Ecological Feedbacks to Global Warming: Extending Results from Plot to Landscape Scale

John Harte

University of California

Berkeley, California

Physics-based climate models predict that during the coming century, global-averaged surface temperatures will rise significantly because of human-caused increases in the concentrations of greenhouse gases in the atmosphere (IPCC, 1996). We know little, however, about the ecological consequences of this warming, and we know even less about the extent to which ecosystem responses to warming will trigger feedback effects on the climate that will either enhance (positive feedback) or suppress (negative feedback) the warming. To reduce these uncertainties, since 1990 Harte and colleagues have been observing the biogeochemical and vegetational effects of heating a subalpine meadow.

Warming is achieved with overhead electric radiators designed to continuously mimic the model-predicted warming (see Figure 1.14). The five 3 meter by 10 meter heated plots and the five similar-sized control plots contain habitat ranging from a mixed shrub-steppe and grassland vegetational community along a semi-arid ridge down to a moist swale containing a diverse assemblage of forbs. The team has been monitoring soil temperature and moisture (every two hours at 5, 15 and 25 cm depths), floral productivity, phenology, and diversity, changes in net carbon storage above and below ground, soil mesofaunal biomass and species diversity, methane consumption rates, nitrogen pool sizes and turnover rates, and plant water stress.

We know little about the ecological consequences of climate warming, and we know even less about the extent to which ecosystem responses to warming will trigger feedback effects on the climate.

Among the major findings:

(1) heated-plot soils average 2°C hotter and 5 to 25% drier than controls and there is a sharp diurnal cycle in the temperature difference (up to 6°C warmer in midday) (Harte et al., 1995);

(2) the snowfree season is about 1 month longer in the heated plots (Harte et al., 1995);

(3) heating shifts the flora from forbs to shrubs such as sagebrush (Harte and Shaw, 1995);

(4) carbon is lost from the heated plots relative to the controls; the cause of this loss is a decline in litter input to soil rather than an increase in the soil decomposition rate in the heated plots (Saleska, Harte and Torn, 1997);

(5) heating enhances mesofaunal diversity and biomass in a cool, wet year and reduces them in a hot dry one (Harte, Rawa and Price, 1996);

(6) soil drying reduces methane consumption under relatively dry ambient soil conditions and enhances it under more moist ambient conditions (Torn and Harte, 1996).

These findings portend several important ecological feedbacks to climate change:

(1) A climate-change-induced alteration of the rate of methane consumption by soil microbes will alter the atmospheric concentration of this greenhouse gas. Results indicate that this feedback can be either positive or negative, depending on ambient soil moisture conditions.

(2) Because the albedo of vegetation differs from species to species, and from forbs to shrubs, the observed shift toward sagebrush dominance will alter surface albedo; radiometric measurements at the site indicate that this will result in lower albedo and thus the climate-induced shift in the composition of the vegetation community will result in a positive feedback to warming.

(3) The observed loss of stored ecosystem carbon in the heated plots implies a positive feedback to warming.

To extend the spatial and temporal generality of these findings, the critical question of scale must be addressed. In particular: How, and using what criteria for success, can results of manipulation experiments carried out at the spatial scale of experimental plots (~10-100 m²) and the temporal scale of NSF-funded studies (a few years) be extrapolated to the scales of concern (landscapes to global and decades to centuries)?

Naive extrapolation from plots to landscape entails simple multiplication by area. This may not be possible for either of two reasons:

(a) Edge effects resulting from the small size of experimental plots may render plot-scale results inapplicable to larger areas.

(b) Responses to climate change differ from one site to another within the landscape.

To extend the spatial and temporal generality of these findings, the critical question of scale must be addressed.

If reason (a) applies, no means of extrapolation may be possible, naive or otherwise; plot-scale results would be artifactual and irrelevant to larger areas. Harte and colleagues have developed and applied tests to show that the patterns of ecosystem response they have observed at the plot scale do not result from edge effects (see, e.g., Harte et al., 1996). The more interesting and likely impediment is reason (b); in this situation, extrapolation may still be possible, but rules for doing so have to be identified and tested. Confidence in those rules will be enhanced to the extent that a mechanistic understanding of patterns of response is developed.

Extrapolation of plot scale results to larger areas may be possible, but rules for doing so have to be identified and tested.

