DESIGN AND IMPLEMENTATION OF CUAHSI WATERML AND WATERONEFLOW WEB SERVICES

WaterOneFlow is a term for a group of web services created by and for the Consortium of Universities for the Advancement of Hydrologic Science, Inc. (CUAHSI) community. CUAHSI is an organization representing more than 100 US universities which is supported by the National Science Foundation to develop infrastructure and services for the advancement of hydrologic science. CUAHSI web services facilitate the retrieval of hydrologic observations information from online data sources using the SOAP protocol. CUAHSI WaterML (below referred to as WaterML) is an XML schema defining the format of messages returned by the WaterOneFlow web services. 

Background

Beginning in 2005, the CUAHSI HIS project team implemented a variety of web services providing access to large repositories of hydrologic observation data, including the USGS National Water Information System (NWIS), and the US Environmental Protection Agency's STORET (Storage and Retrieval) database of water quality information.  The services gave access to station and variable metadata, and observations data stored at these sites.  As the services were written for each data source individually, their inputs and outputs were different across data sources.  The services developed in this ad hoc manner do not scale well.  As the number and heterogeneity of data streams to be integrated in CUAHSI's hydrologic data access system increased, it would become more and more difficult to develop and maintain a growing set of client applications programmed against the different signatures and keep track of data and metadata semantics of different sources. As a result, WaterML was developed to provide a systematic way to access water information from point observation sites.

Point Observations Information Model

In parallel with web service development, CUAHSI has been developing an information model for hydrologic observations that is called the Observations Data Model (ODM).  Its purpose is to represent observation data in a generic structure that accommodates different source schemas. While based on the preliminary set of CUAHSI web services, WaterML was further refined through standardization of terminology between WaterML and ODM, and through analysis of access syntax used by different observation data repositories, including USGS NWIS, EPA STORET, NCDC ASOS, Daymet, MODIS, and NAM12K

Figure 1 near here

According to the information model, a Data Source operates one or more observation networks; a Network is a set of observation sites; a Site is a point location where water measurements are made; a Variable describes one of the types of measurements; and a time series of Values contains the measured data, wherein each value is characterized by its time of measurement and possibly by a qualifier which supplies additional information about the observation. Figure 1 demonstrates the main components of the model, and respective web services.

WaterML concepts

The goal of the first version of WaterML was to encode the semantics of hydrologic observations discovery and retrieval and implement WaterOneFlow services in a way that creates the least barriers for adoption by the hydrologic research community. In particular, this implied maintaining a single common representation for the key constructs returned on web service calls, related to observations, features of interest, observation procedures, observation series, etc.

An observation is considered an act of assigning a number, term or other symbol to a phenomenon, and a result of such assignment. Hydrologic observations are performed against many different phenomena (properties of different features of interest), and shall be associated with time measurements (time points or time intervals). The features of interest common in hydrologic observations may include points (gauging stations, test sites), linear features (steams, river channels), or polygon features (catchments, watersheds). Spatial properties of the features of interest may be further expressed in 2D or 3D, in particular via vertical offsets against common reference features. The observations are made in a particular medium (water, air, sediments) using a procedure. The procedure may represent a multi-step processing chain including an instrument (sensor), algorithms for transforming the initially measured property (e.g. “partial pressure of oxygen in the water” may be transformed into a measure of “dissolved oxygen concentration”), and various techniques for censoring and quality-controlling of the value assignment, including multiple scenarios for assignment of no value. Individual observations are organized into observation series (regular sequences of observations of a specific variable made at a specific site), which are in turn referenced in series catalogs. A series catalog is an element of an observation network, which represents a collection of sites where a particular set of variables is measured. A responsible organization can maintain one or more observation networks.

In addition to point measurements described in the ODM specification, hydrologic information may be available as observations or model outcomes aggregated over user-defined regions or grid cells. While USGS NWIS and EPA STORET exemplify the former case, sources such as MODIS and Daymet are examples of the latter. In this latter case, as in the case of other remote sensing products or model-generated grids, the observation or model-generated data are treated as fields, and sources of such data are referenced in WaterML as datasets, as opposed to sites.

The practice of hydrologic observations provides ample evidence of complications beyond this general treatment. These complications are related to huge, complex and incompatible vocabularies used by several federal hydrologic observation systems, to different and not always documented contexts of measurement and value assignment, to often ambiguously defined features of interest, to complex organizational contexts of hydrologic measurement, transformation and aggregation, etc. It is in response to this complexity that the CUAHSI WaterML is primarily designed. While some of this complexity may be captured within the standards being developed under the OGC's Sensor Web Enablement (SWE) activity, the flexibility inherent in such standards may itself be a barrier to adoption when the target audience is not computer scientists.

Implementation Context

WaterML is primarily designed for relaying fundamental hydrologic time series data and metadata between clients and servers, and to be generic across different data providers. Different implementations of WaterOneFlow services may add supplemental information to the content of messages. However, regardless of whether or not a given WaterML document includes supplemental information, the client shall be sure that the portion of WaterML pertaining to space, time, and variables will be consistent across any data source.

Depending on the type of information that the client requested, a WaterOneFlow web service will assemble the appropriate XML elements into a WaterML response, and deliver that to the client.  The core WaterOneFlow methods include:

·         GetSiteInfo – for requesting information about an observations site. The returned document has a root element of SiteInfoResponse type.

·         GetVariableInfo – for requesting information about a variable. The returned document has a root element of VariableResponse type.

·         GetValues – for requesting a time series for a variable at a given site or spatial fragment of a dataset. The returned document has a root element of TimeSeriesResponse type.

The provisional services are available from http://river.sdsc.edu. The CUAHSI WaterML description has been submitted as a discussion paper to the Open Geospatial Consortium, and is available from the OGC portal. The project web site is http://www.cuahsi.org/his .

 

Figure 1.  CUAHSI Point Observations Information Model, and corresponding web service methods.