Ecology

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.
Bhatt, S. et al. The effect of malaria control on Plasmodium falciparum in Africa between 2000 and 2015. Nature 526, 207–211 (2015).
CAS  Article  ADS  Google Scholar 
2.
World Health Organization. World Malaria Report 2019 (World Health Organization, Geneva, 2019).
Google Scholar 

3.
Sinka, M. E. et al. A global map of dominant malaria vectors. Parasit. Vectors 5, 69 (2012).
Article  Google Scholar 

4.
Killeen, G. F. et al. Measuring, manipulating and exploiting behaviours of adult mosquitoes to optimise malaria vector control impact. BMJ Glob. Health https://doi.org/10.1136/bmjgh-2016-000212 (2017).
Article  PubMed  PubMed Central  Google Scholar 

5.
Bayoh, M. N. et al. Anopheles gambiae: Historical population decline associated with regional distribution of insecticide-treated bed nets in western Nyanza Province, Kenya. Malar. J. 9, 62. https://doi.org/10.1186/1475-2875-9-62 (2010).
Article  PubMed  PubMed Central  Google Scholar 

6.
Mwangangi, J. M. et al. Shifts in malaria vector species composition and transmission dynamics along the Kenyan coast over the past 20 years. Malar. J. https://doi.org/10.1186/1475-2875-12-13 (2013).
Article  PubMed  PubMed Central  Google Scholar 

7.
Russell, T. L. et al. Impact of promoting longer-lasting insecticide treatment of bed nets upon malaria transmission in a rural Tanzanian setting with pre-existing high coverage of untreated nets. Malar. J. 9, 20 (2010).
Article  Google Scholar 

8.
Killeen, G. F. Characterizing, controlling and eliminating residual malaria transmission. Malar. J. https://doi.org/10.1186/1475-2875-13-330 (2014).
Article  PubMed  PubMed Central  Google Scholar 

9.
Sherrard-Smith, E. et al. Mosquito feeding behavior and how it influences residual malaria transmission across Africa. Proc. Natl. Acad. Sci. 116, 15086–15095. https://doi.org/10.1073/pnas.1820646116 (2019).
CAS  Article  PubMed  Google Scholar 

10.
Knox, T. B. et al. An online tool for mapping insecticide resistance in major Anopheles vectors of human malaria parasites and review of resistance status for the Afrotropical region. Parasit. Vectors 7, 76. https://doi.org/10.1186/1756-3305-7-76 (2014).
CAS  Article  PubMed  PubMed Central  Google Scholar 

11.
Moyes, C. L. et al. Analysis-ready datasets for insecticide resistance phenotype and genotype frequency in African malaria vectors. Sci. Data 6, 121. https://doi.org/10.1038/s41597-019-0134-2 (2019).
CAS  Article  PubMed  PubMed Central  Google Scholar 

12.
Russell, T. L., Beebe, N. W., Cooper, R. D., Lobo, N. F. & Burkot, T. R. Successful malaria elimination strategies require interventions that target changing vector behaviours. Malar. J. https://doi.org/10.1186/1475-2875-12-56 (2013).
Article  PubMed  PubMed Central  Google Scholar 

13.
Govella, N. J. & Ferguson, H. Why use of interventions targeting outdoor biting mosquitoes will be necessary to achieve malaria elimination. Front. Physiol. https://doi.org/10.3389/fphys.2012.00199 (2012).
Article  PubMed  PubMed Central  Google Scholar 

14.
Killeen, G. F. & Chitnis, N. Potential causes and consequences of behavioural resilience and resistance in malaria vector populations: A mathematical modelling analysis. Malar. J. https://doi.org/10.1186/1475-2875-13-97 (2014).
Article  PubMed  PubMed Central  Google Scholar 

15.
Gatton, M. L. et al. The importance of mosquito behavioural adaptations to malaria control in Africa. Evolution https://doi.org/10.1111/evo.12063 (2013).
Article  PubMed  PubMed Central  Google Scholar 

16.
Ranson, H. & Lissenden, N. Insecticide resistance in African Anopheles mosquitoes: A worsening situation that needs urgent action to maintain malaria control. Trends Parasitol. 32, 187–196. https://doi.org/10.1016/j.pt.2015.11.010 (2016).
CAS  Article  PubMed  Google Scholar 

17.
Pates, H. & Curtis, C. Mosquito behavior and vector control. Annu. Rev. Entomol. https://doi.org/10.1146/annurev.ento.50.071803.130439 (2005).
Article  PubMed  Google Scholar 

18.
Gordicho, V. et al. First report of an exophilic Anopheles arabiensis population in Bissau City, Guinea-Bissau: Recent introduction or sampling bias?. Malar. J. 13, 423. https://doi.org/10.1186/1475-2875-13-423 (2014).
Article  PubMed  PubMed Central  Google Scholar 

19.
Kitau, J. et al. Species shifts in the Anopheles gambiae complex: Do LLINs successfully control Anopheles arabiensis?. PLoS One 7, e31481 (2012).
CAS  Article  ADS  Google Scholar 

20.
Smith, A. The preferential indoor resting habitats of Anopheles gambiae in the Umbugwe area of Tanganyika. East Afr. Med. J. 39, 631–635 (1962).
CAS  PubMed  Google Scholar 

21.
Govella, N., Chaki, P. & Killeen, G. Entomological surveillance of behavioural resilience and resistance in residual malaria vector populations. Malar. J. 12, 124 (2013).
Article  Google Scholar 

22.
Coluzzi, M. & Sabatini, A. Chromosomal differentiation and adaptation to human environments in the Anopheles gambiae complex. Trans. R. Soc. Trop. Med. Hyg. 73, 483–497 (1979).
CAS  Article  Google Scholar 

23.
Main, B. J. et al. The genetic basis of host preference and resting behavior in the major African Malaria vector, Anopheles arabiensis. PLOS Genet. 12, e1006303. https://doi.org/10.1371/journal.pgen.1006303 (2016).
CAS  Article  PubMed  PubMed Central  Google Scholar 

24.
Lindblade, K. et al. Impact of sustained use of insecticide-treated bednets on malaria vector species distribution and culicine mosquitoes. J. Med. Entomol. 43, 428–432 (2006).
CAS  Article  Google Scholar 

25.
Russell, T. et al. Increased proportions of outdoor feeding among residual malaria vector populations following increased use of insecticide-treated nets in rural Tanzania. Malar. J. 10, 80 (2011).
Article  Google Scholar 

26.
Norris, L. C. & Norris, D. E. Heterogeneity and changes in inequality of malaria risk after introduction of insecticide-treated bed nets in Macha, Zambia. Am. J. Trop. Med. Hyg. https://doi.org/10.4269/ajtmh.11-0595 (2013).
Article  PubMed  PubMed Central  Google Scholar 

27.
Tirados, I., Costantini, C., Gibson, G. & Torr, S. J. Blood-feeding behaviour of the malarial mosquito Anopheles arabiensis: Implications for vector control. Med. Vet. Entomol. 20, 425–437. https://doi.org/10.1111/j.1365-2915.2006.652.x (2006).
CAS  Article  PubMed  Google Scholar 

