Structural equation modelling in root research: a focus on 0-tillage systems

During its downward growth a plant root experiences a combination of internal and external stimuli that shape its architecture. The extent to which roots respond to the surrounding media, termed “plasticity”, governs plant’s ability to capture resources in variable soil environments. Identifying direct and indirect relationships involved in and triggered by root spatial arrangement is crucial for assessing crop response to changing agricultural practices (i.e. conversion to reduced tillage systems). We analyze data collected in one season from a long term tillage trial (conventional versus 0-tillage). Within a randomized block design with 3 replicates, data on roots biomass and soil physico-chemical parameters were sampled along vertical depth profiles. The main purpose is to evaluate if the tillage practice affects the distribution of roots along the vertical depth profile and if the latter has an influence on wheat yield and quality. Due to root and soil samples being repeated along a depth gradient, data are suited to be analyzed in a variety of ways, such as growth models, multilevel models and random regression/mixed-effects models. Here, to allow characteristics of the roots trajectory vary across individuals, the slope and intercept parameters are modeled as latent variables, thus leading to latent growth curve models (LCMs) characterized by great flexibility to examine roots change over depth. LCMs are framed in the context of structural equation modelling providing a framework for the assessment of causal relationships in complex inter-correlated data with several applications in root research.