Data-Driven Investigation of the Relationships Between Atlantic Tropical Cyclones and Landfalling Atmospheric Rivers in Northern Europe
Advisor: Ms. Covert
Internship Location: [Hybrid/Remote] NOAA CWPC - 5830 University Research Ct, College Park, MD 20740
Onsite Mentor: Vijay Tallapragada
Atmospheric rivers (ARs) are responsible for significant impacts across the subtropical regions, subjecting affected areas to heavy rains and strong winds. Over 30-50% of the total annual precipitation totals for Northern California, for example, are the result of atmospheric rivers. Atmospheric rivers that impact Europe, however, are a bit less studied. These atmospheric rivers originate in the Atlantic Ocean and affect the western coast of Europe, including the Iberian Peninsula, the British Isles, Norway, and France. Not only do Atlantic atmospheric rivers affect land, they also affect key transatlantic shipping routes. A noticeable portion of atmospheric rivers in the Atlantic are the result of active tropical cyclones (TCs), also known as hurricanes. Using datasets mostly available to the public, this project will form an understanding of the circumstances under which atmospheric river formation is linked to the presence of a tropical cyclone. I will conduct a case-by-case study of historical events determined through a visual analysis of satellite and model initialization data. I will analyze each case visually and computationally using Python, focusing on metrics like water vapor flux, relative humidity, and temperature aggregates. I ultimately hope to produce some sort of algorithmic framework to be utilized in existing forecast models to improve the predictability of atmospheric rivers.
Project Blog
Week 0: Introduction
Hi everyone, and welcome to my blog! Over the next few weeks, you’ll see me post weekly updates on my project, and learn a bit about atmospheric science, especially the science of atmospheric rivers and tropical cyclones, during the process!
Since childhood, I have been passionate about the study of earth science, particularly meteorology. During the pandemic, I spent a significant portion of my time studying tropical cyclones across the world, particularly in the Atlantic Ocean. After realizing that meteorology was something I enjoyed learning, I spent around a year and a half learning about everything about the atmosphere.
During the winter of 2023, a series of powerful atmospheric rivers impacted the northern coast of California, including the bay area. Torrential downpours and strong winds would bring down trees, flood roads, and cut power to thousands of residents across the region.
Prompted by this event and their local effects, I wanted to learn more about atmospheric rivers, particularly their characteristics and climatology. As a result, I began a research project analyzing the basic climatology of Northern California atmospheric rivers from 1980 to 2023, measuring seasonal variables including, but not limited to, duration, frequency, and strength. I would take this research to the International Atmospheric Rivers Conference in San Diego in 2024. There, I was exposed to the concept of “compound events”—meteorological events that consist of two unique weather phenomena in quick succession (or simultaneously). Almost immediately, I formed a mental connection between atmospheric rivers and tropical cyclones, two fields of meteorology that I have experience with.
The Bay Area is no stranger to the effects of atmospheric rivers. However, atmospheric rivers (ARs) are distributed globally, impacting regions like Chile, Western Europe, and Oceania, bringing similar effects of flooding, damaging winds, and mudslides. The powerful forces exhibited by ARs are even more defined in the open ocean, threatening both marine ecosystems and global shipping routes.
In this project, I want to analyze the relationships between ARs and tropical cyclones. Particularly, I want to determine the circumstances under which an AR is formed as a result of the presence of a tropical cyclone. Due to my increased experience in tracking Atlantic Ocean tropical cyclones, this research will be limited to ARs in the North Atlantic Ocean—ARs that impact Scandinavia, the British Isles, France, and the Iberian Peninsula.
This investigation will be facilitated by numerous different datasets, including atmospheric river catalogs, tropical cyclone reanalysis, and general reanalysis datasets. The data will be parsed and analyzed with Python.
I will work with Dr. Vijay Tallapragada, a Senior Scientist at NOAA’s National Center for Environmental Prediction, mostly virtually. He will help me access more datasets and provide key guidance to my research.
I’m so excited to take you along in this research journey, and I hope that you learn something in the process!
Week 1: Getting Started, and a Thing or Two About Tropical Cyclones
Hurricane Ike. 2008. Credit NASA Earth Observatory.
