Skip to main content

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

Contrasting physiological and structural vegetation feedbacks in climate change simulations


Anthropogenic increases in the atmospheric concentration of carbon dioxide and other greenhouse gases are predicted to cause a warming of the global climate by modifying radiative forcing1. Carbon dioxide concentration increases may make a further contribution to warming by inducing a physiological response of the global vegetation—a reduced stomatal conductance, which suppresses transpiration2. Moreover, a CO2-enriched atmosphere and the corresponding change in climate may also alter the density of vegetation cover, thus modifying the physicalcharacteristics of the land surface to provide yet another climate feedback3,4,5,6. But such feedbacks from changes in vegetation structure have not yet been incorporated into general circulation model predictions of future climate change. Here we use a general circulation model iteratively coupled to an equilibrium vegetation model to quantify the effects of both physiological and structural vegetation feedbacks on a doubled-CO2 climate. On a global scale, changes in vegetation structure are found to partially offset physiological vegetation–climate feedbacks in the long term, but overall vegetation feedbacks provide significant regional-scale effects.


The Sheffield University vegetation model simulates global vegetation under steady-state conditions of climate and atmospheric CO2 (ref. 7). It models the physiological processes of nutrient uptake, photosynthesis, respiration and stomatal limitation of transpiration, and uses these to determine the vegetation structural character in terms of foliage density. The outputs of this model are: (1) leaf area index (LAI), the area of leaf surface per unit area of ground; and (2) daytime mean canopy conductance (gc), the net transpirational conductance of all stomata integrated (numerically) over the canopy depth. LAI is purely a structural variable, whereas gc contains both structural and physiological contributions. The contemporary vegetation simulation has been validated against point measurements7.

The Hadley Centre general circulation model (GCM) used here is a simplified version of that used for climate change prediction8,9, consisting of an explicit representation of the global atmospheric circulation and a thermodynamic ‘mixed-layer’ ocean model with prescribed heat transports to represent ocean currents10. The simulations presented here neglect the relative cooling effect of increased sulphate aerosol concentrations8; this omission and that of explicit ocean current modelling means that the results cannot be regarded as a state-of-the-art prediction. Instead, they demonstrate the effect of vegetation feedback on climate sensitivity to atmospheric CO2 concentrations.

The GCM land surface scheme is of moderate complexity11, with the land surface state defined by seven prognostic variables: root-zone soil moisture, lying snow, intercepted canopy water, and the temperatures of four soil layers in the vertical. The surface energy partitioning, evapotranspiration, runoff and snowmelt are parametrized using driving variables from the atmosphere model and seven vegetation-specific land surface parameters3. The main parameters are: root depth, determining the depth of soil from which water can be extracted for transpiration; snow-free and deep-snow albedos, determining the fraction of incident solar radiation reflected from the surface; roughness length, determining the aerodynamic resistance for turbulent transfers; and surface conductance, determining the additional resistance for water vapour transfers in drought-free conditions. Over vegetated surfaces, the latter accounts for the control of transpiration by stomata, but is a prescribed vegetation-specific constant in this version of the scheme.

The GCM and vegetation model were coupled by iterating between the two models, each providing boundary conditions for the other. The GCM supplied climatological monthly means to the vegetation model, which returned the global distributions of LAI and gc. The latter were used to redefine the GCM land surface parameters for the next iteration, with surface conductance incorporating gc directly, and the remaining structural parameters being derived semi-empirically from LAI (Fig. 1).

Figure 1: Derivation of GCM land surface parameters from leaf area index (LAI).

a, Vegetated fraction is the fractional ground area covered by vegetation; this determines the relative weighting of values appropriate to vegetation and soil for the other variables. An empirical fit with the literature data16,17,18 shows that vegetated fraction increases with LAI, saturating at higher values. b, The mean root depth is found empirically16,17,18,19,20 to increase with LAI in the grid-box mean, and is also assumed to be linearly correlated with the maximum infiltration rate at the soil surface. c, Roughness length increases with LAI at low leaf areas, but decreases with LAI at higher values as the canopy becomes closed21. d, Deep-snow albedo17 is the upper limit of surface reflectance in deep-snow conditions, and this decreases with LAI as more of the snow is masked by the darker vegetation. Snow-free albedo shows a similar but less pronounced relationship.

