Air quality scenario______________________________
The initial aim is generating a near real time and a short-term prediction model for the estimation of air pollution and climate.
The final user is expecting an augmented reality application for mobile devices and computers, which needs this information and will use a mobile device to visualize it and decide whether to go or not to go outside, considering risk for his health due to air pollution.
From the datastore in GEOSS, a scientist would need to extract a weather dataset, most probably based on ground measurements directly obtained from meteorological stations and an air pollution concentration dataset , most likely also provided as measurements at specific locations but it would also be possible to find a RS atmospheric product. An important step here would be the importation of the datasets, and compatible formats, transferring quality measurements at every step of the processes. Dealing with different data formats is not always an easy task but however it is something commonly encountered when working with different datasets coming from diverse institutions. Thus a dataloader tool allowing for data integration and metadata links accounting for this fact would be an interesting contribution at this point.
Depending on whether the sensor measurements consist of continuous records (e.g. raster layers) or not, interpolation methods are applied in order to obtain estimations for a surface. Air pollutant concentrations depend on weather conditions (e.g. pressure, rain or temperatures). This is an example of situations in which several variables combine in models. The result of these processes is an air pollution climate dependent model. In truth, two models should be taken into account: a near-real time one and a short-term predictive model. The user needs a near-real time prediction but will surely appreciate to have some estimation of the plausible situation in at least some hours time or even that day in order to plan not just the immediate moment but for instance the evening or what will be the situation on the spot he is planning to go to in a few hours time. This allows for playing with temporal components.
In the health risk assessment, an expert would determine the thresholds for the different air pollutants to be a risk to human health. From this stage, a diagnostic expert system will be derived. Ideally, it would be a personalized one, in which the user would provide some basic information on symptoms or other data and the system would adapt the threshold to his circumstances and that of that day's weather and air pollution concentrations. The main idea is to deal with categorization of quantitative variables and fit-for-purpose variable uncertainties , or decision trees.
The corresponding uncertainty assessment or quality issue is collected in the overall uncertainty integration, which will be related to the producer quality model in GeoViQua, in particular to quality associated to quantitative variables .
Finally, all these efforts converge, together with an input on current location (and the corresponding positional accuracy), in the development of the augmented reality browser application that will display the information to the final user, with its associated uncertainty. This system can be more or less complex and if successful the application can be included GEOSS as a downloadable service. Maybe the uncertainty is too high and the system can't help the user in his decision. Else, he can decide, and when going outside, he can generate a feedback on the system, from which the model can be readjusted if needed at several points. Depending on the feedback, some steps can be validated and even the quality model can incorporate some components to the quality records in the datastore. The geosearch is tested both ways, and the user feedback/producer quality model in GeoViQua are integrated in GEOSS.