Location: Soil Management ResearchTitle: A comparative study between nonlinear regression and nonparametric approaches for modelling Phalaris paradoxa seedling emergence
|GONZALEZ-ANDUJAR, JOSE - Ministry Of Science And Innovation, Csic|
|FRANCISCO-FERNANDEZ, M - University Of Coruna|
|CAO, R - University Of Coruna|
|REYES, M - University Of Coruna|
|URBANO, J - University Of Sevilla|
|BASTIDA, F - Universidad De Huelva|
Submitted to: Weed Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/3/2016
Publication Date: 10/1/2016
Publication URL: http://handle.nal.usda.gov/10113/5497853
Citation: Gonzalez-Andujar, J.L., Francisco-Fernandez, M., Cao, R., Reyes, M., Urbano, J.M., Forcella, F., Bastida, F. 2016. A comparative study between nonlinear regression and nonparametric approaches for modelling Phalaris paradoxa seedling emergence. Weed Research. 56(5):367-376.
Interpretive Summary: Computer models and software that predict the emergence of weed seedlings is valuable for making management decisions, especially when to apply certain herbicides. However, the traditional approach to the development of such models, called parametric non-linear regression (PNR), is subject to strict statistical requirements, and these requirements typically are not formally met by most research projects using this technique. Our research sought to overcome these statistical limitations by employing an alternative approach using non-parametric estimation techniques, which have far fewer statistical limitations. To test these ideas we used three years of data on seedling emergence of a grassy weed called hooded canary grass, along with associated data for soil temperature and soil water on a daily basis. The non-parametric approach consistently proved to be more accurate and more reliable than the traditional PNR approach. Consequently, weed researchers and modelers, the primary users of the information we generated, should consider using non-parametric estimation techniques in future weed emergence models.
Technical Abstract: Parametric non-linear regression (PNR) techniques commonly are used to develop weed seedling emergence models. Such techniques, however, require statistical assumptions that are difficult to meet. To examine and overcome these limitations, we compared PNR with a nonparametric estimation technique. For the analyses we used seedling emergence data for Phalaris paradoxa, collected over three growing seasons in Spain, and associated soil hydrothermal time data. Non-parametric estimates consistently provided superior accuracy compared to PNR predictions, regardless of whether Logistic, Gompertz, or Weibull functions were used, although the Weibull function was the best of the PNR models. Accordingly, the reliability of weed seedling emergence models may be higher if developed from nonparametric estimation techniques than the more traditional PNR approaches.