28.
Pates, H. & Curtis, C. Mosquito behavior and vector control. Ann. Rev. Entomol. 50, 53–70 (2004).
Article  Google Scholar 

29.
Gillies, M. & Meillon, B. The Anophelinae of Africa south of the Sahara (Ethiopian Zoogeographical Region). S. Afr. Inst. Med. Res. 20, 20 (1968).
Google Scholar 

30.
Meyrowitsch, D. W. et al. Is the current decline in malaria burden in sub-Saharan Africa due to a decrease in vector population?. Malar. J. 10, 188 (2011).
Article  Google Scholar 

31.
Kaindoa, M. N. et al. Interventions that effectively target Anopheles funestus mosquitoes could significantly improve control of persistent malaria transmission in south–eastern Tanzania. PLoS One 12(5), e0177807. https://doi.org/10.1371/journal.pone.0177807 (2017).
CAS  Article  PubMed  PubMed Central  Google Scholar 

32.
Port, G. R. & Boreham, P. F. L. The effects of bednets on feeding by Anopheles gambiae Giles (Diptera: Culicidae). Bull. Entomol. Res. 72, 20 (1982).
Article  Google Scholar 

33.
Lefevre, T. et al. Beyond nature and nurture: Phenotypic plasticity in blood-feeding behavior of Anopheles gambiae s.s. when humans are not readily accessible. Am. J. Trop. Med. Hyg. 81, 1023–1029 (2009).
Article  Google Scholar 

34.
Moiroux, N. et al. Changes in Anopheles funestus biting behavior following universal coverage of long-lasting insecticidal nets in Benin. J. Infect. Dis. https://doi.org/10.1093/infdis/jis565 (2012).
Article  PubMed  Google Scholar 

35.
Petrarca, V. & Beier, J. C. Intraspecific chromosomal polymorphism in the Anopheles gambiae complex as a factor affecting malaria transmission in the Kisumu area of Kenya. Am. J. Trop. Med. Hyg. 46, 20 (1992).
Article  Google Scholar 

36.
Faye, O. et al. Impact of the use of permethrin pre-impregnated mosquito nets on malaria transmission in a hyperendemic village of Senegal. Med. Trop. (Mars) 58, 355–360 (1998).
CAS  Google Scholar 

37.
Cuzin-Ouattara, N. et al. Wide-scale installation of insecticide-treated curtains confers high levels of protection against malaria transmission in a hyperendemic area of Burkina Faso. Trans. R. Soc. Trop. Med. Hyg. 93, 473–479. https://doi.org/10.1016/S0035-9203(99)90343-7 (1999).
CAS  Article  PubMed  Google Scholar 

38.
Ilboudo-Sanogo, E. et al. Insecticide-treated materials, mosquito adaptation and mass effect: Entomological observations after five years of vector control in Burkina Faso. Trans. R. Soc. Trop. Med. Hyg. 95, 353–360. https://doi.org/10.1016/S0035-9203(01)90179-8 (2001).
CAS  Article  PubMed  Google Scholar 

39.
Mathenge, E. et al. Effect of permethrin-impregnated nets on exiting behavior, blood feeding success and time of feeding of malaria mosquitoes (Diptera: Culicidae) in western Kenya. J. Med. Entomol. https://doi.org/10.1603/0022-2585-38.4.531 (2001).
Article  PubMed  Google Scholar 

40.
Renggli, S. et al. Design, implementation and evaluation of a national campaign to deliver 18 million free long-lasting insecticidal nets to uncovered sleeping spaces in Tanzania. Malar. J. 12, 20 (2013).
Article  Google Scholar 

41.
Kramer, K. et al. Effectiveness and equity of the Tanzania National Voucher Scheme for mosquito nets over 10 years of implementation. Malar. J. 16, 255. https://doi.org/10.1186/s12936-017-1902-0 (2017).
Article  PubMed  PubMed Central  Google Scholar 

42.
Schmidt, C. A., Comeau, G., Monaghan, A. J., Williamson, D. J. & Ernst, K. C. Effects of desiccation stress on adult female longevity in Aedes aegypti and Ae. albopictus (Diptera: Culicidae): Results of a systematic review and pooled survival analysis. Parasit. Vectors 11, 267 (2018).
Article  Google Scholar 

43.
Kalra, B. & Parkash, R. Effects of saturation deficit on desiccation resistance and water balance in seasonal populations of the tropical drosophilid Zaprionus indianus. J. Exp. Biol. 219, 3237. https://doi.org/10.1242/jeb.141002 (2016).
Article  PubMed  Google Scholar 

44.
Lwetoijera, H. C. et al. Increasing role of Anopheles funestus and Anopheles arabiensis in malaria transmission in the Kilombero Valley, Tanzania. Malar. J. 331, 20 (2014).
Google Scholar 

45.
Mayagaya, V. et al. The impact of livestock on the abundance, resting behaviour and sporozoite rate of malaria vectors in southern Tanzania. Malar. J. 14, 17 (2015).
Article  Google Scholar 

46.
Mayagaya, V. The impact of livestock on the ecology of malaria vectors and malaria transmission in the Kilombero Valley. Tanzania MSc thesis, University of Dar es Salaam (2010).

47.
Corbel, V. et al. Combination of malaria vector control interventions in pyrethroid resistance area in Benin: A cluster randomised controlled trial. Lancet Infect. Dis. https://doi.org/10.1016/s1473-3099(12)70081-6 (2012).
Article  PubMed  Google Scholar 

48.
Ngowo, H., Kaindoa, E., Matthiopoulos, J., Ferguson, H. & Okumu, F. Variations in household microclimate affect outdoor-biting behaviour of malaria vectors [version 1; referees: 1 approved, 1 approved with reservations]. Vol. 2 (2017).

49.
Russell, T. et al. Impact of promoting long-lasting insecticide treatment of bednets upon malaria transmission in a rural Tanzania setting with pre existing high coverage of untreated nets. Malar. J. 9, 187 (2010).
Article  Google Scholar 

50.
Katharina Sophia, K. et al. Impact of ENSO 2016–17 on regional climate and malaria vector dynamics in Tanzania. Environ. Res. Lett. 20, 20 (2019).
Google Scholar 

51.
Kessler, S. & Guerin, P. M. Responses of Anopheles gambiae, Anopheles stephensi, Aedes aegypti, and Culex pipiens mosquitoes (Diptera: Culicidae) to cool and humid refugium conditions. J. Vector Ecol. https://doi.org/10.3376/1081-1710(2008)33[145:roagas]2.0.co;2 (2008).
Article  PubMed  Google Scholar 

52.
Chaccour, C. & Killeen, G. F. Mind the gap: Residual malaria transmission, veterinary endectocides and livestock as targets for malaria vector control. Malar. J. 15, 24 (2016).
Article  Google Scholar 

53.
Lyimo, I. N., Kessy, S. T., Mbina, K. F., Daraja, A. A. & Mnyone, L. L. Ivermectin-treated cattle reduces blood digestion, egg production and survival of a free-living population of Anopheles arabiensis under semi-field condition in south-eastern Tanzania. Malar. J. https://doi.org/10.1186/s12936-017-1885-x (2017).
Article  PubMed  PubMed Central  Google Scholar 

