Comparison of scales of climate and soil data for aggregating simulated yields of winter wheat in Denmark

Crop growth models are essentially site-based, and use of such models for a ssessing regional productivity of crops requires methods for aggregating ov er space. Different method for aggregating simulated county and national cr op yields for winter wheat (Triticum aestivum L.) in Denmark were tested us ing a crop simulation model (CLIMCROP), which was run with and without irri gation for a range of soil types and climatic conditions. The aggregated co unty or national yield was calculated by summing simulated yield of each ca tegory multiplied by the area, they represent. Ten different combinations o f scales of climate and soil data were used. The wheat area was distributed between the different soil types using either a uniform distribution or a distribution that gave preference to soils with high water-holding capacity . The simulated results were compared with Danish county and national yield statistics for winter wheat from the period 1971-1997. There was, in gener al, a poor relationship between simulated and observed yields when the obse rved yields had been detrended to remove the technology effect. A larger fr action of the inter-annual variability was captured by the model on the loa my soils compared with the sandy soils. The model was able to capture most of the spatial variation in observed yields, except at the coarsest resolut ions of the soil data. The finest resolution of soil and climate data gave a better fit of simulated to observed spatial autocorrelation in yield. The results indicate that upscaling of simulated productivity of crops for Dan ish conditions requires a spatial resolution of soil data of 10 x 10 km(2) or finer. A single climate station may be sufficient if only national yield s are estimated, but more stations are required, if regional yields are to be estimated. Consideration should also be given to the distribution of cro p area on the different soil types, (C) 2000 Elsevier Science B.V. All righ ts reserved.