Computer models are the workhorse of an environmental data processing workflow. Moving this workflow to the cloud requires models to be cloud-enabled. In the cloud, data may come from different sources with different meanings and accuracy. Environmental models must therefore be able understand the data and their limitations, and also be flexible enough to work with a large variety of types, resolutions and quality. This work package experimented with cloud-enabled environmental models. Existing standards for data exchange and web processing were tested and improved. In addition, flexible models that can cope with heterogeneous data and data scarcity were required.
Environmental models as web-services were developed. They focused on the rainfall/river flow relationship, for which a large variety of traditional, desktop-based computer models exist. Those available models have very specific data requirements and application conditions, and need careful and laborious setup. As such, they are not suitable to operate in a cloud environment, which is characterised by heterogeneity of data quality and availability, and the need to flexibly and automatically combine model components. A component-based solution was implemented in which a model can be constructed ad-hoc for a specific application and data availability. This modelling system was delivered as a web-service using an existing open standard, WPS, developed by the Open Geospatial Consortium. However, WPS is not designed to deal with time series data, which are very common in hydrological analysis, and therefore had to be adapted and extended for this application. Additionally, the standards that are used to encode and document hydrological data such that they can conveniently be stored and exchanged in a cloud environment are at a very preliminary stage. They were tested and extended in several directions, such as the inclusion of information about uncertainties in the data and modelling results.
The web-based modelling services can be accessed freely over the internet using the web processing service. By feeding it with rainfall data and other river basin characteristics, it enables prediction of the river flow at the outlet of the basin. As a web-service, the software can be used without the need for any local installation or data processing from the users’ perspective. This makes data processing and prediction much more convenient, but also allows it to be used in complex data processing workflows that combine different modelling web-services that are geographically distributed and offered and maintained by different modelling teams.
Combining cloud-enabled data sources and models makes it possible to construct more complex and faster data processing workflows than ever before. This allows more efficient processing of massive datasets, such as satellite imagery, to predict environmental processes more accurately, and to evaluate different management options in more detail.