54.
Meza, F. C. et al. Mosquito electrocuting traps for directly measuring biting rates and host-preferences of Anopheles arabiensis and Anopheles funestus outdoors. Malar. J. 18, 83. https://doi.org/10.1186/s12936-019-2726-x (2019).
Article  PubMed  PubMed Central  Google Scholar 

55.
Iwashita, H. et al. Push by a net, pull by a cow: Can zooprophylaxis enhance the impact of insecticide treated bed nets on malaria control?. Parasit. Vectors 7, 52 (2014).
Article  Google Scholar 

56.
Tirados, I., Gibson, G., Young, S. & Torr, S. Are herders protected by their herds? An experimental analysis of zooprophylaxis against the malaria vector Anopheles arabiensis. Malar. J. 10, 68 (2011).
Article  Google Scholar 

57.
Donnelly, B., Berrang-Ford, L., Ross, N. A. & Michel, P. A systematic, realist review of zooprophylaxis for malaria control. Malar. J. 14, 313. https://doi.org/10.1186/s12936-015-0822-0 (2015).
Article  PubMed  PubMed Central  Google Scholar 

58.
Killeen, G. F. et al. Developing an expanded vector control toolbox for malaria elimination. BMJ Glob. Health 2, e000211. https://doi.org/10.1136/bmjgh-2016-000211 (2017).
Article  PubMed  PubMed Central  Google Scholar 

59.
Ayala, D., Ullastres, A. & Gonzalez, J. Adaptation through chromosomal inversions in Anopheles. Front. Genet. https://doi.org/10.3389/fgene.2014.00129 (2014).
Article  PubMed  PubMed Central  Google Scholar 

60.
Carrasco, D. et al. Behavioural adaptations of mosquito vectors to insecticide control. Curr. Opin. Insect Sci. 34, 48–54. https://doi.org/10.1016/j.cois.2019.03.005 (2019).
Article  PubMed  Google Scholar 

61.
Matowo, N. S. et al. Fine-scale spatial and temporal heterogeneities in insecticide resistance profiles of the malaria vector, Anopheles arabiensis in rural south-eastern Tanzania. Wellcome Open Res. 2, 20 (2017).
Article  Google Scholar 

62.
Okumu, F. et al. Comparative field evaluation of combinations of long-lasting insecticide treated nets and indoor residual spraying, relative to either method alone, for malaria prevention in an area where the main vector is Anopheles arabiensis. Parasit. Vectors 6, 46 (2013).
Article  Google Scholar 

63.
Briët, O. J. T. et al. Applications and limitations of Centers for Disease Control and Prevention miniature light traps for measuring biting densities of African malaria vector populations: A pooled-analysis of 13 comparisons with human landing catches. Malar. J. 14, 1–13. https://doi.org/10.1186/s12936-015-0761-9 (2015).
Article  Google Scholar 

64.
Clark, G. G., Seda, H. & Gubler, D. J. Use of the “CDC backpack aspirator” for surveillance of Aedes aegypti in San Juan, Puerto Rico. J. Am. Mosq. Control Assoc. 10, 119–124 (1994).
CAS  PubMed  Google Scholar 

65.
Kreppel, K. S. et al. Comparative evaluation of the Sticky-Resting-Box-Trap, the standardised resting-bucket-trap and indoor aspiration for sampling malaria vectors. Parasit. Vectors 8, 462. https://doi.org/10.1186/s13071-015-1066-0 (2015).
Article  PubMed  PubMed Central  Google Scholar 

66.
Allen, R., Pereira, L. S., Raes, D., & Smith, M. Crop evapotranspiration—guidelines for computing crop water requirements. Food and Agriculture Organization (FAO); United Nations, FAO, Irrigation and Drainage Paper 56 (1998).

67.
Funk, C. et al. The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes. Sci. Data 2, 150066. https://doi.org/10.1038/sdata.2015.66 (2015).
Article  PubMed  PubMed Central  Google Scholar 

68.
Scott, J., Brodgon, W. & Collins, F. Identification of single specimens of Anopheles gambiae complex by polymerase chain reaction. Am. J. Trop. Med. Hyg. 49, 520–529 (1993).
CAS  Article  Google Scholar 

69.
Kaindoa, E. W. et al. New evidence of mating swarms of the malaria vector, Anopheles arabiensis in Tanzania. Wellcome Open Res. 2, 88. https://doi.org/10.12688/wellcomeopenres.12458.1 (2017).
Article  PubMed  PubMed Central  Google Scholar 

70.
Lee, Y., Weakley, A. M., Nieman, C. C., Malvick, J. & Lanzaro, G. C. A multi-detection assay for malaria transmitting mosquitoes. J. Vis. Exp. https://doi.org/10.3791/52385 (2015).
Article  PubMed  PubMed Central  Google Scholar 

71.
Cameron, A. C. & Trivedi, P. K. Regression-based tests for overdispersion in the Poisson model. J. Econom. 46, 347–364. https://doi.org/10.1016/0304-4076(90)90014-K (1990).
MathSciNet  Article  Google Scholar 

72.
R-Core-Team. R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, 2016).
Google Scholar 

73.
Pinheiro, J., Bates, D., DebRoy, S. & Sarkar, D. Linear and nonlinear mixed effects models. R Packag. Vers. 3, 57 (2007).
Google Scholar 

74.
Brooks, M. E. et al. Modeling zero-inflated count data with glmmTMB. BioRxiv 132753, 20 (2017).
Google Scholar  More

1.
Ho A, Di Lonardo DP, Bodelier PLE. Revisiting life strategy concepts in environmental microbial ecology. FEMS Microbiol Ecol. 2017;93:fix006.
Google Scholar 
2.
Poindexter JS. Oligotrophy. In: Alexander M, editor. Advances in microbial ecology. Springer US, Boston, MA: Springer US; 1981. pp. 63–89.

3.
Madigan MT, Bender KS, Buckley DH, Sattley WM, Stahl DA. Brock Biology of Microorganisms, 15th Global edition. Boston, US: Benjamin Cummins. 2018.

4.
Navarro Llorens JM, Tormo A, Martínez-García E. Stationary phase in gram-negative bacteria. FEMS Microbiol Rev. 2010;34:476–95.
PubMed  Google Scholar 

5.
Wood TK, Knabel SJ, Kwan BW. Bacterial persister cell formation and dormancy. Appl Environ Microbiol. 2013;79:7116–21.
CAS  PubMed  PubMed Central  Google Scholar 

6.
Klotz A, Georg J, Bučinská L, Watanabe S, Reimann V, Januszewski W, et al. Awakening of a dormant cyanobacterium from nitrogen chlorosis reveals a genetically determined program. Curr Biol. 2016;26:2862–72.
CAS  PubMed  Google Scholar 

7.
Setlow P, Wang S, Li Y-Q. Germination of spores of the orders Bacillales and Clostridiales. Annu Rev Microbiol. 2017;71:459–77.
CAS  PubMed  Google Scholar 

8.
Song S, Wood TK. ppGpp ribosome dimerization model for bacterial persister formation and resuscitation. bioRxiv. 2019. https://doi.org/10.1101/663658.

