Nov 30-Dec 1 2010 HSLC post-mortem thread

According to the current SPC log, there were widespread severe wind reports (at least in the Piedmont), but no tornadoes in GA/SC/NC/VA.  So… is this a null case?  

While the episode is fresh in everyone’s minds (as of this typing, Wakefield, Sterling, and Morehead City may not be 100% out of the woods yet), I would be interested to know more about the procedures followed at each office.  What were you looking at in making warning decisions?  Were there specific clues about the less-than-prolific tornado production that you were able to identify operationally?  Clearly, ingredients were at least sufficient for the continued hoisting of tornado watches, so I am curious about how this shook out from a WFO perspective. 

All the best,


P.S. For those who are interested in the HSLC organizational teleconference (we are targeting mid-December), stay tuned.  I will be contacting people on the HSLC e-mail list in a day or two.  The 11/30-12/1 case ight serve as a nice ice-breaker for our discussions.

This entry was posted in CSTAR, High Shear Low Cape Severe Wx. Bookmark the permalink.

17 Responses to Nov 30-Dec 1 2010 HSLC post-mortem thread

  1. nws-pat moore says:

    Not so fast, my friend! I’m not sure we can call this a null event.

    We’ve got survey crews out right now in Easley (Pickens County, SC) and near Gray Court (Laurens County, SC) looking at damage. Both storms had excellent rotational signatures, so we might have at least two confirmed tornadoes.

    There is also the damage down in Gwinnett County, Georgia, that might have been produced by a tornado. On the FFC radar, the storm that was responsible had the appearance of an embedded supercell.

    As for procedures and warning decision-making, I’ll get back to the blog later today. I was the guy on the hot-seat yesterday, putting out the warnings, but I’m on shift right now and have to get to the forecast. I also need to start archiving the event. I can tell you that it was a frustrating evening, like most of these HSLC events.


  2. Chris Wamsley says:

    We basically were looking for notches in the line itself and some tightening to the signature. Line feeders ahead of the main line never really phased into the line segment to allow for a decent spin to dvlp aloft enough to spin twrds the sfc. Therefore we continued to basically warn for svr with ready mode (polygons) for areas that if tightened were to be raised to a tor wrng. Problem that resided was the needed sbcape needed was over ne nc/se va…well ahead of the main line and never matched up well for enhanced lift/forcing to allow for a decent spin up. Plus…the gulf stream started to become convective late in the watch which robbed us of additional moisture/instability as the line moved twrds the coast.

  3. nws-pat moore says:


    The damage near Gray Court, South Carolina (Laurens County), has been confirmed as having been caused by a tornado. A Public Information Statement should be available on our web page. Here are some statistics…
    damage rated EF1
    path length 2.25 miles
    path width 200 yards
    time 0243 UTC 1 December (943 PM).

  4. nws-pat moore says:


    The damage that occurred yesterday evening in Easley, South Carolina (Pickens County), has also been confirmed as tornadic. A preliminary Public Information Statement is available on our web page. Here are some preliminary stats…

    damage rated at EF1
    path length 2.5 miles
    path width 125 yards
    time 0043 UTC 1 December (743 PM).


  5. nws-pat moore says:

    Looks like the Peachtree City NWS office has confirmed the damage in Buford, Georgia, was caused by a tornado, rated at EF2. They also have a Public Info Statement issued with the survey report. That damage might have occurred from a storm for which there was no warning in effect. That would make it a good candidate for further study.


  6. nws-pat moore says:


    Now that I have some extra time…I will address your questions. What follows is my opinion based on my experience and does not necessarily reflect any standard operating procedure at WFO Greenville-Spartanburg.

    1. I have found that situational awareness of the HSLC environment is half the battle. I evaluate the model data up to 84 hours in advance to look for bulk shear in the 0-3km layer of greater than 40 kt and ANY amount of CAPE. As the event develops, we evaluate the mesoanalysis on the SPC page. We believe that storm relative helicity (SRH) is not important for brief tornado production in QLCS because the vorticity is generated by a descending rear inflow jet tilting crosswise vorticity. That doesn’t mean we don’t necessarily look for SRH and are not impressed by it. In fact, the higher the environmental SRH, the more we expect embedded supercell-like structures. It is just that SRH is not necessary because streamwise vorticity is not as important. As for CAPE, it is not necessary to have much if any surface-based CAPE. We suspect that dynamically induced upward accelerations inside the QLCS are more important than any updraft supported by buoyancy.

