The study of afforestation is crucial to monitor land transformations and represents a central topic in sustainable development procedures, in terms of climate change, ecosystem services monitoring, and planning policies activities. Although surveying afforestation is important, the assessment of the growing forests is difficult, since land cover has different durations depending on the species. In this context, remote sensing can be a valid instrument to evaluate the afforestation process. Nevertheless, while a vast literature on forest disturbance exists, only a few studies focus on afforestation and almost none directly exploits remote sensing data. This study aims to automatically classify non-forest, afforestation, and forest areas using remote sensing data. To this purpose, we constructed a reference dataset of 61 polygons that suffered a change from non-forest to forest in the period 1988-2020. The reference data were constructed with the Land Use Inventory of Italy and through photointerpretation of orthophotos (1988-2012, spatial resolution 50 × 50 cm) and very high-resolution images (2012-2020, spatial resolution 30 × 30 cm). Using Landsat Best Available Pixel composites time-series (1984-2020) we calculated 52 temporal predictors: four temporal metrics (median, standard deviation, Pearson's correlation coefficient R, and slope) calculated for 13 different bands (the six Landsat spectral bands, three Spectral Vegetation Indices, and four Tasseled Cap Indices). To verify the possibility of distinguishing afforestation from non-forest and forest, given the differences between them can be minimal, we tested four different models aiming at classifying the following categories: (i) non-forest/afforestation, (ii) afforestation/forest, (iii) non-forest/ forest and (iv) non-forest/afforestation/forest. Temporal predictors were used with random forest which was calibrated using random search, validated using k-fold Cross-Validation Overall Accuracy (OAcv), and further using out-ofbag independent data (OAoob). Results illustrate that the distinction of afforestation/forest reaches the largest OAcv (87%), followed by non-forest/forest (83%), non-forest/afforestation (75%) and non-forest/afforestation/forest (72%). The different OA values confirm that the difference in photosynthetic activity between forest and afforestation can be analysed through remote sensing to distinguish them. Although remote sensing data are currently not exploited to monitor afforestation areas our results suggest it may be a valid support for country-level monitoring and reporting.

In recent years, deep learning (DL) algorithms have been widely integrated for remote sensing image classification, but fewer studies have applied it for land consumption (LC). LC is the main factor in land transformation dynamics and it is the first cause of natural habitat loss; therefore, monitoring this phenomenon is extremely important for establishing effective policies and sustainable planning. This paper aims to test a DL algorithm on high-resolution aerial images to verify its applicability to land consumption monitoring. For this purpose, we applied a convolutional neural networks (CNNs) architecture called ResNet50 on a reference dataset of six high-spatial-resolution aerial images for the automatic production of thematic maps with the aim of improving accuracy and reducing costs and time compared with traditional techniques. The comparison with the National Land Consumption Map (LCM) of ISPRA suggests that although deep learning techniques are not widely exploited to map consumed land and to monitor land consumption, it might be a valuable support for monitoring and reporting data on highly dynamic peri-urban areas, especially in view of the rapid evolution of these techniques.

Land cover monitoring is crucial to understand land transformations at a global, regional and local level, and the development of innovative methodologies is necessary in order to define appropriate policies and land management practices. Deep learning techniques have recently been demonstrated as a useful method for land cover mapping through the classification of remote sensing imagery. This research aims to test and compare the predictive models created using the convolutional neural networks (CNNs) VGG16, DenseNet121 and ResNet50 on multitemporal and single-date Sentinel-2 satellite data. The most promising model was the VGG16 both with single-date and multi-temporal images, which reach an overall accuracy of 71% and which was used to produce an automatically generated EAGLE-compliant land cover map of Rome for 2019. The methodology is part of the land mapping activities of ISPRA and exploits its main products as input and support data. In this sense, it is a first attempt to develop a high-update-frequency land cover classification tool for dynamic areas to be integrated in the framework of the ISPRA monitoring activities for the Italian territory.