ABSTRACT
Summary Dynamic computer simulation models of forage growth can be useful in interpreting physical and biological limitations on production and in designing management systems for testing in the field. The explanatory detail of such models makes them both more reliable and more expensive to use than simple empirical formulations. Simple simulation models reduce the expense. They can also serve as reference points for designing, evaluating, or selecting more complex versions and thus provide a more rational basis for using models in grassland research and management. Our objectives were to develop and to test a simple dynamic simulation model of alfalfa (Medicago sativa L.) production using a combination of explanatory and empirical functions. Yield and forage quality in terms of digestibility and crude protein concentrations were predicted, and model testing was used to establish a reference point for evaluating more complex alfalfa models, several of which are available. The rationale of model development was to include the primary factors that cause variation in alfalfa yield. Genetic yield potential, moisture supply, temperature effects, and cutting management were selected as the most important variables. Cutting management was assumed to operate through its effect on root reserves in the control of regrowth potential and winter survival. Appropriate functional relationships were taken from other models or derived from data in the literature. Forage-quality prediction was based on an empirical relationship with time or growing degree days as the independent variable. The model was named ALSIM 1 (LEVEL 0). The LEVEL 0 model simulated expected patterns of dry-matter accumulation and soil moisture supply for the northeastern U.S.A. It also indicated that temperature stress is an important element of yield determination in alfalfa and that the phenological development of at least some cultivars must depend on day length as well as on temperature. Critical tests of LEVEL 0 involved simulations of 60 observations from three harvest schedules over 2 years at three locations with differing supplies of soil moisture. The linear regression of observed yields on predicted yields rendered the equation Y = 0.844 + 0.726X with r2 = 0.62 and S.E.y = 0.870 metric tons/ha. There is considerable room for statistical improvement, but LEVEL 0 will be adequate for preliminary study of harvesting options. Corresponding analyses for digestibility and crude protein concentration rendered r2 = 0.25 and 0.13, respectively, with S.E.y on the order of 3 to 5 percentage units. The empirical, time-based quality model is not adequate.
