Skip to main content
University of Wisconsin–Madison

AOS 405 Career Blogs
Forecasting jobs in Atmospheric Science

Introduction

Forecasting is the cornerstone of atmospheric science. Forecasts are created and specifically curated for different customers. The range of customers can involve general public, airports, shipping, farmers, energy companies, insurance companies, etc. Forecasters use sophisticated model data, adjust the information accordingly, and make end products appropriate to specific customers. The end product will vary from customer to customer. For example, using the weather app on your phone will provide a forecast for the daily high and low temperature as well as sky cover and precipitation. This forecast is catered specifically to a casual user that may want to know whether or not to wear Bermuda shorts or khaki pants. Conversely, the forecast for an insurance company will provide the same information but give a probabilistic forecast so they can adjust rates according to the probability of a severe weather hail event. I.e. an insurance company would want to know whether there was a 50% chance of hail vs. 98% chance of hail, rather than just hearing it’s going to hail. A very important aspect of a forecasters job is communicating the forecast to their customer. Without good communication and understanding between forecaster and customer the forecast doesn’t mean anything.

The science behind forecasting is another discipline that in itself. The majority of the responsibility is on models. Models are simulated environments that will predict future weather according to physical parameters. Models for weather prediction will have a variety of ranges of size and complexity. Some parameters that are taken into account when creating models are incoming solar radiation, moisture content, soil temperature, topographical features, etc. Models in their current capacity can make relatively accurate forecasts for up to about a week. Lastly, after models forecast the environmental condition of an area, the forecaster can adjust the end product according to their own knowledge of general circulation, QG-theory, or mesoscale dynamics. Some medium range forecast models (~ week) are the Global Forecasting System (GFS) and the European Centre for Medium Range Forecast (ECMWF). These models will be used for example by farmers to adjust irrigation for the coming week. Some shorter-range models (~day) are the High Resolution Rapid Refresh (HRRR) and the North American Mesoscale Forecasting System (NAM 3km CONUS). These models will provide much higher resolution and would be used, for example, in forecasting severe weather for the general public.  

Return to Top

Helpful Resources

Occupational Outlook Handbook For labor statistics and other general information in the industry https://www.bls.gov/ooh/Life-Physical-and-Social-Science/Atmospheric-scientists-including-meteorologists.htm

Learn how to code online

Pivotal Weather To explicitly see model forecasts and learn typical weather patterns http://www.pivotalweather.com

Storm Prediction Center To see end products predicting chances for severe weather, archived events, and explicit discussion about specific events by real forecasters http://www.spc.noaa.gov

General Information on Meteorology Careers Article written by NOAA with further information https://www.weather.gov/media/bro/outreach/pdf/CareerOpportunitiesMeteorology.pdf

Return to Top

Matt Rydzik

Meteorological applications developer for Commodity Weather Group, LLC. M.S. in Atmospheric Sciences from UW Madison.

Picture of Matt Rydzik

What is your job and how do your day to day operations vary?

“Right now, I work for a company called Commodity Weather Group. We provide energy and agriculture consulting services. So, some of our clients range from individual farmers from throughout the Midwest and around the globe to big banks and New York city trading natural gas futures. I do a lot of the IT support work. I work closely with our forecasters to produce new products. So, day to day it’s very variable, most days involve some coding project or a meeting, discussing how we can better communicate the data we’re giving to our clients or dealing with whatever the current weather situation is at hand. For instance, we’ve had a couple snowstorms creeping up the east coast of the United States over the last week or two and we spend a lot of time in the office communicating the risk of power outage, the impact winds are going to have on energy clients and people living in the communities across the coastline.

With Weather Commodity Group does the overall company make money off of trading different commodities and investing in them or do you just provide insurance information to clients on whether or not weather is going to disrupt their business?

We’re more of the latter. We try to stay out of the trading, we provide more of the consultation services to people that are doing that. There have been a couple other vendors that do similar work we do and they try to play both sides of the card, trading themselves and offering their opinions to other clients who are also doing trading. This never ends up working well because the clients accuse the vendor of having information before they do and it creates a conflict of interest if you’re playing both sides.