    As for warning decision-making…that is a bit more difficult to answer. First, I cannot stress enough how important it is to have a conceptual model. Just about everything we do with severe weather warnings relies on some sort of conceptual model, be it a supercell, multicell, or single cell. This would be one important outcome of our study…that is…conceptual models of HSLC storms.

    I have found that in HSLC situations, all the interesting stuff is happening so low in the storms that I only need to evaluate the lowest three or four elevation scans, unless the QLCS is very close to the radar. That cuts down dramatically on the amount of radar data I have to evaluate, which saves time that can be used for carefully examining the lowest two or three tilts. (The radar should be in VCP 12). I evaluate the reflectivity data for any signs of a line echo wave pattern (or the “S” before it breaks). I will issue a tornado warning if a break in the line segment is seen. Breaks some time begin at mid levels and continue downward. Attention is also drawn to inflow notches, weak echo regions, and rear-flank downdraft signatures that might indicate an embedded supercell. I am very suspicious of weak echo channels and areas of low (or no) reflectivity that develop in the trailing stratiform region and then move toward the high reflectivity in the QLCS. To track these features, I have found that “Feature Following Zoom” is an excellent way to see what is going on in a storm-relative view. As for storm-relative velocity, the lowest elevation scan is generally the most important for evaluating storm-relative motion, because many of the mesocyclones form at low levels and develop upward in the storms. This is unlike the supercell model where mesocyclones form at mid-levels and develop downward. I keep a sharp eye on the 0.5 and 0.9 degree SRM and use the VR-Shear tool to find the strongest shear along the line that corresponds to some kind of low level reflectivity feature, like an inflow notch or “S”. When the shear (typically gate-to-gate) gets above about 0.03 sec-1, then I issue a Tornado Warning. I have studied several events that indicate Vr is not a good indicator, but shear is.

    As for warnings, I try to issue a larger Severe Thunderstorm Warning ahead of the line to cover any straight line winds (60 minute duration), and then issue smaller embedded Tornado Warnings (30 minute duration) for any g-t-g shear or velocity signatures or Broken-S signatures. A wise person once told me to issue tornado warnings for rotating storms when you are in a tornado environment.

    2. As for the second question, my short answer is “no.” You’ll notice that I issued several Tornado Warnings that did not verify. If I am on the fence, I would rather warn and have a false alarm then NOT warn and miss an event. Action is rewarded, inaction is penalized. Any discussion about “game-playing” or philosophical questions about the value of over-warning for brief and minor events is left as an exercise for the readers of this blog.

    3. You’ll have to get the SPC perspective on the Watches. I was glad to have them.

    Feel free to comment. I await your rotten tomatoes.


  7. ctccbc says:

    I was working the mid shift on December 1. While the main line of convection was well to the west over the Central and Eastern Piedmont, a few discrete cells and line segments began to develop over Southeast North Carolina. At about 2:30 am, I noticed a classic broken-S signature over ILM’s CWA, about 2 counties away from our southernmost county, Duplin. The broken-S was accompanied by a strong gate-to-gate shear and ILM issued a tornado warning. While no verification occurred with that storm, broad circulation continued into Sampson, where RAH issued a Severe Thunderstorm Warning. The storm continued into our Duplin County, with fairly broad rotation, where I issued a Severe Thunderstorm Warning. Some power lines were blown down, so we were able to get lead time. Other than this one cell, no severe weather occurred in our CWA. Despite the high shear values, it seems that we normally need at least some marginal CAPE values (i.e. at least 500 j/kg) to get severe weather near the North Carolina coast. I have some screen captures of the broken-S reflectivity and velocity that I would like to share, but am not sure how to post them on here.


  8. Michael Cammarata says:

    Our radar operator provided me with this description in answer to Matt’s questions.


    Based on the strong low level jet…50-60 kts…strong vertical wind shear…limited (weak surface instabilities) and low top storms (30kt ft or less).

