Ventos extremos como mecanismo de distúrbio em florestas da Amazônia Central: padrões de mortalidade de árvores e variação sazonal e topográfica de rajadas de vento extremo

Abstract

Storms propagating extreme winds are an important mechanism of tree mortality in Amazonian forests, influencing structure, species composition, biomass and carbon. Remote sensing and forest inventory have been combined for decades to access forest damage patterns in disturbances ranging from individual trees to large catastrophic events. Landsat has been widely used due to its large spatial and temporal coverage, favoring landscape-scale analysis. However, its medium spatial resolution makes it difficult to detect small groups of impacted trees. In the last decade, high-resolution satellite launches have improved the detection of small-scale forest damage caused by wind. However, their images have limited spatial and temporal coverage, are expensive and have not yet been tested on large-scale disturbances. This in turn limits our understanding of the benefits of spatial resolution in damage estimates that transcend local to regional scales. On the other hand, much of what we know about potentially destructive extreme wind patterns is evidenced by their damage to trees, as few locations have meteorological records. This implies that the patterns of occurrence and actual severity of extreme winds remain poorly understood, which hampers our ability to reliably predict their occurrence and severity at local and regional scales. This thesis is structured in two chapters. Chapter 1 aims to assess the effect of the spatial resolution of different optical satellites in estimating tree mortality in a major catastrophic event. Chapter 2 aims to document the local patterns of wind gusts with destructive potential (WPD) for trees along a topographic and seasonal gradient and quantify how rainfall intensity affects these patterns using meteorological data at canopy height. In chapter 1, Landsat 8 (30 m spatial resolution) presented the most reliable estimate in relation to mortality recorded in the field compared to Sentinel 2 (10 m) and WorldView 2 (2 m), respectively. Landsat 8 had a resolution closer to the level of detail of the damage usually recorded by forest inventories. High-resolution images can improve mortality estimates, but it is necessary for crown damage and the rapid growth of natural regeneration to be quantified along the entire gradient of disturbance severity. In Chapter 2, during 24 months of monitoring, 424 WPD were recorded in every month of the year, but these bursts were more frequent, faster and longer-lasting in the hottest hours of the day and in the transition from the dry to the rainy season. Higher and more exposed areas were particularly vulnerable to WPD. The 87th percentile of rainfall (~0.7 mm min-1) showed the strongest association with the frequency and duration of WPD, highlighting the role of extreme storms in the spread of destructive winds. This research shows that forest damage varies locally, which implies that reliable estimates of tree mortality and the establishment of robust landscape-scale WPD patterns are not trivial. This highlights the importance of local disturbance patterns in landscape-scale damage estimates, improving our knowledge of the processes that regulate forest dynamics and carbon balance in the Amazon.