Iterative nearest neighbour age-height curve adjustment: Addressing the impact of spatial heterogeneity on longitudinal forestry provenance trial data

Abstract

Spatial heterogeneity in long term forestry genetics trials can obscure genetic variation and influence conclusions made based on the data. Nearest neighbour adjustment methods have been employed to account for spatial patterns, however such methods are limited in their applicability to longitudinal provenance trial data because of the need to account for repeated measures as well as the interest in preserving among-site variation. In this study, a novel approach combining nearest neighbour adjustment techniques with longitudinal data analysis concepts was developed to adjust lodgepole pine (Pinus contorta var. latifola Douglas) provenance trial data for spatial heterogeneity. Individual height-age logistic growth curves were fit to each tree to age 35, and the horizontal asymptote parameters of the height-age curves were adjusted for spatial patterns using iterative nearest neighbour adjustments. The methods developed in this study successfully removed positive spatial correlation and reduced the interaction between block and provenance, thus reducing changes in population ranking among blocks within sites. The adjustments applied caused large shifts in population ranks at many sites, indicating that seed selection decisions based on data not adjusted for spatial heterogeneity could lead to selecting non-optimal populations for a given site. This research provides a methodological framework for adjusting longitudinal genetics trial data for underlying site conditions and spatial patterns. Additionally, a spatially adjusted version of a large and extensively utilized provenance trial dataset was created for use in future analyses.