Philippa Kaur

1.
van Klink, R. et al. Science 368, 417–420 (2020).
PubMed  Google Scholar 
2.
Goulson, D. J. Appl. Ecol. 50, 977–987 (2013).
Article  Google Scholar 

3.
Goulson, D., Nicholls, E., Botías, C. & Rotheray, E. L. Science. 347, 1435–1445 (2015).
CAS  Article  Google Scholar 

4.
Siviter, H., Brown, M. J. F. & Leadbeater, E. Nature 561, 109–112 (2018).
CAS  Article  Google Scholar 

5.
Hendrichs, J., Kenmore, P., Robinson, A. S. & Vreysen, M. J. B. in Area-Wide Control of Insect Pests (eds Vreysen, M. J. B. et al.) 3–33 (Springer, 2007).

6.
Boyd, I. L. Nat. Ecol. Evol. 2, 920–921 (2018).
Article  Google Scholar 

7.
Goulson, D., Thompson, J. & Croombs, A. PeerJ 6, e5255 (2018).
Article  Google Scholar 

8.
Sánchez-Bayo, F. Science 346, 806–807 (2014).
Article  Google Scholar 

9.
PUS STATS (Fera Science Limited, accessed 12 May 2020); https://secure.fera.defra.gov.uk/pusstats/index.cfm

10.
Krieger, R. (ed.) Hayes’ Handbook of Pesticide Toxicology 2nd edn, Vol. 1 (Academic Press, 2001).

11.
Desneux, N., Decourtye, A. & Delpuech, J. Annu. Rev. Entomol. 52, 81–106 (2007).
CAS  Article  Google Scholar 

12.
Rundlöf, M. et al. Nature 521, 77–80 (2015).
Article  Google Scholar 

13.
Woodcock, B. A. et al. Science 356, 1393–1395 (2017).
CAS  Article  Google Scholar 

14.
Wang, B., Gao, R., Mastro, V. C. & Reardon, R. C. J. Econ. Entomol. 98, 2292–2300 (2005).
CAS  Article  Google Scholar 

15.
Poland, T. M. et al. J. Econ. Entomol. 99, 383–392 (2009).
Article  Google Scholar 

16.
He, Y. et al. Int. J. Biol. Sci. 9, 246–255 (2013).
Article  Google Scholar 

17.
Michaelides, P. K. & Wright, D. J. Crop Prot. 16, 431–438 (1997).
CAS  Article  Google Scholar 

18.
Pedigo, L. Annu. Rev. Entomol. 31, 341–368 (1986).
Article  Google Scholar 

19.
Allee, W. C. The Social Life of Animals (W.W. Norton & Company, 1938).

20.
Berec, L., Angulo, E. & Courchamp, F. Trends Ecol. Evol. 22, 185–191 (2007).
Article  Google Scholar 

21.
Zubrod, J. P. et al. Environ. Sci. Technol. 53, 3347–3365 (2019).
CAS  Article  Google Scholar 

22.
Clements, J. et al. Sci. Rep. 8, 13282 (2018).
Article  Google Scholar 

23.
Rohr, J. R., Kerby, J. L. & Sih, A. Trends Ecol. Evol. 21, 606–613 (2006).
Article  Google Scholar 

24.
Nathan, C. Nat. Rev. Microbiol. 18, 259–260 (2020).
CAS  Article  Google Scholar 

25.
Whalon, M. E., Mota-Sanchez, D. & Hollingworth, R. M. Global Pesticide Resistance in Arthropods (CABI, 2008); https://doi.org/10.1079/9781845933531.0000

26.
Alkassab, A. T. & Kirchner, W. H. Ecotoxicology 25, 1000–1010 (2016).
CAS  Article  Google Scholar 

27.
Suchail, S., Guez, D. & Belzunces, L. P. Environ. Toxicol. Chem. 19, 1901–1905 (2000).
CAS  Article  Google Scholar 

28.
Chahbar, N., Chahbar, M. & Doumandji, S. Int. J. Zool. Res. 4, 29–40 (2014).
Google Scholar 

29.
Wu-Smart, J. & Spivak, M. Environ. Entomol. 47, 55–62 (2018).
CAS  Article  Google Scholar 

30.
Basley, K., Davenport, B., Vogiatzis, K. & Goulson, D. PeerJ 6, e4258 (2018).
Article  Google Scholar 

31.
Tomé, H. V. V., Martins, G. F., Lima, M. A. P., Campos, L. A. O. & Guedes, R. N. C. PLoS ONE 7, e38406 (2012).
Article  Google Scholar 

32.
Horowitz, A. R., Mendelson, Z., Weintraub, P. G. & Ishaaya, I. Bull. Entomol. Res. 88, 437–442 (1998).
CAS  Article  Google Scholar 

33.
Hunter White, W. et al. J. Econ. Entomol. 100, 155–163 (2007).
Article  Google Scholar 

34.
Wasserberg, G. et al. J. Vector Ecol. 36, S148–S156 (2011).
Article  Google Scholar 

35.
Ramakrishnan, R., Suiter, D. R., Nakatsu, C. H. & Bennett, G. W. J. Econ. Entomol. 93, 422–428 (2000).
CAS  Article  Google Scholar 

36.
Klein, S., Cabirol, A., Devaud, J. M., Barron, A. B. & Lihoreau, M. Trends Ecol. Evol. 32, 268–278 (2017).
Article  Google Scholar  More

To expand on previous BONCAT findings, we initially performed an experiment with E. coli, which demonstrated that translational activity as measured by BONCAT can only be detected when cultures were grown with glucose or sorbitol as the sole carbon and energy source but not with sucrose (Supplementary Fig. 1). This is consistent with the notion that E. coli is genetically incapable of sucrose utilization. Addition of chloramphenicol, an antibiotic targeting ribosome function, to the culture decreased the BONCAT fluorescence intensity to background levels, i.e., fluorescence values observed when cells were grown with HPG but without growth substrate. These results demonstrated that BONCAT can be used to study cell activity responses to substrate amendment and suggested that it could be used to study complex microbiomes.
For this benchmark study, we selected a high temperature (74 °C), alkaline (pH 8.2) hot spring (Five Sisters 5, FS5) in the Lower Geyser Basin of YNP (Fig. 1). Hot springs harbor lower complexity microbial communities compared to other environments, such as soils and marine sediments, making them an ideal system for the development of novel approaches. As indicated by 16S rRNA gene amplicon sequencing, hot spring FS5 was dominated by the archaeal candidate phylum Aigarchaeota (41.9%), with the next most abundant community members belonging to bacterial candidate phylum Fervidibacteria (9.57%) and phylum Deinococcus-Thermus (7.98%) (Fig. 2). Members of the Aigarchaeota have been detected in many high temperature (65–88 °C) hot springs over a wide pH range (2.9–9.3) [25]. Despite multiple studies describing their versatile metabolic potential, Aigarchaeota remain recalcitrant to cultivation and no experimental data on their functional activity are currently available. Reconstructed genomes suggest the capacity for aerobic respiration or anaerobic respiration with nitrate as an electron acceptor [25]. In addition, a recent study found multiple endoglucanases and β-glucosidases that might be involved with degradation of cellulose and cellobiose in an Aigarchaeota metagenome bin [26]. One of the goals of our study was to test these functional predictions using BONCAT-FACS.
Fig. 2: 16S rRNA gene relative abundances in averaged incubations.

Top panel, unamended FS5 community at the time of sampling (T0). Middle panel, community composition of presort samples, representing the extractable microbial community, after the incubation experiments. Bottom panel, composition of the sorted active cell fraction after incubation with substrate amendment. Only taxa that were represented above 1% in at least one sample are shown. Additional taxa were combined into “Other” category. Taxon level represents highest taxonomic resolution for each taxa (Silva 128). Some replicates did not pass the quality control steps of our bioinformatics pipeline (1% relative abundance of the total extractable community, ASVs collapsed to the genus level) were active in at least one condition, and most taxa were active under several conditions (Fig. 2). This demonstrates that BONCAT can be applied to a wide variety of phylogenetic groups, which is consistent with previous reports [5,6,7,8,9, 27]. The sorted, active fraction from incubations with different treatments contained no statistically significant difference when compared to the HPG-only control based on Bray–Curtis dissimilarity (MANOVA, p = 1) (Fig. 2). This result indicates that the microbial activity response to any treatment, as captured by BONCAT, was not large enough or consistent enough among replicates to significantly change the overall active community when compared to the HPG-only control.
Although the overall composition of the active communities did not vary significantly in response to incubation conditions, changes within the richness and evenness of the communities were detected. To further describe the community composition, the Shannon’s diversity index of each incubation was calculated (Fig. 3). For each treatment, the bulk, presort fraction (representing the total, extractable cell community), and the sorted, active cell fraction were analyzed and compared to the respective HPG-only control. Overall, the variation in the presort fraction of Shannon’s diversity indices was less variable than the sorted fraction from the same treatment incubation (Fig. 3). This demonstrated that the BONCAT-FACS approach could detect changes in the diversity of the active community due to individual cell responses prior to shifts in cell abundance occurring in presort communities. We expected little variation in the presort community because the limited time of incubation should not have allowed for an overall shift in community composition on a bulk level. However, Fervidibacteria were observed in higher proportion in most presort populations as compared to the original T0 bulk sample. This could be attributed to either favorable growth of this yet uncultured lineage or be a result of preferential cell extraction or cell lysis during freeze–thaw cycles as compared to bulk sample DNA extraction. Alternatively, their increase in relative abundance could have resulted from sample cooling (72 to 55 °C) during transit from the field to the laboratory (4 h).
Fig. 3: Change in Shannon’s diversity indices of each treatment standardized to HPG-only controls.