Hi everyone, and welcome back to my blog! This week, I not only refined my project goals and methodology, but also began my process of literature review.
First, let me go over the categorization of tropical cyclones in the North Atlantic ocean—the portion of the Atlantic that is north of the Equator.
- Tropical Storm: 37mph to 73mph
- Hurricane: 74+ mph
- One thing to keep in mind is that tropical disturbances (which I’ll cover later in this blog) can have winds as strong as a hurricane, but if it doesn’t have the structure of a mature tropical cyclone, it will remain a disturbance until it develops tropical cyclone characteristics.
Most tropical cyclones are the result of an initial tropical disturbance that is formed by warm ocean waters and atmospheric instability. This fuels convection (uplift of water vapor), leading to a low-pressure system with organized rotation. As the system’s vorticity (rotation) increases, and the eye forms, the system strengthens into a tropical storm or hurricane. (Keep in mind that the “eye” is used to refer to the center of the storm where winds are near zero and the pressure is lowest. Eyes can be covered in clouds, though in stronger tropical cyclones, eyes can be fully see-through.) As the storm develops into a tropical storm, and later hurricane, it is considered “mature”.
The mature structure of a tropical cyclone involves a defined low-pressure center at the “eye” of the storm, with updrafts around this center forming an eyewall, which is where the strongest winds and precipitation occur. Circling the cyclone radially outwards are rain bands that still contain strong winds and precipitation, and get progressively weaker the further you go from the center of the system. At the same time, these rain bands pull moisture into the storm. The circulation of the storm around its center is the result of outflow, which is basically air that rises from the center of the eye into an upper-level anticyclone (ULAC) (the air rotates in the opposite direction of the cyclone’s rotation). This ULAC allows for the cyclone to maintain its intensity and not fall apart.
As the storm decays, the core of the storm weakens. This is the result of many possible factors, including the storm travelling over land, strong wind shear that restricts outflow, or cooler waters that weaken the convection process. Either of these circumstances either result in the cyclone completely fading out of existence, or a transition into an extratropical cyclone, which is something I’ll talk about next week when I expand on my knowledge of atmospheric rivers.
I started this week by reading chapter 2 and 3 of Richard A. Anthes’ 1982 book “Tropical Cyclones: Their Evolution, Structure, and Effects.” Although 1982 is a long time ago, the core processes described in the book remain relevant to this day. Chapter 2 of the book covered key facets of the life cycle of tropical cyclones, as well as their structural elements, while Chapter 3 expands on this.
In addition, I read “Tropical Cyclones of the North Atlantic Ocean: Tracks and Frequencies of Hurricanes and Tropical Storms, 1871-1963” by George W. Cry. Despite being over 50 years old, the text covers key facets of tropical cyclone climatology that are still relevant to this day. In the interest of your reading time, below is a chart that summarizes many things that Cry’s report covers, particularly the climatology of tropical cyclones. Credit: NOAA NHC
Most Atlantic tropical cyclones develop between August and October in the Main Development Region, which is circled in red. ~65% of all North Atlantic tropical cyclones develop in this region, which has low wind shear and warm ocean temperatures during hurricane season. This is NOAA’s chart for tropical cyclone frequency in the month of September.
During the early/late parts of hurricane season, most tropical cyclones are formed outside the Main Development Region, usually in the more northern portions of the Atlantic, as well as the Gulf of Mexico.
Finally, this week I decided on the datasets I will use for this research:
ERA5 (ECMWF Reanalysis v5, basically historical weather reanalysis for a variety of variables)
HURDAT2 (Data for the tracks of tropical cyclones)
Guan and Waliser Catalogs (Historical global atmospheric river catalogs)
For my project, I will focus a bit more on the decay phase of tropical cyclones, as they move northwards into the midlatitudes and encounter cooler waters and stronger shear. Atmospheric rivers are very heavily concentrated in the midlatitudes, and analyzing tropical cyclone behavior as they go past 25-30 degrees north will be a key step to reaching my research goal.
Overall, significant progress was made this week, and I’m excited to talk about my further progress next week!
Week 2: Atmospheric Rivers
You’re here again! Welcome back to my blog!
