Detection of Mortality in Tropical Forests Using Remote Sensing: From Treefall Gaps to Large Disturbances

The frequency, severity, and intensity of natural disturbances in tropical forests continually re-shape forest structure. At small scale, branch or tree-falls gaps and subsequent recovery are important mechanisms for carbon cycling. At landscape scale, large disturbances (blow-downs) may also play a role on the structure and composition of tropical forests. Quantitative studies of natural disturbances across the occurrence spectrum (branch fall-gaps to blow-downs) are rare for the Amazon. Remote sensing coupled with intense field work data collection provides the means to analyze the dynamic of tropical forests at multiple scales. In this dissertation three aspects of natural disturbances were examined: (1) formation and detection of small scale disturbances investigated in the field and with high resolution remote sensing; (2) mapping and spatial analysis of large disturbances (blow-downs) caused by convective cloud drafts; and (3) a quantitative characterization of the large spectrum of natural disturbances in tropical ecosystems. For small scale disturbances, two large plots of 114 and 53 ha were established and surveyed in unmanaged tropical forest of the Amazon. Data of gap area, canopy openness (CO), leaf area index (LAI), coarse woody debris (CWD) and tree mortality were collected in both plots. The relation between CO and LAI of gaps coupled with high resolution satellite images IKONOS-2 was investigated using geostatistics. Based on field plot measurements, tree-fall gaps account only about 30% of the flux of annual tree mortality. Most mortality does not result in gap formation. On average, gap formation accounted for a minor proportion of the stocks (about 5% of the total fallen CWD) and fluxes (about 23%) of CWD carbon. There was no significant correlation between remote sensing products and variables of CO and LAI in both large plots, probably due to high shadow fraction in high-resolution images. For large scale disturbances, a spatial pattern analysis of blow-downs apparently caused by severe storms was discovered using 27 Landsat images and daily precipitation from NOAA satellite data. In this image mosaic from 1999 to 2000, there are 279 patches (from 5 ha to 2,223 ha) characteristic of blow-downs. A total of 21,931 ha of forest were disturbed. There was a strong correlation between occurrence of blow-downs and frequency of heavy rainfall (Spearman\u27s rank, r2=0.84, p\u3c0.0003). We provided the first spectrum analysis of disturbances, at several scales of severity by combining data collected for this dissertation with previously published information from a variety of sources: (1) Large plot surveys; (2) mapped blow-downs; (3) RAINFOR permanent plots; (4) historical blow-downs ≥ 30 ha; and (5) the published mean above-ground biomass map of the Amazon. We found two disjointed disturbance regimes---small-scale tree-fall gaps and larger-scale blow-down disturbances---suggesting that there may be other missed disturbance mechanisms in the Amazon may also play a role in the dynamics of tropical forest ecosystems