    1. Much of the event involved a broken line of showers (very little lightning).
    2. Looked for segments along the line that indicated bowing.
    3. Monitored bowing segments for “comma head” type developments for tornado potential.
    4. Some of the “comma heads” did begin to show some rotation, however they dissipated after a few volume scans or moved into GSP’s CWA.
    5. Monitored 0.5 degree Base Velocity Data (looking for low level strong winds)
    6. Monitored 4 panel combinations of Reflectivity Data/SRM data, looking for developing areas of rotation.
    7. However…most of the warnings we issued were based on bows that formed along the line of showers and thunderstorms or individual cells that began to bow. The VWP was useful in verifying the model forecast of the significant winds that were expected to move across the region during the evening/overnight hours. Almost any significant shower/thunderstorm had the potential to mix down some of the stronger winds, that were just above the surface (3-5k ft.).

    Extra…many of the showers/storms did not appear to be very significant, but the experience of the radar operator (Steve) of past similar events (low top convection/weak surface instability/strong low level jet/little lightning) added in the issuing of warnings for the CWA. It also appeared the the best focus for development, especially in terms of tornado development, was nearer to the remnants of the surface warm that was located across the upstate/North Carolina…possibly better directional wind shear. Jeff noticed a pressure rise/fall couplet that was located from Alabama northeast across northern GA, upstate of SC into western VA/NC.

  9. nws-pat moore says:


    Excellent comments, especially in your extra section.

    Yes, experience has taught us to not let our guard down for lines of showers or storms that do not appear to be very significant in the radar imagery. We pay attention to the lightning data because any amount of cloud-to-ground lightning is indicative of a stronger updraft, but a lack of lightning does not factor into the warning decision. I have seen too many of these events where a tornado occurs with a complete absence of cloud-to-ground lightning.

    I would agree that it would appear that the best tornado potential occurred in a corridor from the Atlanta metro area to the west side of the Charlotte metro area. We also noted the pressure fall/rise couplet that moved across our forecast area. Some remnant of the warm front may have also been located close to the Blue Ridge in South Carolina late in the afternoon as well.


  10. mdparker says:

    Hi All,

    This is fantastic! Keep it going! I am learning a lot from reading all of these accounts, and hopefully it is useful to other forecasters to read procedures and war stories as well. I am going to continue to stand back and absorb as much as I can from this thread.

    In the mean time, I plan to send out a scheduling questionnaire for the HSLC telecon later today.


  11. nws-pat moore says:

    …And another thing…

    Just a few more comments to add to what I posted yesterday afternoon….

    I can’t underestimate the value of experience (most of it bad, unfortunately) in dealing with these HSLC-QLCS-type events. It has taught me to have the courage to issue tornado warnings when there is no cloud-to-ground lightning and no surface-based CAPE. Without that experience, I’m not sure I would have issued a Tornado Warning for the Easley storm or the Gray Court storm. As it is, we had tornado warnings in effect for both of our tornado events, albeit with very little lead time.

    Most of the clues in these events are subtle, so I keep an eye out for anything that looks suspicious or out-of-the-ordinary, such as tightening inbound-outbound radial velocity couplets on the lowest two or three elevation scans, inflow notches, “breaking-S” echoes, weak echo channels behind the line, small curlicues at the back of the high reflectivity, large echo overhangs, and others. All of these things contribute to what some of us refer to as the “funky factor”. I tend to have a more itchy trigger-finger, so the weirder a storm gets, I’m more likely than others to issue a warning. I’ve learned to play my hunches. Perhaps that leads to overwarning. Seven of nine of our tornado warnings did not verify with a tornado. However, we don’t have any missed events for the time being.



  12. nws-pat moore says:

    Yeah, Matt, it’s a neat event for me, personally. Here I am, involved with a HSLC severe storm study, and I was the lucky (or, unlucky?) one to be the Warning Forecaster for this event. So, I get to relate my experience directly to the group, which is most convenient. If we had the time and resources, we could probably go through the KGSP radar data from scan to scan and I could tell you exactly what I was thinking and why I put warnings out (or didn’t) when I did.

    I’ll bet we could run ten forecasters through this event on the Weather Event Simulator and get ten different responses, some of which would probably be very different. This is unlike simulations of classic supercells, where we would probably get ten very similar responses. This is why I say we should shoot for better conceptual models of these HSLC storms to get all the forecasters thinking the same when they are interrogating radar data in HSLC environments. That would be a very positive outcome of the CSTAR research.