This week was all about atmospheric rivers. I spent the week reading relevant portions of “Atmospheric Rivers” by Ralph et al., particularly chapters 1-5 and 7. I placed an increased focus on the structure of atmospheric rivers, the detection of atmospheric rivers, and the climatology of atmospheric rivers. Also, from this point onwards, I’ll sometimes refer to atmospheric rivers as simply ARs. According to the book, an AR is defined as:
“a long, narrow and transient corridor of strong horizontal water vapor transport that is typically associated with a LLJ stream ahead of the cold front of an extratropical cyclone. The water vapor in atmospheric rivers is supplied by tropical and/or extra-tropical moisture sources. Atmospheric rivers frequently lead to heavy precipitation where they are forced upward, e.g., by mountains or by ascent in the warm conveyor belt. Horizontal water vapor transport in the mid-latitudes occurs primarily in atmospheric rivers and is focused in the lower troposphere.”
Yeah, that was a mouthful. An AR is just what its name implies—a river in the atmosphere. Just like rivers, they are longer than they are wide, and transport water from one place to another. In the case of ARs, though, this water is in its gaseous form. These rivers are usually linked with the low-level jet. You may know about the jet stream—the upper-air corridor of fast moving air that makes planes flying from New York to London faster than the planes on the return journey—and the low-level jet is a corridor of fast-moving air in the lower levels of the atmosphere (in meteorology, this is generalized to the 850 millibar pressure level which has a geopotential height of just under a mile above the surface).
An extratropical cyclone (can be interchangeable with mid-latitude cyclone) is a relative of the tropical cyclone, in that both of them rotate around an area of low pressure, and both of them harbor wind and rain that can be devastating. What differs these systems from tropical cyclones are the presence of frontal systems, which tropical cyclones do not possess, as well as area of development. Most importantly, though, tropical cyclones have a warm core, meaning that the strength of the system is fueled by latent heat from warm (sea) surface temperatures while extratropical cyclones have a cold core that is more the result of interacting air masses (hence the frontal systems). Most tropical cyclones form in the tropics (0-30N/S), while most extratropical cyclones form in the midlatitudes (30N/S to 60N/S). For the sake of your time (and mine too, frankly), this diagram from the AMS summarizes an extratropical cyclone pretty well.
You as the reader don’t really need to pay attention the the “embedded convection” illustrated in this diagram, but basically that’s just increased (thunder)storm activity. Now, you see that strip of green along the cold front in the bottom left (triangles on the bar)? That’s where the majority of atmospheric rivers are situated. Obviously, there are some exceptions, but almost all of the atmospheric rivers that affect us here in NorCal are connected in some way to an extratropical cyclone like the one pictured above.
Atmospheric rivers are responsible for significant impacts across the subtropical regions, subjecting affected areas to heavy rains and strong winds. Over 30-50% of the total annual precipitation totals for Northern California, for example, are the result of atmospheric rivers.
Atmospheric river strength is usually quantified by Vertically Integrated Vapor Transport (IVT), defined in units of kg * m-1s-1. Essentially, IVT is a metric that represents water vapor flux in the atmosphere. Integrated Water Vapor, or IVW (mm), is another metric that can be used to identify and quantify ARs. IVW’s calculation is nearly identical to that of IVT, though IVW does not take into account the total vector wind at any given point. This may be a bit confusing right now, but I’ll go over this more when I talk about datasets next week.
A key facet of my research is the climatology of European atmospheric rivers. Just like all atmospheric rivers, they transport water vapor from lower latitudes (including the tropics) to higher latitudes, as can be seen in the below image from the book:
Below is a chart from the same book that gives us an idea of AR activity during various seasons in Europe:
As you can see, AR activity is the highest in January and the winter months, while the least activity is in the spring. There is also an elevated amount of activity during the late summer and fall months, although not as strong as winter. Most European ARs impact France and the Low Countries, the British Isles, and Scandinavia, though Iberia can be impacted too.
Over the next few weeks, I’ll be connecting these two areas of study (ARs and tropical cyclones) in my research. Just before I was writing this blog, I discovered an article that can prove immensely useful in my journey—one about pacific tropical cyclones and atmospheric rivers in close proximity to them. That is something I’ll talk about next week, along with more stuff about data and the technical aspects of the project. Until then, see you later!
