The Earth’s Population Can Reach 14 Billion in the 23rd Century without Significant Adverse Effects on Survivability
Abstract
:1. Introduction
- The premature increase in the world population compared to the increase in productivity of agricultural and natural ecosystems would lead to a decrease in the volume of food per capita. The food deficit is a fact in many areas. Food per person decreases over time and an increase in hungry people is expected [9].
- Global climate change due to the disturbance of cycles of greenhouse gases and water resources leads to a modification of spatial distribution of water resources, including drinking water [12].
- The development of new powerful weapons would contribute additional uncertainties in the problem of human population survivability [13].
- The intensification of both international and regional conflicts would be followed by dramatic changes in the globalization and decentralization processes which would not encourage the improvement of the living conditions of the population [14].
- There would be ecological consequences of mobile communication media including mobile phones [15].
- The authors of the models of the Club of Rome focused their main attention both on global economic processes which connect to separate environmental processes and secondly selecting the demographic block as a key element of the global model.
- Moisseev’s [24] starting position was the research of the biosphere considering the human as an element of the biosphere and that the demographic and economic processes are only taken into account in the systematic analysis of the global ecological evolution.
2. General Description of the GIMS/CNSS Model
3. Description of the GIMS/CNSS Items
ΔTCH4 = 0.019[CH4(t)1/2 − CH4(t*)1/2], ΔTO3 = 0.7[O3(t) − O3(t*)]/15,
ΔTCFC11 = 0.14[CFC11(t) − CFC11(t*)], ΔTCFC12 = 0.16[CFC12(t) − CFC12(t*)].
μ1exp[−ξ2GDP/GDP(t0)] + μ2[1 − exp{−ξ2GDP/GDP(t0)}; μ1exp[−ξ3VG] + μ2(1 − exp[−ξ3VG])}
η1exp[−χ2GDP/GDP(t0)] + η2[1 − exp{−χ2GDP/GDP(t0)]}; η1exp[−χ3VG] + η2(1 − exp[−χ3VG])}
- -
- ΞLD the area occupied by countries with HDI ∈ [0.85,1]
- -
- ΞLM the area occupied by the countries with transition economy (HDI ∈ (0.65,0.85)), and
- -
- ΞLG correspond to the territory of the developing countries (HDI ∈ [0,0.65]).
- Agricultural technologies are the main food producers that can promote food safety and nutrition security. Global agriculture supplies 2940 kcal per person at present with a forecast of up to 3050 in 2030. Existing protein support per person is estimated at 60 g a day when the medical standard is 70 g. The total protein deficit is estimated at 10 to 25 million tons. Nearly half of the world’s population (7.5 billion) suffers from a lack of protein [70].
- The second major source of the food is fishing and cultivation of fish in natural lakes and reservoirs. In 2016 each person consumed about 22 kg of fish production. At present, the ecosystems of the World Ocean and the seas provide about 20% of the world’s needs for proteins of animal origin. Mainly, oceanic biomass is estimated around 150 thousands of the animal species and 10 thousands of the water-plants with a total weight of about 35 billion tons which is sufficient to survive 35 billion people [71].
- Natural plants and forest in the first series can be considered hypothetical sources of food including wild animals and edible plants, hazelnuts, etc. Further development of the food industry and corresponding science allows the expansion of primary use of natural biomass for food production.
4. Simulation Experiments
- the problems arising from the limitation of energy sources will be overcome by 2050;
- the emissions of greenhouse gases will increase by 10% by 2050 compared to 2015 and then begin to fall evenly to 2200 up to 5%;
- agricultural technologies to increase productivity by 100% by 2050 and by 200% by the end of the 22nd century will be production;
- the speed of replacement of forest ecosystems by avifauna will be reduced by 10 times in 2050 compared to 2015 and then the forested pixels will not be disturbed; and
- the contribution of World Ocean resources to food production will increase from 1% in 2015 to 5% in 2050 and then increase steadily to 10% in 2200.
