Vegetation and climate varability : a coupled model study
by
Crucifix, Michel
Hadley Centre, Met Office, Exeter EX1 3PB, United Kingdom
Coauthors: Betts, Richard and Cox, Peter

Vegetation is an important component of the climate system. To date, the majority of the general circulation model studies have considered the impact of prescribed changes of vegetation patterns on the mean climate. The present study explicitly addresses the dynamic coupling between vegetation and climate using a model of the global atmospheric general circulation coupled to a slab ocean (HadSM3) and a dynamic vegetation model (TRIFFID). Two simulations, one with interactive vegetation and another with fixed vegetation, enable us to quantify the impact of seasonal and interannual vegetation changes on the mean climate and its interannual variability. Anthropogenic changes are excluded from the analysis. Vegetation shows substantial interannual variability in central Australia and in the Sahel. Statistical tests on the annual mean precipitation time series yield the conclusion that the impact of interannual vegetation variability on annual mean precipitation is marginally statistically significant in Australia, and below the detection level in the Sahel. This is found to be consistent with the quantitative analysis of the impact of vegetation fraction changes on the energy balance of the boundary layer. We conclude in the HadSM3 - TRIFFID model, vegetation variability is not a prime cause of precipitation variability even in semi-arid areas at the continental scale. This result may be dependent on the model, or on systematic biases in the simulated vegetation. Review of the mechanisms relevant for semi-arid areas Vegetation can influence climate by modifying the surface albedo, water conductance and surface roughness: Vegetation growth decreases the albedo of the surface. This is expected to impact on the surface energy balance and on the radiative transfer through the atmosphere column. Plants use soil water for the photosynthesis and release it in the atmosphere during respiration. Leaves also intercept rainwater, which remains directly available for evaporation. The modulation of the dynamics of the hydrological cycle by plants has particular significance in semi-arid areas where almost all rainwater is recycled within a few months. Evaporation tends to increase the Bowen ratio (the ratio between latent and sensible heat exchange). This increases the relative humidity in the planetary boundary layer, which in turn favours moist convective events instead of dry ones (Sud et al., 1993). Because vegetation has a larger roughness length than bare soil, it increases the turbulent heat exchanges between the surface and the boundary layer, at the expense of infrared emission. Infrared radiation emitted by the surface is partly absorbed by air and clouds, the remainder being transmitted to space. By contrast, turbulent sensible and latent heat delivered by the surface are first mixed in boundary layer were they might possibly be converted into potential energy available for convection (CAPE). Hence, an increase in roughness length is expected to increase the frequency of convective events. Thus, changes in surface properties influence the local vertical profiles of temperature and moisture in the atmosphere, and the consequences for precipitation may be important given the critical role of convective processes in semi-arid areas. Sometimes, arguments based on the large-scale circulation of the atmosphere are invoked. This was the case to explain how the gentle decrease in insolation during the Holocene may have, at some stage, triggered a positive feedback between vegetation and monsoon dynamics (Claussen et al., 1999) and caused the abrupt desertification of Sahara 5, 500 years ago. Experimental design We used HadSM3, the slab-ocean version of the Hadley Centre Climate Model version 3 described by Pope et al., 2000. It is here coupled to the surface scheme MOSES II and the TRIFFID vegetation model (Essery et al., 2003; Cox et al., 2001). We discuss here 465 years of climate simulation forced by pre-industrial boundary conditions. The time evolutions of the globally averaged, tree and grass fractions shown on Figure 1 serve to illustrate the experimental setup. There is first a spin-up of 200 years. The model is run for another 200 years with the atmosphere and vegetation normally coupled. This part of the experiment is referred to as DYN, for dynamic vegetation. Another experiment, called FIX, is initialised with the climate of year 215 of the first experiment, and then run for another 150 years. Vegetation fractions and leaf area indices are fixed to the mean of the years 70 to 100 of DYN, and leaf phenology is disabled. Comparing DYN and FIX will enable us to quantify the impact of interactive vegetation on the mean climate as well as its variability. Climate variability with fixed and dynamic vegetation Detection of variability in HadSM3 - TRIFFID runs Figure 1 shows that in DYN, the annual-mean fraction of C4 grass oscillates with a typical time scale of about one to three decades. In central Australia, for example, which we define as the continental region between 125 E and 145 E, and between 35 S and 15 S, its standard deviation is 10 There are several ways to systematically test for autocorrelation in a time-series. We have selected one. Its principle is based on the fact that the standard deviations of n-year-long means of an unautocorrelated time series behave as , with /2. If the time series is auto-correlated, then is larger. Figure 3 displays the estimates of for continental, annual mean precipitation time series in FIX and DYN (last 150 years). is around -0.5 in most places, both in FIX and in DYN, which indicates that annual mean precipitation over the continents tends to behave as white noise in HadSM3. The main difference between FIX and DYN occurs in central Australia. Estimates for are around -0.5 in FIX, and -0.38 in DYN, that is, for the latter, slightly beyond the 95 Description of the feedback mechanism We assume that if vegetation influences continental-scale precipitation, it does so by encouraging moist convection. Several quantities are helpful to determine whether this is the case. The first one is the sum of the turbulent and radiative sources of equivalent potential temperature in the boundary layer (henceforth, /, r, where the indice t, r stands for turbulent and radiative). We use it as an indicator of the buoyancy forcing of the atmosphere. Annual mean /, r (expressed in K/day) can be decomposed into contributions from turbulent heat fluxes, longwave and shortwave net absorption. It turns out that a vary large part of the variance of these components can be explained by a linear regression of the kind: F(n)=F0+4 C4(n) + P(n) () Which means that the fractional grass cover and annual mean precipitation largely determine these quantities. Estimates of F0, 4 and in Australia are listed in Table 1, along with the percentage of variance explained by P and C4 as obtained by path analysis. For example, variation in grass cover explains 62 The difference between the temperature and the dew point temperature (T-Td) in the first level of the model is an indicator of the likelyhood of moist convective events versus dry ones. The smaller this difference, the lower the lifting condensation level (LCL), and hence the more likely moist convection is to occur. Table 1 shows that grass does not influence neither the Bowen ratio nor the saturation level of the boundary layer (both are controlled at by precipitation, via the soil moisture). This suggests that evapotranspiration has not an important impact on climate in central Australia. The impact of grass growth on the surface albedo is not essential either, because the albedo of grass (0.20) is very close to that of the Australian red soil (0.25) prescribed in the model after Wilson and Henderson-Sellers, 1985. Therefore, we conclude that roughness length is the important parameter. An increase in roughness length contributes to increase the turbulent heat exchanges between the surface and the boundary layer at the expense of infrared emission. In total, a 30

