Development and testing of a computer model for predicting the amount and timing of nitrogen release from soil

Abstract

A computer model has been developed to establish the mineral nitrogen status of the soil throughout the year. The model runs on a weekly time-step using total weekly rainfall and evapotranspiration, and mean weekly air temperature. It simulates the release of nitrogen by mineralization, its loss by denitrification and volatilisation, and its subsequent removal from the soil system through crop uptake, immobilisation and leaching. The model has been designed to specify how much fertiliser nitrogen should be applied to spring or winter cereals growing in a particular field in a particular year and requires only simple inputs. These include the soil type, previous cropping history and the expected yield from the field in question, given an average season and allowing for the likelihood of weed or disease problems. An account is kept of all the nitrogen flows within the model so it can be used, for example, to compare cumulative annual losses under different agronomic systems. Nitrogen may be 'labelled' within the model so that the fate of fertiliser N can also be traced.

The model was tested using data from experiments conducted at Rothamsted, Woburn and Saxmundham experimental farms in which a pulse of 15N-labelled fertiliser was applied to winter wheat. The fate of this labelled N was followed for up to four years. Over this period, the agreement between modelled and measured values of labelled N remaining in the soil and recovered by successive crops was acceptable for all three sites, for N applications ranging from 48 to 192 kg N/ha. The model was also tested against data from experiments involving applications of FYM and 15N-labelled fertiliser to spring barley. Again, model simulations of the fate of residual labelled N in soil agreed well with measurements. Experiments with 15N-labelled fertiliser pose a particularly stringent test of such models, since the model has to mimic the behaviour of both labelled and unlabelled nitrogen.

Comparison of model simulations with measurements of residual labelled nitrogen in inorganic form have indicated that the processes causing loss of fertiliser N in the period immediately following application are complex and are probably highly sensitive to temperature and moisture conditions in the soil. Although the model simulates loss of nitrate by denitrification in wet soil, it appears that other processes, such as the formation of nitrous oxide during nitrification, may cause significant losses of fertiliser in moist but aerobic soil conditions. Further experimental work is needed to clarify the interactions between the factors which govern the loss of fertiliser N.

A related project on the validation of the computer model is reported as Project Report No 114 .