The Shannon’s diversity indices for presort and sorted samples were compared relative to their respective HPG-only controls. The value for HPG-only was set to 0 and the difference for each sample is plotted. Overall, presort samples exhibited less variability than sorted samples. Sorted samples for cellobiose and anoxic conditions (100% N2 headspace) were significantly different from the sorted HPG-only (atmospheric air) control (p  More

1.
Barbier, E. B. et al. The value of estuarine and coastal ecosystem services. Ecol. Monogr. 81, 169–193 (2011).
Google Scholar 
2.
Brander, L. M. et al. Ecosystem service values for mangroves in Southeast Asia: A meta-analysis and value transfer application. Ecosyst. Serv. 1, 62–69 (2012).
Google Scholar 

3.
UNEP. The Importance of Mangroves to People: A Call to Action (United Nations Environment Programme World Conservation Monitoring Centre, Cambridge, 2014).
Google Scholar 

4.
Spalding, M. & Parrett, C. L. Global patterns in mangrove recreation and tourism. Mar. Policy 110, 103540 (2019).
Google Scholar 

5.
Valiela, I., Bowen, J. L. & York, J. K. Mangrove forests: One of the world’s threatened major tropical environments. Bioscience 51, 807–815 (2001).
Google Scholar 

6.
Richards, D. R. & Friess, D. A. Rates and drivers of mangrove deforestation in Southeast Asia, 2000–2012. Proc. Natl. Acad. Sci. 113, 344–349 (2016).
ADS  CAS  PubMed  Google Scholar 

7.
Sloan, S. & Sayer, J. A. Forest Resources Assessment of 2015 shows positive global trends but forest loss and degradation persist in poor tropical countries. For. Ecol. Manag. 352, 134–145 (2015).
Google Scholar 

8.
de Groot, R. et al. Global estimates of the value of ecosystems and their services in monetary units. Ecosyst. Serv. 1, 50–61 (2012).
Google Scholar 

9.
Himes-Cornell, A., Pendleton, L. & Atiyah, P. Valuing ecosystem services from blue forests: A systematic review of the valuation of salt marshes, sea grass beds and mangrove forests. Ecosyst. Serv. 30, 36–48 (2018).
Google Scholar 

10.
Simard, M. et al. Mangrove canopy height globally related to precipitation, temperature and cyclone frequency. Nat. Geosci. 12, 40–45 (2019).
ADS  CAS  Google Scholar 

11.
Spalding, M. D., Kainumu, M. & Collins, L. World Atlas of Mangroves (Earthscan, London, 2010).
Google Scholar 

12.
Ewel, K. C., Twilley, R. R. & Ong, J. Different kinds of mangrove forests provide different goods and services. Glob. Ecol. Biogeogr. Lett. 7, 83–94 (1998).
Google Scholar 

13.
Twilley, R. R., Rovai, A. S. & Riul, P. Coastal morphology explains global blue carbon distributions. Front. Ecol. Environ. 16, 503–508 (2018).
Google Scholar 

14.
Sanderman, J. et al. A global map of mangrove forest soil carbon at 30 m spatial resolution. Environ. Res. Lett. 13, 055002 (2018).
ADS  Google Scholar 

15.
Rovai, A. S. et al. Global controls on carbon storage in mangrove soils. Nat. Clim. Chang. 8, 534–538 (2018).
ADS  CAS  Google Scholar 

16.
Donato, D. C. et al. Mangroves among the most carbon-rich forests in the tropics. Nat. Geosci. 4, 293–297 (2011).
ADS  CAS  Google Scholar 

17.
Koch, E. W. et al. Non-linearity in ecosystem services: Temporal and spatial variability in coastal protection. Front. Ecol. Environ. 7, 29–37 (2009).
Google Scholar 

18.
Baker, R., Sheaves, M. & Johnston, R. Geographic variation in mangrove flooding and accessibility for fishes and nektonic crustaceans. Hydrobiologia 762, 1–14 (2015).
CAS  Google Scholar 

19.
Ward, R. D., Friess, D. A., Day, R. H. & Mackenzie, R. A. Impacts of climate change on mangrove ecosystems: A region by region overview. Ecosyst. Heal. Sustain. 2, e01211 (2016).
Google Scholar 

20.
Balke, T. & Friess, D. A. Geomorphic knowledge for mangrove restoration: A pan-tropical categorization. Earth Surf. Process. Landforms 41, 231–239 (2016).
ADS  Google Scholar 

21.
Spalding, M. D., Brumbaugh, R. D. & Landis, E. Atlas of Ocean Wealth (The Nature Conservancy, Arlington, 2016).
Google Scholar 

22.
Spalding, M., Blasco, F. & Field, C. World Mangrove Atlas (The International Society for Mangrove Ecosystems, Okinawa, 1997).
Google Scholar 

23.
Giri, C. et al. Status and distribution of mangrove forests of the world using earth observation satellite data. Glob. Ecol. Biogeogr. 20, 154–159 (2011).
Google Scholar 

24.
Mahoney, P. C. & Bishop, M. J. Are geomorphological typologies for estuaries also useful for classifying their ecosystems?. Aquat. Conserv. Mar. Freshw. Ecosyst. 28, 1200–1208 (2018).
Google Scholar 

25.
Bunting, P. et al. The Global Mangrove Watch—A new 2010 global baseline of mangrove extent. Remote Sens. 10, 1669 (2018).
ADS  Google Scholar 

26.
Thom, B. G. Coastal landforms and geomorphic processes. In The Mangrove Ecosystem: Research Methods (eds Snedaker, S. C. & Snedaker, J. G.) 18–35 (UNESCO, Paris, 1984).
Google Scholar 

27.
Woodroffe, C. Mangrove sediments and geomorphology. In Tropical Mangrove Ecosystems (eds Robertson, A. I. & Alongi, D. M.) 7–41 (American Geophysical Union, Washington, 1992).
Google Scholar 

28.
Twilley, R. R. & Rivera-Monroy, V. H. Ecogeomorphic models of nutrient biogeochemistry for mangrove wetlands. In Coastal Wetlands: An Integrated Ecosystem Approach (eds Perillo, G. M. E. et al.) 641–684 (Elsevier, New York, 2009).
Google Scholar 

29.
Woodroffe, C. D. et al. Mangrove sedimentation and response to relative sea-level rise. Ann. Rev. Mar. Sci. 8, 243–266 (2016).
CAS  PubMed  Google Scholar 

30.
Reed, D. J., Davidson-Arnott, R. & Perillo, G. M. Estuaries, coastal marshes, tidal flats and coastal dunes. In Geomorphology and Global Environmental Change (eds Slaymaker, O. et al.) 130–157 (Cambridge University Press, Cambridge, 2009).
Google Scholar 

31.
Walsh, J. P. & Nittrouer, C. A. Mangrove-bank sedimentation in a mesotidal environment with large sediment supply, Gulf of Papua. Mar. Geol. 208, 225–248 (2004).
ADS  CAS  Google Scholar 

32.
Swales, A., Bentley, S. J. & Lovelock, C. E. Mangrove-forest evolution in a sediment-rich estuarine system: Opportunists or agents of geomorphic change?. Earth Surf. Process. Landforms 40, 1672–1687 (2015).
ADS  Google Scholar 

33.
Proisy, C. et al. Mud bank colonization by opportunistic mangroves: A case study from French Guiana using lidar data. Cont. Shelf Res. 29, 632–641 (2009).
ADS  Google Scholar 

34.
Nascimento, W. R., Souza-Filho, P. W. M., Proisy, C., Lucas, R. M. & Rosenqvist, A. Mapping changes in the largest continuous Amazonian mangrove belt using object-based classification of multisensor satellite imagery. Estuar. Coast. Shelf Sci. 117, 83–93 (2013).
ADS  Google Scholar 

35.
Murray, N. J. et al. The global distribution and trajectory of tidal flats. Nature 565, 222–225 (2019).
ADS  CAS  PubMed  Google Scholar 

36.
McKee, K. L. Biophysical controls on accretion and elevation change in Caribbean mangrove ecosystems. Estuar. Coast. Shelf Sci. 91, 475–483 (2011).
ADS  Google Scholar 

37.
McKee, K. L. & Vervaeke, W. C. W. C. Impacts of human disturbance on soil erosion potential and habitat stability of mangrove-dominated islands in the Pelican Cays and Twin Cays ranges, Belize. Smithson. Contrib. Mar. Sci. 38, 415–427 (2011).
Google Scholar 

38.
Worthington, T. & Spalding, M. Mangrove Restoration Potential: A Global Map Highlighting a Critical Opportunity. https://doi.org/10.17863/CAM.39153 (2018).
Article  Google Scholar 

39.
Kjerfve, B. et al. Morphodynamics of muddy environments along the Atlantic coasts of North and South America. In Muddy Coasts of the World: Processes, Deposits and Function (eds Healy, T. et al.) 479–532 (Elsevier, New York, 2002).
Google Scholar 

40.
Adame, M. F. et al. Mangroves in arid regions: Ecology, threats, and opportunities. Estuarine Coast. Shelf Sci. https://doi.org/10.1016/j.ecss.2020.106796 (2020).
Article  Google Scholar 

41.
Mahapatro, D., Panigrahy, R. C. & Panda, S. Coastal lagoon: Present status and future challenges. Int. J. Mar. Sci. 3, 178–186 (2013).
Google Scholar 

42.
Gönenç, I. E. & Wolflin, J. P. Introduction. In Coastal Lagoons: Ecosystem Processes and Modeling for Sustainable Use and Development (eds. Wolflin, J. P. & Gönenç, I. E.) 1–6 (CRC Press, London, 2005).
Google Scholar 

43.
Ericson, J. P., Vörösmarty, C. J., Dingman, S. L., Ward, L. G. & Meybeck, M. Effective sea-level rise and deltas: Causes of change and human dimension implications. Glob. Planet. Change 50, 63–82 (2006).
ADS  Google Scholar 

