Modelling the nitrogen surplus and the efficiency of nitrogen use in Flemish dairy farms

In Flemish dairy farms it is very important to decrease the nitrogen surplus and to improve the efficiency of nitrogen use in Flemish dairy farms. This research developed models based on multiple regression in order to have an insight in the nitrogen surplus and the efficiency of nitrogen use in Flemish dairy farms. The dataset used comprised 1511 records from the Farm Accountancy Data Network (FADN) of Flanders. The monitoring period ran from 1989 to 2001. Different methods were used to find the best regression equation: ‘substantive knowledge’ methods (both ascendant and descendant), ‘stepwise’ methods and the ‘best-subsets’ methods. Many models were developed and tested. The main criterion to select a final model was the presence of relevant variables as known from literature. On top of this, models were preferred if the estimators of the regression coefficients were unbiased, if the degree of multicollinearity was low, and if there were no significant interaction terms. Unstandardized and standardized regression coefficients were calculated in order to clarify the relative importance of each independent variable. Finally, one model was selected for the nitrogen surplus and two models for the efficiency of nitrogen use. The two models of the efficiency of nitrogen use were developed with a dataset of records of farms with and without arable crops. The final models recommend similar actions in order to decrease the nitrogen surplus and to improve the efficiency of nitrogen use. These models show that decreasing the nitrogen in fertilizers and concentrates has the strongest potential to decrease the nitrogen surplus. The interpretation also reveals that both extensive and intensive dairy farms can produce in a clean way. There was a negative correlation between the livestock units of dairy cows per hectare and the nitrogen surplus in this dataset. Supported by data from the literature (Kebreab et al. 2001) it could be proven that this relationship was correct