  13. Chris Wamsley says:

    Also interesting to note in our neck of the woods (besides also having no lightning) was most of the higher wind gusts by ASOS sites and home weather stations) were in the area where there was actually SBCAPE (wind gusts of 50-55 mph) far SE VA/NE NC.

    However, the svr tstm warnings that we issued were outside that area where there was no SBCAPE (gnrly along and west of the i-95 corridor) along and with the best H100-85 divergence depicted by local NAM/WRF (less wind gusts 40-45 mph) but breaks/notches in lines (lead to areas of higher winds where we got reports from), which we keyed in on since the intial post by Justin (worked very well). Plus, as PAT stated earlier, you do not need SBCAPE.

    Thinking if this event occured during the daytime hrs…there would of probably been isol’d cells ahead of the main line that could of been tornadic, but never had the chance to dvlp. Plus, the main line also pushed through the area of higher winds over SE VA/NC NC after sunrise. We did have the gulf stream off NC coast light up with lightning as cloud tops over inland areas (where the main line was pushing east of i-95) warmed after our warnings ended before sunrise.

  14. justingsp says:

    I was fortunate to be able to take over warning decision making duties for Pat during the late evening. I say fortunate in that thanks to Pat’s blood, sweat, and tears, I had a pretty good feel for what was going on by then and was able to construct a warning strategy accordingly. My concerns going into the event were (a) broken-S tornadoes, (b) embedded mini-supercells and/or bow echoes, (c) convective mixing of very high momentum air from aloft to the surface.

    By the time I took over, it was obvious that (c) wasn’t much of an issue. 88D Base Velocity when the convective line was near KGSP was unimpressive, and peak wind gusts at ASOS sites were only in the 25-35 kt range. Other than the tornado damage, the reports we were receiving were very marginal (i.e., a tree here, a tree there). Therefore, I decided against “shotgunning” warnings along the entire line, opting instead to issue broad-brush Special Weather Statements (for gusts to 50 mph) and to hone in on storm-scale features for warning purposes. However, there was a bit of a change in the character of the convection during this time. Whereas the QLCS often contained a number of (mainly weak) mesovortices at any given time as it crossed the southern foothills of NC and Upstate SC, this tendency slowed as the line moved into the Piedmont. I only recall identifying one short-lived mesovortex as the QLCS moved over the Charlotte TDWR coverage area. And this was so weak (Vr was about 20 kt at a distance of about 10 nm from the RDA), that I decided against issuing a warning, although I admittedly had my finger on the “Send” button for a TOR. I issued one SVR on a small structure that briefly exhibited bow echo characterisitcs (bookend vortex/pronounced rear-inflow notch), but even verifying this one was like pulling teeth. The only other warning I issued was a TOR for an evolving “broken-S” signature (another subsequent “broken-S” signature moved into the warning polygon a few minutes later). Neither is what I would consider a “textbook” example, but they generally followed our conceptual model. This warning did not verify, but there was some very low-end wind damage reported. There were a couple of other “broken-S”-like features that were not warned, mainly because most of these signatures occurred well within TDWR range, and I felt that even with a non-supercell tornado, I should see at least some degree of significant rotation near cloud base. In reality only very weak (i.e., less than 20 kt) and broad Vr was observed with these reflectivity signatures. Again, even with the broken-S signatures, base velocities were unimpressive. Nevertheless, I probably should have issued SVRs, since even in the absence of a tornado, the “broken-S” region is probably an area in which a heightened threat exists for damaging winds. In other words, I probably got too cute with my warning decisions. As far as we know, none of the unwarned broken-S signatures were associated with even minimal wind damage, but I imagine we’d probably be able to dig up a couple of downed trees if we really put some effort into verification.

  15. nws-pat moore says:


    Thanks for the comments and the kind words.