44.
Syvitski, J. P. M. & Saito, Y. Morphodynamics of deltas under the influence of humans. Glob. Planet. Change 57, 261–282 (2007).
ADS  Google Scholar 

45.
Tessler, Z. D. et al. Profiling risk and sustainability in coastal deltas of the world. Science 349, 638–643 (2015).
ADS  CAS  PubMed  Google Scholar 

46.
Syvitski, J. P. M. et al. Sinking deltas due to human activities. Nat. Geosci. 2, 681–686 (2009).
ADS  CAS  Google Scholar 

47.
Kovacs, J. M., Wang, J. & Blanco-Correa, M. Mapping disturbances in a mangrove forest using multi-date landsat TM imagery. Environ. Manage. 27, 763–776 (2001).
CAS  PubMed  Google Scholar 

48.
Cahoon, D. R. et al. Mass tree mortality leads to mangrove peat collapse at Bay Islands, Honduras after Hurricane Mitch. J. Ecol. 91, 1093–1105 (2003).
Google Scholar 

49.
Lovelock, C. E. et al. The vulnerability of Indo-Pacific mangrove forests to sea-level rise. Nature 526, 559–563 (2015).
ADS  CAS  PubMed  Google Scholar 

50.
Wigand, C. et al. Varying inundation regimes differentially affect natural and sand-amended marsh sediments. PLoS ONE 11, e0164956 (2016).
CAS  PubMed  PubMed Central  Google Scholar 

51.
Lewis, R. R. et al. Stress in mangrove forests: Early detection and preemptive rehabilitation are essential for future successful worldwide mangrove forest management. Mar. Pollut. Bull. 109, 764–771 (2016).
CAS  PubMed  Google Scholar 

52.
Lagomasino, D. et al. Measuring mangrove carbon loss and gain in deltas. Environ. Res. Lett. 14, 025002 (2018).
ADS  Google Scholar 

53.
Goldberg, L., Lagomasino, D., Thomas, N. & Fatoyinbo, T. Global declines in human-driven mangrove loss. Glob. Chang. Biol. https://doi.org/10.1111/gcb.15275 (2020).
Article  PubMed  Google Scholar 

54.
Lacerda, L. D., Borges, R. & Ferreira, A. C. Neotropical mangroves: Conservation and sustainable use in a scenario of global climate change. Aquat. Conserv. Mar. Freshw. Ecosyst. 29, 1347–1364 (2019).
Google Scholar 

55.
Bhargava, R., Sarkar, D. & Friess, D. A. A cloud computing-based approach to mapping mangrove erosion and progradation: Case studies from the Sundarbans and French Guiana. Estuar. Coast. Shelf Sci. 12, 106798. https://doi.org/10.1016/j.ecss.2020.106798 (2020).
Article  Google Scholar 

56.
Hutchison, J., Manica, A., Swetnam, R., Balmford, A. & Spalding, M. Predicting global patterns in mangrove forest biomass. Conserv. Lett. 7, 233–240 (2014).
Google Scholar 

57.
Castañeda-Moya, E. et al. Patterns of root dynamics in mangrove forests along environmental gradients in the Florida Coastal Everglades, USA. Ecosystems 14, 1178–1195 (2011).
Google Scholar 

58.
Twilley, R. R., Rivera-Monroy, V. H., Chen, R. & Botero, L. Adapting an ecological mangrove model to simulate trajectories in restoration ecology. Mar. Pollut. Bull. 37, 404–419 (1998).
CAS  Google Scholar 

59.
Huh, O. K., Coleman, J. M., Braud, D. & Kiage, L. World Deltas Database. Appendix A. The Major River Deltas Of The World. Report. (2004).

60.
Coleman, J. M. & Huh, O. K. Major World Deltas: A Perspective From Space (2003).

61.
Domisch, S., Amatulli, G. & Jetz, W. Near-global freshwater-specific environmental variables for biodiversity analyses in 1 km resolution. Sci. Data 2, 150073 (2015).
CAS  PubMed  PubMed Central  Google Scholar 

62.
Liaw, A. & Wiener, M. Classification and regression by randomForest. R News 2, 18–22 (2002).
Google Scholar 

63.
R Core Team. R: A Language and Environment for Statistical Computing (2019).

64.
Dürr, H. H. et al. Worldwide typology of nearshore coastal systems: Defining the estuarine filter of river inputs to the oceans. Estuaries Coasts 34, 441–458 (2011).
Google Scholar 

65.
Simard, M. et al. Global mangrove aboveground biomass, maximum and basal area weighted canopy heights. https://doi.org/10.3334/ORNLDAAC/1665. (2019).

66.
Pinheiro, J., Bates, D., DebRoy, S., Sarkar, S. & Team, R. C. nlme: Linear and Nonlinear Mixed Effects Models (2019). https://cran.r-project.org/package=nlme.

67.
Zuur, A. F., Ieno, E. N. & Elphick, C. S. A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol. 1, 3–14 (2010).
Google Scholar 

68.
Zuur, A. F., Saveliev, A. A. & Ieno, E. N. A Beginner’s Guide to Generalised Additive Mixed Models with R (Highland Statistics Ltd., Newburgh, 2014).
Google Scholar 

69.
Zuur, A. F., Ieno, E. N., Walker, N., Saveliev, A. A. & Smith, G. M. Mixed Effects Models and Extensions in Ecology with R (Springer, New York, 2009).
Google Scholar 

70.
Lenth, R. emmeans: Estimated Marginal Means, aka Least-Squares Means (2019).

71.
Mangiafico, S. rcompanion: Functions to Support Extension Education Program Evaluation. R package version 2.3.25 (2020). More

Effect of temperature corrections
Figure 1 illustrates the effect of the two adjustments described above on a calculation of annual global ocean-atmosphere fluxes for this period, with calculations starting from the SOCAT v2019 database. To interpolate the SOCAT surface water fCO2 data in space and time we adopt as our standard method the two-step neural network approach described by Landschützer et al.8,18, (see also description below and Methods section). The interpolation was applied to the SOCAT data without modification, after adjusting the data to a subskin temperature and regridding (as described in refs. 12,19, see also Methods section) then additionally after repeating the flux calculation assuming a ΔT across the cool skin of 0.17 K15 salinity increase of 0.1 unit11 and the conservative “rapid transport” scheme of Woolf et al.11 (see Methods section). Each adjustment increases the calculated flux by ~0.4 PgC yr−1 when integrated over the global ocean. For the period ~2000, this approximately doubles the calculated flux into the ocean. Over the 27 years 1992–2018 inclusive, the cumulative uptake is increased from 43 to 67 PgC.
Fig. 1: Effect of near-surface temperature corrections.

Global air–sea flux calculated by interpolating SOCAT gridded data using a neural network technique8, followed by the gas exchange equation applied to the ocean mass boundary layer. The net flux into the ocean is shown as negative, following convention. The uncorrected curve uses the SOCAT fCO2 at inlet temperature as usually done. Correction of the data to a satellite-derived subskin temperature is shown, and the additional change in flux due to a thermal skin assumed to be cooler and saltier than the subskin by 0.17 K15 and 0.1 salinity units11. Excludes the Arctic and some regional seas—ocean regions included are shown in Supplementary Fig. 2.

Full size image

Uncertainty estimates
Ocean-atmosphere fluxes calculated using the gas exchange equation are subject to two broad sources of uncertainty: (1) specification of the gas transfer velocity, which depends on the thickness of the MBL and is usually parameterized as a function of wind speed, and (2) specification of the CO2 concentration difference across the MBL. The recent study by Woolf et al.20 contains a detailed treatment of the uncertainties due to the gas transfer, concluding that a realistic estimate (approximately, a 90% confidence interval) is ±10% when applying this to global data.
The second source of uncertainty, due to the concentration difference, is dominated by that introduced by the interpolation in time and location of surface ocean CO2. This is relatively well constrained in the more densely observed regions such as the North and Equatorial Atlantic and North and Tropical Pacific. However, in more remote regions such as the Southern, South Pacific, and Indian Oceans, the observational coverage is patchier in space and time and often seasonally biased, with few winter measurements (see Supplementary Fig. 3). New sensors and designs of autonomous floats, as now being deployed in the Southern Ocean21, show promise to solve the problem of adequately observing surface CO2 in remote regions22, but for the gap-prone historical data, the interpolation method used can have a substantial influence on the results in these data-poor regions.
To evaluate the uncertainty in flux estimates introduced by the gap-filling procedure, we used three methods for interpolating in space and time, each applied to the global data divided according to three different spatial clustering schemes, for a total of nine mappings. The interpolation methods were as follows: (1) a time series (TS) of fCO2sw data, constructed by a least squares fit to all monthly averaged fCO2 values within the defined region. The model fitted was a seasonal cycle with three harmonics superimposed on a linear trend; (2) simple multilinear regression (MLR) of the fCO2 data on latitude, longitude, and four variables for which continuous comprehensive mappings are available, these being sea-surface temperature (SST), salinity (SSS), mixed layer depth (MLD), and atmospheric CO2 mixing ratio (XCO2); (3) the feed-forward neural network method of Landschützer et al.8,18 (FFN), which also seeks a regression on these four variables. The spatial clustering schemes applied to each of the techniques (shown in Supplementary Fig. 1) were as follows: (a) division into 14 regions along latitude–longitude lines; (b) division into the 17 biogeochemical divisions suggested by Fay and Mckinley23, and (c) division into 16 biomes using a self-organizing map technique employed by Landschützer et al.8.
Where the data are adequately distributed over space and time, the use of multiple mapping techniques and different clustering schemes to estimate uncertainty gives similar results to formal geostatistical techniques, such as kriging7,20. However, in regions of very sparse and uneven coverage, statistically based techniques can underestimate uncertainties because of the assumption that the available data are representative of the true data population over a region, which may not be the case if whole regions or seasons are poorly sampled. In this instance, different mapping techniques can give substantially different results. Altering the clustering of the data by changing the shape of the geographical divisions can also have a major effect, because unsampled areas are assumed to have the same statistical properties as the sampled regions with which they are grouped.
For each combined mapping-and-clustering technique, Table 1 shows the spread and mean of the residuals (the global set of predicted values minus observed values). The neural network FFN mapping method provides a much smaller spread of residuals, giving better agreement with data at a given location and time than do the other methods. This is to be expected given its much greater flexibility, with typically several hundred parameters being adjusted to provide a non-linear fit to each cluster, compared to only 8 and 11 fitted parameters for respectively the TS and MLR methods. Figure 2 shows estimates of global and hemispheric ocean-atmosphere CO2 flux over the period 1992–2018 by the nine interpolations (using a single parameterization of the gas transfer velocity). Despite the difference in the quality of the fits to the individual data as evidenced by Table 1, convergent results are obtained by all the calculations for the Northern Hemisphere over the whole period, and there is a good agreement in the Southern Hemisphere for much of the period after 2000. The average of all the methods is shown, with one and two standard deviations of the nine separate estimates. A few regions are excluded (see Supplementary Fig. 2) to ensure compatibility in the comparison between methods, but these affect the results by More

