Dealing with Uncertainty in Agent‐Based Simulation: Farm‐Level Modeling of Adaptation to Climate Change in Southwest Germany

Abstract

Climate change will most likely confront agricultural producers with natural, economic, and political conditions that have not previously been observed and are largely uncertain. As a consequence, extrapolation from past data reaches its limits, and a process‐based analysis of farmer adaptation is required. Simulation of changes in crop yields using crop growth models is a first step in that direction. However, changes in crop yields are only one pathway through which climate change affects agricultural production. A meaningful process‐based analysis of farmer adaptation requires a whole‐farm analysis at the farm level. We use a highly disaggregated mathematical programming model to analyze farm‐level climate change adaptation for a mountainous area in southwest Germany. Regional‐level results are obtained by simulating each full‐time farm holding in the study area. We address parameter uncertainty and model underdetermination using a cautious calibration approach and a comprehensive uncertainty analysis. We deal with the resulting computational burden using efficient experimental designs and high‐performance computing. We show that in our study area, shifted crop management time slots can have potentially significant effects on agricultural supply, incomes, and various policy objectives promoted under German and European environmental policy schemes. The simulated effects are robust against model uncertainty and underline the importance of a comprehensive assessment of climate change impacts beyond merely looking at crop yield changes. Our simulations demonstrate how farm‐level models can contribute to a process‐based analysis of climate change adaptation if they are embedded into a systematic framework for treating inherent model uncertainty.

Emerging Opportunities for Landscape Ecological Modelling

“…However, models involving more dynamic processes, especially those related to human behaviour, may only be able to achieve reasonable, as opposed to fully Bvalid^, representations [236]. In such cases, model transparency becomes crucial, along with verification of intermediate simulation outputs through modular testing [237], and rigorous exploration of model behaviour through experiments or sensitivity and uncertainty analyses [238].…”

Managing soil functions for a sustainable bioeconomy—Assessment framework and state of the art

“…Recent trends, including emerging pesticide resistance and diversified consumer demand, may lead to increased diversity of crop rotations (DLG-Vorstand, 2017;Last, Buchmann, Gilgen, Grant, & Shreck, 2015); however, more narrow rotations are also possible in some regions, for example, due to climate change (Troost & Berger, 2015).…”

Next generation data systems and knowledge products to support agricultural producers and science-based policy decision making

“…2). Estimation of parameters for TOA-MD and other spatially-explicit impact assessment models requires data from a statistically representative sample of the farm population, as discussed in Antle and Capalbo (2001) for econometric models, and in Troost and Berger (2014) for models based on mathematical programming.

Fig. 2TOA-MD model data inputs and outputs.

…”