Expert systems are much more 'friendly' computer programs than standard databases or spreadsheets. Thea are usually created with ease of use in mind. From numerous examples, I have chosen 11 specifically developed for different branches of horticultural production. They represent a cross-section of the programs available for extension services and growers.
General management
CALEX - is a decision support system for agricultural management.
The program consists of 3 subprograms: an executive, a scheduler, and an expert system shell. The executive takes care of interface with the user, models and disks; the scheduler creates the sequence of production activities through consultation with the expert system; the expert system undertakes the management decisions. Any number of crop specific modules can be connected to this general management program. CALEX was created at the University of California and preliminary tested using two crop specific modules: agricultural for cotton and a horticultural one for peaches.Protected cultivation
The Department of Agricultural Engineering of the Ohio State University, USA, developed a computer decision system for the fertigation of glasshouse crops.
The system comprises of three separate rule bases for: feasible nutrient recipes; parameters to derive probable solar irradiance; and crop responses. The composition of fertigation recipes is derived jointly from probabilities of irradiance and crop response function. The system modifies the parameters of the radiation rule base from one day to the next to optimise its irradiance predictions. The system was tested in cucumber culture using an eight head, single element nutrient injector.
A greenhouse environmental control system based on a crop model and an expert system has been developed by Department of Horticulture of the Shizuoka University and National Institute of Vegetables, Ornamental Plants, and Tea in Japan.
The system controls greenhouse air temperature, humidity and carbon dioxide concentration, and water content in the growing substrate. The input parameters are solar radiation, air temperature and carbon dioxide concentration, and the grower's observations of the crop (colour, shape, size). The output from the system to controlled manipulators is based on the outcomes of models of photosynthesis rate, evaporation rate and water potential of soil and leaf. The system was implemented using 3 computers controlling tomato culture.
MISTING - an expert system for optimal misting strategy.This software was developed at the University of Florida to check knowledge acquisition strategies from an experienced grower. The typical time-clock misting control strategy was modified according to grower's rules of the thumb elicited. The resultant misting algorithm was implemented alongside with conventional time-clock controllers during Peperomia propagation and run 30 days with no software failures successfully following the grower's strategy.
Fruits
Researchers at the Pennsylvania State University developed an expert system which helps growers to select appropriate peach and nectarine cultivars and a second one for the determination of tree spacing in an apple orchard.
The selection of peach and nectarine cultivars is based on such characteristics as harvest date, flower hardiness, pit freeness, fruit size and susceptibility to diseases. In optimising apple tree spacing, the computer takes into consideration soil type, slope, trailing system, cultivar, equipment width and rootstock.
Although not a typical expert system, the software for the diagnosis of macadamia nut problems developed by University of Hawaii certainly needs attention.
It is based on hypermedia techniques, that is, the possibility of communication by text, pictures and sound. Besides a typical text dialogue the user can call up on screen pictures of symptoms. The problems covered by the program are typical for Hawaiian conditions and include insects, diseases, nutritional deficiencies, harvest and post-harvest handling, herbicide injuries, poor flowering and premature nut drop.
Vegetables
Researchers from Purdue University, Indiana, USA created a dynamic production management expert system for muskmelon.
The expert system enables producers to allocate total production resources in an aim to minimise economic risk and maximise net return under competitive market conditions. User interacts with the program in natural language. Horticultural expert recommendations, and the physical and economic parameters for decision-making optimisation form an artificial intelligence environment which enable reasoning and interactive dialogue with the user.
Ornamentals
PALMS (Prototype Application of Language Meaning Structures) - expert database system for palms.The database contains the information on palms for Florida conditions which includes text, computer graphics and images. The backbone of the database system is the data model, knowledge base, language processor and inference engine. User can communicate with the program in natural language or through graphical interface using a pointing device such as a mouse or touch panel. When the language processor has problems understanding a user's query this can be reformulated or the user can switch to graphical browsing. The database system for palms was treated as a model for developing similar knowledge systems for other crops. PALMS was developed at University of Florida, Gainesville, US.
Turfs
TURFPLAN is an expert planning system for turf managers.The program helps in the selection of turfgrass species and cultivars as well as establishment and maintenance procedures during the post-planning period. A portion of the rules for selection of turf grass cultivars was developed using an inductive learning system. During evaluation, the expert system was proven to be superior to human expert in planning home lawns (medium maintenance lawns) but inferior in the case of highway (low maintenance) and golf (high maintenance) turfs. TURFPLAN was designed at University of Illinois for turfs in the north central United States.
Herbicides
HERBASYS is a herbicide advisory system.The system helps predict the degradation of several herbicides in soil and evaluates the potential influences of residues on subsequent crops. After the user enters data concerning site, weather, date and dosage of applied herbicide, the system, through simulation based on parameters from its knowledge base, predicts the breakdown of active ingredients and its effect on any crop in question.
HERBASYS was developed in Germany by a joint effort of the University of Bremen and the Institute of Weed Research of Braunschweig.
Future
These examples represent the main-stream developments in this domain. However, the future belongs to adaptive, dynamic knowledge systems which can control agri-robots and automatic plant factories. In such environments, software must be very fast because it acts in real-time situations and it must have build-in learning capabilities. Many such systems have been under development in Japan. They future learning, pattern recognition, association by analogy, automatic building of knowledge bases from texts, and so on. Such artificial experts will work in the most intensive branches of horticultural production from the turn of the century.