Assessment of Acacia Koa Forest Health across Environmental Gradients in Hawai‘i Using Fine Resolution Remote Sensing and GIS

Abstract

Koa (Acacia koa) forests are found across broad environmental gradients in the Hawai‘ian Islands. Previous studies have identified koa forest health problems and dieback at the plot level, but landscape level patterns remain unstudied. The availability of high-resolution satellite images from the new GeoEye1 satellite offers the opportunity to conduct landscape-level assessments of forest health. The goal of this study was to develop integrated remote sensing and geographic information systems (GIS) methodologies to characterize the health of koa forests and model the spatial distribution and variability of koa forest dieback patterns across an elevation range of 600–1,000 m asl in the island of Kaua‘i, which correspond to gradients of temperature and rainfall ranging from 17–20 °C mean annual temperature and 750–1,500 mm mean annual precipitation. GeoEye1 satellite imagery of koa stands was analyzed using supervised classification techniques based on the analysis of 0.5-m pixel multispectral bands. There was clear differentiation of native koa forest from areas dominated by introduced tree species and differentiation of healthy koa stands from those exhibiting dieback symptoms. The area ratio of healthy koa to koa dieback corresponded linearly to changes in temperature across the environmental gradient, with koa dieback at higher relative abundance in warmer areas. A landscape-scale map of healthy koa forest and dieback distribution demonstrated both the general trend with elevation and the small-scale heterogeneity that exists within particular elevations. The application of these classification techniques with fine spatial resolution imagery can improve the accuracy of koa forest inventory and mapping across the islands of Hawai‘i. Such findings should also improve ecological restoration, conservation and silviculture of this important native tree species.

Precommercial release thinning of potential Acacia koa crop trees increases stem and crown growth in dense, 8-year-old stands in Hawaii

“…Other studies with koa have demonstrated relationships of ARVI with crown dieback or overall koa productivity (Morales et al, 2011;Morales et al, 2012); thus, across diverse stands or landscapes, ARVI is capable of identifying healthier, more productive trees and stands. Posttreatment determination of VIs may indicate improved canopy function, but this requires verification with direct measurements.…”

“…As the number of extracted pixel values per tree was highly variable, single-tree average reflectance was calculated from the visible and NIR bands. This information was used to calculate vegetation indices (VIs) previously used in studies of koa forest canopies (Morales et al, 2011(Morales et al, , 2012) () () () (). The atmospherically reflective vegetation index (ARVI) was chosen as the best VI for distinguishing koa tree crowns.…”

“…Koa is a fast-growing, shade intolerant, nitrogen-fixing tree that grows from near sea level to 2000 m elevation in mesic sites across a range of soil types on several of the main Hawaiian Islands. Mature trees can grow to over 25 m in height with stem diameter at breast height (DBH) greater than 1.5 m (Friday, 2011). The hard-coated seed can remain viable in the soil for many years, even after deforestation (Baker et al, 2009).…”

“…Where seed banks are lacking, koa seedlings are planted at densities of approx. 1000 trees ha -1 (Friday, 2011). Both methods of stand establishment lead to early canopy closure and intense intraspecific competition.…”

Using Remotely Sensed Imagery for Forest Resource Assessment and Inventory

“…While the high spatial resolution allows for delineation of single tree crowns, the multispectral bands allow for determination of variations of canopy greenness within forest stands (Guo et al, 2007). In particular, these satellites have been successfully applied for forest inventory in www.intechopen.com tropical environments and allowed for the mapping of tree crown sizes (Martinez Morales et al, 2008), tree density, species identification, and assessment of temporal changes in individual tree growth and mortality (Clark et al, 2004;Martinez Morales et al, 2011 While high spatial resolution satellite sensors can be used to assess forest structural characteristics, they only collect data on a limited number of spectral bands (blue, green, red, and near-infrared). Hyperspectral remote sensing, or imaging spectroscopy, collects data on hundreds of bands from visible to infrared wavelengths (0.4 to 2.5 µm).…”