IFEVA   02662
INSTITUTO DE INVESTIGACIONES FISIOLOGICAS Y ECOLOGICAS VINCULADAS A LA AGRICULTURA
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
Factors Affecting Weed Community Dynamics in Sugarcane Cropping Systems of Northern Argentina
Autor/es:
FERRARO, DIEGO OMAR; RIVERO, D.E.; GHERSA, C. M.
Lugar:
Vancouver, CANADA
Reunión:
Congreso; V International Weed Science Congress; 2008
Resumen:
Sugarcane production in Argentina is restricted tonorthern areas of the country, originally covered byforests. Yields and management intensity levels arefrequently high. However, weeds are still a major problem,and the hierarchy of factors affecting weed communities isstill poorly understood. This study is aimed at determiningthe magnitude of temporal changes in composition of weedcommunity in sugarcane fields, and environmental andmanagement factors that drive these changes. We surveyed91 crop fields in farms of Northern Argentina during thespring of 2004 and 2005. Change in species compositionwas assessed using the interannual euclidean distance ofeach crop field along the non-metric multidimensionalscaling (NMS) ordination space. Distance values wereclustered in three different groups (i.e. low, medium andhigh change) through a k-means cluster algorithm.Classification and regression trees (CART) were used forpartitioning the clustered groups of weed communitychange into subsets with the highest attainable homogeneitydefined by several explanatory factors. The explanatoryfactors for each crop field were: 1) farm; 2) cropvariety; 3) area; 4) number of ratoon crops; 5) month ofharvest; 6) number of herbicides applied; 7) geographicalcoordinates; and 8) biomass yield. NMS ordination usingpresence-absence and abundance data explained 76% and77 % of weed community data, respectively. Final CARTconfiguration explained 44% of the variation in thepresence-absence weed clusters. The first tree splittingvariable was farm and the model also included latitude,number of ratoon crops and field crop area. When thecontribution of explanatory factors to the construction ofthe whole tree was assessed the number of ratoon cropswas the most important factor. Final CART configurationexplained 49% of the variation in the presence-absenceweed clusters. However, in this case, sugarcane biomassyield was the most important factors for node splitting.Other variables included in abundance CART model werefarm and crop variety. NMS and CART models used inthis study were able to explain almost 50 % of weedcomposition changes due to several explanatory factors.Moreover, CART models suggested different controls forboth changes in presence and abundance in sugarcaneweed communities. However, patterns emerged from thisexploratory study should be further examined using largerdatabases. Although CART are not inference tools, theirresults can be very useful for suggesting futures mechanisticstudies and designing more comprehensive weedmanagement decision for the sugarcane farming systems.