This project has received funding from the European Community's Seventh Framework Program (FP7/2007-2013) under grant agreement nº265178

NEWS
Contact
> Home > Outreach > OGC- WMS-Q ER

OGC-WMS-Q ER________________________________

 

 

The purpose is to develop and apply conventions for conveying information about data quality through the OGC Web Map Service specification. The WMS 1.3.0 currently does not well support the integration of quality information into WMS. For example, if the dataset-level quality information is presented in an unstructured format (e.g. PDF), or encoded using GeoViQua proposed producer model, or encoded using ISO19157, the current WMS 1.3.0 provides no means to integrate them into the WMS. If the pixel-level uncertainty information is presented using NetCDF-U (which defines NetCDF uncertainty conventions), the current WMS does not support how data layer can semantically associate with the corresponding uncertainty layers. Hence, we proposed a WMS-Q specification to address these limitations. This WMS-Q specification was proposed as far as possible within the bounds of the WMS 1.3.0 specification, requiring as few extensions as possible. Discussions of this document has been moved to OGC OWS- 9 and if approved, will become a OGC public engineering report. A description of WMS-Qis presented as follows.

In order to integrate the dataset-level quality information into the WMS, in the WMS-Q, we proposed to expand slightly “Type” attribute of “MetadataURL” element to have “unstructured” and “other-structured” options. We also propose to add a “description”element for the “MetadataURL” element Pixel-level uncertainty information can be encoded using NetCDF Uncertainty Conventions (NetCDF-U)1, which defines some conventions for encoding the uncertainty information following an uncertainty model, namely UncertML2. For example, if the variable “Sea_Surface_Temperature”(SST) follows a normal distribution, it can be characterised using two parameters “mean” and “variance”. NetCDF-U models “Sea_Surface_Temperature” as a “concept without values” and uses a variable attribute “ancillary variables” to associate the “mean” and “variance”. A fragment of capabilities document showing the integration of uncertainty information into WMS is shown in Figure 1 (where uncertainty is modelled as distribution). The proposed WMS-Q specification can be summarised as follows:

  1. The controlled vocabulary for ‘Type' attribute of <MetadataURL> element is expanded. This expansion is compulsory. A ‘<Description>' element for <MetadataURL> element is added, but this extension can be optional.
  2. There could be a top-level non-displayable <Layer> that holds all metadata that apply to all layers in the WMS instance. For example, in ncWMS this includes the coordinate reference systems in which all layers can be returned. In WMS-Q this Layer could hold <MetadataURL>s pointing to quality information that pertains to all Layers.
  3. If the concept of a “Dataset” (i.e. a collection of Variables) is an appropriate grouping in a given situation (e.g. when including multiple NetCDF datasets in a single WMS), they can be modelled as non-displayable Layers directly underneath the top-level Layer. <MetadataURL>s can be associated to convey quality information at the Dataset level.
  4. An uncertain variable can be modelled as a Layer nested inside a relevant Dataset. This Layer may or may not be displayable, depending on the service provider. If it is displayable it needs a Name, and a GetMap request on this name will display some representation of an uncertain variable in a single map image (e.g. a raster field representing the mean, overlain by a contour field representing the variance.) If it is not displayable (i.e. there is no Name) it is simply a container for its children (directly analogous to the NetCDF-U “concept without values”). A Keyword indicates the type of the distribution, from the UncertML vocabulary. Additionally, units of measure could be recorded using a different keyword.
  5. Each component of the uncertain variable (e.g. mean, variance) is modelled as a displayable Layer inside its parent. Each component is tagged using a Keyword that defines which component it represents (e.g. mean or variance). A GetMap request on the component will result in a map that displays a raster field of that component only. <MetadataURL>s can be associated with either the individual components or, more likely, with the parent Layer that represents the uncertain variable as a whole.

Figure 1: “Layer nesting” mechanism: a fragment of capabilities document showing the integration of uncertainty information into WMS.

We have started the development of visualisation tools and components. The development includes the implementation of WMS-Q specification, and the development of associated WMS-Q Viewer client tool. The WMS-Q specification implementation is based on WMS1.3.0 compliant component, namely, ncWMS3, which is widely used in many research projects and institutes (we termed it as ncWMS-Q). At the time of writing this document, we have primarily finished ncWMS-Q GetCapabilities operation and partially finished the GetMap operation. The WMS-Q Viewer includes web-based WMS-Q Viewer and desktop-based WMS-Q viewer. The Web-based WMS-Q viewer is a WMS-Q service client tool that can allow users to visualise data and associated uncertainty information from WMS-Q servers interactively at a pixel level. The Web-based WMS-Q Viewer encompasses at least the “Godiva2” webbased client that is packaged with ncWMS. The desktop WMS-Q Viewer can allow users to visualise data and uncertainty information from users' local disks or remote OPeNDAP servers.

Figure 2. Uncertainty visualisation. ‘variance' is presented as a contour lines overlaid with ‘mean' raster map. Generated with desktop GODIVA-Q WMS-Q Viewer client .

After its publication and as a result of its implementation phase in GeoViQua, the WMS-Q convention was reviewed and a WMS-Q version 2 was elaborated.

     O u t r e a c h