Since 2003 we have performed an annual survey of the distribution and dynamics of invasive aquatic weeds in the Sacramento-San Joaquin Delta, CA, using hyperspectral imagery. The Delta, part of the largest estuary in the Western United States, has undergone long-term changes in ecosystem function, and is currently the focus of a massive, coordinated rehabilitation and restoration program. Target species for classification have been Brazillian waterweed (Egeria densa) and water hyacinth (Eichhornia crassipes). Covering over 2,500 km2 and a base image dataset of over 130 gb per year, this project is one of the largest hyperspectral monitoring efforts ever conducted. Successful target species detection requires classification of several species of three different lifeforms: submerged, emergent, and floating vegetation. Submerged species detection is complicated by varying water conditions (turbidity, depth and specular reflection) and spectrally similar species. Classification challenges for floating and emergent species relate mostly to intraspecies phenological heterogeneity. We discuss field data collection and preprocessing problems and solutions. We present methods and results from the 2003-2007 efforts to classify submerged, floating and emergent aquatic vegetation with a focus on lessons learned and future remote sensing research goals for an improved description of the state and dynamics of aquatic vegetation.