9.
Fenton AK, Kanna M, Woods R, Aizawa S, Sockett RE. Shadowing the actions of a predator: Backlit fluorescent microscopy reveals synchronous nonbinary septation of predatory Bdellovibrio inside prey and exit through discrete bdelloplast pores. J Bacteriol. 2010;192:6329–35.
CAS  PubMed  PubMed Central  Google Scholar 

10.
Makowski Ł, Donczew R, Weigel C, Zawilak-Pawlik A, Zakrzewska-Czerwinska J. Initiation of chromosomal replication in predatory bacterium Bdellovibrio bacteriovorus. Front Microbiol. 2016;7.

11.
Rotem O, Pasternak Z, Jurkevitch E. The genus Bdellovibrio and like organisms. The prokaryotes: deltaproteobacteria and epsilonproteobacteria. 2014. pp. 3–17.

12.
Lambert C, Evans KJ, Till R, Hobley L, Capeness M, Rendulic S, et al. Characterizing the flagellar filament and the role of motility in bacterial prey-penetration by Bdellovibrio bacteriovorus. Mol Microbiol. 2006;60:274–86.
CAS  PubMed  PubMed Central  Google Scholar 

13.
Thomashow LS, Rittenberg SC. Waveform analysis and structure of flagella and basal complexes from Bdellovibrio bacteriovorus 109J. J Bacteriol. 1985;163:1038–46.
CAS  PubMed  PubMed Central  Google Scholar 

14.
Hespell RB, Rosson RA, Thomashow MF, Rittenberg SC.  Respiration of Bdellovibrio bacteriovorus strain 109J and its energy substrates for intraperiplasmic growth. J Bacteriol. 1973;113:1280–8.
CAS  PubMed  PubMed Central  Google Scholar 

15.
Hespell RB, Thomashow MF, Rittenberg SC. Changes in cell composition and viability of Bdellovibrio bacteriovorus during starvation. Arch Microbiol. 1974;97:313–27.
CAS  PubMed  Google Scholar 

16.
Paix B, Ezzedine JA, Jacquet S. Diversity, dynamics, and distribution of Bdellovibrio and like organisms in perialpine lakes. Appl Environ Microbiol. 2019;85.

17.
Varon M, Fine M, Stein A. The maintenance of Bdellovibrio at low prey density. Microb Ecol. 1984;10:95–8.
CAS  PubMed  Google Scholar 

18.
Varon M, Zeigler BP. Bacterial predator-prey interaction at low prey density. Appl Environ Microbiol. 1978;36:11–7.
CAS  PubMed  PubMed Central  Google Scholar 

19.
Chen H, Young S, Berhane TK, Williams HN. Predatory Bacteriovorax communities ordered by various prey species. PLoS ONE. 2012;7.

20.
Rogosky AM, Moak PL, Emmert EAB. Differential predation by Bdellovibrio bacteriovorus 109J. Curr Microbiol. 2006;52:81–5.
CAS  PubMed  Google Scholar 

21.
Jurkevitch E, Minz D, Ramati B, Barel G. Prey range characterization, ribotyping, and diversity of soil and rhizosphere Bdellovibrio spp. isolated on phytopathogenic bacteria. Appl Environ Microbiol. 2000;66:2365–71.
CAS  PubMed  PubMed Central  Google Scholar 

22.
Kandel PP, Pasternak Z, van Rijn J, Nahum O, Jurkevitch E. Abundance, diversity and seasonal dynamics of predatory bacteria in aquaculture zero discharge systems. FEMS Microbiol Ecol. 2014;89:149–61.
CAS  PubMed  Google Scholar 

23.
Pineiro SA, Williams HN, Stine OC, Piñeiro SA, Williams HN, Stine OC. Phylogenetic relationships amongst the saltwater members of the genus Bacteriovorax using rpoB sequences and reclassification of Bacteriovorax stolpii as Bacteriolyticum stolpii gen. nov., comb. nov. Int J Syst Evol Microbiol. 2008;58:1203–9.
CAS  PubMed  Google Scholar 

24.
Chen H, Athar R, Zheng G, Williams HN. Prey bacteria shape the community structure of their predators. ISME J. 2011;5:1314–22.
PubMed  PubMed Central  Google Scholar 

25.
Shatzkes K, Connell ND, Kadouri DE. Predatory bacteria: a new therapeutic approach for a post-antibiotic era. Future Microbiol. 2017;12:469–72.
CAS  PubMed  Google Scholar 

26.
Guo Y, Yan L, Cai J. Effects of Bdellovibrio and like organisms on survival and growth performance of juvenile turbot, scophthalmus maximus. J World Aquac Soc. 2016;47:633–45.
Google Scholar 

27.
Youdkes D, Helman Y, Burdman S, Matan O, Jurkevitch E. Potential control of potato soft rot disease by the obligate predators Bdellovibrio and like organisms. Appl Environ Microbiol. 2020;86.

28.
Sathyamoorthy R, Maoz A, Pasternak Z, Im H, Huppert A, Kadouri D, et al. Bacterial predation under changing viscosities. Environ Microbiol. 2019;21:2997–3010.
CAS  PubMed  Google Scholar 

29.
Hobley L, Fung RKY, Lambert C, Harris MATS, Dabhi JM, King SS, et al. Discrete cyclic di-GMP-dependent control of bacterial predation versus axenic growth in Bdellovibrio bacteriovorus. PLoS Pathog. 2012;8.

30.
Karunker I, Rotem O, Dori-Bachash M, Jurkevitch E, Sorek R. A global transcriptional switch between the attack and growth forms of Bdellovibrio bacteriovorus. PLoS ONE. 2013;8.

31.
Amikam D, Galperin MY. PilZ domain is part of the bacterial c-di-GMP binding protein. Bioinformatics. 2006;22:3–6.
CAS  PubMed  Google Scholar 

32.
Ko J, Ryu K-S, Kim H, Shin J-S, Lee J-O, Cheong C, et al. Structure of PP4397 reveals the molecular basis for different c-di-GMP binding modes by PilZ domain proteins. J Mol Biol. 2010;398:97–110.
CAS  PubMed  Google Scholar 

33.
Wirebrand L, Österberg S, López-Sánchez A, Govantes F, Shingler V. PP4397/FlgZ provides the link between PP2258 c-di-GMP signalling and altered motility in Pseudomonas putida. Sci Rep. 2018;8:1–10.
CAS  Google Scholar 

34.
Shanks RMQ, Davra VR, Romanowski EG, Brothers KM, Stella NA, Godboley D, et al. An eye to a kill: using predatory bacteria to control gram-negative pathogens associated with ocular infections. PLOS ONE. 2013;8:e66723.
CAS  PubMed  PubMed Central  Google Scholar 

35.
Wurtzel O, Dori-Bachash M, Pietrokovski S, Jurkevitch E, Sorek R. Mutation detection with next-generation resequencing through a mediator genome. PLoS ONE. 2010;5.

