Address questions or comments to Charles Merk, CPAC Secretary
cfmerk on (703) 648-5680
In Reply Refer To:
Mail Stop 415
This memorandum establishes a policy for the implementation of quality-assurance measures for real-time water data being served on the World Wide Web (WWW). It builds on the information provided in Water Resources Division (WRD) Memorandum No. 95.19 and No. 97.17. Memorandum 95.19 established a policy that the nature of provisional data available on the WWW would be clearly identified and Memorandum 97.17 provided guidance for use of the rt_www software for serving real-time streamflow data.
Including the WWW in our real-time data delivery system has significantly increased the visibility of these data. Although WRD has been providing real-time streamflow data to selected cooperators for over 25 years, our feedback shows that the public is now using our data in ways we never anticipated. Unfortunately, our ability to quality assure these data has not kept pace with our ability to provide it. Erroneous and spurious data (such as spikes due to transmission errors) that appear on our web pages are problematic for two reasons. First, these data can lead users to make bad decisions involving lives or property. The second is that bad data, even if not used for a decision, undermine the credibility of the vast majority of good data we are serving. When the bad data persists on our web pages for several days, it suggests a lack of attention to quality assurance. Although provisional data statements clearly warn users about potential errors, it must be our goal to avoid serving data that we know to be seriously in error.
Ensuring that the data provided to the public are the best we can offer will require a concerted effort to review data before and after it is available on the web and make appropriate corrections to the displayed data. Our approach to ensuring the quality of our real-time data must be made up of two parts. First, we need to use the software tools available to automatically prevent bad data from being served in the first place. This memorandum describes some of the tools that are currently available for this automatic checking. These tools can be very effective if used correctly and even better tools are being developed to enhance this process. However, even the best automatic process will not eliminate all of the spurious data. Thus, we need to supplement the automatic processes with frequent visual inspection of hydrographs followed-up with rapid deletion of the bad data.
Frequent inspection should be taken to mean at least once per work day although many of our offices already have procedures for checking even on weekends and holidays. However, at times when users most rely on our data (floods and droughts) even more frequent checking may be appropriate. It is far better for us to indicate that the data are missing, and that we are aware of the problem and will work to resolve it, than it is to continue to serve the bad data.
The NWISWeb software, once implemented, will be used to serve our historic and real-time surface water, ground water and water quality data on the WWW. Information on NWISWeb development, requirements documents, and quality-assurance measures associated with the current release of NWISWeb software will always be available from:
The QA measures detailed below apply to the surface water WWW data currently being served using the rt_www software. The NWIS Web software will continue to use these measures.
The rt_www rev 2.0 software will automatically mask from web display any unit-value data that is flagged by ADAPS as exceeding the thresholds for Very-high-value, Very-low-value, Very-rapid-increase, and Very-rapid-decrease. When any data for a day is masked in this way, a small black triangle will appear at the top axis of the graph indicating the condition.
Also included in rt_www 2.0 is a web-based "quick-view" utility which provides data reviewers an easy way to sequentially view stage (and optionally discharge) graphs for monitoring data quality. This utility is especially useful for reviewing graphs from the field using a cellular-capable modem. When viewed on the district LAN, graphs can be sequentially drawn to the screen once every 3 seconds allowing all real-time stations to be reviewed as often as necessary. The visual review of the real-time data using the quick-view utility is useful for catching subtle errors that may pass through the threshold filters.
Memorandum 97.17 established policy for the daily review of data and for making corrections when necessary. The "quick-view" utility can be used to conduct a review of the data once a day under normal conditions. During flood or drought, keep in mind that there is higher visibility of data and errors are potentially compounded, therefore, it is important that data being served be checked and corrected, as necessary, more frequently. A completely spurious data spike during a flood should not be displayed on the WWW for more than 24 hours. During more normal conditions, such a spike should not be displayed for more than a few hours into the next work day.
At a minimum, the ADAPS Very-high-value and Very-low-value must be set for every station with data displayed on the WWW in real time. The values for each of the threshold filters must be defined by individuals that are knowledgeable about the hydrologic and historical flow characteristics of each site.
The Very-high-value should be set to eliminate values that are greater than the highest possible value; therefore, a value that is 0.10 units higher than the highest possible value that can be measured by the sensing and recording equipment at the site would be appropriate. For a streamgaging station, this might be 0.01 feet higher than the bottom of the instrument shelf.
The Very-low-value should be set to eliminate values that are less than the lowest possible value; therefore, a value that is 0.10 units lower than the lowest possible value that can be measured by the sensing and recording equipment at the site would be appropriate. For a streamgaging station, this might be 0.10 feet below the point-of-zero flow of a stable control.
The VERY RAPID INCREASE and VERY RAPID DECREASE threshold can be used by the rt_www software to provide an alert about a possible erroneous jump between successive unit values. Currently, because there is no standard method to determine valid rapid increases and decreases these thresholds are not required, although a knowledgeable streamgager can determine these values for each site by examining the historical unit-values record and then calculating values for both of these thresholds.
The requirement to review data served on the WWW at least once a business day under normal conditions and the establishment of Very-high-value and Very-low-value thresholds in ADAPS are policy of WRD. Other guidelines for quality assuring real-time data were distributed with OSW Technical Memorandum No. 99.07 as part of the addendum to Open-File Report 94.382, "A Workbook for Preparing Surface Water Quality-Assurance Plans for Districts of the U.S. Geological Survey, Water Resources Division."
Using the two-part approach: automatic screening to prevent bad data from appearing, coupled with visual screening and editing at least once per work day, we can significantly lower the amount of erroneous data we serve. This will be of benefit to the public we serve and maintain the high level of credibility that USGS streamflow data rightfully enjoy.
Distribution: A, B, FO, PO, S
Address questions or comments to Charles Merk, CPAC Secretary
cfmerk on (703) 648-5680