Evaluation and application of computer modeling to dairy grazing systems: Pasture intake, pasture selection and whole farm economic evaluatio

GRAZESIM is a decision support system that assists in making decisions regarding dairy grazing systems. Four areas of the model were developed and evaluated: (1) pasture intake prediction model, (2) pasture selection, (3) whole farm economic evaluation, and (4) application of dynamic programming to account for risk and uncertainty. Four algorithms for simulating pasture intake were developed and validated: (1) present algorithm using GRAZESIM, (2) an algorithm accounting for type and amount of supplementation, (3) an algorithm accounting for pasture selection effect, and (4) a combination of algorithms 2 and 3. Pasture intake data from the literature were used to validate the models. The results showed that algorithm 3 had both the lowest variability and bias. Three models were compared and validated to determine the influence of pasture selection using NDF, ADF and CP content in forage consumed by grazing: (1) an empirical model, (2) a model based on a truncated normal distribution, and (3) a model based on an exponential distribution. The models were validated using data from a grazing experiment. The models had the best results when predicting CP. The empirical model resulted in the best predictive ability. The GRAZESIM economic sub-model was validated using actual economic data from Wisconsin dairy farms. Financial data from 972 dairy farms provided information to estimate receipts and costs to be used in GRAZESIM. Actual financial data for 16 farms (8 grazing; 8 non-grazing) were used to validate GRAZESIM. The model simulated receipts and feed, variable, fixed and labor costs of dairy, replacement and crop enterprises for each farm. Simulated dairy feed and variable costs showed moderate correlation and low bias. Simulated crop enterprise costs resulted in significant bias. Three dynamic programming models used to determine supplementation strategies in dairy grazing systems depending on pasture availability were compared: (1) deterministic, (2) stochastic, considering independent probabilities, and (3) stochastic, considering Markov processes. Stochastic models solutions resulted in greater use of concentrates. The solutions for sensitivity to milk price and milk to feed price ratio variation resulted in using less pasture under high milk price conditions or low feed prices.