Tree growth influences forest community dynamics and responses to environmental variations, but currently is not well understood. Tree growth in highly diverse wet tropical forests have been well studied and characterised compared to the species-poor dry tropical forests. Thus, it is not clear if growth rates and community dynamics of dry forests are similar to those of wet forests, given the longer dry season, greater rainfall variability, more open canopy and lower number of species in dry forests. This thesis focuses on identifying important factors that influence tree diameter growth rates in the dry tropical forest at Mudumalai, southern India, and also compares growth patterns at this dry forest with those at moister forests. The thesis thus contributes towards closing the gap in understanding of tree growth patterns across the tropics. An initial analysis involving matrix-based population projections of four common canopy species at Mudumalai showed that variations in diameter growth have the potential to drastically modify population trajectories of dominant species. Thus the main focus of this thesis is aimed at identifying the important intrinsic and extrinsic factors affecting growth in this dry forest, as this information could be useful for future management of the forest. The second important aim of the thesis was to find out if growth rates are influenced by different sets of factors in tropical dry versus moist forests. A large permanent 50ha plot vegetation monitoring plot was set up in 1988-89 in the Mudumalai dry deciduous forest, and was subsequently monitored annually by staff of the Centre for Ecological Sciences. Data used in this thesis represent a 12-year interval between 1988 and 2000. Girth measurements were obtained from all woody tree stems ≥1cm in diameter every four years during this 12 year interval, which provided three census intervals of diameter increment data on >13,000 trees. For the comparison between dry and moist deciduous forests, data were obtained from a similar large plot maintained and monitored at the Barro Colorado Island (BCI) in Panama. Influences of the intrinsic factors, tree size, individual identity, species identity and growth form, were examined using t-tests, Wilcoxon signed ranks tests, linear regressions, analysis of variance (ANOVA), principal components analysis (PCA) and cluster analysis. Among the intrinsic factors tested, species identity explained approximately 20% of growth rates at the community level, while tree diameter explained less of growth variation, and growth form had a minor influence on growth. Growth rates also were examined for variations across the three census intervals, and for relationships with rainfall and survival from fire. Statistical tests included t-tests, Wilcoxon and other non-parametric sign tests, logistic regression and ANOVA. Most species and individuals showed significant reductions of growth in the second census interval (1992-1996), and growth rates of most trees were positively related to rainfall. Growth rate variations generally were not related to survival from fire, and few species were capable of escaping fire mortality by fast growth. Spatial environmental influences were tested in the commonest fifteen species, using five habitat categories, local elevation, slope, aspect, and the biotic neighbourhood variables of local conspecific and heterospecific density. Statistical tests included analysis of covariance, multiple linear regression and redundancy analysis. The tests were quadrat-based or individual-based, and species' growth responses were tested at different levels of distance and spatial scale. Topographic features and habitat categories had ephemeral effects on species growth. Only the most dominant species, Lagerstroemia microcarpa, showed consistent conspecific neighbour density effects. Redundancy analysis using a subset of common species and environmental factors did not reveal common growth responses to spatial environmental factors. Comparison of factors influencing growth at Mudumalai versus at BCI using multiple factor ANOVA and multiple linear regressions showed a similar influence of temporal variation at the two sites, but stronger and more widespread influence of tree size (diameter) at BCI. The greater influence of tree size at BCI may be related to greater light limitation in this dense moist forest. Spatial environmental factors had weak influences at both plots. Species were less differentiated from each other at the more diverse BCI plot compared to the relatively species-poor Mudumalai plot, suggesting that species' growth niches may be weakly related to diversity across tropical forests. Overall the results showed that among the factors tested species identity and census intervals were the most important influences on diameter growth at the Mudumalai dry deciduous forest. Tree diameter was less important and less consistent in affecting growth at the Mudumalai dry forest, contrary to expectations based on moist tropical forests where this relationship has been established previously. When comparing Mudumalai and BCI, the relative importance of different factors was different at the two sites, and the most important difference was a dominant influence of light limitation at the wetter forest in Panama. In terms of management applications, this study showed that fires at Mudumalai might be an inescapable source of mortality for many vulnerable species, and improved fire management is crucial for long term survival of species in this dry forest. At a larger scale, light and other environmental variables were found to influence growth differently at Mudumalai compared to BCI. This suggests that location-specific responses may be important for projections of tree biomass and carbon sequestration, especially under future climatic change scenarios.