    The radar presentation as the QLCS moved across Union County, SC, certainly warranted a Tornado Warning (in my opinion) based on our Broken-S conceptual model. As it turns out, we uncovered some reports of sporadic trees blown down in the east central part of Union County between U.S. Highway 176 and the Chester County line, so that wind damage verified the warning as a Severe Thunderstorm. However, the reports were not compelling enough for us to send someone over to survey the damage. Ordinarily, we will not survey unless we have reports of damage to structures or some sort of concentration or path. Who’s to say that if we had sent someone to scour the countryside, we might have found a path of damage to trees somewhere in that part of the county. Union County is one of our most rural non-mountain counties. Much of it is within the Sumter National Forest and is sparsely populated. Suffice to say there are many square miles of uninhabited forest land that a brief tornado could hit and we might never find out about it. It’s another example of how we need to put an asterisk next to any “false alarm” or “null event” (i.e. unverified warning) in one of these HSLC storms because the damage path is typically so short such that it might not be found. One of our tornadoes just happened to touch down right in the middle of downtown Easley. The other travelled along State Highway 101, which is also relatively populated. If either of these tornadoes happened a few miles away from their actual location, we might not know about any damage.

    You might have already noticed the differences in the warning strategy employed by Justin and me. Two people in the same office doing two slightly different things, yet I think we have generally the same conceptual models. I certainly cannot fault Justin for any of the actions he took or didn’t take. I might have done differently, but not necessarily better or worse. (All of “his warnings” verified with some sort of damage report, only a little over half of “my warnings” verified). I don’t know if we will ever be able to take the human factors out of the warning decision making process. However, I still think there is room to improve our conceptual models for HSLC storms.

  16. mstrickler says:

    Thanks to all participants for the enjoyable and educational thread. I had hoped to post WFO RAH’s perspective sooner, but an extended time off from work (which I’m still enjoying) has delayed this response. Anyhow, I’d like to share a few thoughts with the group before they’re forgotten – thoughts from someone who was tasked with both this portion of the forecast for several days leading up to the event and with working radar during it. Perhaps somewhat of an outline format would be best, with the intent to simply provoke further thought and discussion leading in to our conference call next week.

    …Random observations, hopefully without restating any obvious ones…

    1) Discrete cellular development ahead of the line, depicted by the majority of the convection-allowing models prior to the event, was minimal and likely a function of dry and capped mid levels evident on proximity soundings on the evening of the 1st, and others upstream on the couple of days preceding the event. Does this in part explain the perceived null event, since those storms would have theoretically posed the greatest risk of tornadoes, prior to the arrival of linear forcing from the west and resultant transition to damaging wind dominance?
    2) Individual storm motion was fast (>40kts to the northeast on average), but the eastward translation of line itself was relatively slow. Storms tended to move with mean wind and lacked deviant motion “off the hodograph” and ability to ingest/act upon the high streamwise vorticity present. While important for supercell tornadoes, what is the significance, if any, of this process for QLCS tornadoes, which seem to be induced by downdraft tilting? Pat has already alluded to this idea. Can we investigate and better understand the differences between these two (IE. Supercell vs. QLCS tornadoes and importance of updraft vs. downdraft tilting), and ultimately fine-tune our conceptual model and discover/test radar precursor signatures for the latter?
    3) There was no pronounced STJ, which has been found to be a key component of many previous tornado outbreaks in the Carolinas. What truly is the role of the STJ? Isn’t synoptic-scale lift already sufficient for severe convective storms and tornadoes in these events? If not, perhaps the presence of a significant STJ would have provided adequate forcing for ascent to lift and remove the aforementioned capping aloft that seemed to suppress discrete development in advance of the line in this case.
    4) There was concern over the decrease in lightning activity and overall structure of the convective line (weakening) with time, and degree of destabilization northward owing to low level theta-e advection, which was restricted by southeasterly flow (that was increasingly dry with northward extent due to preceding cP air mass). While the low level flow became increasingly veered with time (to a more southerly component) and resultantly advected northward the reservoir of higher low level theta-e/weak instability into southern and central NC as the QLCS approached, it also induced more unidirectional flow/straightening of hodographs and decreased convergence and storm relative flow along the line, both of which presumably lessened the severe/tornado threat.
    5) While not unique to this particular event, one thing warning forecasters noted on shift was how warning operations for the radar operator(s) has evolved into constant data interrogation, due to the increasing amount of and access to additional data sources, such as TDWR, neighboring 88D, and AMDAR, to name a few. In the not-too-distant past, there was a brief amount of downtime to “catch your breath”, so to speak, as the “home” 88D scanned at higher, less important (particularly so for low CAPE/shallower convective environments) elevation angles.

