_i) is given by$$begin{aligned} z_i= {left{ begin{array}{ll} 0.5times (y_i-{hat{y}}_i)^2, &{}quad text {if } |y_i-{hat{y}}_i| More

  • in

    Linking migratory performance to breeding phenology and productivity in an Afro-Palearctic long-distance migrant

    1.Vickery, J. A. et al. The decline of Afro-Palaearctic migrants and an assessment of potential causes. Ibis (Lond. 1759) 156, 1–22 (2014).
    Google Scholar 
    2.Rosenberg, K. V. et al. Decline of the North American avifauna. Science 366, 120–124 (2019).ADS 
    CAS 
    PubMed 

    Google Scholar 
    3.Knight, S. M. et al. Constructing and evaluating a continent-wide migratory songbird network across the annual cycle. Ecol. Monogr. 88, 445–460 (2018).
    Google Scholar 
    4.Alves, J. A. et al. Costs, benefits, and fitness consequences of different migratory strategies. Ecology 94, 11–17 (2013).ADS 
    PubMed 

    Google Scholar 
    5.van Wijk, R. E., Schaub, M. & Bauer, S. Dependencies in the timing of activities weaken over the annual cycle in a long-distance migratory bird. Behav. Ecol. Sociobiol. 71, 71–73 (2017).
    Google Scholar 
    6.Donald, P. F., Sanderson, F. J., Burfield, I. J. & van Bommel, F. P. J. Further evidence of continent-wide impacts of agricultural intensification on European farmland birds, 1990–2000. Agric. Ecosyst. Environ. 116, 189–196 (2006).
    Google Scholar 
    7.Bowler, D. E., Heldbjerg, H., Fox, A. D., Jong, M. & Böhning-Gaese, K. Long-term declines of European insectivorous bird populations and potential causes. Conserv. Biol. 0, 1–11 (2019).
    Google Scholar 
    8.Harrison, X. A., Blount, J. D., Inger, R., Norris, D. R. & Bearhop, S. Carry-over effects as drivers of fitness differences in animals. J. Anim. Ecol. 80, 4–18 (2010).PubMed 

    Google Scholar 
    9.Emmenegger, T., Hahn, S. & Bauer, S. Individual migration timing of common nightingales is tuned with vegetation and prey phenology at breeding sites. BMC Ecol. 14, 1–8 (2014).
    Google Scholar 
    10.Morrison, C. A., Alves, J. A., Gunnarsson, T. G., Þórisson, B. & Gill, J. A. Why do earlier-arriving migratory birds have better breeding success?. Ecol. Evol. 9, 8856–8864 (2019).PubMed 
    PubMed Central 

    Google Scholar 
    11.Cooper, N. W., Murphy, M. T., Redmond, L. J. & Dolan, A. C. Reproductive correlates of spring arrival date in the Eastern Kingbird Tyrannus tyrannus. J. Ornithol. 152, 143–152 (2011).
    Google Scholar 
    12.Nilsson, C., Klaassen, R. H. G. & Alerstam, T. Differences in speed and duration of bird migration between spring and autumn. Am. Nat. 181, 837–845 (2013).PubMed 

    Google Scholar 
    13.Gow, E. A. et al. Effects of spring migration distance on tree swallow reproductive success within and among flyways. Front. Ecol. Evol. 7, 380 (2019).ADS 

    Google Scholar 
    14.Saino, N. et al. Sex-dependent carry-over effects on timing of reproduction and fecundity of a migratory bird. J. Anim. Ecol. 86, 239–249 (2017).PubMed 

    Google Scholar 
    15.Briedis, M., Hahn, S. & Adamík, P. Cold spell en route delays spring arrival and decreases apparent survival in a long-distance migratory songbird. BMC Ecol. 17, 1–8 (2017).
    Google Scholar 
    16.McKinnon, E. A., Macdonald, C. M., Gilchrist, H. G. & Love, O. P. Spring and fall migration phenology of an arctic-breeding passerine. J. Ornithol. 157, 681–693 (2016).
    Google Scholar 
    17.Woodworth, B. K. et al. Differential migration and the link between winter latitude, timing of migration, and breeding in a songbird. Oecologia 181, 413–422 (2016).ADS 
    PubMed 