5. Conclusions
Author Contributions
Conflicts of Interest
References
- Varotsos, C.A.; Franzke, C.L.; Efstathiou, M.N.; Degermendzhi, A.G. Evidence for two abrupt warming events of SST in the last century. Theor. Appl. Climatol. 2014, 116, 51–60. [Google Scholar] [CrossRef]
- Kondratyev, K.Y.; Varotsos, C. Atmospheric greenhouse effect in the context of global climate change. Nuovo Cimento C 1995, 18, 123–151. [Google Scholar] [CrossRef]
- Varotsos, C.A. The global signature of the ENSO and SST-like fields. Theor. Appl. Climatol. 2013, 113, 197–204. [Google Scholar] [CrossRef]
- Varotsos, C.; Assimakopoulos, M.N.; Efstathiou, M. Long-term memory effect in the atmospheric CO2 concentration at Mauna Loa. Atmos. Chem. Phys. 2007, 7, 629–634. [Google Scholar] [CrossRef]
- Efstathiou, M.N.; Varotsos, C.A. On the altitude dependence of the temperature scaling behaviour at the global troposphere. Int. J. Remote Sens. 2010, 31, 343–349. [Google Scholar] [CrossRef]
- Efstathiou, M.N.; Tzanis, C.; Cracknell, A.P.; Varotsos, C.A. New features of land and sea surface temperature anomalies. Int. J. Remote Sens. 2011, 32, 3231–3238. [Google Scholar] [CrossRef]
- Efstathiou, M.N.; Varotsos, C.A. Intrinsic properties of Sahel precipitation anomalies and rainfall. Theor. Appl. Climatol. 2012, 109, 627–633. [Google Scholar] [CrossRef]
- Varotsos, C.A.; Efstathiou, M.N.; Cracknell, A.P. On the scaling effect in global surface air temperature anomalies. Atmos. Chem. Phys. 2013, 13, 5243–5253. [Google Scholar] [CrossRef]
- Mayhew, R.J. (Ed.) New Perspectives on Malthus; Cambridge University Press: Cambridge, UK, 2016. [Google Scholar]
- Shaw, P. A Treatise of Incurable Diseases; BiblioBazaar: Charleston, SC, USA, 2010. [Google Scholar]
- Okazaki, K. Good-Bye Incurable Diseases! iUniverse, Inc.: Bloomington, IN, USA, 2011. [Google Scholar]
- Weart, S.R. The Discovery of Global Warming; Harvard University Press: Harvard, MA, USA, 2008. [Google Scholar]
- Cimbala, S.J. Nuclear Weapons in the Information Age; Continuum International Publishing Group: London, UK, 2012. [Google Scholar]
- Daun, H. (Ed.) School decentralization in the context of globalizing governance. In International Comparison of Grassroots Responses; Springer: Dordrecht, The Netherlands, 2007. [Google Scholar]
- Sahib, S.S. Impact of mobile phones on the density of honeybees. J. Publ. Admin. Pol. Res. 2011, 3, 131–133. [Google Scholar]
- Anderson, B.A. World Population Dynamics: An Introduction to Demography; Pearson Publ. Ltd.: Cambridge, UK, 2015. [Google Scholar]
- Berthelot, M.; Friedlingstein, P.; Ciais, P.; Monfray, P. Global response of the terrestrial biosphere to CO2 and climate change using a coupled climate-carbon cycle model. Glob. Biogeochem. Cycles 2002, 16, 31. [Google Scholar] [CrossRef]
- Dufresne, J.-L.; Foujols, M.-A.; Denvil, S.; Caubel, A.; Marti, O. Climate change projections using the IPSL-CM5 Earth System Model: From CMIP3 to CMIP5. Clim. Dyn. 2013, 40, 2123–2165. [Google Scholar] [CrossRef] [Green Version]
- Kondratyev, K.Y.; Krapivin, V.F.; Phillips, G.W. Arctic Basin pollution dynamics. In Arctic Environment Variability in the Context of Global Change; Bobylev, L.P., Kondratyev, K.Y., Johannesses, O.M., Eds.; Springer: Chichester, UK, 2003; pp. 309–362. [Google Scholar]
- Kondratyev, K.Y.; Krapivin, V.F.; Varotsos, C.A. Global Carbon Cycle and Climate Change; Springer: Chichester, UK, 2003. [Google Scholar]
- Krapivin, V.F.; Varotsos, C.A. Globalization and Sustainable Development; Springer: Chichester, UK, 2007; 304p. [Google Scholar]
- Krapivin, V.F.; Varotsos, C.A. Biogeochemical Cycles in Globalization and Sustainable Development; Springer: Chichester, UK, 2008. [Google Scholar]