Week 3: An Amazing Research Paper!
Welcome back to my blog! Three weeks have gone by so fast. Time flies.
This week, I made a solid amount of progress, doing mainly literature review while also discussing technical aspects of the project’s research phase with my advisor.
Last week, I mentioned how I found a journal article that was incredibly useful and relevant to my research. That article was titled “The Development and Evolution of Two Atmospheric Rivers in Proximity to Western North Pacific Tropical Cyclones in October 2010,” by Cordeira, Ralph, and Moore.
Right off the bat, you can see how similar this research paper is to my current senior project. However, there are some key differences. First of all, these ARs developed in the Western North Pacific, which is near the Phillippines, Taiwan, and Japan (you get the gist). The Western Pacific has a TC and AR climatology quite different to that of the Atlantic. In addition, this study covers a merged atmospheric river that stems from vapor transport from three different active tropical cyclones. My study, for now, aims to keep things simple, mainly looking at the development of atmospheric rivers as a result of a single Atlantic tropical cyclone.
The paper studied the above scenario with great detail, analyzing practically everything about the water vapor transport channel from the three tropical cyclones depicted in the graphic (Chaba, Megi, and 17W), the precipitation from the merged AR that impacted the Pacific Northwest, and the atmospheric river(s) itself.
So how exactly did these ARs form?
As these tropical cyclones developed, they brought up warm air from the surface (remember that warm sea surface temperatures (>80 degrees Fahrenheit) are the main fuel source for tropical cyclones. Coupled with a high amount of CAPE (basically atmospheric instability), this heat would rise up the troposphere and interact with the North Pacific Jet stream in a diabatic manner (in that heat is exchanged). This warmer air provided more energy to the jetstream.
At the same time, the outflow from the tropical cyclones (remember the ULAC I discussed in blog 1?) interacts with the North Pacific Jet, creating a clear frontal boundary, in a process known as frontogenesis. The water vapor is guided along this frontal boundary, forming an atmospheric river.
I like to visualize things in 2d or 3d, and this schematic from the paper (see below) is a great summary of what I went over this week. Focus on the first four figures (a-d), and the left half of figure (e).
Here's a link to the article for those who dare to read it: https://journals.ametsoc.org/downloadpdf/view/journals/mwre/141/12/mwr-d-13-00019.1.pdf
The information covered in this paper set a clear framework for my project, enabling me to focus on what atmospheric factors I should look at while pursuing my research over the next few weeks—primarily observing jet stream activity, atmospheric instability, and heat exchanges between the tropical cyclone and the air around it.
This week I was also briefly introduced to the means of computing that I will be using for this project. I’m going to be using the HPC2 supercomputer at Mississippi State University (virtually, obviously) for any calculations that cannot be performed on my personal laptop. More on that next week.
The end of this week (and half of next week) marks the transition point between my literature review phase and my “actual research” phase. For most of next week, I’m going to familiarize myself with the HPC2 systems, dataset retrieval, and data assimilation methods. Until then, see you next week!
Week 4: Introduction to the Supercomputers
Hi and welcome back!
It’s week 4 already, and things are starting to get going on the technical side of things. However, this week was a bit of a slow week to prepare for next week.
This week, I was introduced to the means of computing that I will be using for the project. I’ll basically be remotely accessing a supercomputer (well, technically two, but I’ll just be using one) at Mississippi State University. It’s basically got a large stockpile of NOAA’s data that is used by many of its subdivisions like the Center for Weather and Climate Prediction. In addition, it also serves as a computing hub for many of NOAA’s operations.
There are two main supercomputers that fall under the HPC2 (High Performance Computing 2) label, Orion and Hercules. Basically, you can run computing operations and access the servers with the data with either one (it doesn’t matter.)
Okay, now for the more project-oriented stuff. With the help of my external advisor and some of his team members, I was guided on finding the data I’ll be using next week (the AR Catalog data with IVT values and AR presence from 2000-2019), fetching it, and running it through a script that plots the data. This data is in the form of a netCDF file (one of the two file types I’ll be working with, the other being GRIB, which is similar to netCDF), which is a file type that, in one file, stores multidimensional data for multiple parameters including but not limited to:
Time (integer or floating point)
IVT (usually a floating point)
Wind speed (usually a floating point)
AR presence (0 or 1, essentially a binary)
netCDF, along with GRIB, is used heavily in the world of atmospheric science, due to the need to record and access global data relatively quickly.
