Landscape- scale GPP and carbon density inform patterns andimpacts of an invasive tree across wet forests of Hawaii

Abstract

Plant invasion typically occurs within a landscape-scale framework of abioticand biotic conditions, often resulting in emergent feedbacks among environment, ecosystemfunctions, and the dominance of invasive species. Understanding the mechanisms underlyingsuccessful invasions is an important component of conservation and management efforts, butthis has been poorly investigated in a spatially explicit manner. Knowing where and whyinvasion patterns change throughout the landscape enables managers to use context-specificcontrols on the spread of invasive species. Using high-resolution airborne imaging spectroscopy,we studied plant performance in growth within and across landscapes to examine the dominanceand spatial distribution of an invasive tree, Psidium cattleianum (strawberry guava), inheterogeneous environmental conditions of a submontane Hawaiian tropical forest. Weassessed invader performance using the GPP ratio index, which is the relative difference inremotely sensed estimates of gross primary productivity between canopies of guava andcanopies of the invaded plant community. In addition, we used airborne LiDAR data toevaluate the impacts of guava invasion on the forest aboveground carbon density in differentenvironments. Structural equation modeling revealed that substrate type and elevation abovesea level interact and amplify landscape-scale differences in productivity between the invasivespecies and the host plant community (GPP ratio); differences that ultimately control levels ofdominance of guava. We found shifts in patterns of forest carbon storage based on bothgradual increase of invader dominance and changes in environmental conditions. Overall, ourresults demonstrate that the remotely sensed index defined as the GPP ratio provided aninnovative spatially explicit approach to track and predict the success of invasive plants basedin their canopy productivity, particularly within a landscape-scale framework of varyingenvironmental factors such as soils and elevation. This approach may help managers accuratelypredict where invaders of forests, scrublands, or grasslands are likely to exhibit high levels ofdominance before the environment is fully invaded