Climate extremes and ecosystem productivity in global warming simulations

Ecosystem responses to present-day droughts and heat-waves are often considered indicative of future global warming impacts on ecosystems, under the assumption that the temperature above which vegetation experiences heat and drought stress is invariant with changes in climate and carbon dioxide concentration. Understanding how the impacts of temperature extremes on ecosystems can change with climate change is essential for correctly evaluating and developing Earth System Models (ESMs). The Coupled Model Inter-comparison Project (CMIP5) historical and future (RCP8.5) climate predictions were analyzed in this study to illustrate non-stationarity of climate impacts on ecosystems, as evident by changes in the distribution of Gross Primary Production (GPP) as a function of temperature between future and historical climates. These changes consist of (1) a uniform shift in the GPP distribution toward warmer temperatures between future and historical climates, and (2) a proportional increase in GPP at all temperatures, consistent with CO2 fertilization. The temperature at which GPP has a local maximum within a given climate increases with global warming and closely tracks the change in mean temperature for each ecosystem. This maximum GPP temperature can be conceptualized as a stable equilibrium determined by the temperature at which an increase in plant water stress is compensated by a decrease in light stress (decreasing cloud cover) with increasing temperature. Temperature relative to the temperature of maximum GPP is proposed as an improved measure of climate extremes more relevant to ecosystem productivity than absolute temperature. The percentage change in GPP attributed to changes in relative temperature extremes is up to 3% per K (decrease in GPP), and reflects both an increase in the frequency of climate extremes in global warming scenarios and the change in temperature criteria for negative climate impacts on ecosystem productivity. Temperature at GPP maximum as a function of the growing season average temperature, for each ecosystem type. Lengths of crosshairs indicate the standard error of the CMIP5 ensemble average. Both future and historical averages are shown (the temperature of GPP maximum increases monotonically with growing season average temperature between future and historical simulations).