Remote sensing image scene classification consists of classifying images of the Earth surface into scene categories that represent different semantic ones based on the ground objects and their spatial arrangement. Finding the objects within a scene is not trivial, because they can appear in different sizes and mutual positions. An open issue in scene classification with CNNs is understating if the network prediction relies on the clues that human Earth Observation experts consider. A suitable approach for investigating the inference process of neural models relies on Class Activation Maps, which emphasize the areas of an image contributing the most to the classification. This work evaluates CAMs for different CNNs methods, in terms of their capacity to identify the objects that determine the classification of scenes for the illegal landfill detection. Quantitative and qualitative analyses show that ECA-Net has consistent performance across all metrics, resulting the most promising approach to obtain CNNs that focus on the most relevant points with the higher IoU. The illustrated analysis is a step towards the computer-aided study of the variations of scene elements positioning and spatial relations that constitute hints of the presence of illegal waste dumps and opens the way to the application of weakly supervised techniques for training detectors of illegal landfills in large scale remote sensing image repositories.

Open Source Geographical Information Systems, such as OpenStreetMap (OSM), offer a valuable alternative to proprietary solutions for the development of voluntary environment monitoring systems. However, the quantity and quality of information stored in such systems must be carefully evaluated and the contributions of volunteers must be boosted by means of effective engagement methods. This paper reports the results of the assessment of the quality and quantity of OpenStreetMap mountain information: different types of information and world regions have different gaps and improvement requirements. To address this issue, we propose a hybrid approach, in which an open Digital Elevation Model data set is processed with a heuristic algorithm to find candidate mountain information and uncertainty in the automatically extracted candidates is reduced by means of voluntary expert crowdsourcing. The improvement of landform information (not only about mountains, but also about orography and hydrography in general) can support the development of environment monitoring applications.

Mobile Augmented Reality (AR) applications offer a new way to promote the collection of geo-referenced information, by engaging citizens in a useful experience and encouraging them to gather environment data, such as images of plant species or of mountain snow coverage. The distinctive characteristic of mobile AR applications is the overlay of information directly on top of what the user sees, based on the user's context estimated from the device sensors. The application analyzes the sensor readings (GPS position, phone orientation and motion, and possibly also the camera frame content), to understand what the user is watching and enriches the view with contextual information. Developing mobile AR applications poses several challenges related to the acquisition, selection, transmission and display of information, which gets more demanding in mountain applications where usage without Internet connectivity is a strong requirement. This paper discusses the experience of a real world mobile AR application for mountain exploration, which can be used to crowdsource the collection of mountain images for environmental purposes, such as the analysis of snow coverage for water availability prediction and the monitoring of plant diseases.