How much would you say either you or your company differentiates your time between improving the actual physical forecast vs. improving the digestibility of the forecast for your clients

A lot of the work I do is more on the digestibility side of the information. However, we do have 7 forecasters, our company size overall is 12 people. We like to do a lot of niche forecasting, for example the energy commodity is a very specific focus. So, we have three developers, one of which is myself, the rest are forecasters. They [forecasters] do get more into the weeds of where exactly the rain/snow line is going to fall. What exactly is heating demand in Dallas going to be. They’re making final decisions and small little tweaks, they’re directly interfacing with the clients over phone call, text chats online, email … the whole nine yards.

How much does your company deal with forecasting for renewable energies, such as for solar or wind turbine farms?

Yes, definitely that’s become a bigger part of what we do. When I first started here about 5 years ago, we were just starting to implement some wind forecast at that point solar wasn’t even on our radar. At that point, there were a few specialized companies doing it, but it really wasn’t widely being done. Part of that was because it was new and part of that was because it was difficult to do and people were still trying to develop technology to do it better and faster and make it practical for them to do. We’ve created numerous new website visualization tools and our forecasters produce daily forecasts for both wind and solar for many different power regions across the country. That’s probably one of our biggest areas of growth and I imagine that will probably continue to get bigger and bigger as the solar and wind expansion continues, especially in areas like China and Europe as they grow that specific energy sector. That is probably the direction most of our industry is going to go. When my bosses first started the company the big focus in the late nineties was natural gas. It was the big money maker across the US and over the last 15-20 years that has gradually started to diminish as wind and solar has picked up.

Can you elucidate the difficulties in making models and forecasting for solar?

A lot of it was just the information we had available to us. We pull in all the NWS weather models and we buy the full ECMWF model and it’s only within the last 2 or 3 years that they started to provide enough publicly available output for us to even begin to create products to assist our forecasters. A lot of it was also just labor and time intensive for our forecasters so it required a lot of people to parse through a lot of manual data to come down to a forecast. A lot of it is also sensitive, one of the bigger areas in solar is California solar and it is very sensitive for instance to the direction of wind flow off the coast. So, the smallest shift in that can change your cloud pattern and all the sudden you go from a big, record producing day to barely producing anything that day. So, it’s a very risky forecast that changes on the smallest little thing and makes it very difficult to produce a long range forecast. Same day forecast, we usually get them pretty good but once you start getting out into the 4-5 day range it starts getting a little iffy, yet it’s something the clients are still interested in, so we try to do our best to provide them with our forecast, but we also really try to communicate the risk of that forecast. We’re going to be wrong, no forecast is right all the time. So, if we’re going to be wrong how are we going to be wrong and what you should be looking for along the way is to see if it starts to deviate from our [original] forecast to know that the forecast is going to be wrong at that point, to know earlier.

Branching out from Natural Gas to Solar and Wind For example, how do you find new customers. Do they hear about you? Do you look for new emerging industries and try to get in on them from the start?

I would say it’s probably a little bit of both. We do have a full-time sales representative and he is a meteorologist by trade too, so he’s very familiar with the nuts and bolts of forecasting. Some of it is him seeking out companies that are starting out in those fields and he’ll reach out to them and say, “hey this is what we can offer you, are you interested?”. We also get customers hearing about us via our website or word of mouth. There’s a third, seeing another company using us and they think maybe we should be using you too. There’s another third coming from networking at AMS meetings to energy conferences that take place typically down in Texas in the summer and winter months.

At these conferences does your company present on specific research.

It depends on what conference it is, for example AMS it would probably be more research based. For some of the energy conferences like ERCOT, which is a big power region down in Texas, our talks will be focused on the upcoming season forecasts. So, for instance, we have one of our colleagues that just left for Chicago, where there is a grain conference for agriculture commodities. So, he’ll be going out and just presenting on our spring planting forecast. Hopefully, the people who are not our clients at the time see our forecast and find value and are further interested in pursuing our services at that point. We really try to tailor what we’re doing to our audience, without compromising our core values and core goals for the company. We’re pretty small and we have no intentions of making our company huge like AccuWeather, and we want to provide a focused service and not spread ourselves too thin.

For this specific profession did you end up falling into it or we’re interested specifically in this industry before you got this job at the Weather Commodity Group?

