Vegetation-index Models Predict Areas Vulnerable to Purple Loosestrife (lythrum Salicaria) Invasion in Kansas

Purple loosestrife (Lythrum salicaria) constitutes an invasive species detrimental to wetland habitats in North America. To estimate areas vulnerable to it in Kansas, we modeled the potential geographic distribution of the species by using current records in the state, remotely sensed vegetationindex data from the Moderate Resolution Imaging Spectrometer (MODIS), and the Genetic Algorithm for Rule-Set Prediction (GARP). Models built using only localities from northeastern Kansas (the origin of invasion within the state) consistently predicted test localities in other parts of the state with negligible omission. An additional analysis using records from all counties where the species is known showed a similar prediction. All models indicated suitable conditions for purple loosestrife in much of eastern and central Kansas, as well as in riverine and irrigated areas in the western part of the state. The approach presented here might prove useful for assessing the regional colonization potential of other newly detected invasive species before other studies can be undertaken. RESUMEN—Lythrum salicaria constituye una especie invasora perjudicial para los hábitats pantanosos en Norteamérica. Para estimar cuáles son las áreas vulnerables a esta especie en Kansas, modelamos la distribución potencial de la especie utilizando registros actuales del estado, datos de ı́ndices de vegetación tomados remotamente por el Moderate Resolution Imaging Spectrometer (MODIS), y el Genetic Algorithm for Rule-Set Prediction (GARP). Modelos construidos usando sólo las localidades del noreste de Kansas (el origen de la invasión dentro del estado) predijeron consistentemente a las localidades de prueba en otras partes del estado con omisión despreciable. Un análisis adicional usando registros de todos los condados en los cuales se conoce que la especie está presente mostró una predicción similar. Todos los modelos indicaron condiciones apropiadas para la especie en la mayor parte del este y del centro de Kansas, tanto como en áreas ribereñas e irrigadas en la parte occidental del estado. El enfoque presentado aquı́ posiblemente será de utilidad en la evaluación del potencial de colonización regional de otras especies invasoras recientemente detectadas antes de que otros estudios puedan ser realizados. Several techniques now exist for modeling the environmental requirements and geographical distribution of a species based on the environmental characteristics of sites of known occurrence (Graham et al., 2004). Such presence-only occurrence data sets often derive from museum and herbarium collections and represent the only data available for many species. Occurrence records are typically employed in conjunction with digital maps that provide data for long-term climatic patterns, topography, potential vegetation, and other environmental variables that commonly influence the macrodistributions of species (Mackey and Lindenmayer, 2001). Using a Geographic Information System (GIS), a model of the niche requirements of a species (sensu Hutchinson, 1957) in those ecological-environmental dimensions is created and then projected onto geography, thus approximating the potential geographical distribution of the species in the study region (Anderson et al., 2002a). Such techniques have been applied to the study of invasive species, which constitute a pressing economic and scientific problem affecting diverse sectors of society (Ricciardi et al., 2000). In the past, invasive-species management was largely reactive in nature (Smallwood and Salmon, 1992; Sakai et al., 2001). However, THE SOUTHWESTERN NATURALIST 51(4):471–480 DECEMBER 2006