| dc.description.abstracteng | Tropical forests are the world’s most productive terrestrial ecosystems and of central importance for global carbon and water cycles. Global climate projections predict increases in average temperature and an elevated frequency of extreme drought events throughout large parts of the tropics. In response to these changes, increases in mortality rates particularly among large trees have already been reported for many tropical forest ecosystems.
Hence, there is a need for better predictions of the performance of tropical forest trees under more frequent drought conditions, which the present work seeks to address a) by more accurately quantifying how much water plants use and b) by advancing the knowledge about plant traits and mechanisms that control plant water use, growth performance and drought responses. To achieve this, this study is separated in two parts, the first of which aims at methodological improvements of water use and transpiration estimates, while the second part focuses on disentangling the relationship between tree height, wood density and wood anatomical properties, and quantifying their common effect on the productivity and water relations.
The backbone of this thesis is formed by data from a field study on five research sites situated on a rainfall gradient along the Pacific coastline of Costa Rica, which are complemented by additional results from a laboratory-based study of sap flux sensor performance and a large observational dataset from tropical forests in Indonesia.
In Part I, I first present accessory results from a laboratory-based calibration experiment based on 66 stems from five temperate diffuse-porous tree species. Three commonly applied sensor systems, thermal dissipation probes (TDP), heat field deformation (HFD) sensors, and heat ratio method (HRM) sensors, were validated against gravimetrically determined flow rates to compare them in terms of bias, precision and accuracy. Our results indicate a systematic underestimation of true sap flux density by on average 23-45% with the TDP method, and a relatively low precision (but lower bias) with HFD sensors. The best performance was observed for HRM sensors if restricted to low flow ranges. Based on the methods comparison, we conclude that the TDP and HFD methods require species-specific empirical calibration for optimal performance, and that for all methods there is a within-species variability in calibration relationships that puts a limit on accuracy.
In the light of these findings, I then discuss the outcome of a field study of sap flux measurements using the HFD method. In this work, we analyzed a dataset of sap flow measurements from 38 trees belonging to eight tropical dry forest tree species from Costa Rica. Based on a Bayesian hierarchical modelling approach, we developed a model for radial sap flux profiles that allowed to propagate model uncertainty when predicting the shape of HFD-based radial profiles onto new trees and new tree species, and describe how to integrate these model predictions with single-point sensor readings from other sensor systems in order to improve their accuracy. We found that tree height had a credible effect on both the shape of radial profiles and whole-tree water use, with larger trees having the bulk of flow closer to the bark and reaching higher transpiration rates. Compared to water use estimates based on radial profiles, estimates that assumed constant flow over the entire sapwood overestimated water use by 26% on average.
In Part II, I first show results from a dataset comprising trait averages from 99 tropical forest tree species from Sumatra and Sulawesi (Indonesia). In this study, we used structural equation models (SEM) to analyze the relationships between tree size, wood density, wood anatomical traits related to hydraulic efficiency, empirically determined sap flux density, biomass productivity and tree water use, and compared the results to simple bivariate associations. We found a strong correlation between water use and growth, which was completely explained by their common dependence on tree size and sap flux density. While wood hydraulic traits were closely associated with growth and water use, our model suggested that this relationship was mainly driven by a confounding size effect. After accounting for tree size, only a relatively small effect of wood properties remained
that was mediated by sap flux density.
I then present a second SEM-based study that builds upon data from 201 tropical rainforest trees belonging to 40 species distributed along the rainfall gradient in Costa Rica. In this study, we found a strong dependence of biomass increment from canopy position and tree diameter, while the effects of wood density and wood hydraulic traits diminished after controlling for size effects. Notably, differences in growth along the rainfall gradient were completely explained by the effect of annual precipitation on canopy height. We further found trees belonging to species that are more affiliated to drier habitats to have smaller sapwood nonstructural carbohydrate concentrations and to be more common in the upper canopy.
Supplementary, unpublished results from an analysis of vulnerability curves measured from Costa Rican tropical rainforest trees indicate that the strong size effect in growth, water use and wood hydraulic trees surprisingly was mirrored by a size dependence in embolism resistance, with the highest embolism resistance in the largest and most fast-growing species. In addition, we found embolism resistance to be strongly associated with stem sapwood properties, with a significantly higher embolism resistance for species with harder wood and lower vessel diameters.
In summary, the present work provides a set of methodological refinements to sap flow measurement methodology that has the potential to significantly improve the accuracy of tree level transpiration estimates. In addition, it adds to the growing body of evidence indicating that tree size and/or canopy position are important covariates that have to be controlled for when studying relationships between plant traits. In particular, we show that observed positive correlations of biomass increment and water use with wood properties can largely be attributed to a confounding size effect, which suggests that the functional importance of wood anatomical traits may often be overstated. | de |