36.
Pasternak Z, Njagi M, Shani Y, Chanyi R, Rotem O, Lurie-Weinberger MN, et al. In and out: an analysis of epibiotic vs periplasmic bacterial predators. ISME J. 2014;8:625–35.
CAS  PubMed  Google Scholar 

37.
Pletnev P, Osterman I, Sergiev P, Bogdanov A, Dontsova O. Survival guide: Escherichia coli in the stationary phase. Acta Nat. 2015;7:22–33.
CAS  Google Scholar 

38.
Kazmierczak MJ, Wiedmann M, Boor KJ. Alternative sigma factors and their roles in bacterial virulence. Microbiol Mol Biol Rev. 2005;69:527–43.
CAS  PubMed  PubMed Central  Google Scholar 

39.
Paget MS. Bacterial sigma factors and anti-sigma factors: structure, function and distribution. Biomolecules. 2015;5:1245–65.
CAS  PubMed  PubMed Central  Google Scholar 

40.
Avidan O, Petrenko M, Becker R, Beck S, Linscheid M, Pietrokovski S, et al. Identification and characterization of differentially-regulated type IVb pilin genes necessary for predation in obligate bacterial predators. Sci Rep. 2017;7:1–12.

41.
Barembruch C, Hengge R. Cellular levels and activity of the flagellar sigma factor FliA of Escherichia coli are controlled by FlgM-modulated proteolysis. Mol Microbiol. 2007;65:76–89.
CAS  PubMed  Google Scholar 

42.
Rendulic S, Jagtap P, Rosinus A, Eppinger M, Baar C, Lanz C, et al. A predator unmasked: life cycle of Bdellovibrio bacteriovorus from a genomic perspective. Science. 2004;303:689–92.
CAS  PubMed  Google Scholar 

43.
Nyström T. Stationary-phase physiology. Annu Rev Microbiol. 2004;58:161–81.
PubMed  Google Scholar 

44.
Browning AP, Sharp JA, Mapder T, Baker CM, Burrage K, Simpson MJ. Persistence is an optimal hedging strategy for bacteria in volatile environments. bioRxiv. 2019. https://doi.org/10.1101/2019.12.19.883645.

45.
Ratcliff WC, Denison RF. Individual-level bet hedging in the bacterium Sinorhizobium meliloti. Curr Biol. 2010;20:1740–4.
CAS  PubMed  Google Scholar 

46.
Zhang X-X, Rainey PB. Bet hedging in the underworld. Genome Biol. 2010;11:137.
PubMed  PubMed Central  Google Scholar 

47.
Franklin RB, Mills AL. Multi-scale variation in spatial heterogeneity for microbial community structure in an eastern Virginia agricultural field. FEMS Microbiol Ecol. 2003;44:335–46.
CAS  PubMed  Google Scholar 

48.
Manderscheid B, Matzner E. Spatial heterogeneity of soil solution chemistry in a mature Norway spruce (Picea abies (L.) Karst.) stand. Water Air Soil Pollut. 1995;85:1185–90.
CAS  Google Scholar 

49.
Ranjard L, Lejon DPH, Mougel C, Schehrer L, Merdinoglu D, Chaussod R. Sampling strategy in molecular microbial ecology: Influence of soil sample size on DNA fingerprinting analysis of fungal and bacterial communities. Environ Microbiol. 2003;5:1111–20.
CAS  PubMed  Google Scholar 

50.
Jenal U, Reinders A, Lori C. Cyclic di-GMP: second messenger extraordinaire. Nat Rev Microbiol. 2017;15:271–84.
CAS  PubMed  Google Scholar 

51.
Boehm A, Kaiser M, Li H, Spangler C, Kasper CA, Ackermann M, et al. Second messenger-mediated adjustment of bacterial swimming velocity. Cell. 2010;141:107–16.
CAS  PubMed  Google Scholar 

52.
Paul K, Nieto V, Carlquist WC, Blair DF, Harshey RM. The c-di-GMP binding protein YcgR controls flagellar motor direction and speed to affect chemotaxis by a ‘Backstop Brake’ mechanism. Mol Cell. 2010;38:128–39.
CAS  PubMed  PubMed Central  Google Scholar 

53.
Dattner I, Miller E, Petrenko M, Kadouri DE, Jurkevitch E, Huppert A. Modelling and parameter inference of predator–prey dynamics in heterogeneous environments using the direct integral approach. J R Soc Interface. 2017;14:20160525.
PubMed  PubMed Central  Google Scholar 

54.
Hol FJH, Rotem O, Jurkevitch E, Dekker C, Koster DA. Bacterial predator–prey dynamics in microscale patchy landscapes. Proc R Soc B Biol Sci. 2016;283:20152154.
Google Scholar 

55.
Gabel CV, Berg HC. The speed of the flagellar rotary motor of Escherichia coli varies linearly with protonmotive force. Proc Natl Acad Sci USA. 2003;100:8748–51.
CAS  PubMed  Google Scholar 

56.
Gadkari D, Stolp H. Energy metabolism of Bdellovibrio bacteriovorus. I. Energy production, ATP pool, energy charge. Arch Microbiol. 1975;102:179–85.
CAS  PubMed  Google Scholar 

57.
Shioi JI, Galloway RJ, Niwano M, Chinnock RE, Taylor BL. Requirement of ATP in bacterial chemotaxis. J Biol Chem. 1982;257:7969–75.

58.
Fang X, Gomelsky M. A post-translational, c-di-GMP-dependent mechanism regulating flagellar motility. Mol Microbiol. 2010;76:1295–305.
CAS  PubMed  Google Scholar 

59.
Varon M. Interaction of Bdellovibrio with its prey in mixed microbial populations. Microb Ecol. 1981;7:97–105.
CAS  PubMed  Google Scholar 

60.
Kessel M, Shilo M. Relationship of Bdellovibrio elongation and fission to host cell size. J Bacteriol. 1976;128:477–80.
CAS  PubMed  PubMed Central  Google Scholar 

61.
LaMarre AG, Straley SC, Conti SF. Chemotaxis toward amino acids by Bdellovibrio bacteriovorus. J Bacteriol. 1977;131:201–7.
CAS  PubMed  PubMed Central  Google Scholar 

62.
Chauhan A, Williams HN. Response of Bdellovibrio and like organisms (BALOs) to the migration of naturally occurring bacteria to chemoattractants. Curr Microbiol. 2006;53:516–22.
CAS  PubMed  Google Scholar 

63.
Feng S, Tan CH, Constancias F, Kohli GS, Cohen Y, Rice SA. Predation by Bdellovibrio bacteriovorus significantly reduces viability and alters the microbial community composition of activated sludge flocs and granules. FEMS Microbiol Ecol. 2017;93.