1.
Nai C, Meyer V. From axenic to mixed cultures: technological advances accelerating a paradigm shift in microbiology. Trends Microbiol. 2018;26:538–54.
PubMed  Google Scholar 
2.
Powers MJ, Sanabria-Valentín E, Bowers AA, Shank EA. Inhibition of cell differentiation in Bacillus subtilis by Pseudomonas protegens. J Bacteriol. 2015;197:2129–38.
PubMed  PubMed Central  Google Scholar 

3.
Sztajer H, Szafranski SP, Tomasch J, Reck M, Nimtz M, Rohde M, et al. Cross-feeding and interkingdom communication in dual-species biofilms of Streptococcus mutans and Candida albicans. ISME J. 2014;8:2256–71.
PubMed  PubMed Central  Google Scholar 

4.
Trejo-Hernández A, Andrade-Domínguez A, Hernández M, Encarnación S. Interspecies competition triggers virulence and mutability in Candida albicans–Pseudomonas aeruginosa mixed biofilms. ISME J. 2014;8:1974–88.
PubMed  PubMed Central  Google Scholar 

5.
Garbeva P, Silby MW, Raaijmakers JM, Levy SB, de Boer W. Transcriptional and antagonistic responses of Pseudomonas fluorescens Pf0-1 to phylogenetically different bacterial competitors. ISME J. 2011;5:973–85.
PubMed  PubMed Central  Google Scholar 

6.
Yoshida S, Ogawa N, Fujii T, Tsushima S. Enhanced biofilm formation and 3-chlorobenzoate degrading activity by the bacterial consortium of Burkholderia sp. NK8 and Pseudomonas aeruginosa PAO1. J Appl Microbiol. 2009;106:790–800.
PubMed  Google Scholar 

7.
Beliaev AS, Romine MF, Serres M, Bernstein HC, Linggi BE, Markillie LM, et al. Inference of interactions in cyanobacterial–heterotrophic co-cultures via transcriptome sequencing. ISME J. 2014;8:2243–55.
PubMed  PubMed Central  Google Scholar 

8.
Ren D, Madsen JS, Sørensen SJ, Burmølle M. High prevalence of biofilm synergy among bacterial soil isolates in cocultures indicates bacterial interspecific cooperation. ISME J. 2015;9:81–9.
PubMed  Google Scholar 

9.
Al-Shabib NA, Husain FM, Ahmad I, Khan MS, Khan RA, Khan JM. Rutin inhibits mono and multi-species biofilm formation by foodborne drug resistant Escherichia coli and Staphylococcus aureus. Food Control. 2017;79:325–32.
Google Scholar 

10.
Liu W, Jacquiod S, Brejnrod A, Russel J, Burmølle M, Sørensen SJ. Deciphering links between bacterial interactions and spatial organization in multispecies biofilms. ISME J. 2019;13:3054–66.
PubMed  Google Scholar 

11.
Li M, Wei Z, Wang J, Jousset A, Friman VP, Xu Y, et al. Facilitation promotes invasions in plant-associated microbial communities. Ecol Lett. 2019;22:149–58.
PubMed  Google Scholar 

12.
Goyal A, Dubinkina V, Maslov S. Multiple stable states in microbial communities explained by the stable marriage problem. ISME J. 2018;12:2823–34.
PubMed  PubMed Central  Google Scholar 

13.
Zhalnina K, Louie KB, Hao Z, Mansoori N, da Rocha UN, Shi S, et al. Dynamic root exudate chemistry and microbial substrate preferences drive patterns in rhizosphere microbial community assembly. Nat Microbiol. 2018;3:470–80.
PubMed  Google Scholar 

14.
Faust K. Microbial consortium design benefits from metabolic modeling. Trends Biotechnol. 2019;37:123–5.
PubMed  Google Scholar 

15.
Liang J, Bai Y, Men Y, Qu J. Microbe–microbe interactions trigger Mn(II)-oxidizing gene expression. ISME J. 2017;11:67–77.
PubMed  Google Scholar 

16.
Xu X, Zarecki R, Medina S, Ofaim S, Liu X, Chen C, et al. Modeling microbial communities from atrazine contaminated soils promotes the development of biostimulation solutions. ISME J. 2019;13:494–508.
PubMed  Google Scholar 

17.
Kong W, Meldgin DR, Collins JJ, Lu T. Designing microbial consortia with defined social interactions. Nat Chem Biol. 2018;14:821–9.
PubMed  Google Scholar 

18.
Kelvin Lee KW, Hoong Yam JK, Mukherjee M, Periasamy S, Steinberg PD, Kjelleberg S, et al. Interspecific diversity reduces and functionally substitutes for intraspecific variation in biofilm communities. ISME J. 2016;10:846–57.
PubMed  Google Scholar 

19.
Aharonovich D, Sher D. Transcriptional response of Prochlorococcus to co-culture with a marine Alteromonas: differences between strains and the involvement of putative infochemicals. ISME J. 2016;10:2892–906.
PubMed  PubMed Central  Google Scholar 

20.
Kim W, Levy SB, Foster KR. Rapid radiation in bacteria leads to a division of labour. Nat Commun. 2016;7:10508.
PubMed  PubMed Central  Google Scholar 

21.
Venturelli OS, Carr AV, Fisher G, Hsu RH, Lau R, Bowen BP, et al. Deciphering microbial interactions in synthetic human gut microbiome communities. Mol Syst Biol. 2018;14:e8157.
PubMed  PubMed Central  Google Scholar 

22.
Little AE, Robinson CJ, Peterson SB, Raffa KF, Handelsman J. Rules of engagement: interspecies interactions that regulate microbial communities. Annu Rev Microbiol. 2008;62:375–401.
PubMed  Google Scholar 

23.
West SA, Diggle SP, Buckling A, Gardner A, Griffin AS. The social lives of microbes. Annu Rev Ecol Evol Syst. 2007;38:53–7.
Google Scholar 

24.
Foster KR, Bell T. Competition, not cooperation, dominates interactions among culturable microbial species. Curr Biol. 2012;22:1845–50.
PubMed  Google Scholar 

25.
Lee KWK, Periasamy S, Mukherjee M, Xie C, Kjelleberg S, Rice SA. Biofilm development and enhanced stress resistance of a model, mixed-species community biofilm. ISME J. 2014;8:894–907.
PubMed  Google Scholar 

26.
Zengler K, Zaramela LS. The social network of microorganisms—how auxotrophies shape complex communities. Nat Rev Microbiol. 2018;16:383–90.
PubMed  PubMed Central  Google Scholar 

27.
Russel J, Røder HL, Madsen JS, Burmølle M, Sørensen SJ. Antagonism correlates with metabolic similarity in diverse bacteria. Proc Natl Acad Sci. 2017;114:10684–8.
PubMed  Google Scholar 

28.
Nielsen AT, Tolker-Nielsen T, Barken KB, Molin S. Role of commensal relationships on the spatial structure of a surface-attached microbial consortium. Environ Microbiol. 2000;2:59–68.
PubMed  Google Scholar 

29.
Hansen SK, Rainey PB, Haagensen JAJ, Molin S. Evolution of species interactions in a biofilm community. Nature. 2007;445:533–6.
PubMed  Google Scholar 

30.
Leinweber A, Fredrik Inglis R, Kümmerli R. Cheating fosters species co-existence in well-mixed bacterial communities. ISME J. 2017;11:1179–88.
PubMed  PubMed Central  Google Scholar 

31.
Fazzino L, Anisman J, Chacón JM, Heineman RH, Harcombe WR. Lytic bacteriophage have diverse indirect effects in a synthetic cross-feeding community. ISME J. 2020;14:123–34.
PubMed  Google Scholar 

32.
Gao CH, Zhang M, Wu Y, Huang Q, Cai P. Divergent influence to a pathogen invader by resident bacteria with different social interactions. Micro Ecol. 2019;77:76–86.
Google Scholar 

33.
Molina-Santiago C, Udaondo Z, Cordero BF, Ramos JL. Interspecies cross-talk between co-cultured Pseudomonas putida and Escherichia coli. Environ Microbiol Rep. 2017;9:441–8.
PubMed  Google Scholar 

34.
Mallon CA, Le Roux X, van Doorn GS, Dini-Andreote F, Poly F, Salles JF. The impact of failure: unsuccessful bacterial invasions steer the soil microbial community away from the invader’s niche. ISME J. 2018;12:728–41.
PubMed  PubMed Central  Google Scholar 

