UC Davis

Invasive weeds threaten the biodiversity and forage productivity of grasslands worldwide. However, management of these weeds is constrained by the practical difficulty of detecting small-scale infestations across large landscapes and by limits in understanding of landscape-scale invasion dynamics, including mechanisms that enable patches to expand, contract, or remain stable. While high-end hyperspectral remote sensing systems can effectively map vegetation cover, these systems are currently too costly and limited in availability for most land managers. We demonstrate application of a more accessible and cost-effective remote sensing approach, based on simple aerial imagery, for quantifying weed cover dynamics over time. In California annual grasslands, the target communities of interest include invasive weedy grasses (Aegilops triuncialis and Elymus caput-medusae) and desirable forage grass species (primarily Avena spp. and Bromus spp.). Detecting invasion of annual grasses into an annual-dominated community is particularly challenging, but we were able to consistently characterize these two communities based on their phenological differences in peak growth and senescence using maximum likelihood supervised classification of imagery acquired twice per year (in mid- and end-of season). This approach permitted us to map weed-dominated cover at a 1-m scale (correctly detecting 93% of weed patches across the landscape) and to evaluate weed cover change over time. We found that weed cover was more pervasive and persistent in management units that had no significant grazing for several years than in those that were grazed, whereas forage cover was more abundant and stable in the grazed units. This application demonstrates the power of this method for assessing fine-scale vegetation transitions across heterogeneous landscapes. It thus provides means for small-scale early detection of invasive species and for testing fundamental questions about landscape dynamics.