Discriminating between Native Norway Spruce and Invasive Sitka Spruce—A Comparison of Multitemporal Landsat 8 Imagery, Aerial Images and Airborne Laser Scanner Data

Marius Hauglin, Hans Ørka
2016 Remote Sensing  
Invasive species can be considered a threat to biodiversity, and remote sensing has been proposed as a tool for detection and monitoring of invasive species. In this study, we test the ability to discriminate between two tree species of the same genera, using data from Landsat 8 satellite imagery, aerial images, and airborne laser scanning. Ground observations from forest stands dominated by either Norway spruce (Picea abies) or Sitka spruce (Picea sitchensis) were coupled with variables
more » ... from each of the three sets of remote sensing data. Random forest, support vector machine, and logistic regression classification models were fit to the data, and the classification accuracy tested by performing a cross-validation. Classification accuracies were compared for different combinations of remote sensing data and classification methods. The overall classification accuracy varied from 0.53 to 0.79, with the highest accuracy obtained using logistic regression with a combination of data derived from Landsat imagery and aerial images. The corresponding kappa value was 0.58. The contribution to the classification accuracy from using airborne data in addition to Landsat imagery was not substantial in this study. The classification accuracy varied between models using data from individual Landsat images. stands dominated by the introduced species. In addition to the planted stands, scattered occurrences also exist, most likely due to natural dispersion from the planted stands. The direct spread of non-native species will typically occur within a given distance from an initial location, with the distance determined by the characteristics of the specific species, wind, and other factors. From a management perspective, it is desirable to map occurrences of non-native species and to establish systems to monitor further expansion. Reliable mapping of individual stands or single trees through manual field surveys can be time-consuming and costly; the development of methods using remote sensing data is therefore suggested [2] . To be able to identify non-native species through remote sensing, it is required that there are features which set the non-native vegetation apart from the native vegetation, and that these features are directly or indirectly contained in the remote sensing data. One typical-and important-example is spectral information; how vegetation reflects light and other types of electromagnetic radiation depends on a range of factors, including species or species composition [3] . The spectral information may therefore differ between species-such as between Norway spruce and Sitka spruce-and spectral data from aerial or satellite imagery can be used to map vegetation (e.g., in [4]). Three-dimensional remote sensing data-such as data from airborne laser scanning (ALS)-contains information on the spatial structure of the vegetation and can further contribute to a discrimination between species or between vegetation types [5] . In the case of non-native species, this could, for example, mean the ability to identify which species have a diverting crown shape or which form stands with an atypical spatial structure. Since the 2000s and onward, there have been several studies on the detection of non-native and invasive species using remote sensing data. There have also been studies which use similar methods to map and classify different native species and vegetation types. A review of some of these studies is as follows. Carter et al. [6] used multispectral (Landsat 5) and hyperspectral (Hyperion) medium spatial resolution images to classify tamarisk (Tamarix spp.) in North America. They concluded that the high spectral resolution of Hyperion provided an increase in the accuracy of percentage points over the multispectral alternative. The accuracies obtained for this classification were between 80% and 88% in terms of overall accuracy, but commission (false positive) errors were also high at 62%-83%. Fuller [7] used spectral features derived from multispectral satellite images with a spatial resolution of 4 m to detect areas dominated by melaleuca (Melaleuca quinquenervia)-an invasive tree species-in Florida. The results showed that large and dens stands of the invasive species were reliably detected, whereas the method and data were to a lesser degree suitable for detection of smaller groups or single trees. This demonstrates the important relationship between data resolution and the size of detectable objects. Bradley [2] notes-based on results from the reviewed studies-that "detection of more heavily invaded areas seems to be most promising." Asner et al. [8] combined data from airborne ALS and hyperspectral imagery to identify an invasive tree species (Morella faya) in Hawaii. In that study, they found that the spectral signature of the non-native tree differed from the native vegetation. This enabled an identification of areas with occurrences of the invasive species. Classification can be enhanced by acquiring remote sensing data at specific phenological stages where the non-native species can be separated from native vegetation. For example, Resasco et al. [9] evaluated Landsat imagery from different time periods over the year and found better classifications of under-story shrub using leaf-off imagery from a specific time period. A detailed introduction and discussion of the use of remote sensing data for identification of invasive trees and plants can be found in Bradley [2] and Huang and Asner [10] . Both airborne and spaceborne remote sensing data have been used to map and classify vegetation. The main advantage of spaceborne remote sensing data is the availability-in terms of both coverage and costs. Satellite imagery such as the Landsat products are freely available and cover most areas of the globe with multiple acquisitions annually. The disadvantage of satellite imagery of this type is a
doi:10.3390/rs8050363 fatcat:by5dackkjfbwpkzfz7x6l7kcsi