Cristina Tarantino and Maria Adamo, IIA-CNR, Italy
A review of the ECOPOTENTIAL paper:C. Tarantino, F. Casella, M. Adamo, R. Lucas, C. Beierkuhnlein, P. Blonda. (2018). “Ailanthus altissima mapping from multi-temporal very high resolution satellite images”, ISPRS Journal of Photogrammetry and Remote Sensing, 147, 90-103, https://doi.org/10.1016/j.isprsjprs.2018.11.013.
Alien plants, also termed non-native or allochthon plants, can modify biodiversity and functioning of ecosystems causing their degradation by diminishing both abundance and survival of native species. On-going climatic and anthropic changes are making invasive species spreading a global issue whose monitoring and management would require powerful remote sensing data and techniques.CNR-IIA, in the framework of the ECOPOTENTIAL project, studied newsolutions for the automatic detection of Ailanthus altissima (Mill.) Swingle,one of the most widespread and harmful invasive plants in both the USA and Europe.
The study site is located in “Murgia Alta”, a Protected Area in Southern Italy which belongs to the European Natura 2000 network: here the invader, A. altissima,can grow mainly in semi-natural and natural grasslands fields due to shepherds’ practices abandonment or at the edges of cultivated fields, mainly herbaceous areas. Within such fields, the invader is generally controlled by the farmers through regular ploughing or other agricultural practices. Occasional fires occurring in this area can also be favoured by the proliferation of this plant.
The study introduces a novel investigation approach of the A. altissimaspecies by analysing multi-spectral and multi-temporal Very High Resolution (VHR) satellite data (i.e., WV-2, characterized by 8 spectral bands in the visible/near infrared electromagnetic spectrum and 2 meters spatial resolution) as an alternative to the unavailability of hyperspectral data from aerial campaigns (characterized by a finer spectral resolution essential for the discrimination at species level). The technique used relied on a two-stage hybrid classification process to obtain the A. altissimamapping: the first stage applied a knowledge-driven learning scheme to provide a land cover map (LC), including deciduous woody vegetation and other classes, without the need of reference training data; the second stage exploited a data-driven classification to: i) discriminate pixels of the invasive species found within the deciduous vegetation layer of the LC map (two-classes problem); ii) determine the most favourable seasons for such recognition.
Since training data to discriminate A. altissimafrom other deciduous plants are required only in the second stage, the use of the first stage LC mapping as pre-filter can reduce not only classification complexity but also time and costs involved by in-situ reference data collection.
The best encouraging Overall Accuracy (OA) value of 97.96% for the A. altissimamapping was obtained considering the July and October WV-2 images as input to the Support Vector Machine classifier in the second stage.
Although the methodology proposed and the data used would require further applications for the mapping of A. altissimaand other invasive species in different sites, the use of multi-temporal VHR data and the hybrid classification approach may offer new opportunities for invasive plant monitoring and follow up of management decisions.