Well, a little bit of both. I’ve always had an interest in weather and computing. As I went through my undergrad degree it became apparent that some of my best talents and abilities would be to apply computers to meteorology. At the time after undergrad I still wasn’t sure what I wanted to do and then that’s when I was presented the opportunity to work with Dr. Desai at Wisconsin for a graduate degree. The way I got the job was then networking with someone I knew from my undergraduate degree. We got talking, I sent in a resume and went through the interview process and it sort of fell into my lap, but it went back to my networking I did as an undergrad and staying in touch with people. You never know who you run into in previous years of schooling that you’re going to run into in the industry. So I shouldn’t be bullying any kids in my undergraduate program? Yes

You said you were very interested in computing, did you start coding in college or was it something that came later?

I think I’m kind of an anomaly, I did some coding and programming when I was in high school. I was pretty comfortable. I knew a few languages when I was going into grad school, and coming into the industry now. My two coworkers they were later bloomers. Didn’t start coding until undergrad junior and senior year. A lot of them picked up a lot of the bulk of their programming skills in their grad school years. One thing we look for, especially when we’re hiring people here, is not so much knowing a programming language, or an exact skill, it’s determining the ability of the candidate to learn how to do a specific skill quickly. For instance, none of me or my development colleagues knew python coming in. It’s something that all three of us picked up quickly on the job because we realized that was the most appropriate tool for it and it makes our lives easier in the end.

You mentioned your work with Dr. Desai earlier, do you want to talk about how your graduate experience shaped your interest, and even discussing your specific research if you remember it?

We worked on relating midlatitude cyclone trajectories to North American snow cover. We were able to demonstrate that there was a preferential region for cyclone path just north of the snowpack line. We were able to parse through an entire reanalysis data set and build a statistical relationship to show that yes, they are more likely to follow that path. Yeah that definitely provided me some great experience learning how to program more and program different things. Learning things on my own, multitasking, taking classes, doing research dealing with other grad school responsibilities. So, it was definitely a great opportunity to expand myself in the process.

Would you say that most of the people you work with went to grad school, or do you have a mix of grad and undergrad?

About 6 of our employees graduated, the younger generation, between the years of 2002 and 2015-16. All of them but one has graduate degrees. A lot of the older generation tended to just have the undergrad degree. We’ve definitely seen a lot more of our applicants come across with graduate degrees. But when it boils down to it we want to see what their experience is, on the IT side and computing side. We tend to see more graduate students with more programming experience because they tend to do more research projects with data analysis, so they have more opportunity to learn more. We’ve had some undergrads go pretty far in the interview process because they’ve either done undergrad research or they did programming projects or website development on their own in their spare time.

Try to figure out what makes you unique and what you can bring to the table, we have so many applicants that look exactly the same. I took this class and that class, it’s really hard to parse through those and see who stands out. The ones that we do typically move on are the ones who have taken those classes but also did something in their free time to demonstrate their forecasting ability or they’re programming ability.

Do you get many applicants that are not right out of school, i.e. switching from another job?

We’ve had both. We have some applicants that come in with a lot of industry experience. My one colleague worked a couple one-off jobs for a couple years after he graduated and had a hard time finding a job. A couple people come directly out of school, but yes, it’s definitely mixed.

(Fun Question) If you could pick a dream job, what would it be?

I think I might have already stumbled upon it with this job right here. Computer talent and weather forecasting, being a small company, and the founding partners are still very involved in the day to day operations, all four of them are forecasters so they’re very in tune with what we’re doing. There’s no real hierarchy here, everyone’s opinion is respected and encouraged, everyone is encouraged to share their opinions. Something, pretty close to what I’m doing now to have the freedom to innovate and develop new products and we’re actually encouraged to do it. It’s exciting to be on the forefront of new communication methods and new ways to process model data and being one of the first people to see model data as it comes across is very exciting. Just being in tune with the on the go weather industry is very exciting.

When you take different model data do weight the different models according to known biases when giving your end user forecast?  

Definitely, we look a lot at the past performance. In general, the European model kicks the butt of the American model, one of my coworkers was talking about the verification of the American vs. the European. Over the last week verifying an 11-15 day forecast, I’m looking at one now and the American model is winning, but over a longer period of time we tend to lean towards the European. One of the areas we’re really exploring right now is ways to figure out which one is going to verify better in the future. So, trying to correlate the previous forecast to make the call today. “Hey, the model over 11-15 days is saying that the weather is going to do this, should we believe it or not?”. That’s where the human side comes in where you look at the European model shows this pattern and the GFS shows this pattern, but the GFS pattern makes more sense given what I know about general flow or what it’s been doing over the last couple of weeks. We do rely heavily on the forecast but there still is that human aspect of the forecast, to ask if the model is doing something logical. ‘

There’s still a need for humans. Especially, a computer’s never going to be able to call a client up and describe the forecast or answer a client’s trading question on the floor, or a farmer out in the Midwest. Having the ability to communicate the forecast to the client is key.

