WCHR Home WCHR 1 WCHR 2 WCHR 3 WCHR 4 WCHR 5 Search
Evelyne COSTES
INRA - ENSAM Arboriculture Fruitiere
2, Place Viala 34060 - Montpelliiere
2, Place Viala 34060 - Montpellier Cedex 01
In the present economic context, fruit crops are submitted to competitiveness and to consumer's demand which dictate an improvement of products quality and a reduction of inputs. This context defines new objectives for researchers and necessitate to develop new approaches among which modelling is a major research tool to improve orchard management. An exhaustive review of models developed on horticultural context has recently be made by Gary et al. (1998) The present paper focuses on a brief review of the main models developed in the last years in fruit crop context. Models are classified as suggested by Gary et al. (1998) according different objectives : decision support, education or knowledge conceptualization. A particular emphasis will be put on plant scales taken into account in the models.
Since most of fruit species are perennials, their structure complex and changing with time, until now only few species and a limited number of process of growth and development have been modelled. A first class of models focusing on a specific process involved in tree edification can be considered. Most of them deal with a particular physiologic process and a particular organs in the plant. This is the case of models for bud dormancy (Erez et al, 1990), and for leaf photosynthesis (Buwalda, 1991; Grossman and Dejong, 1994). Obviously, many models deal with f Dejong, 1994). Obviously, many models deal with fruit development including different processes such as water import (Bussières, 1994) or sugar content (Génard and Souty, 1996). Only some models deal with the heterogeneity within the plant using stochastic methods (Agostini and Habib, 1996, for flowering and pollination on kiwi fruit; Hall and Gandar, 1996, for fruit growth).
A second class of models accounts for several organs simultaneously. A first example is related to tree structure edification and model the distribution of axillary branches along one-year-old shoots using stochastical processes (Costes et Guédon, 1997). This approach aims at evaluating, in the nursery, the ability of young apple trees to be conducted in the most current training systems used in intensive orchards, such as vertical axis or solaxe. A similar approach has been applied to the distribution of axillary flowers in prunus species, taking into account the vegetative growth associated to fruit development (Fournier et al., 1997; Costes and Guédon, 1996). Models dealing with interactions between fruits and close vegetative shoots have been recently proposed by Bruchou and Génard (1996). In this case, the bearer shoot is represented by different compartments, one being the source (leafy shoot) the other one being sinks (fruits).
Different attempts have been made to account for whole plant. Most of the modelling apprnt for whole plant. Most of the modelling approaches necessitate strong simplifications on the tree structure representation in order to focus on particular functional process. Compartment representation are used for models of carbon and nitrogen partitioning either in aerial part of the trees or in the whole plant, including the rootsystem (Dejong and Grossman, 1992; Habib et al, 1990; Fisher, 1995). When focus is on light interception or interaction with climatic environment, tree structure is assimilated to set of randomly distributed leaves or to geometrical forms such as cones or to successive layers (Gijzen and Goudriaan, 1989). Buwalda (1996) showed that such simplifications can result to significant errors in the estimation of radiation interception and contributed to take into account the spatial distribution of leaves within the canopy (Smith et al., 1992).
Different technics were simultaneously experimented to improve plant structure measurement, in particular spatial distribution of plant entities (Smith, 1995, Sinoquet et al., 1991, 1997). Recently, a model has been proposed to represent plant at different scales of observation (Godin and Caraglio, 1997), taking into account botanical entities resulting from plant architecture and morphology as they were described in 1970's by Halle et coworkers to define architectural models (Hallé et al., 1970). This model, including a strategy for encoding plant topology and ng a strategy for encoding plant topology and a querying language to access to the database (Godin et al., 1997) provides a standardized and generic representation of plant structure.
Most of the models presented above are dedicated to knowledge conceptualization without providing practical outputs. Let us now consider models built with a more applied objective, in particular yield simulation. Several models used the counting of plant constituents to calculate yield estimation. On kiwi fruit, Testolin and Costa (1992) simulated yield and fruit size distribution using the proportion of fruit-bearing buds, the number of fruits per cane and fruit weight distribution as input. More recently, Habib et al. (1996) proposed an estimation of kiwi fruit yield components relying on stochastic estimation of flowering and pollination. In different approaches yield is predicted using an evaluation of the climatic environment. This has been done by Wagenmakers (1996) who used light interception and temperature to simulate potential of apple production under different weather climate.
Simulations lead, in some cases, to decision support, decisions being taken at industries level (Atkins et al., 1992) or at orchardist's level. In this latter case, the support can concern many types of decisions, all included in the term orchard management (Hodson et al., 1992; Alvisi et al, 1992; Lescourret and Habib, 1996). The support can also concern a spebib, 1996). The support can also concern a specific decision such as control of infection periods (Travis et al, 1992; Blaise and Gessler, 1992) or chemical thinning (Crassweller et al., 1992).
Only a few examples exist in which models and software have been combined for teaching objective : PEACH developed by Dejong and coworkers (Grossman and Dejong, 1994) and softwares dedicated to pruning simulation on apple trees using rule-based models (Atkins et al., 1996).
In conclusion and according Gary et al. (1998), diversity of species,
heterogeneity, quality and decision making are some of the present
challenges of Horticulture which necessitate cooperation between
scientific disciplines and progressive integration of knowledge.
Crop modelling is one of the major tool to help in this task and
need to develop towards more generic models, the adoption of standards
and the use of procedures of software quality assurance. Presently,
several integrative approaches are being developed which rely
on different assumptions, simplifications and methods to simulate
plant edification and physiology (Le Dizes, 1995; De Reffye et
al., 1997, Prusinkiewicz et al., 1997).