Skip to:

August 2011 - July 2015

Quantification and reduction of uncertainties in projections of climate impacts on drought and agriculture for North America


Justin Sheffield
Principal Investigator
Princeton University
Principal Investigator
Senior Fellow
  • Associate Professor, Earth System Science
  • William Wrigley Fellow at the Freeman Spogli Institute for International Studies and Stanford Woods Institute for the Environment

Introduction to the Problem: Agricultural productivity is highly dependent on climate variability and is thus susceptible to future changes including temperature extremes and drought. The latter is expected to increase in frequency regionally over this century. However, the uncertainty in projections of drought and its impacts on agriculture is high due to emission scenarios, climate model differences, uncertainty in initial/boundary conditions, and translation to regional scales. Climate models are unanimous in projecting future warming but differ in the magnitude and even sign of regional precipitation changes. They also differ in terms of extremes of temperature, precipitation and other meteorology. When projecting future impacts on crop productivity, these uncertainties are compounded because current crop models often use simplified treatments of climate response and do not include comprehensive treatments of water availability. Therefore, projections of regional climate change, variability and its impacts on water availability and agriculture are highly uncertain and reduction of uncertainties requires attention to all levels in the climate-water-agriculture continuum.

Rationale: Given the uncertainties in future agricultural production and the complex relationships between climate, hydrology and crop development, there is pressing need to make improved estimates of future changes in climate change and crop yields. We propose to evaluate the uncertainties in estimates of future changes in climate, water availability and agricultural production, and make improved estimates by incorporating state of the art knowledge of the relationships between climate, hydrology and agriculture into modeling and downscaling. This has ramifications for disaster preparedness and mitigation, policy making and the political response to climate change, and intersects with fundamental science questions about climate change, extremes and hydrologic cycle intensification. It is central to the mission of the Climate Program Office’s MAPP program to “enhance the Nation’s capability to predict variability and changes of the Earth’s System” and directly addresses its priorities to evaluate and reduce uncertainties in climate projections. This work will leverage from the PIs’ experience and ongoing activities in large-scale climate analysis and hydrologic modeling, particularly in changes in drought historically and under future climates, and agricultural modeling and relationships between climate and crop productivity.

Summary of work to be completed:

  1. Quantify the relationships between hydroclimate variables and the implications for water, drought and agriculture based on observational data. 
  2. Evaluate sensitivities of hydrologic and crop models to changes in climate and drought. Differences in climate variability, land-atmosphere coupling and hydrologic persistence will lead to differences in key metrics of water and agriculture which will form the basis for evaluation of the uncertainties in future projections. 
  3. Evaluate current climate models in how they replicate these observed relationships using the CMIP5 long-term and decadal predictions. 
  4. Estimate uncertainties in future projections of climate, drought and agriculture using a cascade of climate, downscaling, hydrologic and crop models with strategic sampling to decompose sources of uncertainty. 
  5. Implement a set of methods to reduce uncertainties in future projections based on observational constraints including merging of climate model predictions, bias correction and scaling of climate model output, and improvements to impact models.