35.
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 

36.
Wickham H. Ggplot2: elegant graphics for data analysis. New York, USA: Springer Publishing Company; 2009.

37.
Hall T. BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucleic Acids Symp Ser. 1999;41:95–8.
Google Scholar 

38.
Ren D, Madsen JS, de la Cruz-Perera CI, Bergmark L, Sørensen SJ, Burmølle M. High-throughput screening of multispecies biofilm formation and quantitative PCR-based assessment of individual species proportions, useful for exploring interspecific bacterial interactions. Micro Ecol. 2014;68:146–54.
Google Scholar 

39.
Madsen JS, Røder HL, Russel J, Sørensen H, Burmølle M, Sørensen SJ. Coexistence facilitates interspecific biofilm formation in complex microbial communities. Environ Microbiol. 2016;18:2565–74.
PubMed  Google Scholar 

40.
Hibbing ME, Fuqua C, Parsek MR, Peterson SB. Bacterial competition: surviving and thriving in the microbial jungle. Nat Rev Microbiol. 2010;8:15–25.
PubMed  PubMed Central  Google Scholar 

41.
Szamosvári D, Rütschlin S, Böttcher T. From pirates and killers: does metabolite diversity drive bacterial competition? Org Biomol Chem. 2018;16:2814–9.
PubMed  Google Scholar 

42.
Burmølle M, Ren D, Bjarnsholt T, Sørensen SJ. Interactions in multispecies biofilms: do they actually matter? Trends Microbiol. 2014;22:84–91.
PubMed  Google Scholar 

43.
Hansen LB, Ren D, Burmølle M, Sørensen SJ. Distinct gene expression profile of Xanthomonas retroflexus engaged in synergistic multispecies biofilm formation. ISME J. 2017;11:300–3.
PubMed  Google Scholar 

44.
Solden LM, Naas AE, Roux S, Daly RA, Collins WB, Nicora CD, et al. Interspecies cross-feeding orchestrates carbon degradation in the rumen ecosystem. Nat Microbiol. 2018;3:1274–84.
PubMed  PubMed Central  Google Scholar 

45.
Freilich S, Zarecki R, Eilam O, Segal ES, Henry CS, Kupiec M, et al. Competitive and cooperative metabolic interactions in bacterial communities. Nat Commun. 2011;2:589.
PubMed  Google Scholar 

46.
Goldford JE, Lu N, Bajić D, Estrela S, Tikhonov M, Sanchez-Gorostiaga A, et al. Emergent simplicity in microbial community assembly. Science. 2018;361:469–74.
PubMed  PubMed Central  Google Scholar 

47.
Røder HL, Sørensen SJ, Burmølle M. Studying bacterial multispecies biofilms: where to start? Trends Microbiol. 2016;24:503–13.
PubMed  Google Scholar 

48.
Munna MS, Zeba Z, Noor R. Influence of temperature on the growth of Pseudomonas putida. Stamford J Microbiol. 2015;5:9–12.
Google Scholar 

49.
Gonthier A, Guerin-Faublee V, Tilly B, Delignette-Muller ML. Optimal growth temperature of O157 and non-O157 Escherichia coli strains. Lett Appl Microbiol. 2001;33:352–6.
PubMed  Google Scholar 

50.
Van Elsas JD, Semenov AV, Costa R, Trevors JT. Survival of Escherichia coli in the environment: fundamental and public health aspects. ISME J. 2011;5:173–83.
PubMed  Google Scholar  More

Based on the labels of DeepFish, we consider these four computer vision tasks: classification, counting, localization, and segmentation. Deep learning have consistently achieved state-of-the-art results on these tasks as they can leverage the enormous size of the datasets they are trained on. These datasets include ImageNet6, Pascal7, CityScapes5 and COCO24. DeepFish aims to be part of these large scale datasets with the unique goal of understanding complex fish habitats for the purpose of inspiring further research in this area.
We present standard deep learning methods for each of these tasks. Shown as the blue module in Figure 4, these methods have the ResNet-5013 backbone which is one of the most popular feature extractors for image understanding and visual recognition. They enable models to learn from large datasets and transfer the acquired knowledge to train efficiently on another dataset. This process is known as transfer learning and has been consistently used in most current deep learning methods22. Such pretrained models can even recognize object classes that they have never been trained on29. This property illustrates how powerful the extracted features are from a pretrained ResNet-50.
Therefore, we initialize the weights of our ResNet-50 backbones by pre-training it on ImageNet following the procedure discussed in6. ImageNet consists of over 14 million images categorized over 1,000 classes. As a result, the backbone learns to extract strong, general features for unseen images by training on such dataset. These features are then used by a designated module to perform their respective computer vision task such as classification and segmentation. We describe these modules in the sections below.
To put the results into perspective, we also include baseline results by training the same methods without ImageNet pretraining (Table 3). In this case, we randomly initialize the weights of the ResNet-50 backbone with Xavier’s method11. These results also illustrate the efficacy of having pretrained models over randomly initialized models.
Table 3 Comparison between randomly initialized and ImageNet pretrained models.
Full size table

Classification results
The goal of the classification task is to identify whether images are foreground (contains fish) or background (contains no fish). We use accuracy to evaluate the models on this task which is a standard metric for binary classification problems3,8,9,15,27. The metric is computed as

$$begin{aligned} { ACC}=({ TP}+{ TN})/{ N}, end{aligned}$$

where ({ TP}) and ({ TN}) are the true positives and true negatives, respectively, and ({ N}) is the total number of images. A true positive represents an image with at least one fish that is predicted as foreground, whereas a true negative represents an image with no fish that is predicted as background. For this task we used the FishClf dataset for this task where the number of images labeled is 39,766.
The classification architecture consists of a ResNet-50 backbone and a feed-forward network (FFN) (classification branch of Fig. 4). FFN takes as input features extracted by ResNet-50 and outputs a probability for the image corresponding to how likely it contains a fish. If the probability is higher than 0.5 the predicted classification label is foreground. For the FFN, we use the network presented in ImageNet which consists of 3 layers. However, instead of the original 1,000-class output layer, we use a 2-class output layer to represent the foreground or background class.
During training, the classifier learns to minimize the binary cross-entropy objective function28 using the Adam16 optimizer. The learning rate was set as (10^{-3}) and the batch size was set to be 16. Since FFN require a fixed resolution of the extracted features, the input images are resized to (224times 224). At test time, the model outputs a score for each of the two classes for a given unseen image. The predicted class for that image is the class with the higher score.
In Table 3 we compare between a classifier with the backbone pretrained on ImageNet and with the randomly initialized backbone. Note that both classifiers have their FFN network initialized at random. We see that the pretrained model achieved near-perfect classification results outperforming the baseline significantly. This result suggests that transfer learning is important and that deep learning has strong potential for analyzing fish habitats.
Figure 4

Deep learning methods. The architecture used for the four computer vision tasks of classification, counting, localization, and segmentation consists of two components. The first component is the ResNet-50 backbone which is used to extract features from the input image. The second component is either a feed-forward network that outputs a scalar value for the input image or an upsampling path that outputs a value for each pixel in the image.

Full size image

Counting results
The goal of the counting task is to predict the number of fish present in an image. We evaluate the models on the FishLoc dataset, which consists of 3,200 images labeled with point-level annotations. We measure the model’s efficacy in predicting the fish count by using the mean absolute error. It is defined as,

$$begin{aligned} { MAE}=frac{1}{N}sum _{i=1}^N|hat{C}_i-C_i|, end{aligned}$$

where (C_i) is the true fish count for image i and (hat{C}_i) is the model’s predicted fish count for image i. This metric is standard for object counting12,23 and it measures the number of miscounts the model is making on average across the test images.
The counting branch in Fig. 4 shows the architecture used for the counting task, which, similar to the classifier, consists of a ResNet-50 backbone and a feed-forward network (FFN). Given the extracted features from the backbone for an input image, the FFN outputs a number that correspond to the count of the fish in the image. Thus, instead of a 2-class output layer like with the classifier, the counting model has a single node output layer.
We train the models by minimizing the squared error loss28, which is a common objective function for the counting task. At test time, the predicted value for an image is the predicted object count.
The counting model with the backbone pretrained on ImageNet achieved an MAE of 0.38 (Table 3. This result corresponds to making an average of 0.38 fish miscounts per image which is satisfactory as the average number of fish per image is 7. In comparison, the counting model initialized randomly achieved an MAE of 1.30. This result further confirms that transfer learning and deep learning can successfully address the counting task despite the fact that the dataset for counting (FishLoc) is much smaller than classification (FishClf).
Localization results
Localization considers the task of identifying the locations of the fish in the image. It is a more difficult task than classification and counting as the fish can extensively overlap. Like with the counting task, we evaluate the models on the FishLoc dataset. However, MAE scores do not provide how well the model performs at localization as the model can count the wrong objects and still achieve perfect score. To address this limitation, we use a more accurate evaluation for localization by following12, which considers both the object count and the location estimated for the objects. This metric is called Grid Average Mean absolute Error (GAME). It is computed as

$$begin{aligned} GAME = sum _{i=1}^4 { GAME}(L),quad { GAME}(L) = frac{1}{N}sum _{i=1}^Nleft( sum _{l=1}^{4^L}|D^l_i – hat{D}^l_i|right) , end{aligned}$$

