In the field of environmental modelling, there is a huge potential in the combination of existing resources, such as data and models, exposed as web services. Currently, there are several R&D lines being conducted to facilitate the integration of complex resources into intricate model workflows. There are possibilities for a wide range of fields such as climate change monitoring, water cycle management, air pollution monitoring and forecasting, biodiversity monitoring and management, flood preventing, forest monitoring, marine environment monitoring,
natural disaster management and many others. However, all the models and data sources have approximations and errors. Currently, there is no of-the-shelf method for estimating and propagating the error associated to these complex model workflows. This limits the possibilities of making decisions based on complex models. Therefore, complex models derived from multi-resources integration should consider this uncertainty. This will allow understanding the validity of the output at the end of the chain, and thus a rational way to make decisions.
The current project addressed these issues by developing mechanisms, standards, tools and case studies to enable uncertainty management in an interoperable model web context. With a consortium consisting of 8 partners, only RTD centres and academia, the project was funded under FP7_ICT and coordinated by Aston University (United Kingdom). The project started in February 2010 and final conclusions will be obtained in January 2013.
The project provided a holistic solution to uncertainty management in the context of models made interoperable through web services. The outcomes can be classified in three main categories: (i) probabilistic encoding, (ii) uncertainty management software and (iii) demonstrators.
First, the existing UncertML (Uncertainty Markup Language) was extended to provide a complete probabilistic encoding for uncertainty. Moreover, an application programmers interface (API) was developed to easy process UncertML documents and existing interoperable web service upgraded to support UncertML.
Second, web-based software tools were developed to allow people to manage uncertainty and thus use the technology:
- Elicitation tool for assisting the expert elicitation of uncertainties.
- Aggregation tool of uncertainties for coupling models with different spatial and temporal supports.
- Visualization of uncertainty.
- Uncertainty source identification for uncertainty analysis and sensitivity analysis and the use of meta-models.
Third and last, case studies to ensure the usefulness of the technology:
Four demonstrators were conducted to test the behavior of the framework with complex multi-resource models with different characteristics: biodiversity and climate change, land-use response to climatic and economic change, short term uncertainty-enabled forecasts for local air quality and individual activity in the environment.
In general, the developed technology can be applied to manage the uncertainty of any kind of model derived from chaining multiple web resources. In particular, it is more suitable for the integration of resources which consider spatial or temporal aspects. Moreover, the technology performs better when the resources have few parameters to handle, i.e., few inputs and outputs. The maturity of the outcomes is different depending on each outcome. The elicitation tool, the sensitivity and emulation tool, and the visualization tool are the more mature. Therefore, the Technology Readiness Level for the outcomes as a whole is estimated at 5 on the TRL scale. The outcomes shall be used mainly by the scientific community and are open source, as they are conceived to be shared. The developed uncertainty management ensures that the users of complex models can now make rational decisions based on the model outputs.