Collaborative Geovisual Analytics
The PhD defence of Gustavo García Chapeton will take place (partly) online and can be followed by a live stream.
Gustavo García Chapeton is a PhD student in the Department of Geo-information Processing. Supervisors are prof.dr. M.J. Kraak and dr.ir. R.A. de By, co-supervisor is dr. F.O. Ostermann, all from the Faculty of Geo-Information Science and Earth Observation (ITC).
Several advancements in information and communication technologies and geospatial technologies have led to an unprecedented abundance of geodata. This data abundance presents novel opportunities to improve our understanding of natural and artificial processes. However, it challenges analysts who need to make sense of such large, heterogeneous, and multivariate data sets to address complex analysis situations. Two main challenges are (i) the limited capacity of humans to work with large amounts of data, and (ii) the multi-disciplinarity and complexity of data sets that renders analysis by a single person infeasible.
Geovisual Analytics (GVA) enables a synergy of human analytical skills with computer storage and processing power, addressing the first challenge, and facilitates collaboration of multiple analysts through interactive user interfaces, addressing the second challenge. To date, research in this field has focused on developing data transformation algorithms, visualization techniques, and interaction methods. However, the support for collaborative analysis has received less attention. This research project addresses the challenge of supporting collaborative analysis in GVA systems and produced four main outcomes, which are described below.
First, a literature review investigated the state-of-the-art of collaborative analysis in GVA. Chapter 2 reports the results, which include thirteen GVA systems with functions to support collaborative analysis, six distinct collaboration techniques, and three research challenges. The review revealed that the most common collaboration scenario is asynchronous and distributed, because it removes the constraints of time and place for participation. Further, increasing support for multiple types of devices is enabled by cloud-based infrastructures, which improve the scalability of and accessibility to the system. Although the identified collaboration techniques support the whole data analysis process, functions are missing to synthesize analytical contributions and summarize the level of agreement regarding evidence and conclusions. The identified research challenges are the lack of support for (i) collaboration scenarios supporting collocated and distributed participants or synchronous and asynchronous interactions, (ii) collaboration across multiple types of devices, and (iii) time-critical and long-term analysis scenarios.
The second main outcome is a software reference architecture for collaborative geovisual analytics (CGVA) systems, proposed in Chapter 3. A software architecture is an abstract high-level description of the fundamental structures of a system, their relationships, and properties of both. Software architectures can have different goals and scopes. Therefore, it is not designed for one specific system, but it is a template design for a group of similar systems. The architecture design criteria are based on the literature review. The proposed architecture uses the client-server and layered architectural patterns. An architectural pattern is a reusable solution with well-understood properties for a commonly occurring problem in software architecture design. In the proposed architecture, the client-server pattern allows assigning the processing and storage to the server side, enabling participants to work from low-end devices. Additionally, the layered pattern enables separation of concerns, meaning that software designers can address design problems - such as user interface, data processing, security, data storage, and the communication and coordination between the components that perform each function - in isolation.
The third outcome is the approach of Spatiotemporal Analysis Space (STAS), which proposes a philosophy for long-term distributed asynchronous collaborative analysis of spatiotemporal data, described in Chapter 4. STAS was specifically designed to support long-term analysis processes, which the application case of this research - agricultural pest management - requires. The main assumption in the STAS design is that in data sets with large spatial and temporal extents, events of interest such as patterns and outliers occur in diverse locations and times. The central concept of STAS is the analysis space, which is a container for a data subset that contains an event of interest and analytical contributions to make sense of it. The analysis space focuses the analyst’s attention on an event of interest to elicit sensible contributions and generate meaningful knowledge. Additionally, the approach allows creating links between different analysis spaces to promote knowledge-building from previous contributions.
Lastly, a case study demonstrates the applicability and feasibility of the software reference architecture and the collaborative approach in a real-world scenario. Described in Chapter 5, the case study is the monitoring and control process of the Olive Fruit Fly (Bactrocera oleae) in olive groves in Andalusia, Spain. In this context, a CGVA prototype implementing the STAS approach was designed, developed, and tested with case study stakeholders. The stakeholders are one authority representative, one researcher, and five field technicians. They used the prototype to analyze monitoring and control data and the outputs of two statistical models. The results show that the architecture can accommodate the case-specific requirements and that the STAS approach enables collaborative analysis among the stakeholders. In the post-evaluation discussions, all the stakeholders mentioned that they would like to see the prototype becoming a production system to support their pest management activities.
To conclude, this thesis presents research that identifies remaining research challenges for CGVA, followed by the development of required software architecture and collaborative analytical approach (STAS), successfully implemented in a prototype in a real-world case study with stakeholder involvement.