Page 113 - Climate Change and Food Systems

■ Spatial analysis (agent-based models)
Agent-based models (ABMs) refer to a modelling approach that builds on the tradition of Recursive Farm Programming Models, but simulate all individual farms, their spatial interactions and
the natural environment (Berger et al., 2006). It has been argued that ABMs can be very useful
for adaptation analysis and for policy guidance. According to Wreford et al. (2010), evaluating climate change impact on agriculture requires examining scenarios at a more localized levels to develop estimates for local area adjustment costs, for cost-benefit analysis for adaptation plans,
and for building robust climate resilience policies. The proponents of ABMs argue that climate change is expected to have location-specific impacts, but most models are too aggregated
to provide guidance for more targeted policy interventions. Statistical models or Ricardian type models are only useful in capturing human- environment interaction at “coarse resolution” and may not have the necessary detail to allow for a more refined assessment of climate change in agriculture. ABMs have been suggested
as complementary tools for assessing farmer responses to climate change in agriculture and how these are affected by policies (Moss et al., 2001; Patt and Siebenhuner, 2005; Troost and Berger, 2014).
ABMs can address spatial heterogeneity
and effects in a system of distributed, but interdependent and hierarchical, decision-making. While ABMs are not predictive, they are well suited to counter-factual experiments, “what-if”- type analyses and policy discussion. According
to Berger and Troost (2014), the ABM has the capacity to meet a number of critical requirements for analysing farm-level adaptation options due to climate change. Among these requirements are:
(i) Incorporating fine technical and financial detail at the farm level.
(ii) Facilitating policy analysis based on modelling farmers’ decisions, including expectations, learning and risk behaviour.
(iii) Accounting for the role of farmers’ cumulative experience, capacity to learn from neighbours and exchange of information.
(iv) Allowing for spatial interactions in decision, including issues of local land competition among alternative uses.
(v) Accounting for environmental interactions and feedback and ability to account for events like the occurrence of flooding events or invasion of new pests and diseases.
(vi) Performing sensitivity tests using Monte-Carlo or other methods.
Malanson et al. (2014) examine the effects
of extended climatic variability on agricultural
land use using an Agent-based model applied to villages in the Nang Rong district of northeastern Thailand. The land use decisions are made by each household in a village for which socio-economic data was collected through intensive surveys from 41 villages in 1984,1994 and 2000 across Nang Rong District. The land use change decision in
the ABM has five alternatives: jasmine rice, heavy rice, cassava, sugar cane and unused-by-village. The primary basis for change in a given year is the expected income from the given crops, given their yields and prices, but constrained by labour, assets and a threshold of willingness to change. Climate change is modelled through nine weather scenarios based on timing and amount ([early, normal, late] and [low, normal, high]) of monsoon precipitation. These scenarios are based on data on actual monthly rainfall in Nang Rong from 1900 to 2008.
The household-level land decision-making is presented at the village level to allow for broader inference. Modelled (virtual) villages change their agricultural effort in many different ways. While most “virtual” villages reduce the amount of land under cultivation, primarily with reduction in jasmine rice, others do not. The analysis revealed insights into the role of landscape and society in land use
in scenarios of climate change, but the statistical relations are weak, which limits inference. The authors conclude that while ABMs are able in theory to incorporate the variations to which complex systems are sensitive, they require precision in the
chapter 3: economic modelling of climate impacts and adaptation in agriculture: a survey of methods, results and gaps
93