How does non-random mortality change landscape estimation of biomass lost from catastrophic wind disturbances in the Peruvian Amazon?

Background/Question/Methods

Meteorological downbursts and squall line storms constitute a major form of natural forest disturbance in the northwestern Amazon basin. The blowdowns created by these catastrophic disturbance events span orders of magnitude in both spatial extent, and number of trees killed. In the temperate forests of North America, wind disturbances have been shown to kill trees selectively but there is little evidence regarding selective mortality from wind disturbance in tropical forests. If tree mortality is selective, this may alter landscape carbon loss as mortality is usually modeled as a random process in carbon models. To investigate non-random mortality, we installed 3-ha of forest inventory transect plots across a recent ~300-ha blowdown in the department of Loreto, Perú. A remote sensing metric derived from a time series of Landsat imagery, ΔNPV, has proven to be an effective predictor of the fraction mortality. Previous tree mortality and corresponding necromass estimates from blowdowns have been predicted with a linear model using ΔNPV with a random tree mortality component. We compared this approach with a probabilistic individual tree simulation, where the probability of death was affected by the tree’s diameter, wood density, neighborhood ΔNPV, and elevation.    

Results/Conclusions

Similar to blowdowns in North America, we found that patterns of tree mortality within a large scale blowdown in the Peruvian Amazon were highly non-random with larger trees and trees at higher topographic positions more likely to die. Smaller trees, trees with high wood density, and trees in topographic depressions were least likely to be killed. Small trees tended to be killed only when their immediate neighbors were also killed from the storm. Simulations using selective tree mortality suggest that 23% more dead trees and roughly 50% more necromass were produced from the disturbance in comparison to a model using random mortality. The necromass produced was much higher because of the increased probability of mortality in large trees which have disproportionately more biomass than smaller trees. Apart from the large impact upon necromass estimation, accounting for the identity of the killed trees can predict a more realistic post wind disturbance forest structure which has the potential to strongly alter the subsequent forest succession.