The physiological and structural vegetation feedbacks on CO2-induced climate change were isolated and quantified using the following four coupled simulations. (1) Both climate and vegetation consistent with an atmospheric CO2 concentration of 323 parts per million by volume, p.p.m.v. (1 × CO2; simulation C). (2) The climate at equilibrium under 2× CO2 (646p.p.m.v.) radiative forcing, but with the physiological and structural characteristics of the vegetation held at the 1× CO2 state (simulation R). (3) 2× CO2 radiative forcing and 1× CO2 vegetation structure, but with surface conductance including direct effects of 2× CO2 and the associated climate change on plant physiology (simulation RP). (4) 2× CO2 radiative forcing with both the physiology and structure allowed to reach a new equilibrium state under 2× CO2 and the associated climate (simulation RPS).

The difference between simulations R and C represents the standard GCM sensitivity to CO2 excluding vegetation feedbacks, and the difference between RP and R defines the additional climate change resulting from the direct physiological effects (a comparable experiment to that in ref. 2). Finally, the difference between RPS and R demonstrates the combined effect of physiological and structural vegetation change on the climate sensitivity; this is the main new result of this work.

The radiation-only 2× CO2 sensitivity (R − C) of this version of the GCM was 4.3K, which is at the high end of the IPCC range1. The modelled climate change showed relatively large changes in temperature and precipitation in the tropics (Fig. 2), associated with strong cloud-mediated feedbacks. The physiological response in simulation RP was a general reduction in gc relative to simulation R (Fig. 3a), consistent with increased water-use efficiency under 2× CO2. Some areas with modified hydrological regimes experienced gc increases caused by increased humidity, but the global mean change was a reduction of 20% (Table 1). These caused significant feedbacks on climate (RP − R), with temperature increasing by up to 1K over Northern Hemisphere land (Fig. 3b). The large conductance decreases in the tropical forests produced small temperature changes but appreciable reductions in evapotranspiration (Fig. 3c). The modelled effects of physiology on mean land temperature, evapotranspiration and conductance are all in close agreement with those from a previous study2.

Figure 2: Climate change due to doubling the atmospheric concentration of CO2, neglecting vegetation feedback, expressed as differ.

ences between simulations R and C (see text). a, Change in annual mean temperature, diagnosed at a height of 1.5m above the surface. b, Change in annual mean precipitation.

Figure 3: Physiological and structural vegetation change under doubled atmospheric CO2 concentration (2 × CO2.

/f>), and feedback of each on 2× CO2 climate. a, Change in canopy conductance due to physiological response to 2× CO2 and the associated climate change. b, Effect of physiological feedback on 2× CO2 temperature. c, Effect of physiological feedback on 2× CO2 evaporation. d, Change in leaf area index due to structural response to 2× CO2 and the associated climate change. e, Change in canopy conductance due to both physiological and structural response. f, Combined effect of both physiological and structural feedback on 2× CO2 temperature. g, Combined effect of both physiological and structural feedback on 2× CO2 evaporation. 40% of the land surface experienced vegetation feedbacks on temperature (RPS − R) of at least 5% of the magnitude of the changes due to radiative forcing alone (R − C). 13% of the land showed relative temperature feedbacks of 10% or more. The relative evaporation feedbacks were larger and more widespread, with 74% of the land surface having a relative feedback of 10% or more, and 30% showing a feedback of over 50%. Vegetation-induced evaporation changes were larger than the greenhouse-gas-only changes for 18% of the land surface. Calculation of t-statistics for grid-point annual means showed that most temperature changes of 0.5K or more were significant at the 1% confidence level or better. The exceptions to this were in the polar and sub-polar regions, where significance was reduced by higher interannual variability. In Siberia, temperature changes of 0.5K were significant at 5% or better, while in Antarctica and the Arctic Ocean, little of the temperature change was significant at better than 20%. Almost all land evaporation changes of 0.1mmd−1were significant at 1%.

