NEURAL NETWORK MODELS FOR PREDICTING FLOWERING AND PHYSIOLOGICAL MATURITY OF SOYBEAN D. A. Elizondo, R. W. McClendon, G. Hoogenboom MEMBER MEMBER ASAE ASAE ABSTRACT. It is important for farmers to know when various plant development stages occur for making appropriate and timely crop management decisions. Although computer simulation models have been developed to simulate plant growth and development, these models have not always been very accurate in predicting plant development for a wide range of environmental conditions. The objective of this study was to develop a neural network model to predict flowering and physiological maturity for soybean (Glycine max L. Mem). An artificial neural network is a computer software system consisting of various simple and highly interconnected processing elements similar to the neuron structure found in the human brain. A neural network model was used because it has the capabilities to identify relationships between variables of rather large and complex data bases. For this study, field-observed flowering dates for the cultivar 'Bragg' from experimental studies conducted in Gainesville and Quincy, Florida, and Clayton, North Carolina, were used. Inputs considered for the neural network model were daily maximum and minimum air temperature, photoperiod, and days after planting or days after flowering. The data sets were split into training sets to develop the models and independent data sets to test the models. The average relative error of the test data sets for date of flowering prediction was-\- 0.143 days (n = 21, R 2 = 0.987) and for date of physiological maturity prediction was +2.19 days (n = 21, R 2 = 0.950). It can be concluded from this study that the use of neural network models to predict flowering and physiological maturity dates is promising and needs to be explored further. Keywords. Neural network, Crop development, Soybean, Weather. A ccurate predictions of plant growth and development are useful in crop management by allowing the grower to optimize the scheduling of field operations and to maximize net returns. The vegetative and reproductive development processes start as early as planting when the seed germinates, and these processes terminate at harvest maturity. The primary weather variable which controls plant development is temperature. In addition, photoperiod or the length of the daily light period can also affect reproductive development of certain species. Current simulation models have difficulty predicting development correctly for diverse locations for which either temperature or photoperiod vary. Most models seem to perform well when they predict development for the same location at which they were developed, but their accuracy tends to decrease when they Article was submitted for publication in May 1993; reviewed and approved for publication by the Information and Electrical Technologies Div. of ASAE in March 1994. Presented as ASAE Paper No. 92-3596. This project was partially funded by Corporacion Suiza para el Desarrollo (COSUDE) through a research assistantship for David Elizondo and by state and federal funds allocated to Georgia Agricultural Experiment Stations Hatch Projects GEO00526 and GEO01456. Trade names and company names are included for the benefit of the reader and do not imply any endorsement or preferential treatment of the product by the University of Georgia, COSUDE, or Centro Agronomico Tropical de Investigacion y Ensenanza (CATIE). The authors are David A. Elizondo, Graduate Student, Artificial Intelligence Programs, Ronald W. McClendon, Professor, Biological and Agricultural Engineering Dept., University of Georgia, Athens; and Gerrit Hoogenboom, Assistant Professor, Biological and Agricultural Engineering Dept., University of Georgia, Griffin. are applied at new locations, especially when either temperature or photoperiod are different. Several algorithmic models have been used for predicting the development of different soybean cultivars. These models were based on either field or controlled environmental data to define the functional relationships between temperature, photoperiod, and soybean cultivar development (Jones et al., 1991). Major et al. (1975) calculated the daily reproductive development rate as a function of both daylength and average daily temperature. Their model used a critical daylength, which defined the daylength at which the photoperiod sensitivity of a cultivar would change. Jones and Liang (1978) developed a model for predicting the start of flowering in soybean which was comprised of three successive phases: planting to primary leaf, primary leaf to flower initiation, and flower initiation to first flower appearance. Hadley et al. (1984) developed the concept that the rate of progress towards flowering becomes linear with photoperiod when the photoperiod is longer than the critical photoperiod. They also observed that all soybean varieties have the same base temperature of 7.8° C, but that the critical photoperiod varies among varieties. The soybean model SOYPHEN simulates daily development as a function of photoperiod, temperature, and drought stress for seven reproductive phases between planting and maturity (Hodges and French, 1985). Detailed crop growth simulation models require accurate development predictions in order to predict crop growth and yield. The soybean crop simulation model SOYGRO predicts growth, development, and yield as a function of weather and soil conditions and crop management (Hoogenboom et al., 1992; Jones et al., Transactions of the ASAE VOL. 37(3):981-988 © 1994 American Society of Agricultural Engineers 0001-2351 / 94/ 3703-0981 981