The computing environment uses a terminal-style Linux input. Although I have had experience with this form of computing, it’s been a really long time (the last time I used powershell/terminal for anything was in 2022) and my experience level basically reset to beginner. Hopefully, I can learn things a bit faster now since many things are faintly familiar to me.
I’m running a bit behind on schedule, so next week will be very intense, and I hope to identify all of the cases of AR formation from a tropical cyclone that I will be using during the next week, with the help of this supercomputer. For now, though, that concludes this week’s blog. See you next week!
Week 5: The Catalogs
Hi everyone! Welcome back to my blog!
This week, I began the main research phase for my project, actually using and visually analyzing data. I spent a solid portion of the first two days setting up the directory, which took an unhealthy amount of time (around 5 hours) due to a mistake I made in the installation of Miniconda in my personal directory. I fixed that issue by simply deleting the junk files in my original directory and re-installing Miniconda.
From there, I was able to fetch atmospheric river catalog data for the years 2000-2019. This was in the form of Rutz AR Catalogs, which is an AR database that I already have experience with in previous research. I modified an existing python script for plotting netCDF files to plot Rutz Catalog data for daily IVT and AR presence, and ran it through the supercomputer.
After this process was completed (it took almost an hour and a half for the daily image plots for both variables to be created), I transferred the files to my local computer so I can easily view them frame by frame. Since there are over 5,000 individual frames, I’m analyzing them based on season. For example, I’m ignoring any plotted maps for January through May, and only looking for maps from June through December.
Here’s a sample AR presence map, depicting Hurricane Alberto on 8/16/2000 (can you spot it?):
Interestingly, Alberto is visible in the AR presence map, which, if you recall, is a binary metric (1 = AR present, 0 = no AR). Obviously a tropical cyclone is not an atmospheric river, and this metric is basically just a check on whether a certain data point satisfies the prerequisite for an AR—250 IVT and a length of over 2000km and width of under 1000km. Here is the IVT map for the same day:
You can see the structure of Alberto a bit more clearly in this image (also you can see typhoon Ewiniar making a cameo appearance near Japan). The first time I ran the plotting code for IVT, the areas with less than 750 IVT were a bit too faint for my liking. I’m going to redo the IVT run, this time setting ~1500 as the topmost color (purple) (anything higher will just show up as purple and it’s not overly necessary to differentiate such high IVTs at this phase of the research right now).
I’m currently in the process of analyzing the lifetimes of every single storm system from 2000, so this will take a while. I expect this analysis to be done no later than Tuesday of next week. Next week, I will have determined the different events that I will be analyzing in this study, and the different criteria I will be testing for. Until then, see you later!
Week 6: Charts Galore
Hi and welcome back to my blog!
This week was a pretty crazy one. I started the week modifying the python program to create new plots of the IVT, limiting the range from 0 to 1500 IVT, as promised in last week’s blog. Then, I was able to finally do a ton of cross-referencing. Not the most efficient way I’d say, but it was the best I could do.
Before I talk about what I cross-referenced all the time, I did run into the issue with the FTP (file transfer protocol) for the supercomputer. I wanted to view the images frame-by frame (each frame is 1 day) by just pressing the left and right arrow keys, yet I was only able to view one image at a time. As a result, I was forced to transfer all the image files to my computer. It was a negligible amount of disk space, just around 3 GB total for both the IVT and AR presence charts (for the record, 3 gigabytes is less than the storage required for the amount of photos I take on a serious photo walk).