- Krapivin, V.F.; Varotsos, C.A.; Soldatov, V.Y. New Ecoinformatics Tools in Environmental Science: Applications and Decision-Making; Springer: London, UK, 2015. [Google Scholar]
- Moisseev, N.N. Mathematics Produces an Experiment; Science Publ.: Moscow, Russia, 1979; 223p. (In Russian) [Google Scholar]
- Forrester, J.W. World Dynamics; Wright-Allen Press: Cambridge, MA, USA, 1971. [Google Scholar]
- Forrester, J.W. World Dynamics; Productivity Press Publishing: New York, NY, USA, 1979. [Google Scholar]
- Meadows, D.H.; Meadows, D.L.; Randers, J.; Behrens, W.W., III. The Limits to Growth; Universe Books: New York, NY, USA, 1972. [Google Scholar]
- Meadows, D.H.; Randers, J.; Meadows, D.L. Limits of Grows: The 30-Years Update; Chelsea Green Publishing: White River Junction, VT, USA, 2004. [Google Scholar]
- Pestel, E. Beyond the Limits to Growth: A Report to Club of Rome; Universe Books: New York, NY, USA, 1989. [Google Scholar]
- Krapivin, V.F. Mathematical model for global ecological investigations. Ecol. Model. 1993, 67, 103–127. [Google Scholar] [CrossRef]
- Sellers, P.J.; Los, S.O.; Tucker, C.J.; Justice, C.O.; Dazlich, D.A.; Collatz, G.J.; Randall, D.A. A revised land surface parametrization (SiB2) for atmospheric GCMs. Part II: The generation of global fields of terrestrial biophysical parameters from satellite data. J. Clim. 1996, 9, 708–737. [Google Scholar]
- Degermendzhi, A.G. New directions in biophysical ecology. In Global Climatology and Ecodynamics; Cracknell, A.P., Krapivin, V.F., Varotsos, C.A., Eds.; Springer: Chichester, UK, 2009; pp. 379–396. [Google Scholar]
- Degermendzhi, A.G.; Bartsev, S.I.; Gubanov, V.G.; Erokhin, D.V.; Shevirnogov, A.P. Forecast of biosphere dynamics using small-scale models. In Global Climatology and Ecodynamics; Cracknell, A.P., Krapivin, V.F., Varotsos, C.A., Eds.; Springer: Chichester, UK, 2009; pp. 241–300. [Google Scholar]
- Krapivin, V.F.; Kelley, J.J. Model-based method for the assessment of global change in a nature-society system. In Problems of Global Climatology and Ecodynamics; Cracknell, A.P., Krapivin, V.F., Varotsos, C.A., Eds.; Springer: Chichester, UK, 2009; pp. 133–184. [Google Scholar]
- Saavedra-Rivano, N. A critical analysis of the Mesarovic-Pestel world model. Appl. Math. Model. 1979, 3, 384–390. [Google Scholar] [CrossRef]
- Vernadsky, W.I. Problems of biogeochemistry II. Trans. Conn. Acad. Arts Sci. 1944, 35, 493–494. [Google Scholar]
- Xue, Y.; Sellers, P.J.; Kinter, J.L.; Shukla, J. A simplified biosphere model for global climate studies. J. Clim. Am. Meteorol. Soc. 1991, 4, 345–364. [Google Scholar] [CrossRef]
- Ondov, J.M.; Buckley, T.J.; Hopke, P.K.; Ogulei, D.; Parlange, M.B.; Rogge, W.F.; Squibb, K.S.; Johnston, M.V.; Wexler, A.S. Baltimore Supersite: Highly time- and size-resolved concentrations of urban PM2.5 and its constituents for resolution of sources and immune responses. Atmos. Environ. 2006, 40, 224–237. [Google Scholar] [CrossRef]
- Xue, Y.; He, X.W.; Xu, H.; Guang, J.; Guo, J.P.; Mei, L.L. China Collection 2.0: The aerosol optical depth dataset from the synergetic retrieval of aerosol properties algorithm. Atmos. Environ. 2014, 95, 45–58. [Google Scholar] [CrossRef]
- Ebel, A.; Memmesheimer, M.; Jakobs, H.J. Chemical perturbations in the planetary boundary layer and their relevance for chemistry transport modelling. Bound.-Lay. Meteorol. 2007, 125, 265–278. [Google Scholar] [CrossRef]
- Chattopadhyay, G.; Chakraborthy, P.; Chattopadhyay, S. Mann-Kendall trend analysis of tropospheric ozone and its modeling using ARIMA. Theor. Appl. Climatol. 2012, 110, 321–328. [Google Scholar] [CrossRef]