    …Procedures followed and warning decision-making at WFO RAH…

    1) We did not employ anything ground-breaking or terribly different from what others have already noted, as we looked for bows, LEWPS/broken-S, and rotational velocity couplets, and also made use of proximity VWP’s.
    2) We did, however, use a couple of other techniques not previously mentioned, such as surveillance of updraft strengthening and resultant increased vorticity stretching/spin-up, and modification of RUC soundings with representative inflow air mass characteristics (per METAR data). In fact, early on in the event when the storms were about to cross the Yadkin River and into our CWA, this use of observational surface data in our sounding analyses indicated around 1000 J/kg SBCAPE in the NC southern Piedmont – >100% larger than that indicated by the SPC meso-page – and accordingly caused us to be more liberal with warning issuances.
    3) We made constant use of the SPC meso-page, most notably those parameters that include measures of low level shear, instability, and CIN.
    4) There were a couple of longer-lived, albeit weak circulations along inflection points of the line, for which we had TOR polygons/text drawn up when they briefly intensified (though still below our “threshold of pain”). However, since tornadoes in these environments often are not preceded by long lead time rotational couplets, if at all, we remained on high alert for updraft intensification above the circulation, which likely would have prompted us to TOR warn.
    5) As it became apparent that damaging wind production was also less-than-prolific as the storms continued eastward toward the Triangle, we became more conservative in our warning issuances and instead elected for mostly SPS’s. Both measured gusts and reports of damage decreased with time and eastward extent. Why? We suspect this at least in part was a function of the disconnect between the strongest deep layer forcing and the convection. Specifically, it appeared that forcing for ascent attending the deeper, synoptic-scale low level frontal zone (and more appreciable surface density gradient) and primary, negatively-tilted s/w trough aloft, lagged the effective cold front/squall line. This became increasingly so for the former owing to the slowing/blocking effect the NC Blue Ridge had on the low level frontogenetical forcing/circulation. Also, the horizontal pressure gradient and storm relative flow attending the passing mesolow and accompanying isallobaric forcing was most pronounced from KAND to KGSP to KINT – a corridor of densest damaging wind reports across the Carolinas. If memory serves correct, the pressure at KINT fell over three millibars in the hour preceding the QLCS passage and 41 knot peak wind, while the pressure fall/rise couplet and peak winds alike were much less across the remainder of central NC (but continued into VA along the track of the mesolow, which happened to become increasingly apparent in radar imagery as it moved northeast of KINT).
    6) Forecast pattern recognition and confidence was enough to highlight the severe threat and raise public awareness through the AFD and HWO beginning six days out.

    …This case, as have many others, again raises some other interesting points and questions, in no particular order of importance…

    1) What is the significance of pre-existing mesocyclones that often form in a higher instability environment upstream; and what are the criteria for their associated dynamical/rotational updraft forcing to overcome a relative lack of buoyant forcing as they progress downstream?
    2) Can we better differentiate damaging wind events that are efficient in convective mixing/momentum transport of strong flow aloft from those that are not? To what degree is this type of damaging wind event dictated by the storm-bearing and boundary layer flow direction incident to the convective line, and the horizontal pressure gradient? What are other factors? Additionally, what caused the November 16th, 2010 event to be a more prolific damaging wind producer (at least for central NC) than this Nov 30-Dec 1 event?
    3) We have noted the frequent existence of two surface frontal zones over the Carolinas/Mid-Atlantic – the wedge front and the maritime tropical warm front, the latter of which typically best delineates the leading edge of surface- (or nearly so) based instability and minimal CINH. If the mT warm front indeed delineates these characteristics and resultant threat of surface-based convection and severe weather, it is paramount it be located and analyzed properly. Surface analyses from HPC and WFO’s frequently fail to accurately portray one or both of these boundaries, though that’s one man’s subjective opinion. To what extent does the merging, or lack thereof, of these boundaries effect the production of severe winds and tornadoes in the Carolinas and Mid-Atlantic? In this particular event, it appeared the mT front overtook/merged (or came very close to) the wedge front from east central AL northeastward to west central SC, but then arced southeastward to the coast and away from the continually retreating wedge front over the NC Piedmont. Was it happenstance that the confirmed tornadoes occurred where the mT front/leading edge of higher theta-e merged with the wedge front?

  17. Pingback: CIMMSE blog – 2010 in review | CIMMSE

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s