64.
Kadouri DE, O’Toole GA. Susceptibility of biofilms to Bdellovibrio bacteriovorus attack. Appl Environ Microbiol. 2005;71:4044–51.
CAS  PubMed  PubMed Central  Google Scholar 

65.
Szabó E, Liébana R, Hermansson M, Modin O, Persson F, Wilén B-MB-MB-M, et al. Comparison of the bacterial community composition in the granular and the suspended phase of sequencing batch reactors. AMB Express. 2017;7:168.
PubMed  PubMed Central  Google Scholar 

66.
Lambert C, Smith MCM, Sockett RE. A novel assay to monitor predator-prey interactions for Bdellovibrio bacteriovorus 109 J reveals a role for methyl-accepting chemotaxis proteins in predation. Environ Microbiol. 2003;5:127–32.
CAS  PubMed  Google Scholar 

67.
Petrenko M, Friedman SP, Fluss R, Pasternak Z, Huppert A, Jurkevitch E. Spatial heterogeneity stabilizes predator–prey interactions at the microscale while patch connectivity controls their outcome. Environ Microbiol. 2019;22:694–704.

68.
Mukherjee S, Brothers KM, Shanks RMQQ, Kadouri DE. Visualizing Bdellovibrio bacteriovorus by using the tdTomato fluorescent protein. Appl Environ Microbiol. 2015;82:1653–61.
PubMed  Google Scholar 

69.
Jurkevitch E. Isolation and classification of Bdellovibrio and like organisms. Curr Protoc Microbiol. 2012;Chapter 7:Unit 7B.1.
Google Scholar 

70.
Peters JM, Koo B-M, Patino R, Heussler GE, Hearne CC, Qu J, et al. Enabling genetic analysis of diverse bacteria with Mobile-CRISPRi. Nat Microbiol. 2019;4:244–50.
CAS  PubMed  PubMed Central  Google Scholar 

71.
Copeland MF, Weibel DB. Bacterial swarming: a model system for studying dynamic self-assembly. Soft Matter. 2009;5:1174–87.
CAS  PubMed  PubMed Central  Google Scholar 

72.
Rotem O, Pasternak Z, Shimoni E, Belausov E, Porat Z, Pietrokovski S, et al. Cell-cycle progress in obligate predatory bacteria is dependent upon sequential sensing of prey recognition and prey quality cues. Proc Natl Acad Sci. 2015;112:E6028–37.
CAS  PubMed  Google Scholar  More

Bioleaching kinetics
Comparison of Li bioleaching by three various types of organisms (Fig. 1) revealed that the leaching kinetics in systems with yeast R. mucilaginosa was the fastest. Presence of Li in solution was detected at 6th day of the process. After initial faster bioleaching within first 6 days (285.5 µg l−1), there was a gradual decrease of Li concentration in solution due to Li bioaccumulation into the biomass up to 13th day and later stable Li concentration in range of 240–250 µg l−1 was observed suggesting that the rate of bioleaching and bioaccumulation were equal.
Figure 1

Kinetics of Li bioleaching from lepidolite by consortium of A. ferrooxidans and A. thiooxidans (bacteria), A. niger (fungi) and R. mucilaginosa (yeast) (A), long-term kinetics of Li bioleaching by bacteria (B) (fungi: initial ore concentration 10 g l−1, t = 21 °C, pH = 5.1, statically, standard medium, yeast: 10 g l−1, t = 21 °C, pH = 5.1, shaking 160 rpm, rich medium and bacteria: 10 g l−1, t = 30 °C, pH = 1.5, statically, poor medium).

Full size image

The lowest amount of Li was bioleached by fungi A. niger. Under this bioleaching conditions Li was for the first time observed in solution after 26 days of the process. Its concentration gradually increased later on. Again bioaccumulation was observed affecting the amount of Li in the solution.
In the case of bacteria, medium composition was the most important for Li bioleaching. In nutrient rich medium for acidophilic chemoautotrophic acidithiobacilli which contained energy sources (Fe2+ ions and S0) no Li bioleaching was observed during the whole process time. However, in the medium with limited amount of nutrients and energy sources containing just sulphuric acid and elemental sulphur, Li+ ions presence was observed at 21st day for the first time. Bacteria were probably forced to utilize nutrients necessary for their life directly in the leached material. During the first 77 days the lithium bioleaching kinetics was very slow but this stage was followed by the sharp increase of bioleaching rate (400 times increase of the bioleaching rate was observed) resulting in 11 mg l−1 of solubilised Li at the end of the bioleaching experiments (after 336 days). The rapid change in the bioleaching rate might be attributed to the changes of mineral structure due to bacterial activity. No Li was found in control experiments using the media without microorganisms addition.
Kinetic analysis
To kinetically interpret the heterogeneous non-catalytic reaction for lepidolite bioleaching the shrinking core model (SCM) was used. The assumptions to use the model are based on the three facts—(i) mixed lepidolite particles are considered as nonporous particles, (ii) ore grains gradually shrank and (iii) the product layers form around the unreacted grains20. The development and verification of the model were previously described in details by several authors20,21.
Experimental data obtained for all three studied bioleaching systems were substituted into both equations of SCM model. In the case of bacterial bioleaching a plot of 1−(1−X)1/3 versus time (Fig. 2) was found a straight line suggesting that chemical reaction and outer diffusion are the rate controlling steps of the process of bacterial bioleaching. Changes of rate constant, kr, (apparent from slopes of the plots) can be visible, as well. The linear relationship was obtained in the initial stage of bioleaching (R2 = 0.9944) and later at the day 77 the rate of the process changed but still showed the good fitting obtained by plotting 1−(1−X)1/3 versus time (R2 = 0.9991). This changes are very well visible also in the previous Fig. 1 showing the increase of Li+ ion concentration within the experimental period.
Figure 2

Plot of 1−(1−X)1/3 versus time for Li recovery by consortium of bacteria (initial ore concentration 10 g l−1, t = 30 °C, pH = 1.5, statically, poor media).

Full size image

However, the SCM model did not fit to the bioleaching data of two other bioleaching systems, using fungi and yeasts. Obviously, parallel bioaccumulation of Li+ ions into the biomass was responsible for considerably different bioleaching behaviour.
Changes of pH
Conditions of bioleaching experiments (pH, medium composition) were adjusted according the type of the microorganism used. Independently of conditions, the decrease of pH (Fig. 3) was recorded in all three bioleaching system. The most obvious decrease in pH occurred in bioleaching by microscopic fungi A. niger, with a pH decrease from 5.1 to 3 within first 12 days, followed by slow decrease to 2.5 until the end of the experiment. According to various authors22,23, it can be suggested that organic acids, considered the main fungal bioleaching agents, were produced. In the control medium a small increase in pH (from 5.2 to 5.6) was observed.
Figure 3

Changes of pH during bioleaching of lepidolite by consortium of A. ferrooxidans and A. thiooxidans (bacteria), A. niger (fungi) and R. mucilaginosa (yeast) (fungi: initial ore concentration 10 g l−1, t = 21 °C, pH = 5.1, statically, standard medium, yeast: 10 g l−1, t = 21 °C, pH = 5.1, shaking 160 rpm, rich medium and bacteria: 10 g l−1, t = 30 °C, pH = 1.5, statically, poor medium).

