Modeling near-term climate change

Near-term climate change: Strategies for agricultural modeling


A clear need exists for localized climate information relevant to agriculture and extending over the next few decades. Such information can be regarded as complementary to the much longer range climate projections that are the focus of the IPCC, for example.  Climatic fluctuations on decadal time scales are more likely to arise from internal or “unforced” processes intrinsic to the system and are more likely to dominate anthropogenic trends on local-to-regional spatial scales. Ideally, predictions of regional near-term variability would be obtained from forecast systems based on climate models, but the viability of such predictions has not been demonstrated at the regional scale, particularly for agriculturally relevant variables such as precipitation.  In the absence of regional decadal forecasts, information useful for decision-making may still be obtained via stochastic decadal simulations, synthetic data sequences that mimic the properties of the observed decadal variability.  This work is designed to test some new methods for generating stochastic decadal simulations, from an agricultural impacts perspective.


The objective of the work is to advance and validate methodologies for generating stochastic decadal simulations, using Bayesian vector autoregressive moving-average models, among others.


The work will involve experimenting with a number of statistical models for the joint simulation of precipitation and minimum and maximum temperatures, for the Berg River watershed, Western Cape province, South Africa.  Additional procedures will be utilized to combine the generated simulations with the sub-annual variations required to complete the spectrum of climate variability.  The resulting sequences can then be used to drive a complex of agricultural and economic models that are used to assess vulnerability.


Outputs will include a technical report of the work, and computer code and guidance for one or more of the statistical methods evaluated.  The report will also include guidance as to possible future developments and opportunities in this field.


International Research Institute for Climate and Society (IRI), Columbia University


September 2010 – April 2011


Method development and testing at a site in South Africa