    Google Scholar 
    18.Saino, N. et al. Ecological conditions during winter predict arrival date at the breeding quarters in a trans-Saharan migratory bird. Ecol. Lett. 7, 21–25 (2004).
    Google Scholar 
    19.Norris, D. R., Marra, P. P., Kyser, T. K., Sherry, T. W. & Ratcliffe, L. M. Tropical winter habitat limits reproductive success on the temperate breeding grounds in a migratory bird. Proc. R. Soc. B Biol. Sci. 271, 59–64 (2004).
    Google Scholar 
    20.Bearhop, S., Hilton, G. M., Votier, S. C. & Waldron, S. Stable isotope ratios indicate that body condition in migrating passerines is influenced by winter habitat. Proc. R. Soc. London B Biol. Sci. 271, S215–S218 (2004).
    Google Scholar 
    21.Ockendon, N., Leech, D. & Pearce-Higgins, J. W. Climatic effects on breeding grounds are more important drivers of breeding phenology in migrant birds than carry- over effects from wintering grounds. Biol. Lett. 9 (2013).22.Arbeiter, S., Schulze, M., Tamm, P. & Hahn, S. Strong cascading effect of weather conditions on prey availability and annual breeding performance in European bee-eaters Merops apiaster. J. Ornithol. 157, 155–163 (2016).
    Google Scholar 
    23.Harrison, X. A. et al. Environmental conditions during breeding modify the strength of mass-dependent carry-over effects in a migratory bird. PLoS ONE 8, e77783 (2013).ADS 
    CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    24.Swift, R. J., Rodewald, A. D., Johnson, J. A., Andres, B. A. & Senner, N. R. Seasonal survival and reversible state effects in a long-distance migratory shorebird. J. Anim. Ecol. 89, 2043–2055 (2020).PubMed 

    Google Scholar 
    25.Brust, V., Bastian, H. V., Bastian, A. & Schmoll, T. Determinants of between-year burrow re-occupation in a colony of the European bee-eater Merops apiaster. Ecol. Evol. 5, 3223–3230 (2015).PubMed 
    PubMed Central 

    Google Scholar 
    26.Lessells, C. M. & Krebs, J. R. Age and breeding performance of European bee-eaters. Auk 106, 375–382 (1989).
    Google Scholar 
    27.Pârâu, L. G. et al. Dynamics in numbers of group-roosting individuals in relation to pair-sleeping occurrence and onset of egg-laying in European Bee-eaters Merops apiaster. J. Ornithol. 158, 1119–1122 (2017).
    Google Scholar 
    28.Hoi, H., Darolová, A., Krištofík, J. & Hoi, C. The effect of the ectoparasite Carnus hemapterus on immune defence, condition, and health of nestling European Bee-eaters. J. Ornithol. 159, 291–302 (2018).
    Google Scholar 
    29.Kapun, M., Darolová, A., Krištofik, J., Mahr, K. & Hoi, H. Distinct colour morphs in nestling European Bee-eaters Merops apiaster: Is there an adaptive value?. J. Ornithol. 152, 1001–1005 (2011).
    Google Scholar 
    30.Lessells, C. M. & Avery, M. I. Hatching asynchrony in european bee-eaters merops apiaster. J. Anim. Ecol. 58, 815–835 (1989).
    Google Scholar 
    31.Arbeiter, S., Schulze, M., Todte, I. & Hahn, S. Das Zugverhalten und die Ausbreitung von in Sachsen-Anhalt brütenden Bienenfressern (Merops apiaster). Berichte der Vogelwarte Hiddensee 21, 33–40 (2012).
    Google Scholar 
    32.Dhanjal-Adams, K. L. et al. Spatiotemporal group dynamics in a long-distance migratory bird. Curr. Biol. 28, 2824-2830.e3 (2018).CAS 
    PubMed 

    Google Scholar 
    33.Hahn, S. et al. Range-wide migration corridors and non-breeding areas of a northward expanding Afro-Palaearctic migrant, the European Bee-eater Merops apiaster. Ibis (Lond. 1859) 162, 345–355 (2019).
    Google Scholar 
    34.Fry, C. H. The bee-eaters. (T & A D Polyser Ltd, 1984).35.Ramos, R. et al. Population genetic structure and long-distance dispersal of a recently expanding migratory bird. Mol. Phylogenet. Evol. 99, 194–203 (2016).PubMed 

    Google Scholar 
    36.Jacobsen, L. B. et al. Annual spatiotemporal migration schedules in three larger insectivorous birds: European nightjar, common swift and common cuckoo. Anim. Biotelem1 5, 1–11 (2017).
    Google Scholar 
    37.Åkesson, S., Klaassen, R., Holmgren, J., Fox, J. W. & Hedenström, A. Migration routes and strategies in a highly aerial migrant, the common Swift Apus apus, revealed by light-level. Geolocators. PLoS One 7, e41195 (2012).ADS 
    PubMed 

