I do, however, want to apologize to any twitter readers. I stuck heavily to my forecast without considering the dynamics from hour to hour. I was convinced we would see something, and I became invested in seeing my forecast come true. I do believe that this hindered me from giving accurate predictions as the day went on. I hope that this is a mistake I can learn from; being rigid in my forecast is not always a good thing, and it is not always a bad thing to alter views as time goes on.

Why did I get it wrong? I wish I could answer that. I didn’t have much time to watch the event unfold, so I don’t know what features changed and how. The main culprit, however, was predicting the speed of the cold front. This could have been altered by a shift in direction by the shortwave trough responsible for directing the low, or it could have been due to issues predicting the cutoff low over Mexico. Either way, the end result was a very nice, pleasant, warm, sunny day, with no severe weather. I cannot complain too much – I got to ride around with the windows down again!

As a parting word, I do encourage you to use plenty of other sources for weather. My blog and twitter are not designed to be one-stop-shops for weather forecasting. I just don’t have the time or energy to do something like that! However, there are plenty of great resources, including the National Weather Service in Nashville, as well as @NashSevereWx on Twitter. The truth is, there will always be times where I am either wrong or unable to post forecasts. I should not be a primary source; just consider me a second opinion.

As humbling as this was, I cannot wait for next week, where I can practice my forecasting once again! Until then,
Kelton Halbert.

I can tell you right now that a decent portion of aspiring meteorologists and forecasters have never heard of doing hand analysis. I know, because until recently, I was one of those people. However, when I was introduced to it, I was amazed. You mean to tell me that humans can do what the weather models do, with a colored pencil and half a brain? So, I started learning. Then I sought help, and that help was a grad student at the University of Oklahoma. Not only did he teach me how to properly contour the analyses, but he taught me the rules of thumb once used to make forecasts on these things. Then, using maps over a period of time, he demonstrated how the weather models had failed to analyze several shortwave features over the week, features that our hand analyses caught. Things went down hill from here. (BELOW: One of my many analyses for 500mb)

Hand analysis isn’t just a method of forecasting. If a computer can pull off such advanced calculations that often say the *exact* same thing as a hand analysis, then why bother? Hand analysis isn’t just a part of the science of forecasting; it is an art. Getting involved with the data at every individual point gives you a perspective on weather that no weather model can give you. It teaches you how the data was gathered, what the limitations of the data are, where the data is wrong, and lastly, how to interpret the data. It introduces you to the art of pattern recognition. A weather model can spit out scenarios at you all day long that may never come true. However, an experienced forecaster that knows the data, how to interpret it, the limitations of the data, and plenty of experience observing that data, he can beat the pants of of a computer model ANY DAY OF THE WEEK. How? He recognizes this pattern and knows how it should generally resolve. Meteorology isn’t just a mess of calculus inside of numerical models. It is about recognizing patterns, and those patterns are *EVERYWHERE*! Who do you think came up with the idea of a lee trough in the Rockies? Someone who did hand analysis and noticed that there was a near permanent feature there that seemed to spit out cyclones! There was a pattern, and someone investigated that pattern.

Looking at weather models does not teach this. Future meteorologists are becoming married to models like the RUC, NAM, and GFS from an early age, and it sticks with them through their professional career. When this happens, those meteorologists are often robbed of the opportunity to gain experience and observe patterns. I’m not saying that we need to get rid of numerical models; not at all! I just think that the importance of them needs to be tempered with the knowledge that they are incredibly flawed. I use hand analysis as a way to make short term forecasts. When I’m looking long term, of course I use a weather model, but I check that model with a hand analysis. Is this model ignoring a feature that could otherwise break the forecast? This is how hand analysis can become incredibly powerful. This is how you can make an accurate forecast when the weather model pukes on you and develops a cyclone out of nowhere! I’ve seen that happen! Hand analysis is dying, and with it, the art of reasoning and pattern recognition. People now see forecasts without understanding, which is extremely detrimental for when something goes wrong and the forecaster doesn’t know why.

