The Human METAR

So after 10 weeks of work, and approximately 200 escorts in and out of the building, today is my last day here on the Cape Canaveral side of NASA. Tomorrow is the official last day over at KSC, where I will be presenting some of my thesis results. Then it’s one more weekend down here and I head back to Mass on monday with the hopes of arriving Tuesday night. It has been an enjoyable experience down here and I have many people to thank for letting me come down here and do thesis work.

Speaking of that, it’s only proper to give everyone a little taste of results (afterall, that’s why I am here). So here we go:

PROBLEM: Forecasting for convective winds down here are tough, due to the driving physics of the area.  Nowcastng these events with Radar is even trickier, because it is hard to determine which cell produces the gust, where and when it will occur, and what the maximum expected peak gust is.

PREVIOUS WORK: Work done by a former Plymouth student showed some promise in this area through regression based wind gust equations. Additionally a relationship between the height of the maximum reflectivity and peak wind gust was theorized:

• If the maximum reflectivity is located ABOVE the freezing level, then there is a higher probability of the formation of hail. Melt water from hailstones has been shown to enhance evaporative cooling, thus generating stronger downdrafts. If this case is true, generally the wind gust is greater than the 45 WS threshold of 35 knots
• If the maximum reflectivity is BELOW the freezing level, then the wind gust is less than 35 knots

ANOTHER PROBLEM: While these results look great, they were generated using a small dataset of 44 cases. Also it would be ideal if these relationships could be correlated at earlier volume scans to provide a longer lead time to the forecaster.

RESULTS: Using an updated dataset from 2003 – 2008 (~ 310 cases for onset) the previous regression based models do not appear to be as promising, even with earlier volume scans. However some promise can be made out with the hail potential chart:

Two important things can be shown here (These results are consistent with ALL volume scans):

1. Whenever the height difference is positive, this means the maximum reflectivity is above the freezing level, and the wind gust will be > 35kts. While the probability of detection is high, there is also a chance for a high false alarm ratio
2. There tends to be a linear increase in the maximum reflectivity as the peak wind gust increases

NEW STATSITICAL METHODS: Many new methods will be introduced in the future part of thesis work, as they are not complete today, however I would like to show a classification tree for determining a convective event:This tree was generated with values from scan4 (depending on the radar’s VCP, these are values approximately 16 to 25 minutes PRIOR to peak wind gust occurence). The tree tests to see if the maximum reflectivity is greater than 55.5 dBZ. If it is, then a wind gust greater than 35 knots will occur. If not, then it checks the cell’s echo top. If it is greater than 27,700 feet, then a strong wind gust will occur. And vice versa.

Testing this against an independent dataset, the probabilty of detection is 75% and the false alarm ratio is 40%

THE END!

So there is a quick summary of some of the work I have been doing down here. My hope in the fall is to readjust the CART algorithms using techniques such as boostrapping and bagging. So far I have about 25 pages written on my thesis, and that’s just intro / data and methodology. I hope to double that by the end of september.

But I’m not thinking about that until I go back to Plymouth

Good Morning! Just about three weeks into the internship down at Cape Canaveral, and I have already made some progress.

Currently I have 5 years of NCDC Storm Structure data for all the convective events between May and September on the KSC/CCAFS Complex. The Storm Structure data basically tells the user information about an individual convective cell on RADAR. Work performed by Andrew Loconto a few years back showed that one can possibly derive the peak wind gust from this information, however hehad a limited dataset. My next step is to go through all the files (about 4500) and determine which cell is closest to the wind tower that recorded the peak wind gust, and extract specific information and place it into a nice dataset for analysis. I am currently working on that via a perl script.

I have also seen a lot down on the complex. Yesterday alone we got to see a weather balloon go up (KXMR for you weather geeks), as well as visit the Cape Canaveral Light House, and a close view of the launch pad that contains GOES-O. Speaking of that, the Delta rocket will bring GOES-O up to space tomorrow evening, if weather permits. If all works well, GOES-O will become GOES-14 and become a backup to the weather satellites in operation. For more information, go here

and Finally, for the radar nerds (….me). Our vortex page got a new addition to the weather data database. We now have Terminal Doppler Weather Radar data for most of the United States. They work just like WSR-88D’s, but their wavelength is smaller, and so is their range. However it is “high-res” on velocity data. This is why most of the TDWR’s are located near airports, to detect low level wind shear and any turbulence in the area.

The data can be found here

One of the tools weather forcasters use for nowcasting tornadoes is the use of Radar. The Radar images you see on the news most of the time is the base reflectivity, which essentially shows where it is precipitating. But the WSR-88D can produce many different types of products. One of the other main products is radial velocity, which determines if the generic flow is moving towards or away from a radar.

The generic rule of thumb is the following:

Green: Inbound Velocity; flow towards the radar
Red: Outbound Velocity; flow away from the radar

Using this concept, one can look at the radial velocity and see the synoptic flow. But if looked at very closely, one can depict small scale phenomena. A velocity couplet is an area where there is strong inbound and outbound velocity near each other. If positioned correctly, we can see cyclonic convergence. This indicates that the flow is counter clockwise, and converging into one spot. This can be one of the indicators for a TVS, or a tornadic vortex signature. When this is seen, there is a moderate chance that a tornado has, or will form.

Why am I bringing this up? Well we have had a severe weather day in the great state of Oklahoma. I plan on doing a case study sometime tomorrow (stay tuned!) But for now take a look at the image below (SOURCE: GRLEVEL3) :

The area over Pawnee, Oklahoma is experiencing a cyclonic converging event. It is important to note that the radar is SW of this image. Using that concept we can see that the inbound velocity is to the left (towards the radar, so a NE wind), and the outbound velocity to the right (away from the radar, so a SW wind). This indicates an area of cyclonic flow. Also notice how due south of Pawnee, there appears to be an area of red and green close together. This is an area of convergence. And finally, because this is all in the same spot, it is cyclonically convergent.

Notice the red box? That is a tornado warning that was issued by the National Weather Service earlier this evening. I guess they agree that this was a cyclonic converging cell capable of producing tornadoes.