Neotropical lowland forests along environmental gradients

Neotropical lowlands feature an extraordinary display of vegetation types. This is especially the case for Bolivia where three biogeographical regions, Amazonian, Brazilian-Paranaense and Gran Chaco meet in the lowland areas, providing thus an ideal setting to study vegetation-environment relationships. Understanding spatial patterns of tropical forests and the environmental factors determining these patterns is important for forest management and for predicting responses of forests to climate change. Thus, the main objective of this dissertation was to evaluate how environmental factors shape tropical lowland forests in Bolivia. Specifically it assessed how climatic and edaphic factors affect 1) forest structure, 2) floristic composition, 3) tree growth rates, and 4) species distribution. Additionally, it assessed how disturbance factors affect tree growth rates. For this research, I used a network of 220 1-ha permanent sample plots distributed along environmental gradients. For each plot, all stems  10 cm diameter were identified, evaluated and monitored; climatic data were interpolated from weather stations and soil samples were collected. In lowland Bolivia, seasonality increased from north to south. Rainfall decreases, and dry season length increases, along this gradient. Although in general the drier forest had more fertile soils than moister forests, some plots in moister forests were also fertile. A set of five climatic variables [annual temperature, annual precipitation, total precipitation of the three driest months, length of the dry period (# months < 100 mm), and length of the drought period (# months < 50 mm)] and 12 edaphic variables [cation exchange capacity, cations (Ca2+, Mg2+, K+ and Na+), Olsen phosphorous, organic matter, nitrogen, acidity and percentage of different particles (sand, silt and clay)], were summarized into four environmental gradient axes using a Principal Component Analysis (PCA). I used the two main climatic PCA axes (named after the most important factors they represent as “rainfall” and “temperature”) and the two soil PCA axes (named as “soil fertility” and “soil texture”) to reduce the number of highly correlated variables and to have composite variables that summarize the main environmental gradients. Rainfall axis, for example, represents a seasonality gradient. Finally, a stepwise selection approach was used to determine how the climatic and edaphic PCA axes affected the plant community. Structural attributes of the tropical forests differ along gradients. In Chapter 2, I described how forest structure varies among forests across lowland Bolivia. I considered 15 forest structural variables based on height, crown position, diameter, and liana infestation of each stem. I tested the hypothesis that stem density and basal area of trees and palms will increase with water availability and liana density will increase in drier forests. My results showed that tree maximum height, palm density and basal area increased with rainfall while lianas decreased with rainfall. While forest height and lianas were more affected by soil texture, palm density was negatively affected by soil fertility. Surprisingly, tree basal area was not affected by environmental factors. I found that rainfall, temperature and soil texture were more important drivers of forest structure than soil fertility. Thus climatic and edaphic factors have strong effects on variation in forest structure at the landscape-scale. Only recently researchers have started to examine the influence of environmental factors on species composition on a regional scale. In Chapter 3, I evaluated patterns in floristic composition using abundance and presence-absence data of 100 plant species. I predicted that climate is a more important factor than soil in shaping floristic composition. In line with my hypothesis the climatic gradient shaped the floristic variation more strongly than the edaphic gradient. Detrended Correspondence Analysis ordination divided lowland Bolivia primarily into two major groups (Southern Chiquitano versus Amazonian regions) and Multiple Response Permutation Procedure distinguished five floristic regions: Northern Amazon, Western Preandean, Eastern Amazon - Bajo Paraguá, Eastern Amazon - Guarayos, and Southern Chiquitano. In addition, all the environmental variables tested were significantly different among the floristic regions. Of the 100 species, 10 occurred in only one floristic region and 90 occurred in two or more floristic regions. I also distinguished 92 strong indicator species, which had significant environmental preferences for one floristic region, so these species can be indicators of environmental conditions. Given the predicted decreases in rainfall and increases in temperature for lowland forests, our gradient approach suggests that species composition may drastically shift with climate change. Most of the forest dynamic studies have evaluated the effects of temporal variation of rainfall on tree growth rates, but spatial variation on growth rates along gradients is less known. Thus, in Chapter 4, I described the variation in tree growth by examining growth rates at individual level (average diameter growth) and at stand level (basal area growth) across 165 plots, of which 85 were affected by logging. I expected that growth rates would be higher in humid than in dry sites, higher in nutrient-rich than nutrient-poor forests, and higher in logged than non-logged forests because of an increase in water, nutrients and light, respectively. I found positive basal area increases at the stand level which agrees with the generally reported biomass increases in tropical forests. Multiple regression analysis demonstrated that environmental and disturbance factors significantly explained the high variation in growth rates. While rainfall and temperature had positive effects on tree growth, no clear effects with soil fertility were found. Probably our soil fertility was not large enough to detect effects on tree growth, or nutrients may also be available for plant growth from other sources than soil alone. Growth rates increased in logged plots, especially those which had a high logging impact. Future decreases in rainfall and increases in temperature due to climate change can also affect growth rates. The negative effects of increased seasonality, however, may be partly offset by the positive effects of temperature on tree growth. Forest managers should take into account the high variation in growth rates occurring in the lowland forests of Bolivia. Based on the results I advocate that management practices to be developed are specific to each forest and are in line with its characteristics and conditions. Ecologists have found different response curve shapes for species distribution, mostly for temperate species, but data in tropical species responses are surprisingly scarce. Therefore, in Chapter 5, I analyzed the distribution patterns of the 100 selected species and constructed response curves for each species against each of the four environmental gradient axes using logistic regression analysis on presence-absence data. I hypothesized that species frequency and abundance would be positively correlated, that the majority of the species would show unimodal response curves and that species would respond stronger to climate than to soil effects. I indeed found a positive trend between species abundance and occurrence but some abundant species were also narrowly distributed. While 25 species showed unimodal response curves to the rainfall gradient, and 10 species to temperature, only three species showed such response to soil fertility and none to soil texture. Probably, the sampled environmental gradient is not sufficiently large to find a higher number of unimodal responses. In line with my hypothesis, 91% of the species were affected by climatic factors and only 47% of the species were affected by soil factors. These results agree with the notion that species response types to environmental gradients will differ among species and among factors considered. Thus, multiple, rather than single, environmental factors must be used to explain the species distribution in tropical forests. In conclusion, this dissertation documented the high variation of tropical lowland forests in Bolivia and indicated that climate (i.e. rainfall and temperature) was the most important factor shaping forest structure, composition and dynamics. The high variation of forests and the ecological differences among regions have to be taken into account when developing forest-specific management plans. Finally, the results of the gradient approach suggest that with future decreases in rainfall and increases in temperature, due to climate change, drastic shifts can be expected in forest structure, composition and dynamics in these tropical lowland forests.