An evaluation of four crop : weed competition models using a common data set

Authors

W. Deen, University of Guelph
R. Cousens, University of Melbourne
J. Warringa, University of Melbourne
L. Bastiaans, Wageningen Agricultural University
P. Carberrys, Agricultural Production Systems Research Unit, CSIRO Tropical Agriculture, PO Box 102, Toowoomba, Queensland 4350, Australia
K. Rebel, Cornell University
S. Riha, Cornell University
C. Murphy, Office of the Gene Technology Regulator, Department of Health and Ageing, PO Box 100, Woden ACT 2602, Australia
L. R. Benjamin, Rothamsted Research, Harpenden, Hertfordshire, AL5 2JQ, UK
C. Cloughley, New Zealand Institute of Crop and Food Resources
J. Cussans, Rothamsted Research, Harpenden, Hertfordshire, AL5 2JQ, UK
F. Forcella, USDA-ARS, North Centre for Soil Conservation Research Laboratory, Morris MN
T. Hunt, University of Guelph, Guelph, Ontario N1G 2W1, Canada
P. Jamieson, New Zealand Institute of Crop and Food Resources, Private bag 4704, Christchurch, New Zealand
John Lindquist, University of Nebraska-LincolnFollow
E. Wangs, Agricultural Production Systems Research Unit, CSIRO Tropical Agriculture, PO Box 102, Toowoomba, Queensland 4350, Australia

Document Type

Article

Date of this Version

3-29-2003

Comments

Published in Weed Research 2003 43, 116–129.

Abstract

To date, several crop : weed competition models have been developed. Developers of the various models were invited to compare model performance using a common data set. The data set consisted of wheat and Lolium rigidum grown in monoculture and mixtures under dryland and irrigated conditions. Results from four crop : weed competition models are presented: ALMANAC, APSIM, CROPSIM and INTERCOM. For all models, deviations between observed and predicted values for monoculture wheat were only slightly lower than for wheat grown in competition with L. rigidum, even though the workshop participants had access to monoculture data while parameterizing models. Much of the error in simulating competition outcome was associated with difficulties in accurately simulating growth of individual species. Relatively simple competition algorithms were capable of accounting for the majority of the competition response. Increasing model complexity did not appear to dramatically improve model accuracy. Comparison of specific competition processes, such as radiation interception, was very difficult since the effects of these processes within each model could not be isolated. Algorithms for competition processes need to be modularized in such a way that exchange, evaluation and comparison across models is facilitated.