Now, I had three windows open: the file directory for the IVT charts, the AR presence charts, and the HURDAT2 Best Track information for each tropical cyclone season (HURDAT is just “hurricane database”). Looking at the different storm lifetimes, I first opened the IVT chart for the days the cyclone in question was active. Since the storm is quite easily discernible in the IVT chart, it allowed me to get an idea of the location of the storm. Then I opened the AR presence chart and compared it to the IVT chart. If you understood what I just said, I’m impressed. If you didn’t, maybe this screenshot will make things a bit easier to understand:
This is a split screen setup with IVT(left) and AR presence(right) charts for Hurricane Fabian on September 8, 2003. As you can see, Fabian is very clearly visible on the IVT chart just southeast of Newfoundland, but not very discernible on the AR presence chart. You can also see the clear AR linkage from Fabian in the IVT chart. I ran this setup, doing everything manually until the year 2019, until the data was exhausted. It wasn’t as bad as I thought it’d be, since I was just skipping over a lot of storms that I knew had dissipated over, say, the Southeastern U.S.
Initially, I was taking note of every single TC-linked AR, but I then realized I was lost in my train of thought and failed to mark only the ARs that impacted Europe, like the above example for Fabian (not so fun fact: it was a name which was retired due to the hurricane’s adverse effects on Bermuda). As a result, I had to go back and erase all the TCs that I had noted that did not have a linked AR that was landfalling in Europe. Wasn’t that tedious.
Although the process described just now is done, I still have a lot of case studies I can do (my list has over 50 examples 😭). Since ~5 weeks is not enough to analyze all of them, I’m going to choose 5 different case studies to analyze in-depth, while I will use the rest of the case studies for some other base-level statistical analysis.
Next week, I’ll be travelling to Washington, DC, to work in person with Dr. Vijay Tallapragada’s team, and I will discuss with them about which case studies I should choose and which ones I should not choose. I will also determine the methodologies I will use to analyze each individual case study. Right now, I’m hoping to look at tropical cyclone location, age in its life cycle, its intensity quadrants during which an AR is linked, and the presence of any subtropical jets, I see myself using all of these criteria, but the final list of criteria will be in next week’s blog. Until then, see you later!
Week 7: A Week at NOAA's Center for Weather and Climate Prediction
Hi everyone and welcome back to my blog!
This week, I travelled to the other side of the country to work in-person with my External Advisor and his team in College Park, Maryland at the NOAA Center for Weather and Climate Prediction. I’m so incredibly thankful to Dr. Vijay Tallapragada (my EA) and his team (Dr. Murali, Dr. Keqin, and Dr. Xingren) for helping me on this project in person and giving me the opportunity to work at the CWCP office for a week.
I also met people across the NOAA’s National Center for Environmental Prediction and Environmental Modelling Center, including people who pioneered forecast model dynamics, development, and verification, including the creator of the United States Global Forecast System (US GFS) as well as the acting director of the NCEP himself. I genuinely never saw myself being in this position even only a year ago, and I’ve met so many new people and made so many new connections.
Okay, let’s talk about what I did this week. Last week, I viewed the IVT and AR maps with daily intervals. However, there’s a problem with that. You see, the other databases like HURDAT2, GDAS (or ERA5) reanalysis, et cetera, have 6-hour time intervals, in that data is recorded every 6 hours. As a result, I was forced to redo the entire process of creating the image plots, transferring them to my computer, and reanalyzing them. Rather than going the inefficient route of doing this on the terminal manually, I just modified an existing shell script to create the plots without me having to keep the computer open. 17 gigabytes later, I was able to pack all of these image files into a “tarball” (linux version of a zip file) to quickly (in this case 30 minutes is quick) transfer them to my computer, where I unpacked them. That seems like an easy process, but it’s the opposite. Packing the tarball itself was a faster process than what I was doing before, yet it still took over an hour due to the sheer amount of image files for each data variable. Then the transfer took about another hour. All of this was somehow still faster than what I did before (keep in mind, last time was daily intervals, this time was 6-hourly intervals).
Now that that was out of the way, I installed an application called IrfanView (basically a photo viewer app but better) that allowed me to quickly skip through the frames in the folder (meanwhile, windows photos was lagging the entire time), and I finally began the “last dance” of my analysis. To make things more efficient, I pulled up this monstrosity of a split screen (not depicting a TC-linked AR event per se, but it shows four active cyclones feeding into a moisture channel, including Hurricane Ike, which is pretty cool).
