Douglas-fir hydraulic architecture and relationships among xylem properties at multiple scales via a Bayesian analysis

Graduation date: 2006This study investigated variation in xylem anatomy, hydraulic properties, and the relationship between anatomy and properties within Douglas-fir trees at multiple scales. The hierarchical scales in the study included fertilization treatments (fertilized and unfertilized), trees within the treatments, and positions within the trees. Tracheid diameter, tracheid length, percent latewood, number of pits per cell, wood density, and specific conductivity (K[subscript s]) were measured on seven positions within each of 16 fertilized and unfertilized trees: the stem at cambial age 54 (breast height), 25, and 5; a branch at cambial age 20 and 7; and a root at cambial age 42 and 22. Vulnerability to embolism, sap velocity, and sap temperature were also measured on the oldest of the above-listed parts of each position. A Bayesian hierarchical regression model (HM) was developed that partitioned variation among scales. When compared to a simple linear regression model that assumed independence among observations, the HM consistently gave more precise estimates for regression coefficients and provided a more informative and precise estimate of variance. For any anatomical feature or xylem property there was little variation between treatments, however there was great variation among positions. Relationships between xylem anatomy and xylem properties were evident at larger scales (between treatments and among positions); however at a smaller scale usually only the branches and roots maintained the relationship. Tradeoffs among xylem properties were also different at each scale, often existing at larger scales but only within a few positions or positions within treatments at smaller scales. The fact that the relationships analyzed were dependent upon scale suggests that factors driving the relationships are likewise dependent on scale and that wood structure is adapted not only for whole tree functioning but also for small scale, position specific functional demands