Close Window

Landscape Disturbance from Hyperspectral Imagery and Field-Based Tree Mortality Estimates for a Central Amazon Forest

Jeffrey Q Chambers, Tulane University, chambers@tulane.edu (Presenting)
Giuliano Guimar„es, INPA, gg77@ig.com.br
Amanda Sesser, Tulane University, asesser@tulane.edu
Vilany Carneiro, INPA, matillav@hotmail.com
Niro Higuchi, INPA, niro@inpa.gov.br

Tree mortality events impact a number of ecosystem attributes in tropical forests including tree species recruitment dynamics and landscape carbon balance. Mortality rates in the Central Amazon vary from the slow death of standing tree to the instant razing of large areas from microburst blowdown events associated with strong convective storms. Most tropical forest mortality studies are logistically limited to relatively small spatial scales, with observed intensity rarely exceeding 0.5% (stems ha/yr), and event (gap) size generally limited to < 0.10 ha per event. This study takes advantage of a number of scattered blowdowns that occurred in a Central Amazon forest in 2005. A 5000 ha area from the impacted region was extracted from a Hyperion image, and spectral mixture analysis (SMA) methods were used to classify the landscape into five disturbance classes. A total of 30 plots of 20x20 m each, distributed among the disturbance classes, were randomly located across a large portion of this area. Tree mortality rates within these plots were quantified in the field and varied across a gradient of most to least disturbed plots. Tree mortality rates were correlated with independently calculated SMA disturbance metrics. Scaling this relationship across the 5000 ha area allowed development of a unique landscape-scale tree mortality rate map. This disturbance map was used to develop a probability distribution function to aid in modeling landscape carbon balance.

Science Theme:  CD (Carbon Dynamics)

Session:  3C: Vegetation Structure and Disturbance

Presentation Type:  Oral

Abstract ID: 36

View presentation

Close Window