Source DocumentPrevious PageTable Of ContentsNext Page

Devel : a system for crop development modelling

D.P. Holzworth and G.L. Hammer

QDPI/CSIRO Agricultural Production Systems Research Unit, PO Box 102, Toowoomba QLD 4350

Accurate predictions of timing of phenological stages and their rates of development are essential in crop simulation models. Poor prediction of ontogenetic development can lead to large errors in simulation studies. Thus this paper presents a system (DEVEL) for deriving relationships to predict rate of crop development from observations of the timing of ontogenetic stages and associated daily meteorological data.

DEVEL accounts for the two major environmental factors known to significantly affect crop development, temperature and photoperiod. Both are used to predict crop development from the following equation :

Rate of development = Ropt x f(T) x f(PP)

where R.p, is the daily rate of development under optimal temperature and photoperiod, f(T) is a function of daily temperature and f(PP) is a function of photoperiod. The rate of development is the proportional increment in the chosen phenological period (1/day). The phenological stage is predicted to occur when the accumulated daily rate of development reaches 1.0. For example, if a phenological stage has a duration of 50 days, the average daily rate of development is 0.02.

The system provides nine temperature and seven photoperiod functions allowing users to choose the most appropriate response for their applications. Once initial estimates for the parameters of the functions have been chosen, and weather and observation files provided, DEVEL will then find a "best fit" for the combined temperature/photoperiod model. As many of the available functions are not linear, DEVEL uses an optimising algorithm hat is capable of minimising or maximising a given objective function. It works by iteratively changing the initial estimates of the parameters in an attempt to minimise the sums of squares of residuals between the actual observation data and the corresponding predicted data. The algorithm used is the simplex method of optimisation, which is a commercially available routine.

The system has been designed to allow easy analyses across sites and genotypes with a minimum of data manipulation. Complex genotype by environment analyses can be performed by minor alteration of the parameter file. Once the use of DEVEL has been mastered, complex yet intuitive phenological responses can be as easy to develop as simple linear ones. Crop phenology for a given genotype can then be predicted by using the derived model parameters.

Previous PageTop Of PageNext Page