where (D^l_i) is the number of point-level annotations in region l, and (hat{D}^l_i) is the model’s predicted count for region l. ({ GAME}(L)) first divides the image into a grid of (4^L) non-overlapping regions, and then computes the sum of the MAE scores across these regions. The higher L, the more restrictive the GAME metric will be. Note that ({ GAME}(0)) is equivalent to MAE.
The localization branch in Fig. 4 shows the architecture used for the localization task, which consists of a ResNet-50 backbone and an upsampling path. The upsampling path is based on the network described in FCN826 which is a standard fully convolutional neural network meant for localization and segmentation, which consists of three upsampling layers.
FCN8 processes images as follows. The features extracted with the backbone are of a smaller resolution than the input image. These features are then upsampled with the upsampling path to match the resolution of the input image. The final output is a per-pixel probability map where each pixel represents the likelihood that it belongs to the fish class.
The models is trained using a state-of-the-art localization-based loss function called LCFCN21. LCFCN is trained using four objective functions: image-level loss, point-level loss, split-level loss, and false positive loss. The image-level loss encourages the model to predict all pixels as background for background images. The point-level loss encourages the model to predict the centroids of the fish. Unfortunately, these two loss terms alone do not prevent the model from predicting every pixel as fish for foreground images. Thus, LCFCN also minimizes the split loss and false-positive loss. The split loss splits the predicted regions so that no region has more than one point annotation. This results in one blob per point annotation. The false-positive loss prevents the model from predicting blobs for regions where there are no point annotations. Note that training LCFCN only requires point-level annotations which are spatial locations of where the objects are in the image.
At test time, the predicted probability map are thresholded to become 1 if they are larger than 0.5 and 0 otherwise. This results in a binary mask, where each blob is a single connected component and they can be collectively obtained using the standard connected components algorithm. The number of connected components is the object count and each blob represents the location of an object instance (see Fig. 5 for example predictions with FCN8 trained with LCFCN).
Models trained on this dataset are optimized using Adam16 with a learning rate of (10^{-3}) and weight decay of 0.0005, and have been ran for 1,000 epochs on the training set. In all cases the batch size is 1, which makes it applicable for machines with limited memory.
Table 3 shows the MAE and GAME results of training an FCN8 with and without a pretrained ResNet-50 backbone using the LCFCN loss function. We see that pretraining leads to significant improvement on MAE and a slight improvement for GAME. The efficacy of the pretrained model is further confirmed by the qualitative results shown in Fig. 5a where the predicted blobs are well-placed on top of the fish in the images.
Figure 5

Qualitative results on counting, localization, and segmentation. (a) Prediction results of the model trained with the LCFCN loss21. (b) Annotations that represent the (x, y) coordinates of each fish within the images. (c) Prediction results of the model trained with the focal loss25. (d) Annotations that represent the full segmentation masks of the corresponding fish.

Full size image

Segmentation results
The task of segmentation is to label every pixel in the image as either fish or not fish (Fig. 5c,d). When combined with depth information, a segmented image allows us to measure the size and the weight of the fish in a location, which can vastly improve our understanding of fish communities. We evaluate the model on the FishSeg dataset for which we acquired per-pixel labels for 620 images. We evaluate the models on this dataset using the standard Jaccard index5,7 which is defined as the number of correctly labelled pixels of a class, divided by the number of pixels labelled with that class in either the ground truth mask or the predicted mask. It is commonly known as the intersection-over-union metric IoU, computed as (frac{TP}{TP + FP + FN}), where TP, FP, and FN are the numbers of true positive, false positive, and false negative pixels, respectively, which is determined over the whole test set. In segmentation tasks, the IoU is preferred over accuracy as it is not as affected by the class imbalances that are inherent in foreground and background segmentation masks like in DeepFish.
During training, instead of minimizing the standard per-pixel cross-entropy loss26, we use the focal loss function25 which is more suitable when the number of background pixels is much higher than the foreground pixels like in our dataset. The rest of the training procedure is the same as with the methods trained for localization.
At test time, the model outputs a probability for each pixel in the image. If the probability is higher than 0.5 for the foreground class, then the pixel is labeled as fish, resulting in a segmentation mask for the input image.
The results in Table 3 show a comparison between the pretrained and randomly initialized segmentation model. Like with the other tasks, the pretrained model achieves superior results both quantitatively and qualitatively (Fig. 5).
Ethical approval
This work was conducted with the approval of the JCU Animal Ethics Committee (protocol A2258), and conducted in accordance with DAFF general fisheries permit #168652 and GBRMP permit #CMES63. More

1.
Hutchins DA, Fu F. Microorganisms and ocean global change. Nat Microbiol. 2017;2:17058.
CAS  PubMed  Google Scholar 
2.
Shade A, Peter H, Allison SD, Baho DL, Berga M, Bürgmann H, et al. Fundamentals of microbial community resistance and resilience. Front Microbiol. 2012;3:417.
PubMed  PubMed Central  Google Scholar 

3.
Allison SD, Martiny JBH. Resistance, resilience, and redundancy in microbial communities. Proc Natl Acad Sci USA. 2008;105:11512–9.
CAS  Google Scholar 

4.
Eppley RW. Temperature and phytoplankton growth in the sea. Fish Bull. 1972;70:1063–85.
Google Scholar 

5.
Hawkes CV, Keitt TH. Resilience vs. historical contingency in microbial responses to environmental change. Ecol Lett. 2015;18:612–25.
PubMed  Google Scholar 

6.
Stegen JC, Bottos EM, Jansson JK. A unified conceptual framework for prediction and control of microbiomes. Curr Opin Microbiol. 2018;44:20–7.
PubMed  Google Scholar 

7.
Kling JD, Lee MD, Fu F, Phan MD, Wang X, Qu P, et al. Transient exposure to novel high temperatures reshapes coastal phytoplankton communities. ISME J. 2020;14:413–24.
CAS  PubMed  Google Scholar 

8.
Izem R, Kingsolver JG. Variation in continuous reaction norms: quantifying directions of biological interest. Am Naturalist. 2005;166:277–89.
Google Scholar 

9.
Thomas MK, Kremer CT, Klausmeier CA, Litchman E. A global pattern of thermal adaptation in marine phytoplankton. Science. 2012;338:1085–8.
CAS  PubMed  Google Scholar 

10.
Lomas M, Hopkinson B, Ryan JLD, Shi D, Xu Y, Morel F. Effect of ocean acidification on cyanobacteria in the subtropical North Atlantic. Aquat Microb Ecol. 2012;66:211–22.
Google Scholar 

11.
Dutkiewicz S, Morris JJ, Follows MJ, Scott J, Levitan O, Dyhrman ST, et al. Impact of ocean acidification on the structure of future phytoplankton communities. Nat Clim Change. 2015;5:1002–6.
CAS  Google Scholar 

12.
Hunt DE, Ward CS. A network-based approach to disturbance transmission through microbial interactions. Front Microbiol. 2015;6:1182. http://www.frontiersin.org/Journal/Abstract.aspx?s=53&name=aquatic_microbiology&ART_DOI=10.3389/fmicb.2015.01182.

13.
Hennon GM, Morris JJ, Haley ST, Zinser ER, Durrant AR, Entwistle E, et al. The impact of elevated CO2 on Prochlorococcus and microbial interactions with ‘helper’ bacterium Alteromonas. The ISME J. 2017;12:520–31.
Google Scholar 

14.
von Scheibner M, Dörge P, Biermann A, Sommer U, Hoppe HG, Jürgens K. Impact of warming on phyto‐bacterioplankton coupling and bacterial community composition in experimental mesocosms. Environ Microbiol. 2014;16:718–33.
Google Scholar 

15.
Toseland A, Daines SJ, Clark JR, Kirkham A, Strauss J, Uhlig C, et al. The impact of temperature on marine phytoplankton resource allocation and metabolism. Nat Clim Change. 2013;3:979–84.
CAS  Google Scholar 

16.
Ward CS, Yung C-M, Davis KM, Blinebry SK, Williams TC, Johnson ZI, et al. Annual community patterns are driven by seasonal switching between closely related marine bacteria. ISME J. 2017;11:1412–22.
PubMed  PubMed Central  Google Scholar 

17.
Johnson ZI, Zinser ER, Coe A, McNulty NP, Woodward EMS, Chisholm SW. Niche partitioning among Prochlorococcus ecotypes along ocean-scale environmental gradients. Science. 2006;311:1737–40.
CAS  PubMed  Google Scholar 

18.
Fuhrman JA, Steele JA, Hewson I, Schwalbach MS, Brown MV, Green JL, et al. A latitudinal diversity gradient in planktonic marine bacteria. Proc Natl Acad Sci USA. 2008;105:7774–8.
CAS  PubMed  PubMed Central  Google Scholar 

19.
Doo SS, Kealoha A, Andersson A, Cohen AL, Hicks TL, Johnson ZI, et al. The challenges of detecting and attributing ocean acidification impacts on marine ecosystems. ICES J Mar Sci. 2020. https://doi.org/10.1093/icesjms/fsaa094.