    Google Scholar 
    38.Carneiro, C., Gunnarsson, T. G. & Alves, J. A. Faster migration in autumn than in spring: seasonal migration patterns and non-breeding distribution of Icelandic Whimbrels Numenius phaeopus islandicus. J. Avian Biol. 50 (2019).39.Sapir, N. et al. Migration by soaring or flapping: numerical atmospheric simulations reveal that turbulence kinetic energy dictates bee-eater flight mode. Proc. R. Soc. B Biol. Sci. 278, 3380–3386 (2011).
    Google Scholar 
    40.Lemke, H. W. et al. Annual cycle and migration strategies of a Trans-Saharan migratory songbird: a geolocator study in the great reed warbler. PLoS ONE 8, e79209 (2013).ADS 
    CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    41.Briedis, M. et al. A full annual perspective on sex-biased migration timing in long-distance migratory birds. Proc. R. Soc. B Biol. Sci. 286, 20182821 (2019).
    Google Scholar 
    42.Fransson, T. Timing and speed of migration in North and West European populations of Sylvia warblers. J. Avian Biol. 26, 39–48 (1995).
    Google Scholar 
    43.Briedis, M., Hahn, S., Krist, M. & Adamík, P. Finish with a sprint: evidence for time-selected last leg of migration in a long-distance migratory songbird. Ecol. Evol. 8, 6899–6908 (2018).PubMed 
    PubMed Central 

    Google Scholar 
    44.Alerstam, T. Strategies for the transition to breeding in time-selected bird migration. Ardea 94, 347–357 (2006).
    Google Scholar 
    45.Arizaga, J., Willemoes, M., Unamuno, E., Unamuno, J. M. & Thorup, K. Following year-round movements in Barn Swallows using geolocators: could breeding pairs remain together during the winter?. Bird Study 62, 141–145 (2015).
    Google Scholar 
    46.Tøttrup, A. P. et al. Drought in Africa caused delayed arrival of European songbirds. Science 338, 1307 (2012).ADS 
    PubMed 

    Google Scholar 
    47.Smith, R. J. & Moore, F. R. Arrival timing and seasonal reproductive performance in a long-distance migratory landbird. Behav. Ecol. Sociobiol. 57, 231–239 (2005).
    Google Scholar 
    48.IPMA. Climate bulletin, June 2017, Portugal. http://www.ipma.pt/resources.www/docs/im.publicacoes/edicoes.online/20170719/bXUzZOgrqXmTjnUVRtro/cli_20170601_20170630_pcl_mm_co_pt.pdf (2017).49.Kearney, M., Shine, R. & Porter, W. P. The potential for behavioral thermoregulation to buffer ‘cold-blooded’ animals against climate warming. Proc. Natl. Acad. Sci. U.S.A. 106, 3835–3840 (2009).ADS 
    CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    50.Cunningham, S. J., Martin, R. O., Hojem, C. L. & Hockey, P. A. R. Temperatures in excess of critical thresholds threaten nestling growth and survival in a rapidly-warming Arid Savanna: a study of common fiscals. PLoS ONE 8 (2013).51.Cruz-Mcdonnell, K. K. & Wolf, B. O. Rapid warming and drought negatively impact population size and reproductive dynamics of an avian predator in the arid southwest. Glob. Chang. Biol. 22, 237–253 (2016).ADS 
    PubMed 

    Google Scholar 
    52.Shukla, P. R. et al. Technical summary. IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems (2019).53.Persson, C. Age structure, sex ratios and survival rates in a south Swedish Sand martin (Riparia riparia) population, 1964 to 1984. J. Zool. 1, 639–670 (1987).
    Google Scholar 
    54.Costa, J. S., Rocha, A. D., Correia, R. A. & Alves, J. A. Developing and validating a nestling photographic aging guide for cavity-nesting birds: an example with the European Bee-eater (Merops apiaster). Avian Res. 11, 1–8 (2020).
    Google Scholar 
    55.Lisovski, S., Wotherspoon, S. & Sumner, M. TwGeos: Basic data processing for light-level geolocation archival tags. R package version 0.1.2. (2016). 56.Lisovski, S. et al. Geolocation by light: accuracy and precision affected by environmental factors. Methods Ecol. Evol. 3, 603–612 (2012).
    Google Scholar 
    57.Wotherspoon, S., Sumner, M. & Lisovski, S. R package SGAT: solar/satellite geolocation for animal tracking (2016).58.Lisovski, S. et al. Light-level geolocator analyses: a user’s guide. J. Anim. Ecol. 89, 221–236 (2019).PubMed 

    Google Scholar 
    59.Lisovski, S. & Hahn, S. GeoLight—processing and analysing light-based geolocator data in R. Methods Ecol. Evol. 3, 1055–1059 (2012).
    Google Scholar 
    60.Mazerolle, M. J. AICcmodavg: model selection and multimodel inference based on (Q)AIC(c). R package version 2.2–2. (2019).61.Team, R. C. R: a language and environment for statistical computing. (2017). More

  • in

    More than skin deep

    Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
    the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
    Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
    and JavaScript. More