If it works, it works. No complaints. I’d show the results from the analysis if I hadn’t written them down instead of typed them. Basically I followed my previous methodology, this time for only six-hour intervals. In addition, I only looked at AR events that lasted at least 24 hours, such as from 00z UTC on September 19 to 00z UTC on September 20. The longer the duration, the better. So far, the longest TC-driven AR event that I have found was a 3-day, 12-hour long AR event that impacted northern France and the British Isles, driven by Tropical Storm/Hurricane Cristobal in August 2014.
For the sake of simplicity, I only “counted” an event in my logbook if there was a clear and active IVT channel from the source cyclone to the region being impacted (Europe). Here’s an example of the IVT chart for one frame during Cristobal’s AR:
You can see the signature of an extratropical cyclone (look at the rotation) near the British Isles carrying all that atmospheric moisture from Tropical Storm Cristobal (which is near Cuba in this image). I’ve found that this setup exists for virtually every case of a TC-linked AR. The real question lies in how intense the extratropical cyclone needs to be to be able to channel all that moisture from thousands of miles away.
The intensity of the present extratropical cyclone (ExTC) is one of the variables I hope to look at. To do this, I will be using GDAS/ERA5 reanalysis data obtained from my external advisor. In addition to intensity, I will also look at things like the lower-level jet (wind at a 700 millibar pressure level) and precipitable moisture.
Right now, I’m still indecisive on what reanalysis dataset to use, but I’m leaning towards using ERA5 due to its simplicity of data retrieval. I’ll make this decision on Monday, though the analysis tools remain the same. Next week is spring break, and I’m still not sure if I want to work on this project next week, but I probably will. Until then, see you later!
Week 8: Final Preparations
Hi everyone and welcome back to my blog!
Last week was spring break, so I took a refreshing break from this project. This week, I resumed working on the project, and I have decided to use ERA5 (ECMWF Reanalysis v5) reanalysis for data obtainment.
ERA5 has a series of useful variables that I will be using, including zonal and meridional wind (u and v wind vectors), precipitable moisture (this one’s important, since it’s the main facet of an atmospheric river), temperature, relative humidity (also important), MSLP (mean sea level pressure), and cloud cover. I discussed these variables in week 7 with the reanalysis data specialist at NCEP (Dr. Tallapragada’s colleague, Dr. Bentley), and she gave me some helpful advice on how to visualize and process this data, including the MetPy tools in Python.
I will be doing five case studies using the ERA5 reanalysis data for the following five tropical cyclone(s) which fed into landfalling atmospheric rivers in Europe.
Alberto 2000
Allison 2001
Wilma + Alpha 2005
Cristobal 2014
Arthur 2014
I chose these ones for the sake of simplicity, and since there are literally two weeks left. In the future, I will work on this project (during the summer if possible) and that phase of the project will give me more time to take a look at more complex events like:
2008: Ike + Josephine + Hanna feed a complex moisture channel covering the entire Atlantic that impacts Europe
Jerry 2013 contributed to a huge conveyor belt of moisture transport involving two massive extratropical cyclones
2004: Jeanne + Karl + Lisa feed another complex moisture channel consisting of multiple atmospheric rivers. This one looked so confusing I genuinely didn’t know what was even going on 😂
However, I still have my list of all the atmospheric river events attributed to tropical cyclones (including the ones above) in my notebook, which I will produce some more large-scale data charts. That would include: cyclone stage of life, when during the timeline of the hurricane season it took place (think early or late season), duration of events (minimum 24 hours), strength of the cyclone (I might scratch this one because I’m not exactly sure on its relevance), location of the storm (latitude and longitude), and frequency of events relative to hurricane season overall activity.
For that last one, I’m pretty sure that there’s going to be a lot of variation with no visible trend, but that’s just a hypothesis.
Next week will be a doozy, since I’ll be going head-first into the case study and reanalysis, with merely a week and a half remaining. See you then!
Week 9: Finishing up, part 1
Hi everyone, and welcome back to my blog. This will be a quick, short one because I’m super busy trying to get everything done.
This week, I dove into the bulk of the data analysis and case study portion of my project. I started off the week downloading NetCDF files for the five case studies I presented in last week’s blog. However, when I opened up the charts for Alberto 2000, I was not seeing any signature of an atmospheric river.
