Easier surveillance of climate-related health vulnerabilities through a Web-based spatial OLAP application

Abstract

Background: Climate change has a significant impact on population health. Population vulnerabilities depend on several determinants of different types, including biological, psychological, environmental, social and economic ones. Surveillance of climate-related health vulnerabilities must take into account these different factors, their interdependence, as well as their inherent spatial and temporal aspects on several scales, for informed analyses. Currently used technology includes commercial off-the-shelf Geographic Information Systems (GIS) and Database Management Systems with spatial extensions. It has been widely recognized that such OLTP (On-Line Transaction Processing) systems were not designed to support complex, multi-temporal and multi-scale analysis as required above. On-Line Analytical Processing (OLAP) is central to the field known as BI (Business Intelligence), a key field for such decision-support systems. In the last few years, we have seen a few projects that combine OLAP and GIS to improve spatio-temporal analysis and geographic knowledge discovery. This has given rise to SOLAP (Spatial OLAP) and a new research area. This paper presents how SOLAP and climate-related health vulnerability data were investigated and combined to facilitate surveillance.

Results: Based on recent spatial decision-support technologies, this paper presents a spatio-temporal web-based application that goes beyond GIS applications with regard to speed, ease of use, and interactive analysis capabilities. It supports the multi-scale exploration and analysis of integrated socio-economic, health and environmental geospatial data over several periods. This project was meant to validate the potential of recent technologies to contribute to a better understanding of the interactions between public health and climate change, and to facilitate future decision-making by public health agencies and municipalities in Canada and elsewhere. The project also aimed at integrating an initial collection of geo-referenced multi-scale indicators that were identified by Canadian specialists and end-users as relevant for the surveillance of the public health impacts of climate change. This system was developed in a multidisciplinary context involving researchers, policy makers and practitioners, using BI and web-mapping concepts (more particularly SOLAP technologies), while exploring new solutions for frequent automatic updating of data and for providing contextual warnings for users (to minimize the risk of data misinterpretation). According to the project participants, the final system succeeds in facilitating surveillance activities in a way not achievable with today's GIS. Regarding the experiments on frequent automatic updating and contextual user warnings, the results obtained indicate that these are meaningful and achievable goals but they still require research and development for their successful implementation in the context of surveillance and multiple organizations.

Conclusion: Surveillance of climate-related health vulnerabilities may be more efficiently supported using a combination of BI and GIS concepts, and more specifically, SOLAP technologies (in that it facilitates and accelerates multi-scale spatial and temporal analysis to a point where a user can maintain an uninterrupted train of thought by focussing on "what" she/he wants (not on "how" to get it) and always obtain instant answers, including to the most complex queries that take minutes or hours with OLTP systems (e.g., aggregated, temporal, comparative)). The developed system respects Newell's cognitive band of 10 seconds when performing knowledge discovery (exploring data, looking for hypotheses, validating models). The developed system provides new operators for easily and rapidly exploring multidimensional data at different levels of granularity, for different regions and epochs, and for visualizing the results in synchronized maps, tables and charts. It is naturally adapted to deal with multiscale indicators such as those used in the surveillance community, as confirmed by this project's end-users.

Figure 1

Exploring the data through the drill functionalities. The drill functionalities can be used in any of the three types of displays (cartographic, chart and tabular). Navigating through different LoDs is simply done by clicking directly on an element or a group of elements. The automatic legends have been removed for better clarity. Section A: (top) Spatial drill-down on a specific member. The result shows only the number of people for the sub-regions included in the selected member. (bottom) Spatial drill-down on a specific LoD. The result shows the number of people for all the members of the next LoD, i.e., all the sub-regions. Section B: Drill-down on a specific member, i.e., a bar of the bar chart. The result shows the number of people for the sub-regions included in the selected member. Section C: Drill-down on a specific member, i.e., a row in the table. The result shows the number of people for the sub-regions included in the selected member.

Figure 2

Spatio-temporal evolution of heat waves using automatically built comparative multiple maps. The first map (top left) is obtained from observed data from 1971–2000, the second map (top right) is obtained from prediction data for 2011–2040, and the third map (bottom) is obtained from prediction data for 2041–2070.

Figure 3

Proportion of the population living alone. Spatial representation of the proportion of the population living alone at two different LoDs (upper maps) and distribution of the population living alone by age group (table) for the entire territory.

Figure 4

Two different ways of spatially representing the proportion of the population living alone. The left window displays the proportion for each age group at a specific LoD in a different map, while the right window uses pie charts over the territory to depict the ratio of the two age groups.

Figure 5

Combining two measures on the same map for an enriched analysis. Spatial representation of the proportion of the population living alone (choropleth map), combined with the proportion of the population with low income (symbols).

Figure 6

Representation of the average temperature. The table is used to compare the average temperature at a regional level (health regions), while the map represents this information at a more local level.

Figure 7

Contextual spatial layers can provide additional information. Québec City's public parks and private and public pools have been added to the temperature map at the local level.

Figure 8

The number of buildings located in areas with flooding potential. Comparison of the number of buildings located in flooding areas at the county level (left map), at the regional level, and according to their type (table) and detailed spatial representation of the schools located in flooding areas (right map). The road network has been added as contextual spatial information.

Figure 9

Contextual warnings will be used to provide additional information and avoid data misinterpretation. Contextual warnings can be displayed during the analysis to inform the user about implicit aspects of the data that must be taken into account when conducting the analysis to avoid data misinterpretation.