Variability in Forest Plant Traits along the Western Ghats of India and Their Environmental Drivers at Different Resolutions

Abstract

AbstractIdentifying key environmental drivers for plant functional traits is an important step to understanding and predicting ecosystem responses to a changing climate. Imaging spectroscopy offers great potential to map plant traits at fine resolution across broad regions and then assess controls on their variation across spatial resolutions. We applied permutational partial least-squares regression to map seven key foliar chemical and morphological traits using NASA’s Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) for six sites spanning the Western Ghats of India. We studied the variation of trait space using principal components analysis at spatial resolutions from the plot level (4m), community level (30m and 100m) to the ecosystem level (1000m). We observed a consistent pattern of trait space across different resolutions, with one axis representing the traditional leaf economic spectrum defined by foliar nitrogen concentration and leaf mass per area (LMA) and another axis representing leaf structure and defense defined by fiber, lignin, and total phenolics. We also observed consistent directionality of environment-trait correlations across resolutions with generally higher predictive capacity of our environment-traits models at coarser resolutions. Among the seven traits, total phenolics, fiber, and lignin showed strong environmental dependencies across sites, while calcium, sugar, and nitrogen were significantly affected by site conditions. Models incorporating site as a fixed effect explained more than 50% of the trait variance at 1000m resolution. LMA showed little dependence on both environment and site conditions, implying other factors such as species composition and perhaps site history strongly affect variation in LMA. Our results show that reliable trait-trait relationships can be identified in coarse resolution imagery, but that local scale trait-trait relationships (resolutions finer than 30m) are not sensitive to broad-scale abiotic/biotc factors.