Experimental validation of an N-response model for widely different crops

Abstract

A simulation model, developed from a previously published one generally gave a reasonably good description of the effects of N-fertilizer on plant dry weight and N-content of 12 different vegetable crops grown in 26 experiments on separate sites within the same field.

Apart from standard weather and soil data, the model required as inputs, the maximum yield of dry matter, the fractional recovery of N by the crop (with minimum fertilizer), the plant mass and N content (at the time of planting or drilling), the dates of planting or drilling and of harvest. Simulations always started from the previous autumn. A constant value of soil mineral-N at that time and the same temperature dependent mineralization was assumed for all experiments.

The validity of the model was tested by a range of statistical procedures. In 19 out of the 26 experiments there were no detectable trends in the deviations of the simulated from the measured dry matter yields with increase in fertilizer-N. The model over-estimated the %N in the dry matter of some crops at the highest level of fertilizer-N and was unsatisfactory for one of the crops. With these exceptions, the sum of squares of the differences between measured and simulated %N was 22% of the sum of squares of the measured values above the mean. When N-fertilizer was withheld, the average N-uptake over all experiments was 69 kg N ha⁻¹, whereas that simulated was 59 kg N ha⁻¹; the average difference between simulated and measured uptake for each experiment was 20 kg N ha⁻¹. Simplification of the model by incorporating the same relationship between critical %N and plant weight for all crops did not lead to appreciable loss of accuracy. A user-friendly version of the model has been compiled so that it will run on IBM-compatible microcomputers with outputs that can be coupled with high level graphics packages.