A simple and reliable method for estimating cactus pear biomass production using cladode dimensions
Authors
Angres, Gabriel
Issue Date
2022
Type
Thesis
Language
Keywords
cactus pear , drought stress , indirect modeling , linear regression , productivity modeling
Alternative Title
Abstract
In a warming world, crassulacean acid metabolism (CAM) plants, which exhibit much greater water-use efficiency and tolerance to high temperatures than C3 and C4 photosynthesis plants, are increasingly attractive as food, feed, fiber, and biofuel crops. Hence, as agricultural development of these plants expands, the issue of measuring crop yield becomes more relevant. Traditionally, biomass estimates of a crop are based upon gravimetric assessment; however, this process is expensive, labor-intensive, and time-consuming. Indirect approaches that make accurate predictions from dimensional measurements, images, or simple object counts are simpler, faster, and relatively inexpensive. However, current indirect methods for photosynthetic-stem (cladode) plants, such as cactus pear (Opuntia spp.) require further refinement to be used widely in field settings. An optimal model for indirect measurement must be parsimonious and highly accurate, while also having a cogent supporting rationale for the model’s construction. We conducted statistical analyses on cladode dimensional data using 14 accessions representing five species, including Opuntia cochenillifera, O. crassa, O. ficus-indica, O. robusta, and O. undulata to derive a general model for approximating the fresh weight of a cladode from its physical dimensions, and quantify any statistically significant morphological differences among them. Correlations between fresh weight and four physical measures (pad diameter, width, height, and thickness) and factorial combinations thereof (e.g., height x width vs. fresh weight) were assessed for accuracy (per the coefficient of determination) and parsimony (per the Schwarz-Bayes Criterion) (> 0.86 adj. R^2 minimum, 0.95 adj. R^2 general fit). A second modeling approach used a refined measured area in ImageJ (> 0.85 adj. R^2 minimum, 0.94 adj. R^2 general fit). A third modeling approach attempted an approximate but less logistically intensive area approximation using an ellipse which lost relatively little performance over all accessions (> 0.79 adj. R^2, 0.94 adj. R^2 general fit). These formulae generally met or exceeded the performance of models obtained from prior research efforts on an all-accessions basis. Further theoretical work consists of incorporating the effect of morphological variation and water content into the formula for a particular accession.