20.
Lindh MV, Riemann L, Baltar F, Romero‐Oliva C, Salomon PS, Granéli E, et al. Consequences of increased temperature and acidification on bacterioplankton community composition during a mesocosm spring bloom in the Baltic Sea. Environ Microbiol Rep. 2013;5:252–62.
CAS  PubMed  Google Scholar 

21.
Roy A-S, Gibbons S, Schunck H, Owens S, Caporaso J, Sperling M, et al. Ocean acidification shows negligible impacts on high-latitude bacterial community structure in coastal pelagic mesocosms. Biogeosciences. 2013;10:555–66.
Google Scholar 

22.
Oliver AE, Newbold LK, Whiteley AS, van der Gast CJ. Marine bacterial communities are resistant to elevated carbon dioxide levels. Environ Microbiol Rep. 2014;6:574–82.
CAS  PubMed  Google Scholar 

23.
Newbold LK, Oliver AE, Booth T, Tiwari B, DeSantis T, Maguire M, et al. The response of marine picoplankton to ocean acidification. Environ Microbiol. 2012;14:2293–307.
CAS  PubMed  Google Scholar 

24.
Bergen B, Endres S, Engel A, Zark M, Dittmar T, Sommer U, et al. Acidification and warming affect prominent bacteria in two seasonal phytoplankton bloom mesocosms. Environ Microbiol. 2016;18:4579–95.
CAS  PubMed  Google Scholar 

25.
Bunse C, Lundin D, Karlsson CM, Akram N, Vila-Costa M, Palovaara J, et al. Response of marine bacterioplankton pH homeostasis gene expression to elevated CO2. Nat Clim Change. 2016;6:483–7.
CAS  Google Scholar 

26.
Xia X, Wang Y, Yang Y, Luo T, Van Nostrand JD, Zhou J, et al. Ocean acidification regulates the activity, community structure, and functional potential of heterotrophic bacterioplankton in an oligotrophic gyre. J Geophys Res. 2019;124:1001–17.
CAS  Google Scholar 

27.
Crawfurd KJ, Alvarez-Fernandez S, Mojica KD, Riebesell U, Brussaard CP. Alterations in microbial community composition with increasing fCO2: a mesocosm study in the eastern Baltic Sea. Biogeosciences. 2017;14:3831–49.
CAS  Google Scholar 

28.
Johnson ZI, Wheeler BJ, Blinebry SK, Carlson CM, Ward CS, Hunt DE. Dramatic variability of the carbonate system at a temperate coastal ocean site (Beaufort, North Carolina, USA) is regulated by physical and biogeochemical processes on multiple timescales. PLoS ONE. 2013;8:e85117.
PubMed  PubMed Central  Google Scholar 

29.
Wang Z, Juarez DL, Pan JF, Blinebry SK, Gronniger J, Clark JS, et al. Microbial communities across nearshore to offshore coastal transects are primarily shaped by distance and temperature. Environ Microbiol. 2019;21:3862–72.
CAS  PubMed  Google Scholar 

30.
Dore JE, Lukas R, Sadler DW, Church MJ, Karl DM. Physical and biogeochemical modulation of ocean acidification in the central North Pacific. Proc Natl Acad Sci USA. 2009;106:12235–40.
CAS  PubMed  Google Scholar 

31.
Giovannoni SJ, Thrash JC, Temperton B. Implications of streamlining theory for microbial ecology. ISME J. 2014;8:1553–65.
PubMed  PubMed Central  Google Scholar 

32.
Caldeira K, Wickett ME. Anthropogenic carbon and ocean pH. Nature. 2003;425:365.
CAS  PubMed  Google Scholar 

33.
Clayton TD, Byrne RH. Spectrophotometric seawater pH measurements: total hydrogen ion concentration scale calibration of m-cresol purple and at-sea results. Deep Sea Res Part I. 1993;40:2115–29.
CAS  Google Scholar 

34.
Barber RT, Marra J, Bidigare RC, Codispoti LA, Halpern D, Johnson Z, et al. Primary productivity and its regulation in the Arabian Sea during 1995. Deep Sea Res Part II. 2001;48:1127–72.
CAS  Google Scholar 

35.
Labasque T, Chaumery C, Aminot A, Kergoat G. Spectrophotometric Winkler determination of dissolved oxygen: re-examination of critical factors and reliability. Marine Chemistry. 2004;88:53–60.

36.
Johnson ZI, Shyam R, Ritchie AE, Mioni C, Lance VP, Murray JW, et al. The effect of iron-and light-limitation on phytoplankton communities of deep chlorophyll maxima of the western Pacific Ocean. J Mar Res. 2010;68:283–308.
CAS  Google Scholar 

37.
Marie D, Partensky F, Jacquet S, Vaulot D. Enumeration and cell cycle analysis of natural populations of marine picoplankton by flow cytometry using the nucleic acid stain SYBR Green I. Appl Environ Microbiol. 1997;63:186–93.
CAS  PubMed  PubMed Central  Google Scholar 

38.
Parada AE, Needham DM, Fuhrman JA. Every base matters: assessing small subunit rRNA primers for marine microbiomes with mock communities, time series and global field samples. Environ Microbiol. 2016;18:1403–14.
CAS  PubMed  Google Scholar 

39.
Edgar RC. UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nat Methods. 2013;10:996–8.
CAS  PubMed  Google Scholar 

40.
Eren AM, Morrison HG, Lescault PJ, Reveillaud J, Vineis JH, Sogin ML. Minimum entropy decomposition: unsupervised oligotyping for sensitive partitioning of high-throughput marker gene sequences. ISME J. 2015;9:968–79.
CAS  PubMed  Google Scholar 

41.
Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O’hara R, et al. Package ‘vegan’. Community ecology package, version. 2015;2.

42.
Tsementzi D, Poretsky R, Rodriguez-R LM, Luo C, Konstantinidis KT. Evaluation of metatranscriptomic protocols and application to the study of freshwater microbial communities. Environ Microbiol Rep. 2014;6:640–55.
CAS  PubMed  Google Scholar 

43.
Cox MP, Peterson DA, Biggs PJ. SolexaQA: at-a-glance quality assessment of Illumina second-generation sequencing data. BMC Bioinform. 2010;11:485.
Google Scholar 

44.
Bengtsson‐Palme J, Hartmann M, Eriksson KM, Pal C, Thorell K, Larsson DGJ, et al. METAXA2: improved identification and taxonomic classification of small and large subunit rRNA in metagenomic data. Mol Ecol Resour. 2015;15:1403–14.
PubMed  Google Scholar 

45.
Kopylova E, Noé L, Touzet H. SortMeRNA: fast and accurate filtering of ribosomal RNAs in metatranscriptomic data. Bioinformatics. 2012;28:3211–7.
CAS  PubMed  Google Scholar 

46.
Burge SW, Daub J, Eberhardt R, Tate J, Barquist L, Nawrocki EP, et al. Rfam 11.0: 10 years of RNA families. Nucleic Acids Res. 2012;41:D226–32.
PubMed  PubMed Central  Google Scholar 

47.
Kent WJ. BLAT—the BLAST-like alignment tool. Genome Res. 2002;12:656–64.
CAS  PubMed  PubMed Central  Google Scholar 

48.
Parks DH, Rinke C, Chuvochina M, Chaumeil P-A, Woodcroft BJ, Evans PN, et al. Recovery of nearly 8,000 metagenome-assembled genomes substantially expands the tree of life. Nat Microbiol. 2017;2:1533–42.
CAS  PubMed  Google Scholar 

49.
Tully BJ, Graham ED, Heidelberg JF. The reconstruction of 2,631 draft metagenome-assembled genomes from the global oceans. Sci data. 2018;5:170203.
CAS  PubMed  PubMed Central  Google Scholar 

50.
Rodriguez-R LM, Tsementzi D, Luo C, Konstantinidis KT. Iterative subtractive binning of freshwater chronoseries metagenomes identifies over 400 novel species and their ecologic preferences. Environ Microbiol. 2020;22:3394–412.
CAS  PubMed  Google Scholar 

51.
Tsementzi D, Rodriguez-R LM, Ruiz-Perez CA, Meziti A, Hatt JK, Konstantinidis KT. Ecogenomic characterization of widespread, closely-related SAR11 clades of the freshwater genus “Candidatus Fonsibacter” and proposal of Ca. Fonsibacter lacus sp. nov. Syst Appl Microbiol. 2019;42:495–505.
PubMed  Google Scholar 

52.
Wu CH, Apweiler R, Bairoch A, Natale DA, Barker WC, Boeckmann B, et al. The universal protein resource (UniProt): an expanding universe of protein information. Nucleic Acids Res. 2006;34:D187–91.
CAS  PubMed  Google Scholar 

53.
Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: tool for the unification of biology. Nat Genet. 2000;25–29:25.
Google Scholar 

54.
Overbeek R, Begley T, Butler RM, Choudhuri JV, Chuang H-Y, Cohoon M, et al. The subsystems approach to genome annotation and its use in the project to annotate 1000 genomes. Nucleic Acids Res. 2005;33:5691–702.
CAS  PubMed  PubMed Central  Google Scholar 

55.
Buchfink B, Xie C, Huson DH. Fast and sensitive protein alignment using DIAMOND. Nat Methods. 2015;12:59.
CAS  PubMed  Google Scholar 

56.
Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:1.
Google Scholar 

57.
Wohlers J, Engel A, Zöllner E, Breithaupt P, Jürgens K, Hoppe HG, et al. Changes in biogenic carbon flow in response to sea surface warming. Proc Natl Acad Sci USA. 2009;106:7067–72.
CAS  PubMed  Google Scholar 

58.
Fu F-X, Warner ME, Zhang Y, Feng Y, Hutchins DA. Effects of increased temperature and CO2 on photosynthesis, growth, and elemental rations in marine Synechoccocus and Prochlorococcus (Cyanobacteria). J Phycol. 2007;43:485–96.
Google Scholar 

59.
Riebesell U, Wolf-Gladrow D, Smetacek V. Carbon dioxide limitation of marine phytoplankton growth rates. Nature. 1993;361:249–51.
CAS  Google Scholar 

60.
McCarthy A, Rogers SP, Duffy SJ, Campbell DA. Elevated carbon dioxide differentially alters the photophysiology of Thalassiosira pseudonana (Bacillariophyceae) and Emiliania huxleyi (Haptophyta). J Phycol. 2012;48:635–46.
CAS  PubMed  Google Scholar 

61.
Fuhrman JA, Hewson I, Schwalbach MS, Steele JA, Brown MV, Naeem S. Annually reoccurring bacterial communities are predictable from ocean conditions. Proc Natl Acad Sci USA. 2006;103:13104–9.
CAS  PubMed  Google Scholar 

62.
Baltar F, Palovaara J, Vila-Costa M, Salazar G, Calvo E, Pelejero C, et al. Response of rare, common and abundant bacterioplankton to anthropogenic perturbations in a Mediterranean coastal site. FEMS Microbiol Ecol. 2015;91:fiv058.
PubMed  Google Scholar 