The exercise can be repeated for the Sahel, here defined as the continental area between 15 #730;W to 20 #730;E and 12 #730;N to 16 #730;N. In the previous section, we concluded from the analysis of the precipitation time series that the model does not exhibit any precipitation - vegetation feedback at the interannual scale. Table 2 shows that a same change in C4 grass cover (i.e., looking at 4) has similar consequences in Sahel and in Australia. 4 even tends to be larger in Sahel. In particular, evapo-transpiration probably does play a role.

This is suggested by the large partial regression coefficients between grass and the Bowen ratio, the 1.5-m air temperature and the dew point temperature, even though the uncertainties are fairly large. The large uncertainties are due to the quite small variance in annual mean grass area (three times less than in Australia, compare Figures 2 and 4). Consequently, Sahel grass variability only explains a very small part of the boundary layer heating total variance. The difference between Sahel and Australia can be found in the very zonal structure of vegetation distribution in Sahel. The model suggests here that in South Sahel, plants do not expand much because they already cover most of the ground.

Summary In summary, we conclude that in our model of the large-scale climate and vegetation dynamics (resolution of 3.75 o x 2.75 o) that the dynamic coupling between vegetation and climate is not a prime source of climate variability at the intra-seasonal and inter-annual time scales. This result, which applies to the continental scale, implicitly neglects the possible role of sub-synoptic scale vegetation changes.

Claussen, M., C. Kubatzki, V. Brovkin, and A. Ganopolski, 1999. Simulation of an abrupt change in Saharan vegetation in the mid-Holocene, Geophysical Research Letters, 26, 2037–2040. Cox, P. M., R. A. Betts, C. D. Jones, S. A. Spall, and I. J. Totterdell, 2001. Modelling vegetation and the carbon cycle as interactive elements of the climate system, in Meteorology at the Millennium, edited by R. Pearce, Academic Press. Essery, R. L. H., M. J. Best, R. A. Betts, P. M. Cox, and C. M. Taylor, 2003. Explicit representation of subgrid heterogeneity in a GCM land-surface scheme, J. Hydromet., 4, 530–543. Pope, V. D., M. L. Gallani, P. R. Rowntree, and R. A. Stratton, 2000. The impact of new physical parametrizations in the Hadley Centre climate model – HadAM3, Climate Dyn., 16, 123–146. Sud, Y. C., W. C. Chao, and G. K. Walker, 1993. Dependence of rainfall on vegetation: theoretical considerations, simulation experiments, observations, and inferences from simulated atmospheric soundings, J. Arid Environ., pp. 5–18. Wilson, M. F. and A. Henderson-Sellers, 1985. A global archive of land cover and soils data for use in general circulation climate models, Journal of Climatology, 5, 119–143.

Date received: November 28, 2003

Copyright © 2003 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # camu-14.