Uncertainties in large-scale analysis of agricultural land use—A case study for simulation of nitrate leaching

Simulation studies play an important role in predicting the effects of agri-environmental policies. Management plans for the Water Framework Directive, for example, show how policy measures can be based on both the absolute effects and the probability of the achievement of objectives. Here, comprehensive uncertainty analysis is a valuable modern tool.In the context of the six most important uncertainties in this study, the article explains the overall uncertainties of a nitrate leaching model for agricultural land use. Although the primary input data are ‘natural conditionś and ‘land management́, there are also variances caused by model description and upscaling.The management options in a large number of small farming systems are described as adapted in a particular region. A variance analysis of the simulated nitrate leaching-rates by means of the H-test shows significant differences between locations, meaning that the distribution of results often allows no discrimination between two or several locations in terms of simulation objects. These locations, similar in nitrate leaching behaviour, are summarised to so-called Nitrogen Response Units (NRUs). Within these NRUs, a multiple regression analysis, with the dependent variable ‘nitrate leaching-rate,’ and the independent variables ‘percentage of cereals in the crop rotation’ and ‘livestock,’ shows the regression coefficients of each NRU.A calculation of the mean, or theoretically most probable, nitrate leaching-rate follows: all distribution patterns of livestock and cereals are combined, and the regression equations are applied for each NRU. This reveals the range of the possible nitrate leaching-rates in the area. The mean value, like the average of all possible land use patterns, of this distribution is taken as the most probable figure. The minimum and maximum values set the limits for the management distribution of the uncertainty analysis. The overall uncertainties of the results are determined by the sum of the variances that are calculated by all effects, which are classified as important. The first application of this new approach was a study region of about 1000 km2 in Germany. The nitrate leaching losses were 63 kg NO3/ha on average with a range of uncertainty of +59% and −79%.