63.
Paerl HW, Huisman J. Blooms like it hot. Science. 2008;320:57–8.
CAS  PubMed  Google Scholar 

64.
Yung C-M, Vereen MK, Herbert A, Davis KM, Yang J, Kantorowska A, et al. Thermally adaptive tradeoffs in closely-related marine bacterial strains. Environ Microbiol. 2015;17:2421–9.
PubMed  Google Scholar 

65.
Gao H, Wang Y, Liu X, Yan T, Wu L, Alm E, et al. Global transcriptome analysis of the heat shock response of Shewanella oneidensis. J Bacteriol. 2004;186:7796–803.
CAS  PubMed  PubMed Central  Google Scholar 

66.
van der Veen S, Hain T, Wouters JA, Hossain H, de Vos WM, Abee T, et al. The heat-shock response of Listeria monocytogenes comprises genes involved in heat shock, cell division, cell wall synthesis, and the SOS response. Microbiology. 2007;153:3593–607.
PubMed  Google Scholar 

67.
Hall EK, Singer GA, Kainz MJ, Lennon JT. Evidence for a temperature acclimation mechanism in bacteria: an empirical test of a membrane-mediated trade-off. Funct Ecol. 2010;24:898–908.
Google Scholar 

68.
Lauro FM, McDougald D, Thomas T, Williams TJ, Egan S, Rice S, et al. The genomic basis of trophic strategy in marine bacteria. Proc Natl Acad Sci USA. 2009;106:15527–33.
CAS  PubMed  Google Scholar 

69.
Yutin N, Puigbò P, Koonin EV, Wolf YI. Phylogenomics of prokaryotic ribosomal proteins. PLoS ONE. 2012;7:5.
Google Scholar 

70.
Blazewicz SJ, Barnard RL, Daly RA, Firestone MK. Evaluating rRNA as an indicator of microbial activity in environmental communities: limitations and uses. ISME J. 2013;7:2061–8.
CAS  PubMed  PubMed Central  Google Scholar 

71.
Hunt DE, Lin Y, Church MJ, Karl DM, Izzo LK, Tringe S, et al. Relationship between abundance and specific activity of bacterioplankton in open ocean surface waters. Appl Environ Microbiol. 2013;79:177–84.
CAS  PubMed  PubMed Central  Google Scholar 

72.
Boeuf D, Edwards BR, Eppley JM, Hu SK, Poff KE, Romano AE, et al. Biological composition and microbial dynamics of sinking particulate organic matter at abyssal depths in the oligotrophic open ocean. Proc Natl Acad Sci USA. 2019;116:11824–32.
CAS  PubMed  Google Scholar  More

The videos were imported from digital videotapes using iMovie 4 and iMovie HD 6.0.3. They were deinterlaced using JES Deinterlacer 3.8.4. Images are processed here using QuickTime Player 7.3.3, GraphicConverter 8.8.3, and GIMP 2.10. Within these applications, it is possible to interpolate and adjust brightness, contrast, color, and other parameters. The simple processing applied here is effective for some cases. With advanced processing techniques that involve greater control and analysis of parameters, experts in image processing might be able to extract additional information.
The 2006 video
The first video was obtained from a kayak with a Sony DCR-HC36 standard video camera (which captures interlaced video at 720 × 480 pixels) in the Pearl River swamp in Louisiana on February 20, 2006, in an area along English Bayou where there were five sightings that week; the ‘kent’ calls of the Ivory-billed Woodpecker were heard twice during the same period, once coming simultaneously from different directions. The 2006 video shows a large woodpecker perched on a tree, climbing upward, taking a short flight between limbs, and then taking off into a longer flight. Part of the perch tree, which includes two forks that facilitated scaling, was used in the size comparison in Fig. 2; the bird in the video appears to be larger than a Pileated Woodpecker specimen8. According to Julie Zickefoose, whose paintings of the Ivory-billed Woodpecker have appeared on the covers of the January 2006 issue of the Auk and both editions of Ref.3, the “long but fluffy and squared-off crest,” “extremely long, erect head and neck,” “large, long bill,” “bill to head proportions,” “rared-back pose,” “long and thin” wings, “flapping leap” between limbs, and “ponderous and heavy” flight are suggestive of the Ivory-billed Woodpecker but not the Pileated Woodpecker13.
Figure 2

A pileated Woodpecker specimen is mounted on part of the perch tree. Frames from the 2006 video were scaled using forks in the tree (dashed lines). A meter stick is placed at the point where the flight between limbs occurred. The inset shows Pileated Woodpecker and Ivory-billed Woodpecker specimens that were photographed side by side at the National Museum of Natural History. The bird in the video is partially hidden by vegetation in the image on the lower left, but it is fully in view in the images at the top when it took the flight between limbs.

Full size image

The 2008 video
A short distance up the same bayou, another video was obtained with the same camera on March 29, 2008, from 23 m up a tree that was used as an observation platform for keeping watch for Ivory-billed Woodpeckers flying over the treetops in the distance. A large bird that flew along the bayou and passed below was identified as an Ivory-billed Woodpecker on the basis of two white stripes on the back and black leading edges and white trailing edges on the dorsal surfaces of the wings (those definitive field marks were observed from an ideal vantage point at close range and nearly directly above). The appearance in the video of the bird, its reflection from the still surface of the bayou, and reference objects made it possible to determine positions along the flight path and obtain estimates of the flight speed and wingspan. The bird in the 2008 video folded its wings closed during the middle of each upstroke as illustrated in Fig. 3. The two large woodpeckers are the only large birds north of the Rio Grande that have this distinctive wing motion, which is clearly resolved in the video. Using an approach that he had previously developed and applied to other woodpeckers17, Bret Tobalske, an expert on woodpecker flight mechanics, digitized the horizontal and vertical motions of the wingtips and concluded that the bird in the video is a large woodpecker13. The flap rate of the bird in the video is about ten standard deviations greater than the mean flap rate of the Pileated Woodpecker13.
Figure 3

Illustrations of large woodpeckers in flight. Left: The Pileated Woodpecker typically swoops upward a short distance before landing on a surface that faces the direction of approach; the Ivory-billed Woodpecker has long vertical ascents that allow time for maneuvering and landing on surfaces that do not face the direction of approach. Center: An Ivory-billed Woodpecker takes off with rapid wingbeats into a horizontal flight that quickly transitions into an upward swooping flight. Right: Illustration of a flight in the Pearl River swamp on March 29, 2008, that was viewed from 23 m up in a cypress tree. When the wings are folded closed in flight, the dorsal stripes and the white triangular patch have the same appearance as they do for the perched birds in Fig. 1. As discussed in Movie S6 of Ref.8, the wings of an Ivory-billed Woodpecker in a historical photo and of the bird in the 2008 video have the swept-back appearance of the wings in the middle image.

Full size image

Additional characteristics of the bird in the video that are consistent with the Ivory-billed Woodpecker but not the Pileated Woodpecker are the high flight speed, narrow wings, swept back wings, and prominent white patches on the dorsal surfaces of the wings8,13. There is one characteristic of the bird in the video that was initially thought to be inconsistent with the Ivory-billed Woodpecker. On the basis of historical accounts of a ‘duck-like’ flight, the Ivory-billed Woodpecker was thought to have a duck-like wing motion in which the wings remain extended throughout the flap cycle. In a series of paintings of the large woodpeckers in flight by Zickefoose18, the wings of the Pileated Woodpecker are correctly shown folding closed during the middle of the upstroke; in a proper representation of conventional wisdom at the time, the wings of the Ivory-billed Woodpecker are shown remaining extended throughout the flap cycle (duck-like flaps). An apparent paradox arose during the initial inspection of the video, which revealed an unexpected wing motion. The paradox was resolved after the discovery that a photo from 1939 shows an Ivory-billed Woodpecker in flight at an instant when the wings are nearly folded closed13.
The 2007 video
The other video was obtained with a Sony HDR-HC3 high-definition video camera (which captures interlaced video at 1,440 × 1,080 pixels) that was mounted on kayak paddles8 in the Choctawhatchee River swamp in Florida on January 19, 2007, in an area where an ornithologist and his colleagues had recently reported a series of sightings7. During an encounter with a pair of birds that were identified as Ivory-billed Woodpeckers on the basis of field marks and remarkable swooping flights, the camera captured a series of events that involve flights, field marks, and other behaviors and characteristics that are consistent with the Ivory-billed Woodpecker but no other species of the region. The analysis of the 2007 video is based in part on the fact that the probability of a series of unlikely events becomes extremely small as the number of events increases12. There is a downward swooping takeoff with a long horizontal glide that is consistent with the following account by Audubon15: “The transit from one tree to another, even should the distance be as much as a hundred yards, is performed by a single sweep, and the bird appears as if merely swinging from the top of the one tree to that of the other, forming an elegantly curved line.” There are upward swooping landings with long vertical ascents that are not consistent with the Pileated Woodpecker but are consistent with an account by Eckleberry of an Ivory-billed Woodpecker that “alighted with one magnificent upward swoop”19.
A long vertical ascent allows time for maneuvering, and the bird appears to rotate about its axis during two of the ascents as illustrated in Fig. 3. In a film of the closely related Magellanic Woodpecker (Campephilus magellanicus)20, there is maneuvering during a landing with a long vertical ascent. During and after one of the ascents, a woodpecker in the 2007 video shows field marks and body proportions that are consistent with the Ivory-billed Woodpecker but no other species of the region. There is a takeoff into horizontal flight with deep and rapid flaps that are not consistent with the Pileated Woodpecker but are similar to the deep and rapid flaps during a takeoff of the closely related Imperial Woodpecker (Campephilus imperialis)21. In another event, a woodpecker climbs upward and engages in a series of behaviors that are consistent with the Ivory-billed Woodpecker but no other species of the region, including delivering a blow that produces an audible double knock and taking off with rapid wingbeats into a flight that immediately transitions into an upward swooping flight that is illustrated in Fig. 3. More