MKX Virtual Operations Support Team

(MKX-VOST)

 

Weather Event Data Capture through Social Media

 

Introduction

Since the 1970s the US population has moved across tools like chat windows in private networks to public networks using content rich sites like MySpace to the current/mainsteam social media tools like Facebook, Vine, Twitter, Snapchat, Instagram and LinkedIn, just to name a few.  What makes life more interesting is we know there will be many others yet to come.

 

It is safe to say that most, if not all people reading this page is a user of a social media tool or two and is most likely a user of a smartphone or other small form communication device using some level of paid - access service provided by a wireless carrier.

 

How do people contact others in these tools?  A bit of irony: current day smartphone devices are used, for the most part, just about everything other than phone calls; phone calls have never been a primary use for these devices.  E-mail?  passe’ for daily communications.  The current state is instant messaging in some form.  Posts to social media sites or good old text messaging – that is how it has been done for years.

 

Smartphones have enabled us to become photographers, news reporters, chronicle our lives and do many other things we never thought possible.  Maintaining contact with high school and college friends, sharing information with others who share common interests and even things like work with our medical professionals on diagnosis and treatment of minor health issues - would those be possible without the help of smartphones?

 

To say that five-pack cigar box sized, pancake thick device we cannot live without has transformed most of our lives is a simple understatement.  Ignore that these devices and apps exist and watch one’s relevance in the world fade quickly.

 

Want to engage the public?  Social media is THE way to do so.  But.. why so many tools nowadays?  The short answer is niche.  According to one source, Snapchat has been adopted by the (relatively) younger crowd, Twitter and Facebook by the adult set, Instagram’s user base is predominately female and LinkedIn found a home in professional networking.

 

“Like Us on Facebook”, “Follow Us on Twitter” and many other taglines are common characteristics of our interaction landscape.  Many people “Like”, “follow” and in the case of the National Weather Service Forecast Offices, contribute observed data via these tools.

 

What does that mean for the Weather Service?  The available detail for any event has risen substantially.    But...does the data received contain the content necessary for use?

 

Depending on event there may be one, maybe two useful items out of dozens, or even hundreds of posts.  That is a lot of data to “mine” for the value it provides, and mining of that data takes time away from other things.

 

Another consideration:  what about the posts that never get to the Forecast Office social media pages?  A data mining exercise in 2014 showed there was a significant amount of valuable detail being missed due to several factors.

 

The quality of information was also driven by the tool used.  Free form based tools like Facebook, while valuable and easier to use, promoted verbosity where may not needed (or appropriate) where other tools  focused on content.

 

So, in 2015 the Milwaukee Forecast Office staff responsible for social media outreach determined that because of the “Keep It Simple” short message, available routing features, advanced analysis and filtering tools available, Twitter would be a “lead” tool for reporting of weather information.

 

But...that approach in itself had some drawbacks because of the area (square mileage) involved and the keywords used (weather focused keywords drive some interesting results!)  As a result, the staff approached SulCom with a need.

 

Now for the rest of the story….

 


 

Current State

The Milwaukee Forecast Office uses HootSuite for looking at a 100 mile radius for Twitter posts having keywords like weather, hail, etc.  This is called a “feedwatch.”

 

As can be seen a “Office (MKX) Center” (blue lines) and a “CWA Center” (green lines) area does not drive any significant change in area being canvassed.  Either approach drives areas of capture that are well outside of the CWA (shown in red.)

 

100 Mile Radii and CWA.jpg

 

The above approach has several disadvantages.  Among them being:

1.      Large radius allows for pickup of posts that are well outside of CWA (though this may be a good thing depending on the location of the posts.)

 

2.      Large radius does not allow for focus/granularity of where posts are coming from in an efficient manner.

 

3.      Posts may, or may not be related to desired subject matter based on keywords used.

 

4.      The disadvantages above are not exclusive of each other, meaning that they feed from each other in compounding of the issues noted.

So, the above approach creates significant noise in the office’s Hootsuite feedwatch and requires analysis that does not generate value.

 

 

 

Future State

So, some thought was given to the drawbacks.  All pointed to the size of the feedwatch footprint being the root cause.

 

But...Twitter tools and messaging features still provide for better analysis opportunities than any other mainstream Social Media tool for this purpose.  Short and sweet is really what is required.

 

So, with the above in mind, what can be done?

 

How about making the search areas more granular (smaller) and use volunteers to watch the areas?

 

Some advantages to this:

 

            A individual volunteer can manage more than one feedwatch/area

            Volunteers can be located anywhere

           

Next...what is the process?

 

Plan is to use the Twitter/Tweetdeck accounts created with pre-established guidance to not only be used as team accounts for forwarding reports within the intended scope, but also for Retweeting any other posts that meet criteria.

 

The Milwaukee Forecast Office has a Hootsuite feedwatch that has only the team Twitter/Tweetdeck account handles in them and will parse the reports originating from the team along with the Retweets via the team (as long as the content being retweeted meets the needs of the program being supported.)


 

See below, and the next diagram.

 

TweetDeck Account Radius Set.jpg

 

NOTE:  The map above is only an illustration. There are currently eleven (11) teams in SulCom.

 

 

 

Above is the estimated coverage of the eleven (11) current teams in SulCom.  Gaps can be seen in the NE and SW portions of the CWA.

 

 

Setup

1.   Twitter account creation is handled by the forecast office or program management.  This allows the recipient (in this case the Milwaukee Forecast Office) to understand the general location of the content.

 

2.   Geocoding and keyword search strings for Tweetdeck feeds will be created and used for capturing activity in a given area.  This should adequately parse out the weather observation activity from the general activity.

 

3.   The volunteer’s twitter account will be linked/added to the team twitter account created in 1, above.  This allows the volunteer to communicate on behalf of the team twitter account.’

 

4.   The volunteer will set up feedwatches* in Tweetdeck using the feed parameters developed in Number 2, above.

 

*Why “feedwatches” (plural?)  A volunteer can watch more than one area and be set up to forward on behalf of more than one account.

 

Now for the disadvantage - Tweetdeck is not available for use on mobile devices.

 

Process

1.   The “Feedwatchers” If a message containing appropriate/relevant content is captured, the feedwatcher simply retweets the message.

 

As long as the feedwatcher is retweeting on behalf of the “team” account then the message will appear on the feedwatch monitored by the forecast office.

 

Result

The consumer of the information (the Milwaukee Forecast office in this program) receives relevant content to aid in, or validate the decisions they make for a variety of purposes without the need to sift through irrelevant and sometimes inappropriate content. This translates into the forecasters can spend less time analyzing data and more time doing their necessary tasks.