Full size image

A similar pattern was also observed in bacterial bioleaching, in which fast decrease of pH to 1.2 was observed during first 7 days followed by slow decrease to 0.9. Later the pH was stable in range of 0.9–1.2. Probably bacteria A. thiooxidans were mainly responsible for such pH decrease. In the control without bacteria addition the pH initially decreased from 1.5 to 1.3 and later increased and remained at 1.5.
As shown in Fig. 3 fast pH decrease was observed during first 6 days of bioleaching with yeast R. mucilaginosa from initial 5.1 to 4.1. Later pH did not change until 20th day followed by slow decrease to 3.5 at 30th day. In control media, without microorganisms, pH value slowly increased from initial 5.1 to final 5.5.
Bioleaching mechanisms
According to obtained results different mechanisms can be suggested for lepidolite bioleaching by biological systems studied. Mechanisms of Li bioleaching from lepidolite by A. niger fungus may be attributed to combination of biochemical (due to organic acids production) and biomechanical (due to hyphae penetration) leaching mechanisms. Significant drop of pH values indicates increased concentration of organic acids in the media as the result of high metabolic activity of the A. niger cell what was confirmed by various authors studying bioleaching by the microscopic fungi14,22,23,24,25. However, lepidolite interpenetration by A. niger hyphae growing along cleavages was observed by SEM analysis of solid residue after bioleaching, as well (Supplementary Information, Fig. S1), suggesting that direct biomechanical deterioration of lepidolite was also a part of the whole lithium extraction mechanism. However, according to Gadd26 the biochemical activities of microorganisms play more significant role than mechanical degradation.
Mechanisms of lepidolite bioleaching by bacteria is unknown. However, from abovementioned results it is obvious that no other substance except H+ ions contributed to the dissolution of Li+ ions. These results suggested that Li in lepidolite was dissolved by acid. Probably the mechanisms suggested by Liu et al.20 for leaching of lepidolite in sulphuric acid may be applied to bioleaching by acidophilic bacteria with sulphuric acid as a main bioleaching agent, as well. The main reaction of mixed alkali metal bioleaching may be expressed as follows:

$$ {text{M}}_{{2}} {text{O }} + {text{ H}}_{{2}} {text{SO}}_{{4}} = {text{ M}}_{{2}} {text{SO}}_{{4}} + {text{ H}}_{{2}} {text{O}} $$
(1)

where M presents alkali metals. Metallic elements from lepidolite are dissolved to form metal sulphates and mixed alums in the solution resulting just in partial lepidolite dissolution20. Overal reaction of lepidolite bioleaching in sulphuric acid produced by bacteria may be adopted from Onalbaeva et al.11:

$$ {text{3Li}}_{{2}} {text{O}}cdot{text{2K}}_{{2}} {text{O}}cdot{text{5Al}}_{{2}} {text{O}}_{{3}} cdot{1}0{text{SiO}}_{{2}} cdot{text{2SiF}}_{{4}} + { 2}0{text{H}}_{{2}} {text{SO}}_{{4}} = {text{ 3Li}}_{{2}} {text{SO}}_{{4}} + {text{ 2K}}_{{2}} {text{SO}}_{{4}} + {text{ 5Al}}_{{2}} left( {{text{SO}}_{{4}} } right)_{{3}} + {text{ 11SiO}}_{{2}} + {text{ H}}_{{2}} {text{SiF}}_{{6}} + {text{ 18H}}_{{2}} {text{O }} + {text{ 2HF}} $$
(2)

$$ {text{3Li}}_{{2}} {text{O}}cdot{text{2K}}_{{2}} {text{O}}cdot{text{5Al}}_{{2}} {text{O}}_{{3}} cdot{text{12SiO}}_{{2}} cdot{text{4H}}_{{2}} {text{O }} + { 2}0{text{H}}_{{2}} {text{SO}}_{{4}} = {text{ 3Li}}_{{2}} {text{SO}}_{{4}} + {text{ 2K}}_{{2}} {text{SO}}_{{4}} + {text{ 5Al}}_{{2}} left( {{text{SO}}_{{4}} } right)_{{3}} + {text{ 12SiO}}_{{2}} + {text{ 24H}}_{{2}} {text{O}} $$
(3)

Also Guo et al.27observed that increased H+ concentration catalysed the process of Li leaching from lepidolite via accelerating the protonation of the crystal lattices.
X-ray diffraction analysis
XRD analysis was applied in this study for phase identification and structural changes evaluation of samples before and after bioleaching in all three studied systems. Significant differences in mineralogical composition of leaching residue among the three studied bioleaching systems are visible from XRD spectra comparison (Supplementary Information, Fig. S2) suggesting that different mechanisms can be responsible for bioleaching. While bacterial bioleaching led to the disappearing of muscovite phase from XRD spectrum, the fungal bioleaching led to the appearance of new silicate phase (SiO2) and muscovite was found a dominant phase. According to Liu et al.20 presence of quartz in the spectrum at the end of the process may correspond with alkali metal dissolution from the silicate lattice. Phase changes were observed also after bioleaching by yeast R. mucilaginosa. Reallocation and significant decrease of diffraction peaks intensity was observed and similarly as in case of microscopic fungi muscovite has become a dominant phase while polylithionite phase significantly weakened. Based on the results, it can be suggested that the bioleaching mechanisms of lepidolite by fungi and yeast may be similar, however, in the case of bacteria the mechanisms might be significantly different. Further experiments are necessary to understand the mechanisms behind the lepidolite bioleaching.
Li distribution
Bioaccumulation of lithium into the biomass was observed when heterotrophic microorganisms A. niger and R. mucilaginosa were used (Fig. 4A). No bioaccumulation was found when bioleaching by consortium of acidophilic bacteria was studied. It can be suggested that the process of Li recovery by A. niger and R. mucilaginosa is a combination of two basic processes – initial bioleaching (metal solubilisation) followed by rapid bioaccumulation (intracellular lithium accumulation). It is possible that lithium bioaccumulation could significantly contribute to its solubilisation as released Li+ cations were fast accumulated in the cells and thus “pulled” the equilibrium resulting in the increased efficiency of the Li dissolution.
Figure 4

Distribution of Li between solution and biomass during bioleaching of lepidolite (A) and efficiency of the lepidolite bioleaching (B) by consortium of A. ferrooxidans and A. thiooxidans (bacteria), A. niger (fungi) and R. mucilaginosa (yeast) (fungi: initial ore concentration 10 g l−1, t = 21 °C, pH = 5.1, statically, standard medium, yeast: 10 g l−1, t = 21 °C, pH = 5.1, shaking 160 rpm, rich medium and bacteria: 10 g l−1, t = 30 °C, pH = 1.5, statically, poor medium).

