MODERATO: an object-oriented decision tool for designing maize irrigation schedules

Abstract

The rapidly changing economic, technical and regulatory context of irrigated agriculture, coupled with seasonal variation in precipitation, presents a problem for irrigation management, especially in sub-humid regions. During years of drought, frequent irrigation bans may be applied and shortage of water for crops becomes a critical problem. Simulation models offer the opportunity to optimise production strategies, such as optimal irrigation scheduling. But few biophysical models are designed for decision making. MODERATO is a management oriented cropping system model developed for use at a strategic level by irrigation advisors confronted with the question: ‘How to irrigate maize with a limited amount of irrigation water?’. It includes the main constraints specifically related to irrigation (work time, available amount of water, flow rate, blackout days), simulates the plant-soil system with a dynamic biophysical model (parametrized on a large database) and takes into account within-field variability that results from sequentially irrigating the plots in a block of irrigation. Five elementary irrigation rules are distinguished: (1) a rule to irrigate to facilitate plant emergence; (2) a rule to decide when to start the main irrigation period; (3) a rule to determine when to start a new irrigation cycle; (4) a rule to decide when to stop irrigation; and (5) a rule to delay irrigation due to weather conditions. The elementary rules consist of two boolean conditions which depends respectively on development stage and soil water availability. The details of the rules are input using a graphic user interface. The dynamic biophysical model is based on the well-known interception–conversion process. The model outputs allow one to analyse the consequences of the decision rules for various climatic series and context. MODERATO is the result of 3 years of collaborative research between scientists and irrigation advisors and has been used to calculate optimized starting and ending rules for irrigation on a specific pedoclimate.

Introduction

Water availability plays a key role in yield and quality of most agricultural crops. However water is quite often a limiting resource. Seasonal variation in precipitation presents a problem for irrigation management, specifically in sub-humid regions (Hook, 1994, Cabelguenne et al., 1995). In south-western France, climatic water deficit (potential evapotranspiration minus rainfall) is ∼300 mm for the 5 months from May to September (Cabelguenne and Deumier, 1996) with a large year-to-year variability as the standard deviation for a 49 year series of weather records in Toulouse is ∼120 mm. During years of drought, frequent irrigation bans may be applied, and shortage of water for crops becomes a critical problem. Furthermore, irrigation is a costly operation and quite often accounts for a third of the annual gross margin, due to the rising cost of irrigation pumping, low commodity prices, inadequate irrigation system capacity, limited water supply, etc. Profitability of irrigated crops must be improved by reducing the amount of water used and optimising timing of application (Stockle and James, 1989). On the other hand, if irrigation is badly managed and if too much is applied without regard to soil water depletion or rainfall, irrigation becomes a source of pollution (Leenhardt et al., 1998), leading to nitrate leaching, surface water runoff loaded with pesticide etc. It is therefore necessary to develop a precise knowledge base and accurate irrigation scheduling to assure a profit and avoid environmental damage.

However, the cost of generating the data set and the time needed to experience a broad weather spectrum and develop a comprehensive irrigation scheduling system is prohibitive. In addition, technology changes rapidly (cultivar, chemicals, equipment, etc.) causing any empirical schedule to rapidly become obsolete. Simulation models offer the opportunity to generalise and extrapolate the data set required for analysis of optimum production strategies, such as optimal irrigation scheduling (Whisler et al., 1986, Bryant et al., 1992). Computer simulation models are able to predict the effect of weather, soil properties, plant characteristics and management practices on the soil water balance, nutrient dynamics and growth of crops. Therefore they can further our understanding of cropping system performance under different water and nitrogen regimes (Pala et al., 1996).

Many agricultural models have been developed to analyse irrigation requirements of maize crops: either empirical ones such as PLANGRO (Wenda and Hanks, 1981) or more mechanistic one like CERES-Maize (Howell et al., 1989, Epperson et al., 1993, Hook, 1994), EPIC (Bryant et al., 1992), EPIC-PHASE (Cabelguenne et al., 1997), CROPSYST (Stockle, 1997) or STICS (Tayot et al., 1998, Brisson et al., 1998). Some of these models have been used to test irrigation schedules using a limited amount of water (Epperson et al., 1993, Cabelguenne et al., 1997). Although few biophysical models are designed for decision making, several have been used for this purpose. Models were first created by gathering the existing knowledge and then used to test irrigation schedules with different amounts of irrigation and estimating either yield or net return value from irrigation (Bogess and Ritchie, 1988, Stockle and James, 1989, Martin et al., 1996). However, a given irrigation schedule cannot be generalised, as the irrigation level that maximises economic return varies with several factors including soil, uniformity of application, ratio of commodity prices to production costs, total pumping head, seasonal distribution of the applied water, etc. (Stockle and James, 1989).

