My scientific interests include: large-scale machine learning and pattern recognition, basketball, numerical optimization, and eating pizza with friends.
It seems like yesterday's big dust storm that originated in Nebraska and Colorado (see my previous post) has now traveled a huge distance. The affected states include Texas, Louisiana, Mississippi, and Alabama.
I would risk to say (with no evidence at this moment except by a common sense guess) that the storm came into Texas during the afternoon of the 18 of Oct. 2012. And traveled through Louisiana today early in the morning, and, this picture is captured a little before noon local time. The image shown here is produced by our "beta" Near-Real-Time (NRT) dust aerosol detection system. And is produced using NASA Terra-MODIS multispectral data.
As the dust is traveling through central and south-eastern US, it seems to be weaker and weaker. I would forecast that the dust would end up near the East coast of the US, but very weak.
I promise a complete followup on Sunday night.
--------------------UPDATE------------------
I was right! It moved to the East coast. Check the following video...
Today our system shows high probabilities of dust aerosols present in the states of Nebraska, Colorado, Kansas, and Oklahoma. It seems like a strong winds originated from the Dakotas to Oklahoma and possibly cold fronts provoked a dust storm of a large size. The area monitored by NASA-Terra-MODIS suggests that the dust storm size could cover the state of Oklahoma almost in its entirety.
A report from Tulsa, OK, says that the visibility in one of their highways was of about 10 feet, near black out.
"A massive dust storm swirling reddish-brown clouds over northern Oklahoma triggered a multi-vehicle accident along a major interstate Thursday, forcing police to shut down the heavily traveled roadway amid near blackout conditions. (Oct. 18)" - Associated Press. Here is a video from a helicopter (I guess).
Another official source says the following:
"Strong winds in the Plains today have caused travel disruptions from the Dakotas to Oklahoma. This picture shows several trucks blown over on Interstate 90 near Belvidere, SD. This is just one of several reports of traffic accidents in the region caused by strong winds and/or blowing dust." - U.S. National Weather Service.
Our Near-Real-Time (NRT) Dust Aerosol Detection System (DADS) for the date can be accessed here: NRT DADS. In this site you can download the data granules corresponding to this dust storm, which are 16:55 (UTC) and 18:35 (UTC), both corresponding to multispectral data processed from the NASA Terra-MODIS instrument.
More information (local news) for Oklahoma can be found here and here.
Also for other reports about Kansas can be found here.
From our NRT-DADS system we produced the following images:
On the left is the true color image of the dust storm. On the right the detected probability of dust aerosols. The darker the less probable, the lighter the more probable.
This is the full-size image granule from MODIS 16:55 UTC.
This is the full-size dust aerosol detection system's output.
This is the true color detection with a color coded area where the dust aerosol probability is higher.
Today I had the privilege of being asked by a "Dust-Storm" expert the following:
"There is massive amounts of dust today in New Mexico and El Paso. But
on the satellite I can't easily tell which is dust and which is cloud.
Can you run your algorithm today and see if it differentiates them?" - Dr. Thomas Gill
Unfortunately, the algorithm was "trained" or "taught" to discriminate clouds, that is, wherever there is a cloud, don't bother searching for dust. The original premise by some NASA experts and I was that, since my algorithm uses thermal spectral bands near the infra-red spectrum, any cloud would significantly deteriorate the thermal emission of the dust aerosols; in other words: it would mess up the whole algorithm. So anyway, what I am trying to say is that all clouds are "mapped" to black (non-dusty) aerosols.
Below are the results of my algorithm. On the left we have true color images and on the right we have the output of my algorithm; on the first row is shown MODIS-Terra and on the second MODIS-Aqua, 17:35 UTC and 2050 UTC, respectively.
This dust storm was classified as "severe" and it is probably the worst in the last five years. Here is the full overpass of MODIS-Terra and my algorithm's output...
This is the GOES East animation of the event; as you can see, there is a large number of clouds moving over the area of the incident, making almost impossible to analyze using traditional infra-red imagery.
Finally, here are some pictures of friends and videos uploaded by other people.