Here’s the real kicker. Guess which agency is one of the last to routinely use hand analyses in their forecasts? The Storm Prediction Center. LINK: http://www.spc.noaa.gov/faq/tornado/sfcanls.htm

“Despite the powerful computers, there is no substitute for drawing weather maps by hand for making a forecaster take the time to thoroughly understand the ongoing weather situation. And without knowing the intricate details of what’s happening now, a forecast can suffer. So SPC forecasters routinely draw — by hand — surface and upper air features on printed maps, many times per day. This is a piece of a surface map containing lows and warm fronts (bright red), highs and cold fronts (blue), outflow boundaries (purple dash-dot), pressure troughs and isobars (dark gray), isotherms and warm spots for temperature (dark red), isodrosotherms and moist spots for dew point (green), a dryline (dark brown), and finally, snapshots of wind flow called streamlines (tan). It may look like a jumbled mess, bad food or abstract art; but to a severe storms meteorologist, it is stuffed with useful information.”

If you want to know more about how to conduct hand analysis, I encourage you to email me. I can get you started with some links on how to properly contour the map, but ultimately, it is your own experience and pattern recognition that will teach you how to make a forecast.

To download AWIDS, head over to my GitHub  (link: https://github.com/keltonhalbert/AWIDS ) and click on the ZIP button (highlighted in the image above). Installation instructions, information about the program, and instructions on how to run the program are all included in the README. Be sure to read it thoroughly before emailing me with bugs/issues!

This is an un-documented pre-alpha release. For programmers interested on building on top of this package, I’d recommend waiting before modifying the package. The next release should include thorough documentation, and many other significant changes to code structure as things are reworked to be more efficient. My hope, however, is that a public release will allow for other programmers to add their suggestions, report bugs, and make the program better as a whole. Feel free to contact me with bugs, reports, suggestions, or success stories!

A lot of the deaths appear to boil down to a few factors, but even those seem skewed. People who took the proper action died, and people that took no action lived. Therefore, there is no real finger pointer that wins here. The hope is that by better structuring the tornado warning process, people will be more likely to respond. However, one such speaker at AMS really captured my attention. Kim Klockow, a PhD at the University of Oklahoma really hit hard with what she said. (When the link comes from the AMS talk, I’ll be sure to post it). Klockow said that the first thing we need to do is to stop trying to control the people under a tornado warning. We cannot force them to act a certain way, no matter how hard we try, because it’s their right to act on their own free will, tornado or no tornado. This conclusion came about in her study of interviewing tornado victims and asking about their experience in the warning process. Almost all of the people Klockow interviewed took action, but the interesting thing was the train of verification it took for them to take action (again, the slides will make this easier to see when they come out). Several people only took action after verifying the tornado warning with 3 or more sources; some took an indirect action towards the tornado warning. For example, if the warning was issued and the person heard sirens, they would go inside, turn on the TV to the news station, and get their shelter area prepared and then wait for their warning to be verified on the TV. Klockow argues that this decision process of incremental action is natural, and the people that did this survived.

Therefore, my point to this post is to emphasize what Kim Klockow said in her presentation; it is not right of meteorologists to try to control a person’s actions to achieve a desired outcome. This sounds sinister, but it isn’t. Most meteorologists have very good intentions – they want to save lives in the warning process. However, if we acknowledge that people are smart enough to make their own decisions, then it changes our job from “how do we get people to respond?” to “how do we give people the most accurate and timely information so that they can make critical decisions?” THAT, right there, should be the core focus of the coming discussions about tornado warning effectiveness. In its current state, tornado warnings are good, but not effective enough. We’ve come a long way, but now it is time to go further yet. Hopefully you found this post intriguing, and I hope to get the slides up soon.

As far as my explanation, I am at the American Meteorological Society conference in New Orleans, LA. The image above (this morning’s sounding by LIX) is the fruit of this conference; Patrick Marsh released his new sounding analysis program called SHARPPy. I’ve been here since last Friday, attending talks, meetings, presentations, and the whole lot. I got to meet Reed Timmer, the now retiring Director of the National Hurricane Center, and Dr. Josh Wurman of the Doppler on Wheels (DOW). In other words, I’m in weather nerd heaven and having a great time. I return home Saturday and will be able to resume normal life, if life permits.