Tropical forests are storehouses of more thanhalf of the world‘s biodiversity and play a key role in global carbon, water and energy cycles. However, as a consequence of rapid anthropogenic climate change, biodiversity and climate functions of these forests are under a threat. Climate is changing not only in mean state but its variability is increasing, with extreme events such as droughts, heat waves and storms also rising. Water is fundamental to plants‘ existence, and in the tropics, is a key determinant of plant species‘richness, composition, growth and survival. There is thus an increasing interest in understanding how changing rainfall may cause functional changes in forests or change their species composition. Therefore, the overarching goal of thisdissertation was to understand the impact of water variability on tropical forest tree growth and vulnerability to drought.Forest tree growth along spatial and temporal rainfall gradientsObservational studies that measure whole forest tree growth along spatial or temporal gradients of rainfall are the most common way of formulating forest growth response curves to water availability, when manipulative experiments are cost-prohibitive or impractical (fire or large mammal disturbance). In the tropics, since very few species show anatomically distinct tree rings, estimating tree growth from trunk diameter is the standard practice to obtain growth patterns across species. However, this method—of equating woody growth to diameter change--is susceptible to bias from water-induced stem flexing. In the absence of bias correction, temporal variability in growth is likely to be overestimated and incorrectly attributed to fluctuations in resource availability, especially in forests with high seasonal and inter-annual variability in water. This problem has been largely ignored in the absence of any corrective measure and due to under-appreciation of the magnitude of error. While diameter re-censuses in permanent sampling plots (PSPs) have been most commonly done at 3-5 year scale (using a graduate tape), increasingly they are done at seasonal and annual scales (using band dendrometers) to closely match variation in rainfall, the scales at which hydrostatic bias may be greater in magnitude relative to woody growth. Besides, along a spatial rainfall gradient, inter-annual variability in water may vary, causing systematic differences in the hydrostatic bias for forests along the gradient. Therefore, one broad objective of this thesis was to evaluate the problem of hydrostatic bias in whole forest growth-rainfall relationship at annual and supra-annual scales, for temporal as well as spatial rainfall gradients and propose and test a novel corrective solution.Further, it also examines if growth-diameter relationship vary along the spatial gradient, which it may arise due to differences in light environments and/or disturbance history and species composition. The missing link of Eco-hydrology Differential responses of tree species in terms of growth and survival to variation in water that they can access, the proximate cause is likely shaped through their life-history strategies, the ultimate cause. However, we neither know the depths at which the diverse tree species in a forest draw water from and its dynamics, nor variation in water at those depths vis-à-vis rainfall patterns—for lack of appropriate methods. This has been a key missing link in understanding how water shapes trees‘ life-history strategies, their demographic trade-offs and co-existence, and also our predictive ability to determine species-specific responses to changing rainfall patterns, especially droughts. Since droughts are highly stochastic events and trees‘ responses to their drought ―experiences‖ may be revealed at decadal scales, long-term evaluations are key. Therefore, the second broad objective of this thesis was to develop a framework to determine trees’ water uptake depths, variation in water availability at those depths and trees’ demographic responses over multiple decades. From this, to understand how belowground hydrology shapes drought-vulnerability, demographic trade-offs and coexistence of forest tree species. This thesis titled—Eco-Hydrology of a Seasonally Dry Tropical Forest: Tree Growth, Belowground Water Dynamics and Drought-Vulnerability—is organized as follows: Chapter 1 lays down an introduction to the thesis, followed by a description of the study site and datasets used in the thesis in Chapter 2. This thesis uses a variety of methods and multiple datasets, all of which are from the protected Seasonally Dry Tropical Forests of the Western Ghats in southern India in the Mudumalai and Bandipur National Parks. It is then followed by three data chapters: Chapter 3 describes the seasonal fluctuations in a five year long (1980-1985) tree diameter time series (using dendrometers) of a Seasonally Dry Tropical Forest in Bandipur National Park to illustrate the issue of hydrostatic stem-flexing. It investigates the possibility that band dendrometers may themselves underestimate stem shrinkage at diurnal or seasonal scale. It also evaluates if there could be a best season and time of the day for undertaking forest diameter censuses that can minimize hydrostatic bias. Chapter 4(published in Forest Ecology and Management)measures the hydrostatic bias in a sample of trees in a 50 ha PSP of a Seasonally Dry Tropical Forest in Mudumalai National Park, and proposes a novel way to correct this bias at the whole community level in the 20 year long 4-year interval growth time series. Chapter 5 (in review with Environmental Research Letters) investigates and presents two new confounding factors in growth-rainfall relationships along a spatial rainfall gradient: hydrostatic bias and size-dependency in growth rates. For this it evaluates forest tree growth estimates in seven 1-ha PSPs (~800 trees, 3-year annual time series 9using dendrometers) along a 1000 mm rainfall gradient spanning a mesic savanna-moist forest transition in Mudumalai National Park. Using the period for which seasonal diameter time series was available (2 yrs), it evaluates if the extent of seasonal fluctuations systematically vary along the gradient—most likely due to hydrostatic stem flexing. It also describes the presence of an anomalous size-diameter relationship in the mesic savanna from a large plots (50 ha PSP, diameter records using graduated tape). These observations are then used to draw insights for ―space for time‖ substitution modeling. Chapter 6 (in prep for Nature Plants) analyses belowground water environments of trees over two decades (1992-2012), a period that includes a prolonged and intense drought, in the 50 ha PSP of a Seasonally Dry Tropical Forest in Mudumalai. It uses a locally parametarised dynamic hydrological model in which site rainfall is also a forcing variable. It then develops a novel dynamic growth model and inversely estimates water uptake depths for adult trees of all common species (include ~9000 trees) in the PSP from their above-ground growth patterns over two decades vis-à-vis belowground water availability at multiple depths. It then examines if species‘ water uptake depth obtained thus is a predictor of their drought-driven mortality. Finally, this is used to evaluate the hydrological niche partitioning tree species operate under and how that drives their water uptake strategies, demographic trade-offs, and drought-vulnerability. Summarizes the thesis and suggests future directions