A possible approach to improving naive extrapolation relies on information about natural variation in ecosystem parameters along natural climatic gradients. Under the assumption that ecological variation along natural climate gradients mirrors ecological responses to manipulated climate (i. e. , variation in space is a surrogate for variation in time), data from large-scale gradients can be used to predict large-scale ecological responses to warming. This, too, can fail, however, because of a mismatch in the time scales for response to relatively rapid anthropogenic climatic change and for slower adaptation to natural climatic variation along elevational gradients. Moreover, contingent factors that have little relation to climate variation along natural gradients may render ecological patterns of variation along those gradients irrelevant to anthropogenic climate change.

Click image to see color photo: Marv Waterstone, Danny Harvey, Don Wuebbles, and John Harte at the Gothic, Colorado site (photo by Susan Joy Hassol).

Confidence in these rules for extrapolation will be enhanced to the extent that a mechanistic understanding of patterns of response is developed.

These five graphs (Figure 1.15) illustrate some of the ways in which the space for time assumption can succeed or fail and point to ways in which rules for extrapolation can be identified and tested. The variable, V, is an ecological parameter such as the stock of soil carbon or the rate of sagebrush seedling establishment. T is an environmental variable such as soil temperature or moisture that both varies along a gradient and also is affected by a manipulation and perhaps varies interannually. The arrows describe the effect of the manipulation (here taken to be an increase in T) on V at particular sites. The base of the arrow is the value of V in the control plots and the tip of the arrow is its value in the manipulated plots.

The cases A through E pose a mix of declining opportunity and increasing challenge for scaling. Standard statistical methods such as analysis of covariance permit serial rejection of each of the cases A-D. For cases B-E, it is useful to search for underlying "hidden" variable(s) or contingent factors that explain these more complex patterns.

To apply this conceptual framework to montane meadows, Harte and colleagues have set up field sites and climate manipulations along an elevational gradient spanning a horizontal distance of 12 km and an elevational range of 350 m. As in the illustration, monitoring and climate manipulation are carried out at 3 sites along the gradient . Preliminary results suggest that the timing of the flowering cycle of plants along the gradient resembles cases A and C, depending on the individual species.

Soil carbon exemplifies a more interesting situation. It varies non-monotonically along the elevational gradient (arrow bases in case B), as does the biomass of grasses. Indeed, grass biomass and soil carbon correlate strongly across a range of scales from quadrats within plots, to plots within elevational sites, and across elevational sites. But graminoid biomass is unresponsive to warming, indicating that the long-term controls over soil carbon may bear little relation to the factors influencing soil carbon response to manipulated climate. The latter appears to be most responsive to forb biomass.

Click image to see color photo: The AGCI group visits John Harte's experiment in Gothic, Colorado (photo by Susan Joy Hassol).

Harte concludes that natural climate gradient analyses may not serve as a valid substitute for manipulation experiments in predicting how soil carbon will respond to global warming. Thus determination of the magnitude and even sign of this carbon-mediated ecological feedback to climate change will require further application of manipulation experiments.

References

J. Harte and R. Shaw, 1995. Shifting Dominance Within a Montane Vegetation Community: Results of a Climate Warming Experiment. Science, 267:876-880.

J. Harte, M. Torn, F-R. Chang, B. Feifarek, A. Kinzig, R. Shaw, and K. Shen, 1995. Results from a Global Warming Experiment: Soil Temperature and Moisture Responses in a Subalpine Meadow Ecosystem, Ecological Applications , 5(17):322-150.

M. Torn and J. Harte, 1996. Methane Consumption by Montane Soils: Implications for Positive and Negative Feedback with Climate Change. Biogeochemistry, 32:53-67.

J. Harte, A. Rawa, and V. Price, 1996. Effects of Manipulated Soil Microclimate on Mesofauna Biomass and Diversity, Soil Biology and Biochemistry , 28(3):313-322.

M. Loik and J. Harte, 1996. High Temperature Tolerance for Artemisia tridentata and Potentilla gracilis under a Climate Change Manipulation. Oecologia , 108:224-231.

IPCC, 1996. Climate Change 1995: The Science of Climate Change, contribution of Working Group I to the Second Assessment Report of the Intergovernmental Panel on Climate Change. Intergovernmental Panel on Climate Change. Cam bridge University Press.

Natural climate gradient analyses may not serve as a valid substitute for manipulation experiments in predicting how soil carbon will respond to global warming.

Click image to see color photo: John Harte discusses his experiment with the AGCI group (photo by Susan Joy Hassol).