No severe weather is expected for the Middle Tennessee area over the next few days.

(LEFT: Surface Theta-E from a cold front on December 5th at 18z)

The decoder:

• New approach going from list indexing to loading data into dictionaries. This brings about an incredible speed boost in sifting through thousands and thousands of lines of METAR data.
• Several decoder bugs worked out while sifting through data
• NEW DATA SOURCE: Now reading data from http://metfs1.agron.iastate.edu/data/text/sao/ - This also speeds up the decoder because international stations like Iraq are not in these files, speeding up the sifting process.

GRIDMAKER:

• This new module handles ALL grid making processes, simplifying the plotting process and making the syntax more uniform.
• NEW GRID TYPES: 3 hour tendency (change over time) grids for all scalar outputs, new special grids (divergence, vorticity)
• NEW SCALAR GRIDS: Theta, Theta-E, Mixing Ratio, Relative Humidity, LCL Height in meters, SLP (Sea Level Pressure), wind speed scalar, wind chill, dewpoint depression, 3 hour tendencies of all of the above, and two new special grids: divergence and vorticity, including their 3 hour tendencies.

New Output Grid:

• Changed from a uniform 0.5 degree grid to the RUC 40km grid, the same grid used by the SPC’s mesoanalysis pages (Thanks again Patrick Marsh!)

More Data Observations:

• Canadian METAR sites
• Mexican METAR sites
• Caribbean METAR sites
• Cuban METAR sites

UI Improvements:

• Constructed a command-line UI – GEMPAK veterans need not worry; the command line syntax is based off of the GEMPAK model. I didn’t want to fix what wasn’t broken. This familiar syntax will help seasoned meteorologists and new users alike to be able to easily navigate the program without too much hassle.
• Plans for an improved GUI

And that’s just what has already been done – I haven’t even touched on what I have planned! Things like frontogenesis, advection of scalar quantities, and upper air maps are in the process of being planned out. I also have not forgotten about the meteogram plots; even though no new development has taken place on this program, a lot of the improvements in the mapping program will apply to the meteogram program. How interesting would it be to have meteogram plots of conserved quantities like mixing ratio and theta-e? Or meteograms of values like the lcl height? Could be very interesting!

All of this post is to say, some exciting progress has been made. I figured it would be good to post and update since the last post drew so much of a positive response. However, one VERY important part of the program remains unfinished – THE NAME! I was thinking that in honor of AWIPS/GEMPAK, I would name it PyWIPS: Python Weather Interactive Processing System. It sounds cool, but I don’t think the name makes that much sense. Perhaps I should turn to the interwebs for help! What do you think the program should be named? (RIGHT: Surface Vorticity from December 5th at 18z. Values are multiplied * 10 to the 5th)

So what is Python? Well, I won’t go into too much detail, as any Google search could probably enlighten you better than I. Regardless, Python could be summarized as just another programming language. However, that would not be the truth. The truth is that Python is a very powerful programming language. Confused? Google. They use Python!

So what am I doing with Python, and how is it weather related? Well, I’m doing a lot, actually, and it could be very useful data if the community so desires it. At first I decided to write a program to simply decode METAR (weather station) data from the internet, but then I realized I could do more with that data. So, I am attempting to write a full fledged surface data analysis suite that will include meteograms, contoured plots, station plots – you name it! How is this useful? Well, as far as I am aware, there is no program that shows the raw (contoured) station data on a map (station plots are a different story) – most data you see is from a model analysis, save for some programs that will show the METAR station data, but won’t visualize that data beyond an ordinary plot. The second reason that this is useful is because there is only one program that can come close to doing this kind of stuff, called GEMPAK, which I use often. However, it is a pain in the butt to build, is extremely complex, cumbersome, and quite “buggy”. Where it does succeed is in its high level of customization and its diversity in showing a large array of products. What I am hoping to do with this program is provide a lightweight, fast, and functioning alternative to GEMPAK while keeping that customization intact. Python just happened to fit the bill!