Table 1 Global mean vegetation feedbacks on 2× CO2 climate

The structural response in simulation RPS was a widespread increase in LAI relative to simulation R (Fig. 3d), due to increased productivity and water-use efficiency under the new CO2 concentration and climate. The greatest LAI increases were in regions of increased rainfall. These changes acted to offset the physiological reductions in conductance, and at high latitudes the result was an overall increase in gc (Fig. 3e); this is contrary to the result obtained when allowing physiological change alone (Fig. 3a). Elsewhere, the reduced gc seen in simulation RP also occurred in simulation RPS; the reductions were smaller than in RP, except in regions where significantly reduced rainfall (Fig. 2b) had caused conspicuous reductions in LAI (Fig. 3d). The combined effect of physiology and structure was a reduction of 12% in gc in the global mean (Table 1), which is considerably less than the reduction due to physiology alone.

The combined physiological and structural vegetation feedbacks had significant effects on the climate sensitivity (RPS − R; Fig. 3f, g). Structural changes acted via two competing effects; increased LAI tended to warm the land surface by lowering its albedo4,5,6 and to cool the land surface by enhancing evaporation (and consequently cloud cover12,13) via increases in root depth14, roughness length15 and surface conductance. Similarly, decreased LAI tended to cool the surface via increased albedo, and warm the surface via reduced evaporation. The albedo effect dominated in regions where the vegetation was sparse, or where the underlying surface was much more reflective than the vegetation such as in snow-covered regions4,5,6; however, the evaporation effect dominated elsewhere. Temperature changes (Fig. 3f) were therefore negatively correlated with LAI changes (Fig. 3d), except in sparsely vegetated regions and also northern Siberia, where greater LAI caused a warming via increased masking of snow. The feedback through evaporation was significantly modified by structural changes, especially in the middle- and high-latitude regions (compare Fig. 3c, g). However, transpiration from the tropical rainforests, which experienced negligible changes in LAI, was still significantly reduced compared to simulation R.

It is important to recognize that changes in vegetation structure may lag the physiological response to increased CO2 by several years or even decades. Therefore, the actual effect of vegetation feedback on climate at the time of CO2 doubling is likely to lie somewhere between the results of simulations RP and RPS. A full assessment of this will require a model of vegetation dynamics fully integrated within a GCM. Nevertheless, our results show that changes in land surface properties due to vegetation can provide climatic feedback mechanisms that are both positive and negative in relation to climate change due to radiative forcing alone; furthermore, they demonstrate that the sign of the feedback depends partly on whether local vegetation growth is enhanced or suppressed by increased CO2 concentration and the associated climate change, and partly on the nature of the locally dominant surface–atmosphere interaction. Both physiological and structural characteristics of the vegetation have been shown to be important, with changes in one property often counteracting changes in another. In the global mean, the competing effects of increased water use efficiency and increased LAI cause a small surface evaporation change relative to the climate change simulation with fixed vegetation properties. We conclude that a short-term enhancement of regional climate warming by vegetation physiology may eventually be mitigated by a longer term modification of surface characteristics due to vegetation morphology. As this work does not account for the timescales involved in the full suite of vegetation feedbacks, the next stage should be to include dynamical changes in both vegetation physiology and structure in GCM predictions of future climate change.


  1. 1

    Houghton, J. T. et al. (eds) Climate Change 1995(Cambridge Univ. Press, (1995)).

    Google Scholar 

  2. 2

    Sellers, P. J. et al. Comparison of radiative and physiological effects of doubled atmospheric CO2on climate. Science 271, 1402–1406 (1996).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Lean, J. & Rowntree, P. R. AGCM simulation of the impact of Amazonian deforestation on climate using an improved canopy representation. Q. J. R. Meteorol. Soc. 119, 509–530 (1993).

    ADS  Article  Google Scholar 

  4. 4

    Bonan, G. B., Pollard, D. & Thompson, S. L. Effects of boreal forest vegetation on global climate. Nature 359, 716–718 (1992).

    ADS  Article  Google Scholar 

  5. 5

    Foley, J. A., Kutzbach, J. E., Coe, M. T. & Levis, S. Feedbacks between climate and boreal forests during the Holocene epoch. Nature 371, 52–54 (1994).

    ADS  Article  Google Scholar 

  6. 6

    Gallimore, R. G. & Kutzbach, J. E. Role of orbitally induced changes in tundra area in the onset of glaciation. Nature 381, 503–505 (1996).

