This chapter introduces a decision support system which is capable of predicting the weekly yields of tomatoes in a greenhouse. The development of this system involves a set of Artificial Intelligence based techniques, namely Artificial Neural Networks (ANNs), Genetic Algorithms (GAs), and Grey System Theory (GST). The prediction was performed by an ANN using a set of optimised input variables, chosen from all available environmental and measured yield parameters. The reduction and optimisation of the inputs was done using either GAs or GST and compared in terms of the ANN’s performance. It was shown that the use of artificial intelligence based methods can offer a promising approach to yield prediction and compared favourably with traditional methods.
TopBackground
As mentioned above, tomato growers are sometimes contracted to sell agreed quantities of produce to supermarkets. However, tomato yields often vary from week to week, and so the ability to accurately predict future yields would give them a competitive advantage. If a grower is forecast to have insufficient fruits in a given week they could source additional produce from elsewhere, and if they are predicted to have excess fruits they could look for alternative markets or arrange promotions. As a result there has been considerable interest from growers and researchers for developing tomato yield prediction systems. (S.R. Adams, Cockshull, & Cave, 2001; S.R. Adams, Valdes, Cave, & Fenlon, 2001)
The systems developed take into account various growing conditions and factors affecting plant development. Environmental conditions that influence the growth and productivity of tomato plants include air temperature (day and night), fruit temperature, radiation, CO2 concentration, fruit load, plant density, stress etc. The fluctuation of temperature affects mostly the time of fruit ripening and rate of fruit growth. The research work of Willits and Peet (1998) also suggests that warmer conditions in the greenhouse at night can significantly improve the quality and quantity of tomatoes. The relationship between temperature and yield is complex and studies have shown that the sensitivity of fruits to temperature changes over time, as fruits become more sensitive to temperature as they approach maturity. This explains why raising the greenhouse temperature results in a peak in yield followed a few days later by a yield reduction (Adams & Valdes, 2002; Mulholland, Edmondson, Fussell, Basham & Ho, 2003). However, temperature fluctuations do not significantly influence the overall tomato yields when compared with temperature controlled growing conditions (Koning, 1988, 1990) however they influence the weekly tomato yields in greenhouses. Studies funded by the UK governmental agency DEFRA looked at understanding how the aerial environment affected the pattern of tomato yields. They showed that the primary cause of fluctuations in yield was due to the effect of temperature on fruit ripening and that the variation in ripening time smoothes out the effect on the yield that the fruit set may have (Adams, Cockshull, & Cave, 2001a; Adams, Valdes, Cave, & Fenlon, 2001b). Another important environmental factor is CO2, which is the key in photosynthesis, therefore CO2 enrichment of in greenhouses can significantly increase tomato yields (Nilsen, Hovland, Dons, & Sletten, 1983).