AMSADE – Employing Machine Learning for AMbrosia SAtellite DEtection using Earth Observation Data

Ambrosia is so widespread in Europe that its complete eradication is no longer practical or economically feasible. In the Republic of Croatia, it is especially represented in the area of Zagreb, Poreč and Slavonia. If preventive methods of ambrosia control fail and ambrosia colonizes a new area, early detection is essential for rapid removal and prevention of further spread. It is important to emphasize the benefits of using space technology and remote sensing in support of ambrosia detection, prediction and control.

Ambrosia is an annual allergenic plant, which spreads very quickly and occupies new areas. It is an invasive species, which is today practically spread all over the world and represents a major economic as well as public health problem. It produces a large amount of extremely allergenic pollen grains, which cause significant health problems for an increasing number of people, while also creating significant economic costs. In addition, ambrosia suppresses native flora and reduces natural diversity, and has become a major weed in European agriculture.

Suppression of ambrosia is a difficult and long-term process, and the measures of suppression are limited by the means of classic field inspection and the will of citizens to eradicate it. Its removal must be carried out before the beginning of flowering, and it is often not removed completely and continues to spread. Numerous guidelines, declarations and acts have been adopted at the global and European level for the suppression of ambrosia. They are also accepted in Croatia.

The Ministry of Agriculture, Forestry and Water Management issued the “Order on taking measures for the mandatory removal of ambrosia – Ambrosia artemisiifolia L.” (NN 72/07), according to which the owners, possessors and users of cultivated and uncultivated agricultural land, forests and hunting grounds, are authorized to manage watercourses and canals and areas next to watercourses and canals, authorized persons for the management and maintenance of areas along public roads and railways, authorized persons for the management of parks and other public green areas are obliged to carry out measures to remove ambrosia, wherever they find it, otherwise they face a fine.

To optimize field inspection routes, a system will be developed to detect locations of ambrosia growth. This system will allow planning routes for inspection and removal of ambrosia. Earth observation data, machine learning methods, field data on ambrosia growth locations, ground truth data collected in the field and phenological data will be used.

Such a system would simplify the supervision and control of ambrosia suppression for public bodies and institutions, enabling a timely reaction and taking action against owners or users who did not perform proper removal. With high accuracy and reliability of the results, this solution brings opportunities to reduce the costs of physical verification in the field.

The final result of the project will be a web GIS application that will visualize locations where ambrosia grows in order to automate surveillance and monitoring in the area of ​​Zagreb and Poreč based on satellite data, machine learning models and field data.

The project is financed by the European Space Agency (ESA). The City of Zagreb and Institute for Agriculture and Tourism Poreč will provide historical data on the locations of ambrosia growth in the area of Zagreb and Poreč, whileVesela Motika d.o.o. will conduct field collection of ambrosia growth locations during flowering. The project start date is December 1, 2023, and the expected completion date is December 1, 2024.