- Krapivin, V.F.; Mkrtchyan, F.A.; Van, T.D. Constructive method for the vegetation microwave monitoring. In Proceedings of the International Symposium on Engineering Ecology, Moscow, Russia, 2–4 December 2015; The Russian Sciences Engineering A.S. Popov Society for Radio, Electronics and Communication: Moscow, Russia, 2015; pp. 21–27. [Google Scholar]
- Krapivin, V.F.; Mkrtchyan, F.A.; Nazaryan, N.A. Development of GIMS-technology for environmental monitoring of ocean ecosystems. In Proceedings of the 31st International Symposium on Okhotsk Sea & Sea Ice, Mombetsu, Hokaido, Japan, 21–24 February 2016; The Okhotsk Sea & Polar Oceans Research Association: Mombetsy, Hokkaido, Japan, 2016; pp. 116–119. [Google Scholar]
- Nitu, C.; Krapivin, V.F.; Bruno, A. Intelligent Techniques in Ecology; Printech: Bucharest, Romania, 2000. [Google Scholar]
- Pawłowski, A. Sustainable development as a civilizational revolution. In A Multidisciplinary Approach to the Challenges of the 21st Century; CRC Press: New York, NY, USA, 2011. [Google Scholar]
- Krapivin, V.F.; Varotsos, C.A. Modelling the CO2 atmosphere-ocean flux in the upwelling zones using radiative transfer tools. J. Atmos. Sol.-Terr. Phys. 2016, 150–151, 47–54. [Google Scholar] [CrossRef]
- Krapivin, V.F.; Varotsos, C.A.; Soldatov, V.Y. Simulation results from a coupled model of carbon dioxide and methane global cycles. Ecol. Model. 2017, 359, 69–79. [Google Scholar] [CrossRef]
- Krapivin, V.F.; Vilkova, L.P. Model estimation of excess CO2 distribution in biosphere structure. Ecol. Model. 1990, 50, 57–78. [Google Scholar] [CrossRef]
- Korotaev, A.; Zinkina, J. On the structure of the present-day convergence. Campus-Wide Inf. Syst. 2014, 31, 139–152. [Google Scholar] [CrossRef]
- Nitu, C.; Krapivin, V.F.; Pruteanu, E. Ecoinformatics: Intelligent Systems in Ecology; Magic Print Onesti: Bucharest, Romania, 2004. [Google Scholar]
- Nitu, C.; Krapivin, V.F.; Soldatov, V.Y. Information-Modeling Technology for Environmental Investigations; Matrix Rom: Bucharest, Romania, 2013. [Google Scholar]
- Tarko, A.M. Analysis of Global and Regional Changes in Biogeochemical Carbon Cycle: A Spatially Distributed Model; IR-03-041; IIASA, Inter. Rep.: Laxenburg, Austria, 2003. [Google Scholar]
- Kondratyev, K.Y.; Krapivin, V.F.; Savinykh, V.P.; Varotsos, C.A. Global Ecodynamics: A Multidimensional Analysis; Springer: Chichester, UK, 2004. [Google Scholar]
- Mintzer, I.M. A Matter of Degrees: The Potential for Controlling the Greenhouse Effect; World Resources Institute: Washitong, DC, USA, 1987. [Google Scholar]
- Varotsos, C.A.; Krapivin, V.F.; Soldatov, V.Yu. Modeling the carbon and nitrogen cycles. Front. Environ. Sci. Air Pollut. 2014, 2. [Google Scholar] [CrossRef]
- Kondratyev, K.Y.; Ivlev, L.S.; Krapivin, V.F.; Varotsos, C.A. Atmospheric Aerosol Properties: Formation, Processes and Impacts; Springer: Chichester, UK, 2006. [Google Scholar]
- Krapivin, V.F.; Shutko, A.M. Information Technologies for Remote Monitoring of the Environment; Springer: Chichester, UK, 2012. [Google Scholar]
- Krapivin, V.F. The estimation of the Peruvian current ecosystem by a mathematical model of biosphere. Ecol. Model. 1996, 91, 1–14. [Google Scholar] [CrossRef]
- Kondratyev, K.Y.; Krapivin, V.F.; Phillips, G.W. High Latitude Environmental Pollution Problems; Cankt-Petersburg State University Publ.: Sankt-Petersburg, Russia, 2002. [Google Scholar]
- Krapivin, V.F.; Mkrtchyan, F.A.; Soldatov, V.Y. Simulation model of the Arctic Basin ecosystem. In Proceedings of the 32nd International Symposium on Okhotsk Sea & Polar Oceans, Mombetsu, Hokkaido, Japan, 19–22 February 2017; Okhotsk Sea and Polar Oceans Research Association: Mombetsu, Hokkaido, Japan, 2017; pp. 337–340. [Google Scholar]
