Вход на сайт


За неделю:703
За месяц:2735

Integrated Environment for Information Support, Design and Implementation of Multifactor Computer Experiments

To facilitate and support the AGROTOOL model application for solving a variety of research and practical tasks an integrated information environment is developed at the ARI Laboratory of Mathematical Modelling of Agroecoystems. The structure of the Integrated Environment and interrelation of its components is presented in the following scheme (scheme).

From a formal point of view of the dynamic crop model is complex, but univocal algorithm describing the course of the production process of the selected cultures in terms of the selected indicators that meet a certain set of specified growth conditions (weather, soil, terrain, agricultural machinery, etc.) Accordingly, the simplest external model shape is an interface for a single model run. It display the dynamics of growth and development for the current crop depending on the conditions of a particular calculation scenario defined in the operative database.

In this case, formulating a new scenario for the model calculation does not apply to the model itself, and is entirely dependent on other modules of the integrated environment. The operative database for model calculation can be formed by selection a particular data set from the so-called Stationary Database of Field Trials (SDB). The calculation will be made for a combination of factors that have been observed in reality, that is, according to one or other really occurring experiment or observation over the production crop.

To choose one field experience from all their diversity, stored in SDB, a special software module «Selector» is used. The module is a standalone application (see Fig. 1), in which the user interactively forms requests to SBD to extract specific data and then to write these records to the Operative Database (ODB).

Interaction with the user begins by selecting a particular farm and a field experiment from the available in the SDB. Then, on the next step the data about the seasons (year), cultures and fields for which variants become available is loaded (see Fig. 1).

In parallel, the module forms the description of a chosen variant of field experiment. Only after unambiguous identification of variants, it becomes possible to export data in the prescribed format into the ODB.

Fig. 1. Dialog box of the "Selector" type interface. Selection of the variant of field trials

It should be also noticed that this procedure is carried out not merely for "transfer" of selected information from one data repository to another, but some calculations are carried out simultaneously. In particular, estimates of missing parameters are calculated which values are strictly required for calculating the model, but does not explicitly stored in a database of field experiment.

For example, the model uses the coefficient of attenuation of solar radiation, i.e. the ratio of actually received radiation to its maximum possible value for a given day at the selected latitude, as an input data among daily meteorological required. This value is calculated either directly (in the presence of solar radiation observations) or estimated from indirect data (cloudiness, sunshine duration, and so on). Also, in this mode an interpolation of spatially distributed and measured soil characteristics is carried out to provide their values in those default layers that are used in the model.

Finally, a separate large task is to evaluate the soil hydrophysical parameters (functions) using various indirect indicators that are available, that is, the use of so-called pedotransfer functions. The last term was defined by J. Bouma as “translating data we have into what we need”. Some basic pedotransfer algorithms are already implemented within the «Selector» module. Meanwhile, a standalone application AgroHydrology designed just for that purpose provides much more opportunities for the modeler.

Stationary database SDB is available for delivery in two possible formats, either MS ACCESS or PostgreSQL. A more advanced user interface for desktop application as well as a new Web interface for standard browsers are under development now.

Stochastic simulator of daily weather meteorological variables (so called weather generator) is widely used in computer experiments with agroecosystem model, especially in short-term (operative) and long-term (strategic) forecasts of the dynamics of production process. From mathematical point of view the weather generator is a stochastic modelling algorithm (method Monte-Carlo) which allows for a set of constant parameters describing the weather patterns in the area, get in any number of "synthetic" weather scenario. The Weather generator is developed in the ARI Laboratory of Mathematical Modelling of Agroecoystems by modifying a basic modelling algorithm proposed by Richardson and Wright (1984).

All the parameters of the Weather generator have to be evaluated for each geographic location using the real course of weather data for a number of retrospective years collected in the database. A designated module of parametric identification performs the evaluation procedure (Fig. 2). The module also allows visual testing of quality of calculated estimates check the adequacy of the evaluation by generating multiple test implementations of weather course during the year.

Fig. 2. Parameter identification and visual verification of generated weather scenarios

The resulting set of statistical climate parameters is stored in the Stationary Database. The algorithm that generates virtual weather scenarios is itself implemented in the form of an independent software module, which writes "synthetic" weather information directly to an Operative Database of the current version of the model calculation. The module and is built as an integral part into the System of Polyvariant Analysis, which will be discussed below.

A typical use case of model application in any decision support systems often involves not a single but multiple calculation of the same model with different sets of input parameters in order to analyze and compare the results. Table 1 presents a list of problems and/or practical tasks, resulting in the need for multiple model calculation indicating the source of the variance of the input data for each of them, is presented in

Table 1. Problems and tasks that require poly-variant analysis of crop model.


Problem, task

Source of the variance


Parametric identification of the model

Changing parameters values


Operative forecast of yield during the vegetation period

Weather scenarios


Search for optimized application of agro technological measures

Variety of farming technologies (dates of application and amounts of agro technological measures)


Impact of climate change on agroecosystem

"Synthetic" weather information for future climate scenarios


Precision agriculture

Spatial heterogeneity of soil and plant cover.

Under these conditions, a natural requirement is the existence of some specialized shell or infrastructure for poly-variant analysis that would allow automated data preparation, calculation of multiple model runs and an analysis of the results obtained in batch mode. Moreover, taking into account the universal nature of these modelling operations, and that the different tasks could require application of different models, it is desirable that this shell allows you to perform them not only with some, but also with an arbitrary dynamical model.

Such a specialized shell allowing the user to design and to conduct multifactorial computer experiments with AGROTOOL model is called APEX (Automation of Polyvariant EXperiments). It encapsulates two basic functional components – logger descriptions of external models of the production process and versatile infrastructure of their multiplicity of analysis.

Finally, the model AGROTOOL and its information environment can be also integrated with an automatic weather station, forming a measuring and modeling system for monitoring and control of production process of agricultural crops.