Functional linear models to test for differences in prairie wetland hydraulic gradients

Functional data analysis provides a framework for analyzing multiple time seriesmeasured frequently in time, treating each series as a continuous function of time.Functional linear models are used to test for effects on hydraulic gradient functionalresponses collected from three types of land use in Northeastern Montana at fourteenlocations. Penalized regression-splines are used to estimate the underlying continuousfunctions based on the discretely recorded (over time) gradient measurements. Permutationmethods are used to assess the statistical significance of effects. A method foraccommodating missing observations in each time series is described. Hydraulic gradientsmay be an initial and fundamental ecosystem process that responds to climate change. Wesuggest other potential uses of these methods for detecting evidence of climate change.