- Van Tuyet, D.; Man, N.X.; Van, L.T.T.; Krapivin, V.F.; Mkrtchyan, F.A.; Hung, N.T.; Thanh, L.N. Global model of carbon cycle as instrument of primary agriculture production assessment. In Proceedings of the International Symposium “Some Aspects of Contemporary Issues in Ecoinformatics”, Hochiminh City, Vietnam, 20 March 2015; Institute of Applied Mechanics and Informatics, Vietnam Academy of Science and Technology: Hochiminh City, Vietnam, 2015; pp. 50–58. [Google Scholar]
- Nitu, C.; Dumitrasku, A.; Krapivin, V.F.; Mkrtchyan, F.A. Reducing risks in agriculture. In Proceedings of the 20th International Conference on Control Systems and Computer Science, Bucharest, Romania, 27–29 May 2015; University Politehnica of Bucharest Campus: Bucharest, Romania, 2015; pp. 941–945. [Google Scholar]
- Kaduk, J.; Heimann, M. A prognostic phenology scheme for global terrestrial carbon cycle models. Clim. Res. 1996, 6, 1–19. [Google Scholar] [CrossRef]
- Hatfield, J.L.; Prueger, J.H. Temperature extremes: Effect on plant growth and development. Weather Clim. Extrem. 2015, 10, 4–10. [Google Scholar] [CrossRef]
- Coupel, P.; Ruiz-Pino, D.; Sicre, M.A.; Chen, J.F.; Lee, S.H.; Schiffrine, N.; Li, H.L.; Gascard, J.C. The impact of freshening on phytoplankton production in the Pacific Arctic Ocean. Prog. Oceanogr. 2015, 131, 113–125. [Google Scholar] [CrossRef] [Green Version]
- Burford, M.A.; Rothlisberg, P.C. Factors limiting phytoplankton production in a tropical continental shelf ecosystem estuarine. Coast. Shelf Sci. 1999, 48, 541–549. [Google Scholar] [CrossRef]
- Raymont, J.E.G. Plankton and Productivity in the Oceans; Vol. 1: Phytoplankton; Pergamon Press: New York, NY, USA, 1980. [Google Scholar]
- Alexandratos, N.; Bruinsma, J. World Agriculture Forwards 2030/2050; FAO: Rome, Italy, 2012. [Google Scholar]
- Butler, J.H.; Montzka, S.A. The NOAA Annual Greenhouse Gas Index (AGGI); NOAA Earth System Research Laboratory, Global Monitoring Division. Available online: www.esrl.noaa.gov/gmd/aggi/aggi.html (accessed on 6 August 2017).
- Debertin, D.L. Agricultural Production Economics; Macmillan Publish Company: London, UK, 2012. [Google Scholar]
- Lucas, J.S.; Southgate, P.C. Aquaculture: Farming Aquatic Animals and Plants; John Wiley and Sons: New York, NY, USA, 2012. [Google Scholar]
- Card, D.; Raphael, S. Immigration, Poverty, and Socioeconomic Inequality; Russel Sage Foundation: New York, NY, USA, 2013. [Google Scholar]
- Dixon, A.P.; Faber-Langendoen, D.; Josse, C.; Morrison, J.; Loucks, C.J. Distribution mapping of world grassland types. J. Biogeogr. 2014, 41, 2003–2019. [Google Scholar] [CrossRef]
- Shvidenko, A.Z.; Schepaschenko, D.C.; Nilsson, S.; Buluy, Y.I. Tables and Models of Growth and Productivity of Forests of Major Forest Forming Species of Northern Eurasia; Federal Agency of Forest Management: Moscow, Russia, 2008. [Google Scholar]
- Riahi, K.; Krey, V.; Rao, S.; Chirkov, V.; Fischer, G.; Kolp, P.; Kindermann, G.; Nakicenovic, N.; Rafai, P. RCP-8.5: Exploring the consequence of high emission trajectories. Clim. Chang. 2011, 109. [Google Scholar] [CrossRef]
- Van Vuuren, D.P.; Stehfest, E.; den Elzen, M.G.J.; Kram, T.; van Vliet, J.; Deetman, S.; Isaac, M.; Goldewijk, K.K.; Hof, A.; Beltran, A.M.; et al. RCP2. 6: Exploring the possibility to keep global mean temperature increase below 2 C. Clim. Chang. 2011, 109. [Google Scholar] [CrossRef]
- Li, J.; Mao, J. Changes in the boreal summer intraseasonal oscillation projected by the CNRM-CM5 model under the RCP 8.5 scenario. Clim. Dyn. 2016, 47, 3713–3736. [Google Scholar] [CrossRef]
- Wayne, G.P. The beginner’s guide to representative concentration pathways. Scept. Sci. 2013, 25. Available online: http://denning.atmos.colostate.edu/ats760/Readings/RCP_Guide.pdf (accessed on 6 August 2017).
