SENSITIVITY ANALYSES OF THE ARCWHEAT1 CROP MODEL - THE EFFECT OF CHANGES IN RADIATION AND TEMPERATURE

The sensitivity of predicted final grain yield to changes in the mean and variance of daily temperature and daily log(radiation) was studied for the ARCWHEAT1 crop model of winter wheat. These two climatic vari ables were each simulated by each of two stochastic models, the parame ters of which were estimated from 12 years' data from Rothamsted Exper imental Station. When ARCWHEAT1 was run with systematic changes in dai ly temperature from - 2-degrees-C to + 2-degrees-C in steps of 0.1-deg rees-C, there were abrupt changes in predicted final grain yield which coincided with predicted increases in leaf number. These effects were smoothed out when the stochastic climate models incorporating simulat ed daily variation were used as inputs, in multiple runs Of ARCWHEAT1. In such runs, the response of predicted yield to changes in the means of temperature and log(radiation) was studied both for individual dev elopmental stages and over the entire growing season. Yield was insens itive to changes in temperature and radiation during the stage between sowing and emergence, and to changes in radiation before terminal spi kelet formation. An increase in temperature after emergence and before the beginning of car growth tended to increase yield; an increase bet ween then and anthesis had a neutral effect, and there was an interact ion with radiation; an increase after anthesis had a strong negative e ffect on yield. The final developmental stage, from anthesis to end of grain filling, was the dominant influence on yield, for both temperat ure and radiation. The effect of an increase in radiation was to incre ase photosynthesis and, consequently, yield. When changes were made ov er the entire growing season, the effects of temperature and radiation were independent. Radiation was associated positively with yield and linearly on a logarithmic scale; a 20 % increase in radiation resulted in a 14 % increase in predicted yield. Temperature was associated neg atively with yield overall; a change in mean daily temperature from - 2-degrees-C to + 2-degrees-C caused an 8 % decrease in predicted yield . However, the relationship was less simple, because a change in tempe rature of + 1-degrees-C caused a slight increase in yield. Much of the variation in predicted yield could be explained by a simple empirical model, suggesting that mechanistic crop models may lend themselves to radical simplification for prediction. Quantitatively, there were qui te large differences in predicted yield between the two climatic model s chosen, despite the apparent good fits of each. Sensitivity analyses confirmed that the mean predicted yield was not affected by changes i n the variability of the climatic driving variables. However, the vari ance of yield was substantially and positively associated with increas es in the variance of log(radiation); similar results, to a lesser ext ent, were found for temperature.