Dairy farm nutrient management model. 1. Model description and validation

Intensive dairy farming results in significant phosphorus (P) emission to the environment. Field data indicates that farm‐gate P surplus is highly positive in Finland and strategies to mitigate the surplus are needed. The objectives of this study were to build a P cycle model for dairy farms (1) and to validate the model with independent field data (2). The dairy farm nutrient management model (‘Lypsikki’) described in this paper includes three sub‐models: (1) soil and crop, (2) dairy herd and (3) manure management. The model is based on empirical regression equations allowing estimations of crop and milk yields in response to increased fertilisation and nutrient supply, respectively. In addition, the model includes a dynamic simulation model of the dairy herd structure and calculation of the farm‐gate nutrient surplus. The model was validated with independent annual (average for 1–4years) farm‐gate P surplus data from 21 dairy farms. Model simulations were conducted using two levels of soil productivity, mean (M) and low (L). The model validation indicated a strong relationships between model‐predicted and observed farm‐gate P surplus: (M: R ² =0.77 and L: R ² =0.80). The line bias between the model‐predicted and observed data was negligible and insignificant (P >0.6) suggesting a robustness of the model. The mean biases were relatively high and significant (M: 4.7 and L: 1.8kg/ha, P <0.001), but evidently related to overestimation of crop yields that has to be taken into account when using the model on a single farm. The prediction error of the model (observed minus predicted P surplus) was significantly correlated to the difference between simulated and observed P import in feeds (M: R ² =0.55 and L: R ² =0.51). This suggests either that all the dairy farms did not fully exploit the possibilities in the crop production or that all the model assumptions are not correct. The effects of purchased feed and fertiliser P and exported milk P (per cow or cropping area) on farm‐gate P surplus were of the same magnitude in both observed and simulated data. This implies that the model developed can be used as a management decision tool to find strategies to mitigate P surplus on dairy farms.