Using map algebra to determine the mesoscale distribution of invasive plants: the case of celastrus orbiculatus in Southern Illinois, USA

Document Type

Article

Journal/Book Title

Biological Invasions

Publisher

Springer

Publication Date

11-11-2006

Volume

9

Issue

4

First Page

419

Last Page

431

Abstract

Since its introduction into North America in the late 19th century, Celastrus orbiculatus (Thumb.) has become a serious ecological threat to native ecosystems. Development of a method to accurately map the occurrence of invasive plants, including C. orbiculatus, would greatly assist in their assessment and control. Using an innovative map regression model, we predicted 85% of presence and absence of C. orbiculatus within our study area. We identify environmental characteristics associated with C. orbiculatus and demonstrate the use of this information to predict occurrence of C. orbiculatus across a broad area in Southern Illinois, USA. Presence and absence information were obtained at sample points within discrete areas of C. orbiculatus occurrence. Forest cover, elevation, slope gradient and aspect, soil pH and texture, distance to nearest road, and potential annual direct incident radiation were recorded for invaded and adjacent non-invaded areas. Presence of oak, elevation, slope gradient, soil pH, soil texture, and distance to road were significant factors associated with the presence or absence of C. orbiculatus. Probability of occurrence of C. orbiculatus was highest on gently sloping interfluves with successional forest canopy not dominated by oak, and less acidic, mesic soil. A logistic regression model was developed and extrapolated over a raster GIS data layer using map algebra to predict current invasion throughout the study area. The model correctly predicted at least 85% occurrence of C. orbiculatus. When combined with logistic regression, map algebra is a potentially powerful tool for evaluating the spatial distribution of invasive plants provided sound statistical principles are applied in extrapolating validated regression models.

Link to Full Text

Share

COinS