Obviously I am not starting ENTIRELY from scratch here. The only “from scratch” part is the program that gathers and interprets the data. The rest of it is made possible with MATPLOTLIB via ENTHOUGHT. Matplotlib is what I am using to visualize the data and does a lot of heavy lifting.

So what do I have so far? The two major ones are meteograms and surface contours (with wind barbs). An example of the meteogram can be found above: It’s bare bones right now, only taking in temperature, dewpoint, and wind data. In the future, however, it will also support pressure data and possibly precip. Meteograms are great for viewing data over time for a particular station. The one above is quite interesting, for station KOUN (Norman, Oklahoma), and if you look closely, denotes the passage of a cold front. The other plot I have completed is contouring the surface data on a map. This proved to be quite a challenge. Thankfully I had a bit of help from Patrick Marsh, so a big thanks to him! I won’t get too detailed, but the data is interpolated onto a lon (x) lat (y) grid (like a graph) and then plotted on a map via Basemap, a part of MATPLOTLIB. This is an example image output to your right (the program supports all major image outputs – PNG, GIF, JPG, PS, PDF, etc.). Quite visible in the map is a surface cyclone with a trailing cold front. Again, this is ALL RAW METAR DATA. No model analysis. Pretty cool, huh? The best part is, the program supports interactive viewing, with pan and zoom functions available. However, the user interface is primitive, at best.

So what now? Well, it may look like a lot is done, but this is only the bare guts. The next step (which I am in the process of doing) is to optimize the code and make it faster. After that, the user interface will need to be built, followed by plenty more code revisions. Another huge step is to make it more interactive with the user, taking in options that the user can specify. As you can see, there is a ways to go! But, because the foundation has been lain, I figured it was time to make this public. If you are all intrigued and would like to know more, you are more than welcome to email me or message me on facebook/twitter/google+. I’m going to refrain from posting any code online, but if you’d like to ask questions or even make suggestions (help is always appreciated), give me a holler.

Hopefully this post wasn’t too terribly boring, but I wanted to share the project with the interwebs. You’ll be seeing more posts and photos about this in the future as things continue to progress. It’s really coming together quite nicely.

On a side note, this experiment was part of the NOAA CAS Weather Camp at Jackson State University. We were observing the rotation and rising motion within the tornado. Other campers involved in this experiment were Rachel, Steven, Kevin, Kiya (probably spelled it wrong, my apologies) and Max.

Severe weather forecasts will resume once I return home, if anything of interest appears.

-Kelton Halbert

So, what makes this risk of particular interest? Well the days of May 9th-11th are showing potential with a low pressure system over the plains. It is still a little too far out to rely solely on the GFS, but it is suggesting temperatures over 90 degrees, dewpoints of 80 degrees, CAPE values over 4,000, and some moderate wind shear. Minor details such as timing are still unclear, and the values above are subject to change, but the SPC says it best.

ALTHOUGH THERE ARE DIFFERENCES IN THE GUIDANCE…THEY DO AGREE THAT BEGINNING ON SUN/D6…A LARGER SCALE TROUGH WILL DEVELOP ACROSS THE WRN STATES…WITH RIDGING DEVELOPING AHEAD OF THIS TROUGH OVER THE PLAINS STATES. THIS WOULD PORTEND A GREATER SEVERE THREAT…WITH THE POSSIBILITY OF 3-4 CONSECUTIVE DAYS OF SIGNIFICANT SEVERE DEVELOPING FROM THE PLAINS STATES INTO THE MIDWEST BEGINNING ON MON/D7 INTO TUE/D8 AND LIKELY EXTENDING INTO D9. IF TREND IN MODEL GUIDANCE PERSISTS…SEVERE RISK AREAS WILL LIKELY BE FORTHCOMING IN SUBSEQUENT D4-8 OUTLOOKS.

So, I will continue to keep an eye on the situation for Oklahoma. I will also update the website before I head out, and be sure to keep everyone not only up to date with how the trip goes, but with any significant weather that occurs here in Middle Tennessee. Enjoy the quiet weather while it lasts!