Estimating tree abundance from remotely sensed imagery in semi-arid and arid environments: bringing small trees to the light

Abstract

The analysis of remotely sensed images provides a powerful method for estimating tree abundance. However, a number of trees have sizes that are below the spatial resolution of remote sensing images, and as a result they cannot be observed and classified. We propose a method for estimating the number of such sub-resolution trees on forest stands. The method is based on a backwards extrapolation of the size-class distribution of trees as observed from the remotely sensed images. We apply our method to a tree database containing around 13,000 tree individuals to determine the number of sub-resolution trees. While the proposed method is formulated for estimating tree abundance from remotely sensed images, it is generally applicable to any database containing tree canopy surface area data with a minimum size cut-off.

Characterising forest spatial structure through inhomogeneous second order characteristics

“…Distance-dependent tree-level models (i.e. trees are the basic unit of analysis) not only improve the predictive power of these formulations and permit to analyse spatially explicit silvicultural problems such as plantation and thinning strategies, but also allow the study of complex forest dynamics (Miina et al 1991;Pukkala et al 1998;Moustakas and Hristopulos 2007;Renshaw et al 2008). These models are also necessary to generate spatially explicit forest patterns (realistic synthetic data), to simulate and study realistic silvicultural operations such as thinning and regeneration strategies, and to compare inventory designs.…”