Crop heat stress in the context of Earth System modeling

A little over ten years ago I became familiar with the crop parameterizations of the AgroIBIS dynamic vegetation model (Kucharik and Brye 2003) and introduced these to the Community Land Model (CLM) (Levis et al 2012). CLMcrop belongs to a class of process-based ecological model intended to improve Earth System Model (ESM) climate and carbon cycle simulations through the more accurate representation of land surface characteristics. Unlike in detailed crop models developed to match local to regional crop behaviors by design (e.g., Hodges et al 1987), the representation of agricultural management beyond approximate planting, harvesting, fertilizing, and irrigating exceeded the level of detail considered worth including in CLMcrop given the potential complexity in implementation and uncertainty in global scale representation.

Such considerations have changed a decade later and models like CLMcrop are used to simulate potential effects of projected climate change on human systems (Müller and Robertson 2014). Models like CLMcrop have the advantage that they may run synchronously coupled to the ESM simulating the climate and provide in this way crop yields at large spatial scales while feeding back energy, water, carbon, and nitrogen fluxes to the atmosphere and ocean components of the ESM (Osborne et al 2009).

The shift in attention from land–atmosphere interactions to crop yields brings into focus limitations in the modeling approaches that have served ESMs to date. For example, CLMcrop simulates increasing yields for some crops in an RCP8.5 climate (the most pessimistic scenario of 21st century climate change considered by the Intergovernmental Panel on Climate Change) even when excluding the beneficial effect of CO₂ fertilization to plant productivity (unpublished results). Is this plausible? Do heat extremes affect this model's crops? In CLM's algorithms, optimum leaf temperature for C3 and C4 plant photosynthesis is about 28 °C and 38 °C, respectively, declining for higher and lower leaf temperatures (Bonan et al 2011). CLM's photosynthesis declines also as soils become drier.

Siebert et al (2014) emphasize that heat stress affects crop yield the most during sensitive phases in a crop's growth cycle, sometimes with little chance for recovery later in the season. We do not account for such specific effects of heat in CLMcrop. Siebert et al focus in particular on wheat heat stress during anthesis in Germany. They find that a decline in wheat yield due to heat stress is much lower when calculated from hourly 2 m rather than ground temperature observations (both collected at weather-stations). The calculation involves a GDD-like temperature accumulation (GDD: Growing Degree Days) that Siebert et al refer to as Stress Thermal Time (STT). They define STT as the sum of hourly positive temperature differences (T–T_crit) for 18 days beginning at heading. For wheat, heading precedes anthesis by a few days. T_crit is defined as critical threshold temperature, set to 31 °C according to pot experiments under controlled conditions.

Siebert et al (2014) explain that a crop canopy's thermal gradient can lead to overestimated STT if calculating the STT from the ground temperature and underestimated STT if calculating STT from 2 m air temperature. Yield as a function of STT as reported from controlled experiments best matched the field response when calculating the STT from 0.2 m air temperature. In Germany, only agro-meteorological stations measure 0.2 m air temperature, offering limited data coverage for regional analyses.

Siebert et al recommend that crop model algorithms determining the effect of heat stress on yield should derive a canopy temperature, as equivalent to the 0.2 m air temperature reported by agro-meteorological stations, instead of using 2 m temperature. To avoid this and continue using the 2 m temperature, Siebert et al consider the option of lowering T_crit in crop models. They find reduced agreement to the controlled experiments this way, suggesting that factors such as soil moisture may affect the canopy temperature more than the 2 m temperature, making the 2 m temperature a less reliable predictor of yield even with T_crit set to a lower value.

Siebert et al expose the reader to extensive research published on the potential effects of temperature extremes on crop health and yields. I now see the introduction of heat stress effects on crop yield as a priority for ESM development. The work by Deryng et al (2014) interests me in particular because they document the spring wheat, maize, and soybean heat stress-at-anthesis factor (f_HSA) of the PEGASUS global crop model. Tao and Zhang (2013) document an approach for rice. To include such parameterizations, some ESMs may need to begin calculating canopy temperature.

Still, to introduce appropriate heat stress effects on yield in ESMs, we have long way to go. The results of Siebert et al (2014) apply to wheat in Germany. The results of Asseng et al (2011) apply to wheat in Australia. Sanches et al (2014) focus on maize and rice. Continued fieldwork will help quantify the effects of heat stress on the yield of more crops; the parameterization of such results will help us introduce these effects in ESMs for projects that evaluate 21st century food security (Challinor et al 2014, Rosenzweig et al 2014).