Irrigation is a more complex technical operation than fertilisation, for instance. This is not just a two-fold decision process of when and how much. Different constraints have to be considered. It is thus worth developing alternative irrigation strategies based on decision rules rather than on prescriptive schedules (Aubry et al., 1998, Shaffer and Brodahl, 1998). Rule-based management systems can offer the farmer or the adviser the opportunity to better approximate current or proposed management options, especially if they relate to dynamic conditions in farm fields and across the whole farm (Shaffer and Brodahl, 1998). A few models offer decision rule management for maize irrigation, but fail to consider constraints due to equipment and policy and spatial variability due to irrigation delays. There are few tools which use both dynamic modelling and decision rules for simple applications and testing (Deumier et al., 1995, EEC, 1996).

In 1995, 734 000 ha of crop land were irrigated in southwestern France, of which maize represents 60% (AGPM, 1998). Developing tools to test irrigation rules on maize for a limited amount of irrigation water can therefore play a terse role in increasing water efficiency. Integrating the context (prices, amount of water available, equipment, regulations, etc.) into the rules is necessary in order to enable the tool to take account of variations of this context.

The major improvements in irrigation will come from better strategic decisions, i.e. decisions made before the irrigation campaign has begun (Andersen et al., 1996): what amount and flow for each irrigated area? For a given limited water volume or flow rate, how best to share the available water?. The main enquiries from the advisors regarding the decision rules for irrigation are (Mircovich, 1999): (1) when to start the main irrigation period?; (2) when to stop the irrigation period?; and (3) how to take rainfall into account? Experiments carried out by the research or technical institutes and irrigators practice give some ideas on how to optimize irrigation planning. However, practical knowledge cannot adapt sufficiently quickly to the rapidly changing economic, technical and regulatory context. The use of biophysical models linked with decisional models allows one to perform virtual experiments that take into account the variation of context and thus adapt irrigation planning to this context.

This paper describes MODERATO, a management oriented cropping system model. It has been developed for use by irrigation advisors confronted with the question: ‘How to irrigate maize with a limited amount of irrigation water?’. MODERATO is meant to be used before the irrigation campaign, to develop decision rules taking into account the irrigation context and different aspects of crop performance (yield, water drainage …). These rules can then be tested on a climatic series not only to obtain average results but also to develop risk analyses on the decision rules. MODERATO can be used to simulate the consequences of different rules, but can also help to determine the optimum thresholds for the different rules.

Irrigation decision models based on an isolated plot (i.e. without any inter-plot relationships regarding management) fail to consider some important aspects of irrigation management: (1) they do not include equipment and resource constraints; and (2) they do not take into account the within-field variability due to the time required to irrigate the whole field. On the other hand, irrigation decision models based on a description of the whole farm (LORA, Leroy and Jacquin, 1994; IRMA, Leroy et al., 1997, Labbé et al., 2000) are quite often difficult to work with as they require a detailed description of the farm and more specifically of the irrigation system, and they rarely include a dynamic model to describe the soil and plant behaviour but instead make use of empirical functions. Furthermore, they are not developed to answer our specific question and time and spatial scales are not satisfactory with the time and spatial scales we required to answer the question. We decided to develop a hybrid decision model able to include in the irrigation decision the main constraints for irrigation but requiring fewer detailed data than the whole farm models, simulating the plant-soil system with a dynamic biophysical model and taking into account some within-field variability due to the time required to irrigate a whole block of irrigation. The decision rule will be a simplification of the farmer's complex decision process.

MODERATO is the result of a collaborative work between research institutes (Institut National de la Recherche Agronomique) and technical institutes (Association Générale des Producteurs de Maı̈s, Institut Technique des Céréales et des Fourrages). The questions to be answered with MODERATO were raised by the technical institutes and translated by the research institute. Frequent meetings were held to specify the different keypoints of the work. Data to develop the biophysical part of the tool and used for its parameter estimation were provided by the technical institutes.