Full size image

The highest amount of lithium was accumulated by R. mucilaginosa cells, representing 92% of the total amount of Li recovered from the ore. In the case of microscopic fungi A. niger, produced biomass accumulated 77% of the total solubilised Li. Distribution of Li between solution and biomass of particular microorganisms is shown in Fig. 4A. It is obvious that in both cases (fungi and yeast) bioaccumulation is dominant process of Li recovery and just small amount of Li+ ions remain in solution.
Bioleaching efficiency
The bioleaching efficiency is given as a sum of two processes – Li dissolution and its accumulation in the biomass. The final bioleaching yields for consortium of A. ferrooxidans and A. thiooxidans, fungi A. niger and R. mucilaginosa were found to be 8.8%, 0.2% and 1.1%, respectively. The results suggested that the most efficient among all three studied systems was the consortium of acidophilic bacteria A. ferrooxidans and A. thiooxidans (Fig. 4B) with the final bioleaching yield of almost 9%. On the other hand, very long time (336 days) was necessary for the process. Reichel et al.15 found 11% Li recovery from zinnwaldite using consortium of sulphur-oxidising bacteria, however, authors reported just 14 days for observed Li bioleaching efficiency although they do not found clear explanation of higher bioleaching efficiency in comparison with chemical leaching.
The lowest bioleaching yield was observed when A. niger was used. Rezza et al.13,14 used A. niger for Li bioleaching from spodumene with highest recovery of 0.75 mg l−1 of lithium, they do not reported any bioaccumulation.
Composition of medium had very strong effect on bioleaching efficiency by R. mucilaginosa as in nutrient rich medium due to significantly higher biomass production majority of Li has accumulated into the biomass resulting in 3 times higher final Li recovery. There were also morphological differences observed between yeasts cultivated in nutrient rich and poor environments with spherical shape and thin exopolymer layer of 0.48 µm for yeast from nutrient rich media in comparison with oval cells and thick exopolymer layer (1.8 µm) when cultivated in nutrient poor medium17.
Despite of quite low bioleaching efficiency there is clearly visible potential of all three biological systems for Li recovery from hard rocks. Even with low Li concentration in solution after bioleaching, the lithium concentration in the leaching solution resembles the lithium concentration of sea water (0.1–0.2 mg l−1) and brines (0.1–2 g l−1) considered for economic recovery28,29. That shows that the leaching solution is generally suitable for further processing15.
Due to the expensive separation of Li from leaching liquor, the conventional processing routes are likely not economic. However, ability of fungus A. niger and especially yeast R. mucilaginosa represent advantageous route of Li recovery after bioleaching. Thermal, chemical or microbiological process can be used to Li extraction from the biomass later on.
Metabolic activity and hyphae penetration of microscopic fungi and yeasts resulted in significant structural changes of mineral enhancing the access of lithium by bioleaching agent. Maybe the combination of heterotrophic microorganisms (microscopic fungi or yeast) bioleaching leading to mineral structure changes with consequent bacterial bioleaching could bring better results in the future. More

Sampling of N. barkeri and related species
Phytoseiid mites inhabit a variety of habitats, such as various plants and soil litters. Individuals were collected from plants and those on substrates and soil litters were isolated using Berlese’ funnels and kept in 95% alcohol. Samples were mounted in Hoyer’s medium and softened and cleaned with lactic acid if the mite body was hard. In addition, specimens were deposited at several institutes: GIABR (Guangdong Institute of Applied Biological Resources, Guangzhou, Guangdong, China), HUM (Hokkaido University Museum, Sapporo, Japan), NMNS (National Museum of Nature and Science, Tsukuba, Japan), NTU (Department of Entomology, National Taiwan University, Taipei, Taiwan), TARL (Taiwan Acari Research Laboratory, Taichung City, Taiwan). Female phytoseiid mites were collected, including 250 specimens of N. barkeri, and 262 specimens of 35 non-target species belonging to subfamily Amblyseiinae, in 6 tribes, and 11 genera. The following numbers of these non-target species were collected: 4 of N. baraki, 10 of N. longispinosus, 10 of N. makuwa, 6 of N. taiwanicus, 9 of N. womersleyi, 9 of Amblyseius alpinia, 10 of A. bellatulus, 10 of A. eharai, 10 of A. herbicolus, 2 of A. pascalis, 10 of A. tamatavensis, 10 of Euseius aizawai, 6 of E. circellatus, 7 of E. daluensis, 11 of E. macaranga, 10 of E. ovalis, 6 of E. paraovalis, 3 of E. nicholsi, 6 of E. oolong, 7 of E. sojaensis, 4 of Gynaeseius liturivorus, 3 of G. santosoi, 10 of Okiseius subtropicus, 4 of Paraamblyseius formosanus, 7 of Paraphytoseius chihpenensis, 10 of Parap. cracentis, 3 of Parap. hualienensis, 10 of Parap. orientalis, 6 of Phytoscutus salebrosus, 10 of Proprioseiopsis asetus, 3 of Prop. ovatus, 8 of Scapulaseius anuwati, 10 of S. cantonensis, 10 of S. okinawanus, and 8 of S. tienhsainensis. In addition, specimens of N. barkeri were collected from the United States, China, Israel, Japan, the Netherlands, Taiwan, and Thailand (including intercepted specimens in plant quarantine).
Quantitative measurements of phytoseiid mites
Specimens were examined under an Olympus BX51 microscope, and measurements were performed using a stage-calibrated ocular micrometer and ImageJ 1.4736. Photos were taken using a Motic Moticam 5+ camera attached to the microscope (Figure S1). All measurements were recorded in micrometres (μm). The general terminology used for morphological descriptions in this study conformed to that of Chant and McMurtry20. The notation for idiosomal setae conformed to that of Lindquist and Evans37 and Lindquist38, as adapted by Rowell et al.39 and Chant and Yoshida-Shaul32. Phytoseiid mites exhibit pronounced sexual dimorphism, and female individuals are more crucial for identification because of their distinguishing features and greater prevalence. In the present study, 22 quantitative measurements were collected from the female specimens: dorsal shield length and width; j1, j3, j4, j6, J5, z2, z4, z5, Z1, Z4, Z5, s4, r3, and R1 setae length; ventrianal shield length and width (at ZV2 level); JV5 length; St IV length; spermatheca calyx length, and spermatheca calyx width (Fig. 1, Table 1).
XGBoost training and computing
We used XGBoost to develop a classification system for target mite species and related species based on their morphological features. Among machine learning methods, XGBoost is the most efficient for implementing the gradient boosting decision tree algorithm from multiple decision trees, which are created successively. For each iteration, a tree enhances its predictive power by minimising the unexplained part of the last tree. First, we determined the number of decision trees through cross-validation. The original sample was randomly partitioned into five equally sized subsamples (Table S1). A single subsample and the other subsamples were retained for use as the validation and training data, respectively. Cross-validation was then performed five times, with each subsample used exactly once as the validation data. The number of decision trees allows the same level of performance to be achieved in training and validation. The number of decision trees was then used for the full dataset to create a final model, and key morphological features were selected for their relative importance. Next, we used ICE plots to indicate the determinative roles of these key features in classification. Plots in which one line represents one specimen indicate changes in predictions (of target species) that occur as a morphological feature change. We generated XGBoost and ICE plots by respectively using the R package “xgboost”40 and “pdp”41.
Drawings
Hand-drawn illustrations (Fig. 1) were made under an optic microscope (Olympus BX51). These drawings were first scanned, then processed and digitized with Photoshop CS6 (Adobe Systems Incorporated, USA). More