- Varotsos, C. Solar ultraviolet radiation and total ozone, as derived from satellite and ground-based instrumentation. Geophys. Res. Lett. 1994, 21, 1787–1790. [Google Scholar] [CrossRef]
- Varotsos, C. Climate change problems and carbon dioxide emissions: Expecting ‘Rio + 10’. Environ. Sci. Pollut. Res. 2002, 9, 97–98. [Google Scholar] [CrossRef]
- Cracknell, A.P.; Varotsos, C.A. Ozone depletion over Scotland as derived from Nimbus-7 TOMS measurements. Int. J. Remote Sens. 1994, 15, 2659–2668. [Google Scholar] [CrossRef]
- Cracknell, A.P.; Varotsos, C.A. The present status of the total ozone depletion over Greece and Scotland: A comparison between Mediterranean and more northerly latitudes. Int. J. Remote Sens. 1995, 16, 1751–1763. [Google Scholar] [CrossRef]
- Varotsos, C.; Kalabokas, P.; Chronopoulos, G. Association of the laminated vertical ozone structure with the lower-stratospheric circulation. J. Appl. Meteorol. 1994, 33, 473–476. [Google Scholar] [CrossRef]
- Varotsos, C. The southern hemisphere ozone hole split in 2002. Environ. Sci. Pollut. Res. 2002, 9, 375–376. [Google Scholar] [CrossRef]
- Varotsos, C.; Cartalis, C. Re-evaluation of surface ozone over Athens, Greece, for the period 1901–1940. Atmos. Res. 1991, 26, 303–310. [Google Scholar] [CrossRef]
- Varotsos, C.A.; Cracknell, A.P. Ozone depletion over Greece as deduced from Nimbus-7 TOMS measurements. Int. J. Remote Sens. 1993, 14, 2053–2059. [Google Scholar] [CrossRef]
- Varotsos, C. Airborne measurements of aerosol, ozone, and solar ultraviolet irradiance in the troposphere. J. Geophys. Res.-Atmos. 2005, 110, D09202. [Google Scholar] [CrossRef]
Item | Item Functions |
---|---|
DM | Demographic model [53]. |
CM | Climate model [42,54]. |
CMCM | Coupled model of the carbon dioxide and methane cycles [47]. |
GSCM | Global sulphur cycle model [22]. |
GCOO | Coupled model of global cycles of oxygen and ozone [22]. |
GNCM | Global nitrogen cycle model [55]. |
GPCM | Global phosphorus cycle model [56]. |
RHCM | Regional hydrological cycle model [57]. |
BMSPF | Biocenotic model of the soil-plant formations [48,50]. |
PMTM | Photosynthesis model for the tropical and moderate oceanic zones [58]. |
PMAA | Photosynthesis model for the Arctic and Antarctic zones of the World Ocean [19,59,60]. |
APM | Agriculture production model [61,62]. |
AIFI | Evolutionary algorithm for the indicator calculation of the food industry [44,50]. |
UEM | An upwelling ecosystem model [46]. |
MWEL | Model of the typical water ecosystem on the land [57]. |
AHIS | An algorithm for the human indicator survivability calculation. |
DMEP | Dynamic model of the environmental pollutants [56]. |
GSA | The GIMS structure adaptation to the simulation experiment conditions [23,42]. |
DFM | Database formation and management. |
SS | Synthesis of the scenarios for the interaction of population with the environment. |
SEMC | Simulation experiment management and control. |
Parameter | Symbol | Parameter Evaluation |
---|---|---|
Photosynthesis compensation constant: | Γ, ppmv | |
Equator | 5 | |
Pole | 50 | |
Coefficient reflecting the effect of the CO2 factor on plant production. | aC | 3.226 |
