Predicting nitrous oxide emissions with a random-effects model

The main source of anthropogenic N2O emissions is the application of nitrogen fertilizer to agricultural soils. We present an approach for predicting N2O emissions based on a statistical random-effects model: the N2O emission response to applied nitrogen fertilizer is described by an exponential function, the parameters of which are assumed to vary randomly between locations. One of the advantages of this model is that its parameters are easily adjusted to one or several location-specific N2O measurements. The adjusted model can then be used to predict N2O emissions for nitrogen fertilizer doses other than those applied at the considered location. We evaluated the accuracy of model prediction, with real and simulated data. The use of location-specific rather than average predictions reduced prediction errors in most cases. Location-specific predictions could be used to estimate background emission in on-farm studies.