Toward a Visual Analytics Approach to Support Multi-Sensor Analysis in Remote Sensing Science

Abstract

Multi-sensor analysis is a novel scientific approach in remote sensing science. The basic idea is to enable users to combine various satellite mission data (called scenes) into a common data set. This combination produces millions of high-resolution time series (one time series for each pixel) from which users want to extract potentially interesting spatio-temporal patterns. A challenge of multi-sensor analysis is that users often experience difficulties interpreting the extracted patterns. We use Visual Analytics (VA) to help users understand these patterns. We learned from our interdisciplinary cooperation in the GeoMultiSens project that VA has to support the assessment and selection of scenes suitable for the current application scenario and question to achieve this goal. The contribution of this paper is twofold. First, we describe how we devised a VA approach that supports users in the assessment and selection of remote sensing data based on a user and task analysis. We demonstrate how our VA approach helps users to select and assess scenes to study forest cover change in Europe between 2010 and 2016. The study of forest cover change is an important scientific scenario because the loss of forest cover has negative effects on the environment, such as undermining the capacity of ecosystems to maintain fresh water, loosing the ability to regulate the climate, and poorer air quality. Second, we discuss the Scientific Data Explorer, our research vision for VA to enable users to effectively develop VA approaches for a variety of scientific scenarios.