Constant of the photosynthetic responses to atmospheric CO2 changes. | bC, ppmv | 930.03 |
Coefficient reflecting the impact of solar radiation on plant production. | aE | 1.177 |
Parameter indicating the solar radiation in which the stability of plant production is achieved. | bE, W/m2 | 60.538 |
Coefficient reflecting the effect of precipitation on plant production. | aW | 4.742 |
Parameter indicating the precipitation in which the stability of plant production is achieved. | bW mm/year | 592.357 |
Parameter indicating the maximal rate of growth of plant biomass under temperature change. | aT | 0.56 |
Indicator of declining plant production under temperature change. | bT | 0.42 |
Maximal rate of loss of plant biomass under temperature change. | ρT | 1.214 |
Parameter that controls early delay to achieve a maximal rate of loss of plant biomass due to temperature change. | dT, °C | 5.714 |
Maximal rate of loss of plant biomass under change of soil moisture. | da | 0.0267 |
Parameter that controls early delay to achieve a maximal rate of loss of plant biomass to precipitation change. | db, mm/year | 208.333 |
Ratio coefficient that characterizes phytoplankton rate dependence on temperature. | θW | 0.21 |
Ratio coefficient that characterizes phytoplankton rate dependence on solar energy. | θE | 0.25 |
Constant that determines the characteristics of phytoplankton species dependent on biogenic salts. | γN | 0.1 |
Constant that determines phytoplankton production as a function of its biomass. | γP | 0.25 |
The area of the biosphere. | σ, km2 | 510.1 × 106 |
Model start time. | t0 | 2015 |
Indicator and Type of Soil-Plant Formation | σS | Φ* | Tmin (°C) | Topt (°C) | |
---|---|---|---|---|---|
A—Arctic deserts and tundra | 2.55 | 0.17 | 0.4 | −5 | 40 |
C—Tundra | 2.93 | 0.36 | 1.9 | −5 | 40 |
M—Mountain tundra | 2.33 | 0.38 | 1.9 | −3 | 35 |
L—Forest tundra | 1.55 | 0.65 | 3.8 | −5 | 40 |
F—North-taiga forests | 5.45 | 0.54 | 10 | −5 | 40 |
D—Mid-taiga forests | 5.73 | 0.63 | 22.5 | −5 | 40 |
6.6 | 0.65 | 23.5 | −5 | 40 | |
G—South-taiga forests | 2.12 | 0.87 | 25 | −1 | 43 |
7.21 | 1.25 | 45 | −1 | 43 | |
R—Broad-leaved coniferous forests | 5.75 | 1.72 | 43 | 0 | 43 |
+—Broad-leaved forests | 3.91 | 0.56 | 3.8 | 0 | 43 |
P—Sub-tropical broad-leaved and coniferous forests | 3.72 | 0.74 | 1.9 | 2 | 43 |
U—Xerophytic open woodlands and shrubs | 4.29 | 0.79 | 1.9 | 2 | 43 |
X—Forest-steppes (meadow steppes) | 1.66 | 1.11 | 3.8 | 5 | 45 |
W—Moderately arid and arid (mountain including) steppes | 2.66 | 0.38 | 0.8 | 5 | 45 |
E—Pampas and grass savannas | 2.08 | 0.45 | 0.4 | 5 | 45 |
V—Dry steppes | 2.69 | 0.25 | 0.2 | 5 | 50 |
#—Mangrove forests | 1.99 | 0.35 | 0.8 | 5 | 30 |
S—Sub-boreal and saltwort deserts | 7.16 | 0.12 | 0.1 | 5 | 45 |
&—Sub-tropical semi-deserts | 1.15 | 0.47 | 0.8 | −3 | 10 |
H—Sub-tropical deserts | 3.54 | 0.76 | 1.9 | −3 | 10 |
B—Alpine deserts | 10.4 | 3.17 | 60 | 5 | 50 |
Q—Alpine and sub-alpine meadows | 7.81 | 2.46 | 60 | 5 | 50 |
Z—Humid evergreen tropical forests | 9.18 | 1.42 | 10 | 5 | 50 |
Y—Variably-humid deciduous tropical forests | 17.1 | 1.35 | 0.1 | 5 | 45 |