That’s when I realized that the Rutz AR catalog, although useful for narrowing it down a lot, was still not the most optimized for AR detection, since it had only three main criteria—an IVT of 250 kg m-1 s-1, a width of under 1000 km, and a length of over 2000 km. This meant that it would call the normal tropical wave “train” in the tropics an atmospheric river, which it is not.
As a result, I had to retrieve the more “accurate” Guan and Waliser catalog (basically more constraints) and then review all 48 events I had recorded in my logbook. Four hours later, I was able to cut that list from 48 to 28. In that process my case study list also changed, replacing Alberto 2000 with Joaquin 2015.
I’m currently working on the analysis part of the individual case studies, and the Python program for the charts is in progress. (Refer to last week’s blog if you’re not sure what I’m talking about). For the record, here are the variables that I’m going to be looking at the most for the case studies, as shown by my data requests from the Copernicus Climate Data Store for Allison 2001:
That seems like a lot. However, don’t worry, I’m only really looking at maximum/minimum values over a time frame. Some variables, like total cloud cover or the water vapor flux components, are simply for verification. Other variables, like significant wave height, precipitation, and 10m wind, will give me an idea of the effects of the AR events, simply reinforcing the rationale for this project. The rest, including everything in the first image, will be used to observe the conditions surrounding the tropical cyclone, and I hope to find some sort of trend between the five cases.
Next week will be BUSY. I’m going to be working on the weekend to get things done, and hopefully have a poster ready by Wednesday (Thursday at the latest). See you next week!
Week 10: Finishing up, part 2
Hi everyone! It was the final week of research for my project, and I can confidently say that this was a very hectic week. Due to a very tight deadline, I spent this week doing a lot, most of it centered around creating my poster.
Most of my week was spent working with the Jupyter Notebook, taking in the list of AR-linked TC events and using Tropycal, Cartopy, seaborn, and matplotlib to plot not only the chart below but also “subcharts” that looked at the general location of a tropical cyclone as it was feeding into an atmospheric river.
The main conclusion that can be drawn from this chart is that the general location of TCs during their AR-linked stage usually lies in the open Atlantic ocean. I experimented with different ways to present the data in a, well, presentable manner, and I ultimately settled on that chart. I plotted the category of the tropical system too according to the Saffir-Simpson Hurricane Wind scale too, with blue as tropical depression and red as category three (no higher category in this chart) but ultimately the overall category of the storm doesn’t really matter. One interesting case I can note from here is Hurricane Vince in 2005, which is the track that nears Portugal as a tropical depression. Moisture was being pulled from Vince in the form of an atmospheric river, and was impacting the Iberian Peninsula as such. As the AR moved towards the Mediterranean and Southern France, Vince also neared the Iberian Peninsula, bringing tropical impacts to the region as well.
After making some charts in Python, I then moved to analyzing the case study ERA5 data with Panoply. Initially, I thought that I would have to make a chart using Python, but I realized then that Panoply had a nice feature that would not require me to use Python. I looked at multiple variables and their charts, and after discussing with my External Advisor, I decided to chart four variables: IVT, 250mb wind, 700mb wind, and Mean Sea Level Pressure. After that process and analysis, I ended up with these charts (literally a screenshot from my poster):
As you can see, I took three frames for each variable (left to right: IVT, 700mb wind, 250mb wind, MSLP), depicting the start, middle, and end of the AR event.
After the case study analysis, some more conclusions were drawn. Most significantly was that the alignment of the 250mb and 700mb jet stream helped pull moisture from the TC, and also the presence of an extratropical cyclone enhances this process and creates an AR. Though there are no truly concrete numbers that can characterize this, I did notice that the strength of the extratropical cyclone [ETC] required to create an AR from a TC moisture source depended on its distance from the TC itself. Evidently, ETC’s in closer proximity to the TC would not need to be as strong as one far from it.
Overall, the past ten weeks have been an extremely formative experience, and I’m so grateful to my internal and external advisors for their support. This project is only a stepping stone for a vast array of future research that I hope to do involving this subject. I hope to look at more variables surrounding the AR event that will allow me to zero in on the exact causes of this phenomenon. See you at the symposium!