    ADS  CAS  Article  Google Scholar 

  7. 7

    Woodward, F., Smith, T. M. & Emanuel, W. R. Aglobal land primary productivity and phytogeography model. Glob. Biogeochem. Cycles 9, 471–490 (1995).

    ADS  CAS  Article  Google Scholar 

  8. 8

    Mitchell, J. F. B., Johns, T. C., Gregory, J. M. & Tett, S. F. B. Climate response to increasing levels of greenhouse gases and sulphate aerosols. Nature 376, 501–504 (1995).

    ADS  CAS  Article  Google Scholar 

  9. 9

    Jones, R. G., Murphy, J. M. & Noguer, M. Simulation of climate change over Europe using a nested regional-climate model. I: Assessment of control climate, including sensitivity to location of lateral boundaries. Q. J. R. Meteorol. Soc. 121, 1413–1449 (1995).

    ADS  Google Scholar 

  10. 10

    Senior, C. A. & Mitchell, J. F. B. Carbon dioxide and climate: the impact of cloud parameterization. J. Clim. 6, 393–418 (1993).

    ADS  Article  Google Scholar 

  11. 11

    Chen, T. H. et al. Cabauw experimental results from the project for intercomparison of land-surface schemes (PILPS). J. Clim.(in the press).

  12. 12

    Rowntree, P. R. & Bolton, J. A. Simulation of the atmospheric response to soil moisture anomalies over Europe. Q. J. R. Meteorol. Soc. 109, 501–526 (1983).

    ADS  Article  Google Scholar 

  13. 13

    Beljaars, A. C. M., Viterbo, P., Miller, M. J. & Betts, A. K. The anomolous rainfall over the United States during July 1993: sensitivity to land surface parameterization and soil moisture anomalies. Mon. Weath. Rev. 124, 362–383 (1996).

    ADS  Article  Google Scholar 

  14. 14

    Milly, P. C. D. & Dunne, K. A. Sensitivity of the global water cycle to the water-holding capacity of land. J. Clim. 7, 506–526 (1994).

    ADS  Article  Google Scholar 

  15. 15

    Rowntree, P. R. in Land Surface Evaporation(eds Schmugge, T. J. & Andre, J. C.) 5–30 (Springer, New York, (1991)).

    Book  Google Scholar 

  16. 16

    Wilson, M. F. & Henderson-Sellers, A. Aglobal archive of land cover and soils data for use in general circulation climate models. J. Climatol. 5, 119–143 (1985).

    Article  Google Scholar 

  17. 17

    Sellers, P. J., Mitz, Y., Sud, Y. C. & Dalcher, A. Asimple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci. 43, 505–531 (1986).

    ADS  Article  Google Scholar 

  18. 18

    Lowry, W. P. & Lowry, P. P. Fundamentals of Biometeorology. Interactions of Organisms and the Atmosphere Vol. 1 (Peavine, McMinnville, OR, (1989)).

    Google Scholar 

  19. 19

    Eagleson, P. S. Dynamic Hydrology(McGraw Hill, New York, (1970)).

    Google Scholar 

  20. 20

    Halldin, S., Saugier, B. & Pontailler, F. Y. Evapotranspiration of a deciduous forest. Simulation using routine meteorological data. J. Hydrol. 75, 323–341 (1984).

    ADS  Article  Google Scholar 

  21. 21

    Shaw, R. H. & Pereira, A. R. Aerodynamic roughness of a plant canopy. Agricult. Meterol. 26, 51–65 (1982).

    Article  Google Scholar 

Download references


We thank E. M. Blyth, J. Foley, R. J. Harding, W. J. Ingram, J. E. Lovelock, J. F. B. Mitchell, P. L. Mitchell, P. R. Rowntree, C. A. Senior, W. J. Shuttleworth, S. F. B. Tett and P. J. Valdes for comments, advice and discussion. This work was supported by the NERC TIGER programme and the UK Department of the Environment.

Author information



Additional information

Correspondence and requests for materials should be addressed to R.A.B.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Betts, R., Cox, P., Lee, S. et al. Contrasting physiological and structural vegetation feedbacks in climate change simulations. Nature 387, 796–799 (1997).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing