Accounting for Spatial Variability in Field Experiments on Tea

Abstract

Spatial variability among experimental units is a common problem in field experiments on tree crops such as tea (Camellia sinensis). Spatial variability is partly accounted for by blocks, but a substantial amount remains unaccounted for and this may lead to erroneous conclusions. In order to capture spatial variability in field experiments on tea, six commonly used spatial analysis techniques were investigated: Covariate method with pre-treatment yield as the covariate, Papadakis and the Modified Papadakis nearest neighbour adjustments, Moving means and Modified moving means methods, and Autoregressive method. The data from long-term fertilizer experiments and cultivar evaluation trials, conducted at different locations by the Tea Research Institute of Sri Lanka, were used in the study. Spatial techniques were evaluated by means of their relative efficiency at each location and year. Evaluation of the four neighbour methods analysed in conjunction with the pre-treatment yield, revealed that spatial variability due to both past and current conditions are operative, especially in experiments with large blocks, and could be captured simultaneously. Relative efficiencies averaging 141% clearly indicated that the neighbour techniques in combination with pre-treatment yield would be effective in controlling the experimental error in tea experiments with large blocks (nine plots per block or more). Experiments with small blocks were not affected by spatial variability due to past conditions and only that due to current conditions need to be addressed. Neighbour techniques, on their own, were found to be adequate to capture spatial variability due to current conditions. The modified Papadakis technique was found to be the best with an average relative efficiency of 145%. The techniques investigated in the study can easily be implemented using standard statistical software. The precision of tea experiments could be increased by using covariate analysis with pre-treatment yield and any one of the four nearest neighbour adjustments tested, when the block size is large; and modified Papadakis technique, on its own, when the block size is small.

[No Title Available]

“…Assim, o método de Papadakis, realizado com base na análise de covariância, em que a covariável é determinada como a média das estimativas dos erros nas parcelas vizinhas (variação espacial), mostrou-se eficiente para a comparação da produtividade de grãos de genótipos de trigo (Benin et al, 2013) e de soja (Storck et al, 2008). Estudos relacionados a métodos de análise espacial (Duarte & Vencovsky, 2005), para as culturas de chá (Peiris et al, 2008), feijoeiro (Souza et al, 2000), feijoeiro e milho (Costa et al, 2005) e laranjeira (Maia et al, 2013) foram realizados e, de maneira geral, relataram a melhoria da eficiência na seleção de genótipos.…”

Increased Genomic Prediction Accuracy in Wheat Breeding Through Spatial Adjustment of Field Trial Data

“…These spatial trends can be eliminated with postdata treatment. Different strategies exist, including model variance-covariance matrixes, row-columns, and moving-means ( Cullis et al 1998 ; Peiris et al 2008 ; Müller et al 2010 ; Leiser et al 2012 ).…”

Canopy coverage phenotyping and field spatial variability adjustment as an efficient selection tool in soybean breeding

“…Strategies to eliminate field spatial variation from phenotypic data include modeling variancecovariance matrixes, row-columns design, and moving means (Cullis et al, 1998;Lado et al, 2013;Leiser et al, 2012;B. U. Müller et al, 2010;Peiris et al, 2008). Yield prediction models accounting for spatial field variation decreased the error variance and increased the accuracy of yield selection in early generations (Sun et al, 2015).…”

“…Field spatial variation can be measured as the trait correlation among plots, which can affect the yield ranks and impact the accuracy of selection (Gilmour et al., 1997; Sun et al., 2015); however, appropriate spatial adjustment of the phenotypic data can help to remove these field spatial variation trends and increase selection accuracy (Lado et al., 2013). Strategies to eliminate field spatial variation from phenotypic data include modeling variance–covariance matrixes, row–columns design, and moving means (Cullis et al., 1998; Lado et al., 2013; Leiser et al., 2012; B. U. Müller et al., 2010; Peiris et al., 2008). Yield prediction models accounting for spatial field variation decreased the error variance and increased the accuracy of yield selection in early generations (Sun et al., 2015).…”