PARENT SESSION
Contributed Oral Session 123: Photosynthesis and Water Relations: Leaf Architecture
Thursday, August 11, 8:00 AM - 11:30 AM, Meeting Room 522 A, Level 5, Palais des congrès de Montréal
Estimating leaf area on individual plants by image analysis.
Gutschick, Vincent*,1, Maxwell, Connie2, Linnell, Anna1, Logan, Carla1, Brown, Karen3, 4, Anchondo, Alvaro5, 1 Dept. of Biology, Las Cruces, NM, USA2 USDA-ARS, Las Cruces, NM, USA3 Desert Research Institute, Las Vegas, NV, USA4 Los Alamos National Laboratory, Los Alamos, NM, USA5 Facultad de Ciencias Agricolas y Forestales, Delicias, Chih, Mexico
ABSTRACT- Many methods exist for estimating leaf area, with destructive harvest being the "gold standard." However, many studies require minimal sampling (e.g., FACE sites) or repeated measures, thus disallowing wholesale harvest. We have developed methods to estimate leaf area from digitized whole-plant images taken against natural backgrounds, such as sky, or artificial backgrounds, such as fabric that temporarily isolates plants visually from neighbors. Our algorithms distinguish leaf from non-leaf in a 2-D color space of (G-R, G-B) projected from ordinary RGB space. We apply turbid medium models to estimate leaf area from fractional green-leaf area in view, thus accounting for angled leaf presentation and hidden leaves. The algorithms also filter out pixel noise and color shifts in shadowed areas. The model is applied in small tiles to avoid averaging dense and sparse areas (the sieve effect, causing underestimation). We are making available compiled Fortran programs (for Windows or UNIX) for processing digital images singly or in batch mode. The method has been tested on piecewise-harvested Larrea tridentata. In application to pecan trees the method elucidates control of leaf area development by environmental manipulations. We discuss extensions to landscape-level equilibration of leaf area index and of evapotranspiration.
Key words: leaf area, image analysis, Larrea tridentata, evapotranspiration