Mapping of Eucalyptus in Natura 2000 Areas Using Sentinel 2 Imagery and Artificial Neural Networks

Abstract

<jats:p>Plantations of fast-growing Eucalyptus trees have become a common sight in the western Iberian peninsula where they are planted to exploit their economic potential. Negative side-effects of large scale plantations including the invasive behavior of Eucalyptus trees outside of regular plantations have become apparent. This study uses medium resolution, multi-spectral imagery of the Sentinel 2 satellites to map Eucalyptus across Portugal and parts of Spain with a focus on Natura 2000 areas inside Portugal, that are protected under the European birds and habitats directives. This method enables the detection of small incipient as well as mixed populations outside of regular plantations. Ground truth maps were compiled using field surveys as well as high resolution satellite imagery and were used to train Feedforward Neural Networks. These models predict Eucalyptus tree cover with a sensitivity of up to 75.7% as well as a specificity of up to 95.8%. The overall accuracy of the prediction is 92.5%. A qualitative assessment of Natura 2000 areas in Portugal has been performed and 15 areas have been found to be affected by Eucalyptus of which 9 are strongly affected. This study demonstrates the applicability of multi-spectral imagery for tree-species classification and invasive species control. It provides a probability-map of Eucalyptus tree cover for the western Iberian peninsula with 10 m spatial resolution and shows the need for monitoring of Eucalyptus in protected areas.</jats:p>

Data fusion approach for eucalyptus trees identification

“…Table 4 presents the confusion matrix for this tile along with the overall accuracy, user's accuracy, producer's accuracy, calculated as in (Olofsson et al 2014). These results are at a similar level of the ones reported recently by (Forstmaier, Shekhar, and Chen 2020), using an approach based on Artificial Neural Networks (lower sensitivity: 69.61% compared to 75.70%, but higher specificity and overall accuracy, 99.43% compared to 95.80% and 98.19% compared to 92.50%).…”

“…The most recent work used Artificial Neural Networks to identify eucalyptus trees across Portugal and parts of Spain, directly using all the multispectral bands coming from images acquired by Sentinel 2 with a sensitivity of up to 75.7% as well as a specificity of up to 95.8% (Forstmaier, Shekhar, and Chen 2020). The overall accuracy of the prediction is 92.5% (Forstmaier, Shekhar, and Chen 2020).…”

“…The most recent work used Artificial Neural Networks to identify eucalyptus trees across Portugal and parts of Spain, directly using all the multispectral bands coming from images acquired by Sentinel 2 with a sensitivity of up to 75.7% as well as a specificity of up to 95.8% (Forstmaier, Shekhar, and Chen 2020). The overall accuracy of the prediction is 92.5% (Forstmaier, Shekhar, and Chen 2020). The main novelty of this paper over this recent work is the fusion of the vegetation indices with the information obtained directly from Sentinel 2, while the previously mentioned work (Forstmaier, Shekhar, and Chen 2020) only used the direct information from the images as an input.…”

Capturing the Impact of the 2018 European Drought and Heat across Different Vegetation Types Using OCO-2 Solar-Induced Fluorescence

“…In contrast to NDVI over the four years period, SIF was considerably lower in 2018 compared to the already low mean of the previous three years (2015-2017), which may indicate its higher sensitivity to both physiological and structural drought response from plants [32]. This finding highlights the predominance of SIF over NDVI even if short reference periods are available, for example, SIF available from OCO-2, the upcoming OCO-3, and the TROPOMI in comparison to ESA's SENTINEL-2A [78] (for NDVI).…”

Machine Learning Classification of Mediterranean Forest Habitats in Google Earth Engine Based on Seasonal Sentinel-2 Time-Series and Input Image Composition Optimisation

“…To date, GEE has been widely used for different mapping purposes; primarily to exploit its massive catalogue of images to capture time-series data or derive useful information to analyse phenomena over a long period. The capabilities of GEE, jointly with external software and applications, have been widely explored using different data and algorithms for a wide range of applications, such as forest mapping [26,54,55], LU/LC classification [56,57], fire severity analysis [58], forest disturbance mapping [59], forest defoliation assessment [60], surface water detection [61], mine mapping [62], snow [63] and shoreline detection [64], urban and rural settlement mapping [65,66], and species habitat monitoring [67].…”