Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 11/12/2004
Publication Date: 11/12/2004
Citation: Timlin, D.J., Pachepsky, L., Pachepsky, Y.A., Whisler, F., Reddy, V., Fleisher, D.H. 2004. An evaluation of glycim after nine years of on-farm trials: why does glycim often predict higher than measured yields? [abstract]. Meeting Abstract. Interpretive Summary:
Technical Abstract: The soybean model, GLYCIM, was developed after the cotton model GOSSYM and shares some design components and modules. GLYCIM has highly mechanistic, dynamic representations of plant growth, development and yield, and soil and weather processes. The mechanisms involved in the physical and physiological processes in soybean and its environment are mathematically described in GLYCIM. These processes include light interception, carbon and nitrogen fixation, organ initiation, growth and abscission, and flows of water, nutrients, heat and oxygen in the soil. From 1991 to 2000, GLYCIM has been used by farmers for crop management and input optimization in the Mississippi Valley region of the U.S. and over 150 datasets containing phenology, yield, soils and management data have been collected. The model was being used prior to planting for cultivar selection, row spacing, plant population and planting date, and for post planting decisions such as irrigation scheduling, insect control, harvest timing, and forecasting of final yield (Reddy et al., 1995). Based on the on-farm testing it has been shown that GLYCIM may often overpredict yields rather than underpredict. Phenology, however is simulated relatively well. The cause for the overprediction of yield appears to be due to partitioning of carbon between the pods and the seeds. In this paper we show how this can be partially addressed by reparameterizing the model for only two parameters related to increase in seed weight and seed fill and progress in reproductive (R) stages and constraining the ratio of pod weight to seed weight to a range reported in the literature (about 0.3). We also show how the large database collected during this project can be used to identify strengths and weaknesses in GLYCIM and perform similar analyses for other parameters.