Uncertainty flourishes in a warming climate

A recent study of the projected impacts of differing climate change scenarios on maize and wheat production in South Africa shows that the scientific community is still far from being able to predict the effect of global warming on food production with accuracy.

Lyndon Estes, a researcher with Princeton University, and his co-authors found a vast difference in the results between studies using mechanistic or mathematical models (MM) and those using empirical models (EM), based on data from direct observation.

Researchers generally use either one or the other to make forecasts. The Princeton study used both kinds of models together to forecast how climate change might impact maize and wheat in South Africa in 2055, and found that EMs indicated larger yield losses or smaller gains than MMs.

The EM projected a 3.6 percent reduction in maize yield, while the MM projected a 6.5 percent gain, "mostly as a result of water use efficiency gains from elevated CO2 concentrations and projected rainfall increases".

The findings were in line with those of earlier studies led by eminent researchers like David Lobell of Stanford University, a lead author of the Fifth Assessment report by Working Group II of the Intergovernmental Panel on Climate Change (IPCC), which deals with impact, adaptation and vulnerability, and Martin Parry, the former co-chair of Working Group II. Studies based on EMs, led by Lobell, showed losses that were twice as large as those using MMs, one of which was led by Parry.

"Scientists need input from policy-makers as to what are the most relevant questions and where we should focus our efforts'"

Estes told IRIN that it would be useful to employ both approaches when projecting climate impacts because “[They] have widely differing structures, assumptions, and data sensitivities, so these larger differences can help improve confidence in findings where they agree, while serving to more effectively highlight model and data limitations where they disagree."

Earlier in 2013, Sharon Gourdji, a research scholar at Stanford University, co-wrote a study with Lobell which found that by 2050, maize harvesting area exposed to extreme heat was expected to expand from 15 percent in the early 2000s to 44 percent by 2050.

"The key uncertainty in climate projections is in precipitation, due to high inter-annual variability. This plays a role in feeding into crop model forecasts," she noted.

"In the case of the South Africa study, one reason why the two modelling approaches may differ is because climate projections are for a warmer and wetter climate. If the projections were for warmer and drier, or warmer with no real change in precipitation, you might find both modelling approaches were more in agreement. (Warming is generally bad, but wetter is good, so different models can play off these effects in different ways.)"

Writing to IRIN in his personal capacity, Andrew Challinor, professor of climate impacts at the University of Leeds, UK, and a lead author of the chapter on food security in the Fifth Assessment report by IPCC Working Group II, noted, "EM versus MM comparisons are in their early stages and this kind of study is very important… Both types of model have their pros and cons and as the authors… note, a joint assessment… is an important research direction."

The opposing effects found by the models suggest a potentially greater range of uncertainty about how crops might respond to climate changes. "Given this, methods for effectively communicating uncertainty, for making decisions under uncertainty, and for making policy makers comfortable with uncertainty, should be the area of focus," Estes suggests.

A study that Challinor and Philip Thornton, a senior scientist with the International Livestock Research Institute (ILRI) Thornton co-authored with Julian Ramirez-Villegas of the Colombia-based International Center for Tropical Agriculture and Andy Jarvis of ILRI, considered improvements in global climate models for regional agriculture impact studies. "If effective and appropriate agricultural adaptation is to happen in the next two to four decades, uncertainties and lack of skill in simulated regional climates need to be communicated and understood by agricultural researchers and policy makers."

Dealing with it

Climate scientists have always pointed out how difficult it was to determine whether small changes in climate were the result of natural shorter-term fluctuations, which could be seasonal, or were brought about by global warming.

"Managing known risks may be more important than identifying completely new (and ambiguous) possibilities," Sonja Vermeulen, head of research at the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) pointed out in a recent study co-authored with Thornton.

They call for investment to help farmers widen their portfolio of strategies, both old and new, to manage climatic risks. For instance, in anticipation of possibly drier times, farmers in Sri Lanka have begun to recycle their household wastewater and scale back groundwater use to sustainable levels.

But we still need projections of climate change impact on food security. The intention of the Gourdji-led study is to indicate to policymakers and farmers what kind of adaptation strategies they should consider putting in place now.

Her study put "[An] upper bound on risk in the global maize, wheat, rice, soy markets from future warming due to extreme heat in [the reproductive stage,] a very sensitive period for crops. This study assumes no adaptation by farmers... an unlikely assumption, but it's still useful to have an upper bound of what could happen in a worst case scenario,” Gourdji told IRIN.

“Also, studies at a global scale gloss over localized details, e.g., more heat-tolerant maize varieties in the tropics. I think as scientists we are afraid that people can take our conclusions too literally without an understanding of the nuance or caveats." 

Scientists "also need input from policy-makers as to what are the most relevant questions and where we should focus our efforts,” Gourdji said.

“It's good for readers to keep in mind that uncertainties exist, and there are different classes of models with different strengths and weaknesses.”