Location: Plant Science Research
Title: Optimizing the statistical estimation of the parameters of the Farquhar-von Caemmerer-Berry model of phyotosynthesis Authors
|Dubois, Jean Jacques|
|Flowers, Michael - OREGON STATE UNIVERSITY|
|Reid, Chantal - DUKE UNIVERSITY|
Submitted to: New Phytologist
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: June 11, 2007
Publication Date: October 1, 2007
Repository URL: http://hdl.handle.net/10113/35849
Citation: Dubois, J.B., Fiscus, E.L., Booker, F.L., Flowers, M., Reid, C.D. 2007. Optimizing the statistical estimation of the parameters of the Farquhar-von Caemmerer-Berry model of phyotosynthesis. New Phytologist. 176:402-414. Interpretive Summary: The mathematical model devised by Farquhar, von Caemmerer and Berry is the standard model used in quantifying carbon assimilation by plants. Applying it to plant gas exchange data is far from trivial, and the methods customarily used are both painstakingly laborious, and less than optimal from a statistical perspective. Yet no treatment of these issues has ever been published. We propose a method that is both statistically sound, and greatly enhances the efficiency of processing those data. Abundant details and illustrations using real data are provided, as well as a review of potential difficulties and their solutions. Complete software programs are also provided.
Technical Abstract: The model of Farquhar, von Caemmerer and Berry has long been the standard in relating photosynthetic carbon assimilation and intercellular CO2 concentration. Assimilation is segmented into three regions, each modeled by a distinct function. Disjunct fitting of the component functions is customarily used, requiring preliminary subsetting of data into sets believed to correspond to each region. Where to subset is at the discretion of the investigator, and each parameter is estimated using less than all available information. Added operations also make analyzing large amounts of gas exchange data unduly laborious. We show how multiple segments can be estimated simultaneously, using the entire dataset at once, without predetermination of transitions. The method is shown to be fast, and to lend itself to rapid recursive processing. Investigation of the number of estimable parameters suggests that three parameters at least have excellent statistical properties, but that estimating more than three may be difficult. Illustrations of software programming are provided, and a review of difficulties and their solutions.