Using hyperspectral and LIDAR remote sensing to study the distribution and phenologic stages of Lepidium latifolium

Background/Question/Methods .  Lepidium latifolium (perennial pepperweed) is a noxious Eurasian weed invading riparian and wetland areas of the western US.  Understanding which sites are most susceptible to invasion by Lepidium will allow more efficient management of this weed.  We used high-powered remote sensing data to develop habitat suitability models for Lepidium at the Rush Ranch Open Space Preserve in Suisun Marsh, in California’s San Francisco Bay/Sacramento-San Joaquin River Delta Estuary.  Lepidium distribution was mapped with hyperspectral image data of the site, providing presence/absence data to train and validate habitat models.  A LIDAR dataset of the Bay-Delta was used to create a high-resolution digital elevation model of Rush Ranch and derive predictor environmental variables (distance to channel, distance to upland, elevation, slope, aspect, convexity).  An aggregate decision tree model was developed to identify which of these variables are associated with Lepidium presence and to predict suitable habitat that has not yet been invaded.

Results/Conclusions .  The habitat model found Lepidium to infest two zones:  near the marshland-upland margin and near channels within the marsh.  Topographical data, which is typically strongly correlated to wetland species distributions, was relatively unimportant to Lepidium occurrence.  The map of predicted Lepidium habitat reveals that Rush Ranch contains considerable habitat that it is at risk to continued invasion.  The hyperspectral dataset indicated a gradient of Lepidium phenology across the marsh.  LIDAR-derived environmental variables were used to explore the ecological controls of this phenological variation.