AccuWeather Global Headquarters – February 6, 2017 – This morning AccuWeather passed on a National Weather Service Tsunami Warning that was intended by the NWS to be a test but was miscoded by the NWS as a real warning. AccuWeather has the most sophisticated system for passing on NWS tsunami warnings based on a complete computer scan of the codes used by the NWS. While the words “TEST” were in the header, the actual codes read by computers used coding for real warning, indicating it was a real warning.
The NWS warning also later appeared on other sources such as The Weather Channel and it even appears on some pages of the NWS own website as a real warning. The NWS is the original source of the information and displayed it as a real warning.
Tsunami warnings are handled with the utmost concern by AccuWeather and it has sophisticated algorithms to scan the entire message, not just header words, as from the time of a warning to the actual event can be mere minutes. AccuWeather was correct in reading the mistaken NWS codes embedded in the warning. The responsibility is on the NWS to properly and consistently code the messages, for only they know if the message is correct or not.
As reported by AccuWeather, once discovered that the NWS had incorrectly coded the warning, we sent messages via social channels that no tsunami warning is in effect for the East Coast of the U.S.
This is not the first time legitimate warning coding was embedded erroneously by the NWS and consequently triggered alerts. In October 2014, AccuWeather advised NWS in writing about the potential for this problem to be repeated if not fixed.
AccuWeather’s CEO Barry Myers wrote to the NWS over three years ago:
“We understand the reason for test messages, but we feel that NWS consider fail safe measures for the future to prevent such an occurrence. The issuance did say it was a “TSUNAMI WARNING,” but it was not a tsunami warning, rather simply a test of the system. We note that the method currently used of relying on the “TEST” in the header of the product and a test in the VTEC status, as the identifying device for software coding in numerous programs and systems used by a plethora of companies to identify such messages, has proven to be a less than perfect system.”
This is all a matter of public record.
We are continuing to work with NWS to determine why this coding was improperly embedded in its test alert system.