Predicting infectivity of Arbuscular Mycorrhizal fungi from soil variables using Generalized Additive Models and Generalized Linear Models

Djuuna IAF, Abbott LK, Van Niel K (2010) Predicting infectivity of Arbuscular Mycorrhizal fungi from soil variables using Generalized Additive Models and Generalized Linear Models. Biodiversitas 11: 145-150. The objective of this study was to predict the infectivity of arbuscular mycorrhizal fungi (AM fungi), from field soil based on soil properties and land use history using generalized additive models (GAMs) and generalized linear models (GLMs). A total of 291 soil samples from a farm in Western Australia near Wickepin were collected and used in this study. Nine soil properties, including elevation, pH, EC, total C, total N, P, K, microbial biomass carbon, and soil texture, and land use history of the farm were used as independent variables, while the percentage of root length colonized (%RLC) was used as the dependent variable. GAMs parameterized for the percent of root length colonized suggested skewed quadratic responses to soil pH and microbial biomass carbon; cubic responses to elevation and soil K; and linear responses to soil P, EC and total C. The strength of the relationship between percent root length colonized by AM fungi and environmental variables showed that only elevation, total C and microbial biomass carbon had strong relationships. In general, GAMs and GLMs models confirmed the strong relationship between infectivity of AM fungi (assessed in a glasshouse bioassay for soil collected in summer prior to the first rain of the season) and soil properties.