Page 135 - Climate Change and Food Systems

chapter 4: an overview of climate change impact on crop production and its variability in europe, related uncertainties and research challenges
various approaches of current biophysical impact assessment methodology in relation to information demands are presented in Section 2.2.
Process-based crop growth simulation models were developed in the 1960s (de Wit, 1965).
They differ in complexity – i.e. level of detail in
which biophysical processes (e.g. phenology, photosynthesis, respiration, transpiration and soil evaporation) are simulated – and in which yield constraints are explicitly taken into account – i.e. just crop characteristics, temperature and solar radiation (as needed for simulations of potential production) or including water and nutrient-limited productivities as well (van Ittersum and Rabbinge, 1997).
The more complex cropping system models are able to integrate processes of carbon and nitrogen and detailed water balance components (evapotranspiration, soil moisture, deep percolation, etc.) from planting to maturity, and can provide estimates of final yield and biomass production as well as daily values of crop and soil components during the crop growth cycle (van Ittersum et al., 2003). Only a few models also cover dynamics of phosphorus and estimates of GHG fluxes. Also quite rare are models that cover crop-weed interactions, damage by pests and pathogens (Savary et al., 2006).
In fact, there is no fully deterministic crop simulation model widely used for practical applications such as CC impact assessment at field/farm or higher aggregation levels. The less detailed the process model and – usually – the larger the spatial extent it is targeting, the more empirical relationships are incorporated. If a complex process like photosynthesis by crop canopies is simplified and reduced to such an extent that only a few parameters are sufficient in a crop model to mimick the influence of temperature and irradiation on gross assimilation in a crop model, the model itself becomes less generic and usually requires local or region-specific data for statistically deriving robust values for those key (such as radiation use efficiency, RUE); likewise, the larger the spatial extent (pixel size) of the basic calculation units used in either the crop models
or climate models, the more certain detailed data
(be it soil data or climatic variables) need to be aggregated or generalized - and the more data gaps usually occur that make it then impossible
to retain detailed process descriptions - but
rather resort to simplifications that again result
in replacement of physically based process descriptions by statistically derived empirical relationships. This rule of thumb is not restricted to crop simulation models but also applies to climate models (see Rummukainen, 2010; 2014).
For this reason, we find a wide range of crop models that are semi-empirical – i.e. they combine deterministic elements based on biophysical, chemical and ecophysiological principles with a number of empirical parameterizations (e.g. setting a fixed fraction for run-off instead of simulating its underlying processes, such as infiltration, soil water flows, etc. in detail). Simplifying and parameterizing processes in this way requires crop- or region- specific calibration and validation.
Given the capabilities and specific limitations
of each individual biophysical impact assessment approach – agroclimatic index, statistical model or crop simulation model – it has been suggested that a combination of these different approaches would be fruitful (Challinor, 2011). However, to date, application of such a combination approach is still rare (e.g. Rötter et al., 2013a).
In this paper, we briefly characterize current biophysical models and assessment tools that
are part of state-of-the-art integrated CC impact assessment methodology for agriculture (e.g. Nelson et al., 2014; Figure 3). As mentioned previously, despite their limitations, crop simulation models are the tools most widely used at present as part of integrated CC impact assessments at different scales. Figure 3 illustrates the state-of- the-art methodology applied in a recent global study on CC effects on agriculture (Nelson et al., 2014) by means of a modelling chain comprising global-scale climate models, gridded crop models and economic models to estimate changes in crop yield, cultivation area, food consumption and trade. The various implications of using “imperfect crop models” in the integrated assessment methodology are further discussed in the next
115