Skip to main content
ARS Home » Research » Publications at this Location » Publication #262381
Title: Modeling "habitat suitability" for a herbicide resistant weed using a species distribution model and presence-only data

Author
item Wiles, Lori
item KUMAR, SUNIL - Colorado State University
item DAVIS, VINCE - University Of Illinois
item JOHNSON, BILL - Purdue University

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 1/2/2010
Publication Date: 2/8/2011
Citation: Wiles, L., Kumar, S., Davis, V.M., Johnson, B. 2011. Modeling "habitat suitability" for a herbicide resistant weed using a species distribution model and presence-only data. Meeting Abstract. Feb 7-10 Portland Oregon.

Interpretive Summary:

Technical Abstract: Herbicide resistant weeds are like invasive weeds: prompt management is needed to prevent their spread. For invasive weeds, first reports of a weed's occurrence are often analyzed with species distribution models (SDM) to prioritize detection and treatment. Suitability of other areas as habitat for the weed is estimated using the observed locations, environmental data, and an appropriate SDM for the available data. Compared to invasive weeds, the presence of herbicide resistance weeds is driven more by crop management than environment, yet SDMs might be similarly used. After a resistant weed has been found in a few fields, the risk of resistance in other fields might be estimated from information about the environment and past crop management. We tested this idea with three SDMs and data on presence of glyphosate-resistant (GR) horseweed (Conyza Canadensis) in Indiana corn fields. Prediction of the presence of GR plants from presence-only data using MAXENT was compared with prediction from presence/absence data using either CART or Boosted Regression Trees. The latter two models require data from fields where no GR weeds are present but MAXENT does not. All three models fit the data well and identified ‘visible injury’ and ‘weed abundance’ as the most important predictors of GR. MAXENT was the best at prediction, supporting the use of SDMs and first reports of resistance to advise farmers about which fields are most at risk. Environmental data for this use of SDM are readily available; the challenge will be collecting historical crop management information.