Skip to main content
ARS Home » Midwest Area » Madison, Wisconsin » U.S. Dairy Forage Research Center » Research » Publications at this Location » Publication #142682
Title: THE USE OF SPATIALLY ADJUSTED HERBAGE YIELDS DURING THE ANALYSIS OF PERENNIAL FORAGE GRASS TRIALS ACROSS LOCATIONS

Author
item SMITH, K - AGRIC. VIC., AUSTRALIA
item Casler, Michael

Submitted to: Crop Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/19/2003
Publication Date: 12/23/2003
Citation: SMITH, K.F., CASLER, M.D. THE USE OF SPATIALLY ADJUSTED HERBAGE YIELDS DURING THE ANALYSIS OF PERENNIAL FORAGE GRASS TRIALS ACROSS LOCATIONS. CROP SCIENCE. 2003. v. 44. pp. 56-62.

Interpretive Summary: Forage yield trials are an essential part of cultivar evaluations, plant breeding efforts, and cultivar recommendations to forage producers. Without this information, forage producers and the seed industry cannot make intelligent choices among available forage cultivars. Field trials are typically variable and imprecise. New statistical methods prediction and use spatial variation to adjust cultivar means for environmental effects. We developed and evaluated methods for combining forage trials across years and locations, using the spatial adjustment method of nearest neighbor analysis (NNA). These methods will allow researchers to combine analyses across locations and years, allowing for the study of treatment x environment interactions. We also showed that adjustment of data from individual harvests provided no advantage over adjustment based on the season total yields. This provides a simplification to spatial adjustment methods for forage researchers, allowing a mechanism to control environmental effects without the need for complex experimental designs.

Technical Abstract: Methods were developed to use mixed model analysis to estimate least squares mean yields for forage yield across sites and years using yield data that had been previously adjusted for spatial variation. The method involved the use of nearest neighbor (Papadakis) analysis NNA to adjust plot herbage yields. These adjustments were conducted and compared either on yearly totals of herbage yield or on the yield at individual harvests within a year that were then combined to give total yield. These methods were then validated on the data from 9 diverse forage grass trials that were managed either to a hay-cutting regime or as biomass accumulation trials with one harvest per year. Nearest neighbor analysis (NNA) was shown to consistently improve the precision of cultivar trials managed to simulate a hay cutting regime. NNA of yields across sites and years was shown to have a relative efficiency (RE) between 105 and 135% (mean 121%) compared to RCB regardless of the method used for NNA. The adjustment of plot yields at individual harvests within a year was shown to provide small, but consistent improvements over the adjustment of annual plot yields. The RE of NNA was not significantly associated with location, harvest time or whether trials were in the first or second year of measurements. These results suggest that the RE of NNA on a given trial is related to the spatial heterogeneity within the specific area that the plots are sown. Therefore, when the intrablock heterogeneity happens to be small, then the RE of NNA will also be small compared to RCBD. In contrast, the RE of NNA compared to RCB analysis for herbage biomass trials of switchgrass (mean 105%) was generally lower than that achieved with hay-cutting trials. The reason for this difference is unknown but may be due to differences in trial management and harvesting regime. We suggest that trial operators assess the RE of NNA on the early harvests from all locations within a trial and if the RE are large then they should consider the use of NNA across locations and years when reporting entry means.