MODELING AND MEASUREMENT OF THE NITROGEN-CYCLE IN A VEGETABLE FIELD IN SWITZERLAND .2. VALIDATION OF THE NITROGEN MODEL WITH FIELD DATA

Field experiments were conducted over a period of two years on a commercial vegetable farm for calibration and validation of the nitrogen model (tables 1 and 2). Yield and nitrogen uptake of the crops were measured at least five times during crop growth. Soil water tension and nitrate concentration in the soil solution were assessed by frequent sampling of different soil layers. The volumetric soil water content was calculated from continuously measured water potential by means of a field desorption curve. The model was calibrated using a transfer factor for rapid and slow water transport corresponding to changes in water content observed during a leaching experiment in winter 1987/88. Other required input parameters were taken from existing knowledge about the growth of vegetable crops, an evaluation of the soil profile to assess water transport in the soil, and the results of an incubation experiment for the estimation of the net nitrogen‐mineralisation. Yield and nitrate uptake of different vegetables are predicted with an accuracy of about 15% for yield and of about 20 kg N/ha for nitrate uptake (table 4, figure 1). Measurement of water and nitrate transport under fennel and a fallow plot (figure 2) reveals that the simulated net nitrogen‐mineralisation is overestimated by about 30 percent, and that simulation of rooting depth and evapotranspiration should be improved. Nitrate leaching during summer takes place only after periods of very heavy rain. The accuracy of the agreement between simulated and measured data is within 10% for the water content, and within 25% for the nitrate concentration in the soil solution, while measured nitrate leaching exceeded the simulated values by up to 80 kg N/ha. The precision of the model seems to be satisfactory in view of the large amount of nitrogen and water cycled in a vegetable field (table 5). However with respect to water management improvements are necessary. Further validation of the model on other sites is required before a predictive model for practical purposes can be elaborated. Copyright © 1990 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim