Servicios Públicos y Sociales

The primary aim of the work described in this thesis was to quantify and understand the implications of introducing new sustainable greenhouse systems by developing simulation models to describe the dynamics of the selected pest, disease and biological control agents for different climate management scenarios. For the purposes of this work, sustainable greenhouse systems are defined as those that

15

minimise the need for fossil fuel, water and chemical inputs, make efficient use those that are added and reduce their escape from the system. These models would predict the pest and disease pressures and the extent to which they can be ameliorated by biological control in response to greenhouse temperature and humidity changes throughout the year. Models such as these can be defined as being “a representation in mathematical terms of the behaviour of real devices and objects” (Dym, 2007) and, as the system or device in question is a living system, they are more precisely described here as ecological models (Gillman and Hails, 1997; Hill and Coquillard, 2007). Such models contribute to the understanding of the complex interactions between organisms and the environment (Hill and Coquillard, 2007; Breckling et al., 2011; Jorgensen and DeAngelis, 2011) and their uses include discerning the properties underpinning a system, testing hypotheses and assumptions, and suggesting future research priorities by indicating areas where important information is lacking (Hill and Coquillard, 2007; Breckling et al., 2011). However, ecological systems are usually complex and so the ultimate aim of an ecological model is to suitably simplify a system whilst simultaneously capturing its essence (Breckling et al., 2011).

Ecological models are playing an increasing role in many aspects of crop production (France and Thornley, 1984; van Ittersum et al., 2003; Cao et al., 2009), including for pest and disease control (de Wolf and Isard, 2007; Nietschke et al., 2007; UCIPM, 2012), and provide a tool to economically and expeditiously predict the behaviour of a production system where the experimental alternative may be expensive, onerous and impractical (Nachman, 2001). They can be used to accurately identify areas where further experimentation is needed and, in the case of pest and disease management, where it would be most rewarding in terms of providing control solutions (France and Thornley, 1984). They can also be adapted for use in decision support systems that assist growers in considering when, where and how to apply control strategies (Xia et al., 2007). In this thesis the models provide a means of comparing the impacts of novel climate management systems on pest and disease pressures and their control with more traditional climate management approaches.

There are a number of characteristics that define a model in a mathematical sense (Haefner, 2005). The first is the manner in which the system processes described in

16

the model are mathematically represented. In this regard, models largely fall into two categories; empirical (or phenomenological) and mechanistic models. The former is a relatively simple approach, whereby the system is represented purely through the use of statistical models with few of the underlying processes described and is based entirely on the data. These models may simplify systems but they are still capable of describing complex dynamic processes (Breckling et al., 2011). Mechanistic models describe systems in terms of their underlying mechanisms (though they still largely rely on experimentally-derived empirical data) and can provide clearer insights into the dynamics and processes driving the system that may not be possible with empirical models (Haefner, 2005).

The second characteristic is the temporal aspects of the model. A dynamic model is one in which future states or conditions are explicitly represented. If they are not then the model is described as static. The temporal aspect can be further classified based on whether time can take any value (continuous) or an integer only, e.g., an hour or day (discrete).

The third defining aspect of a model is its representation of space. A spatially homogenous model is one in which space is not represented and a spatially explicit model is one in which it is, be it discretely, in which areas have categorical descriptions (e.g., land, water, plant x, plant y), or continuously, in which each point has its own characteristics.

The final characteristic of a model is the way in which it deals with random events. They can be either stochastic, in which probabilistic events are included, or deterministic, in which they are not. The former imparts a degree of realism to the model, allowing natural variation and uncertainty in parameter values to be accounted for, while in the latter the output will always be the same for a given set of inputs. The inclusion of stochasticity is important if the variability of a system is of interest (Grant et al., 2000) as it “provides a more realistic representation of the population dynamics” and is “a natural way to incorporate process noise” (Skirvin et al., 2002).

The choice between these model types largely depends on the purposes of the model, the level of complexity needed (bearing the parsimony principle in mind), the data available and the current understanding of the system. Three simulation models

17

were constructed in this thesis; a mechanistic, spatially homogenous and deterministic disease epidemiology model for the O. neolycopersici - B. subtilis

system, a mechanistic, spatially explicit and stochastic population dynamics model for the T. urticae - P. persimilis system and an empirical, stochastic greenhouse environment model. All models were dynamic, with discrete time steps. The greenhouse environment model simulated the diurnal and seasonal variation in temperature and humidity in traditional and novel greenhouse systems in Spain and the Netherlands, providing stochastically varied climate conditions, which could be used as driving variables in the pest and disease models. This allowed the sustainable, novel greenhouses to be compared to traditional systems in terms of their impacts on pest and disease pressures and their implications for biological control. Literature sources were reviewed for data to parameterise the models and where this was not available or unsuitable, experimental work was conducted. Details of model choice and construction are given in the respective chapter for each model.