N—Tropical xerophytic open woodlands | 13.52 | 0.18 | 0.4 | 5 | 45 |
J—Tropical savannas | 0.38 | 0.18 | 45 | 4 | 50 |
T—Tropical deserts | 0.9 | 1.96 | 45 | 4 | 45 |
K—Saline lands | 14.6 | 0 | 0 | - | - |
I—Sub-tropical & tropical grass-tree thickets of the tugai type | |||||
*—Lack of vegetation |
Coefficient | Region ΞLD | Region ΞLM | Region ΞLG |
---|---|---|---|
ρ, year−1 | 1.19 | 1.26 | 1.32 |
β, year−1 | 1.21 | 1.23 | 1.25 |
η1 | 0.01 | 0.011 | 0.014 |
η2 | 0.003 | 0.005 | 0.009 |
ξ1 | 0.031 | 0.027 | 0.025 |
ξ2 | 0.012 | 0.011 | 0.009 |
ξ3 | 0.006 | 0.005 | 0.004 |
χ1 | 0.035 | 0.032 | 0.031 |
χ2 | 0.014 | 0.012 | 0.011 |
χ3 | 0.003 | 0.002 | 0.001 |
μ1 | 0.02 | 0.03 | 0.04 |
μ2 | 0.005 | 0.009 | 0.012 |
γ, °C | 34 | 34 | 34 |
ϖ | 0.56 | 0.61 | 0.67 |
Precipitation, WΞ (mm/Year) | Atmospheric Temperature, TΞ (°C) | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
−14 | −10 | −6 | −2 | 2 | 6 | 10 | 14 | 18 | 22 | 26 | 30 | |
3130 | 3.39 | 3.49 | 3.68 | 3.81 | 3.92 | 4.01 | ||||||
2880 | 3.27 | 3.36 | 3.47 | 3.63 | 3.73 | 3.82 | ||||||
2630 | 3.09 | 3.27 | 3.31 | 3.44 | 3.54 | 3.65 | ||||||
2380 | 2.85 | 2.93 | 3.09 | 3.12 | 3.22 | 3.33 | ||||||
2130 | 2.57 | 2.69 | 2.67 | 2.94 | 2.91 | 3.03 | ||||||
1880 | 1.63 | 2.38 | 2.38 | 2.43 | 2.55 | 2.62 | 2.74 | |||||
1630 | 0.39 | 0.62 | 1.34 | 2.04 | 2.14 | 2.12 | 2.26 | 2.35 | 2.42 | |||
1380 | 0.18 | 0.31 | 0.41 | 0.73 | 1.16 | 1.75 | 1.91 | 1.95 | 2.13 | 2.18 | 2.09 | |
1130 | 0.19 | 0.26 | 0.32 | 0.43 | 0.77 | 1.05 | 1.66 | 1.84 | 1.92 | 1.84 | 1.83 | 1.75 |
880 | 0.21 | 0.28 | 0.42 | 0.52 | 0.83 | 0.92 | 1.53 | 1.43 | 1.33 | 1.36 | 1.27 | 1.24 |
630 | 0.28 | 0.29 | 0.53 | 0.57 | 0.89 | 0.91 | 0.92 | 0.85 | 0.84 | 0.73 | 0.72 | 0.71 |
380 | 0.39 | 0.41 | 0.54 | 0.69 | 0.66 | 0.64 | 0.67 | 0.57 | 0.56 | 0.55 | 0.43 | 0.42 |
130 | 0.14 | 0.32 | 0.31 | 0.22 | 0.24 | 0.24 | 0.24 | 0.24 | 0.23 | 0.14 | 0.13 | 0.11 |
© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Krapivin, V.F.; Varotsos, C.A.; Soldatov, V.Y. The Earth’s Population Can Reach 14 Billion in the 23rd Century without Significant Adverse Effects on Survivability. Int. J. Environ. Res. Public Health 2017, 14, 885. https://doi.org/10.3390/ijerph14080885
Krapivin VF, Varotsos CA, Soldatov VY. The Earth’s Population Can Reach 14 Billion in the 23rd Century without Significant Adverse Effects on Survivability. International Journal of Environmental Research and Public Health. 2017; 14(8):885. https://doi.org/10.3390/ijerph14080885
Chicago/Turabian StyleKrapivin, Vladimir F., Costas A. Varotsos, and Vladimir Yu. Soldatov. 2017. "The Earth’s Population Can Reach 14 Billion in the 23rd Century without Significant Adverse Effects on Survivability" International Journal of Environmental Research and Public Health 14, no. 8: 885. https://doi.org/10.3390/ijerph14080885
APA StyleKrapivin, V. F., Varotsos, C. A., & Soldatov, V. Y. (2017). The Earth’s Population Can Reach 14 Billion in the 23rd Century without Significant Adverse Effects on Survivability. International Journal of Environmental Research and Public Health, 14(8), 885. https://doi.org/10.3390/ijerph14080885