Introduction
Grassland ecosystems are sensitive to water stress. Seasonal drought can thus disrupt the normal progression of herbage accumulation, with consequences for the economy of grassland-based farming systems. In many areas of the world, risk from drought is anticipated to increase in the future, as climate change is likely to cause not only a rise in temperature but also a shift in precipitation patterns and a decrease in growing-season precipitation levels. The ability to gauge the impact of these changes on herbage production is central to the discussion of how livestock systems can cope with future climatic conditions (Hopkins and Del Prado, Reference Hopkins and Del Prado2007).
Grassland simulation models are expected to play a key role in translating climate change scenarios into information for adaptation planning (Wilby et al., Reference Wilby, Troni, Biot, Tedd, Hewitson, Smith and Sutton2009). Clearly, the utility of such assessments depends on the consistency and trustworthiness of the predictions. The purpose of this paper is to discuss some of the problems and challenges faced in modelling the impacts of seasonal drought on herbage growth. The discussion draws on results from numerical experiments carried out with three models of increasing mathematical complexity.
Material and methods
The models considered for this study were as follows:
-
∙ ModVege (Jouven et al., Reference Jouven, Carrère and Baumont2006), a simple model for multi-species grasslands operating at the daily time step that adopts a conceptual framework to calculate the mass flow in four structural compartments of above-ground biomass. Potential growth is predicted from intercepted photosynthetic active radiation via prescription of the radiation-use efficiency, whereas environmental controls on growth are expressed as a function of nitrogen availability, temperature, radiation intensity and soil water availability. A standard model setup is available for four functional groups in the classification of Cruz et al. (Reference Cruz, Duru, Therond, Theau, Ducourtieux, Jouany, Al Haj Khaled and Ansquer2002) that span a substantial fraction of the grassland types of Europe.
-
∙ PROGRASS (Lazzarotto et al., Reference Lazzarotto, Calanca and Fuhrer2009; Lazzarotto et al., Reference Lazzarotto, Calanca, Semenov and Fuhrer2010), a dynamic, plot-scale model of intermediate complexity developed for the simulation of the seasonal and inter-annual dynamics of productive grass/clover mixtures. Operating at the hourly time step, the model explicitly accounts for light, water and nitrogen relations, being capable of predicting different co-existence regimes as a function of environmental conditions and management.
-
∙ ecosys (Grant, Reference Grant2001), a comprehensive mathematical model of natural and managed terrestrial ecosystems (including permanent grasslands) that is entirely constructed from algorithms representing basic physical, chemical and biological processes, and capable of evaluation against laboratory data (i.e. independently of the model). ecosys operates with an hourly time step and uses multiple canopy and soil layers to simulate the fluxes of resources (energy, water, nutrients) within the soil–vegetation–atmosphere continuum. It can cope with multiple species, distinguishing (if required) between dominant and non-dominant sward components (Grant, Reference Grant2001). Root growth, an important aspect of the model for predicting the effects of seasonal drought, is formulated following Grant (Reference Grant1998). This approach allows the root profile to continuously adjust to varying soil water conditions.
Monitoring and experimental data collected in Switzerland were used to setup the models and for verification of the model predictions. For all simulations, observed weather data were used as input. Management was specified according to the information available from the field books. The main management information required by the models included the dates of cutting and fertilization, as well as the amount and type of applied fertilizer. In the case of PROGRASS and ecosys, spin-up runs were conducted for proper initialization of the carbon, nitrogen and energy cycles.
Results
All three models were able to simulate the seasonal decline in water availability under drought. The best agreement with the observations was obtained with ecosys, the only one of the tested models that includes an explicit, layered solution of the water transfer equations in the soil–root continuum.
The simulated impact of water stress on above-ground biomass was consistent with the general trends disclosed by the experimental and monitoring data. Nevertheless, ModVege and PROGRASS displayed a tendency to overestimate the effects of water stress, with implications for the simulated rate of recovery from stress after drought. The opposite was true for ecosys. The inclusion of alternate formulations of the water stress function in ModVege and PROGRASS did not improve the results (Calanca et al., Reference Calanca, Deléglise, Martin, Carrère and Mosimann2016).
For PROGRASS and ecosys, considerable differences were found with regard to how water deficit affects root turnover of the individual constituents of the sward. Although both models predicted a decline in root biomass of the clover component in grass/clover mixtures exposed to seasonal drought, only PROGRASS simulated a decrease in the root biomass of the grass component.
Overall, the models performed satisfactorily in predicting the start of the growing season and other cardinal dates in the phenological calendar, in particular the timing of peak growth in spring. In a few cases, however, the start of the growing seasons was set too early in the year. This has implications for the overall model performance under prescription of spring drought.
Discussion
Most grassland models do already include an adequate formulation of the effects of water deficit on photosynthesis and transpiration. At times, this is sufficient to predict the total effect of seasonal drought on herbage growth. In general, however, the models appear to lack the level of realism necessary to reproduce the range of responses observed in the field. The present analysis disclosed deficiencies irrespective of model complexity. It further highlighted a few issues that are worth being considered with priority in upcoming efforts to improve grassland models:
-
∙ Phenology: as with annual crops, grassland responses to water deficit are likely to depend on phenology, namely the developmental stage of the individual constituents of the sward. Conversely, water scarcity is likely to induce shifts in phenology that can eventually distress production (McMaster et al., Reference McMaster, White, Weiss, Baenziger, Wilhelm, Porter and Jamieson2008). Both of these paths of response to seasonal drought are missing in many grassland models, including those examined in this study.
-
∙ Allocation of assimilates and root dynamics: grassland ecosystems can reduce their sensitivity with respect to persistent water deficit through increased allocation of assimilates to the root system and enhanced root growth. The diverging responses produced by PROGRASS and ecosys suggest limited understanding of the underlying processes and scope for improvement. Unfortunately, available experimental evidence is also inconclusive in this regard (Prechsl et al., Reference Prechsl, Burri, Gilgen, Kahmen and Buchmann2015).
-
∙ Community dynamics: interactions at the community level and shifts in the composition of the sward contribute to shaping the overall responses to prolonged water stress. These processes operate on time scales ranging from the seasonal to the decadal, and become therefore particularly important for the assessment of climate change impacts. A better hold on species interactions and long-term ecosystem dynamics is required for informing adaptation.
Acknowledgements
This research contributes to the project ‘MODelling vegetation response to EXTREMe Events’ (MODEXTREME, http://modextreme.org), which has received funding from European Community’s Seventh Framework Programme – FP7 (KBBE.2013.1.4-09) under Grant Agreement No. 613817, 2013-2016. The Swiss Federal Office of Meteorology and Climatology (MeteoSwiss) is kindly acknowledged for providing access to its weather data repository.