Communicating the probabilistic part of the forecast and the error?

Yeah, the European has 51 ensemble members. So out of the 51 members every single weather solution is out there, so in a 15 day forecast it could be 90 degrees or 20 degrees one of the members has one of those. Which one do you believe and where are the majority of members siding?

Environmental conditions where specific models have known biases?

I can’t think of any situations where a particular model is good but I know there’s times where I’ll discount a specific model. For instance, with the GFS, the summer time heat bias over dry land, something they’ve been working on over the past couple of years but we still see it crop up in South America summer right now. The model is putting out in the 5-day forecast that it is going to 125 degrees across central Brazil and it’s doing the same thing in the US over the summer and yeah, I’m pretty sure it’s not going to verify that high. So, we’d discount a little bit of that. The ECMWF does tend to have a cold bias over snowpack in the winter over the US, it tends to run 5-6 degrees too cold. Yeah you do pick up a lot of these little things, thinking why did a particular forecast bust and seeing it was because of a snowpack, so there was too much radiative feedback in the model, so something to look for next time you time you see snow cover pop in.

Thanks Matt! Return to Top

Courtney Obergfell


General Forecaster for National Weather Service (NWS) Sacramento. M.S. in Atmospheric Sciences from UW Madison

Picture of Courtney Obergfell

What is your day-to-day job is like, if you do mostly the same thing every day or if it varies a lot?

It varies day to day at the weather service. As you probably know, we’re open all the time so at my office we have three shifts. We have a day shift, an evening shift, and an overnight shift. On the day shift and overnight shift, we’re responsible for producing a full forecast package. We do short-term, long-term, we do aviation forecasting, and we do fire weather which you may have heard is a big deal out west. That was definitely something I had to learn, living in Indiana my whole life. For the forecast, my office is kind of progressive and one of the leaders in the western region. We do a lot of social media, a lot of graphics, a lot of messaging, as well as getting outside the office and meeting with our partners. I spend a lot of time on the hydrology side of things, dealing with water which is highly controlled and politicized in California, so I meet a lot with emergency managers and some of our state and federal partners as well.

How big of change is it to move to a completely different region like California from the Midwest, specifically considering different forecasting challenges?

Well, it definitely was a challenge. What the weather service does now is use blended models, and so the forecast itself is still a focus and it still has to be done right but we focus a lot on messaging. The major change was just learning a lot of the local processes and certainly fire weather. Out in California all the water is in reservoirs and it’s controlled really well so flooding is controlled, it’s a whole different ballgame. I think it’s always good to get out of your comfort zone and learn something different. It might be hard for me to go back to the Midwest and have to deal with severe weather and snow forecasting and that realm. I certainly enjoyed learning something different and I like it out here now!

You mentioned your office is more progressive. Would you say that’s just like a regional thing and where your office emphasizes social media and outreach more so than other offices?

Yeah, if you’re interested in the National Weather Service, its one National Weather Service but you’ll find that every office is very different, just because everybody specializes in something different. We have some people who really like climate. We have some people who really enjoy the communication side of things, messaging, social media, graphics, etc. Some people just like to do the forecast and that’s it. Each office is very different; it just depends on who you’re working with and what’s going on within your office. There’s certainly opportunities for everyone everywhere but if you’re interested in getting into the National Weather Service you have to be flexible.

Can you explain your experience getting your Master’s degree and how that shaped your future career path?

I went to Perdue for my undergrad and then went to Wisconsin for my graduate school. I actually didn’t do a research assistantship; I got my funding through being a teaching assistant. I T.A’ed for professor Martin and a few other graduate students were there teaching the introductory level courses. I T.A’ed about three class per semester in order to get my funding. I did do a research paper but not having to do research all the time; I wasn’t really interested in doing that after school so I got to T.A. and interact with students a lot more. I just developed what I was interested in and what I focus on now in my office and throughout my career.

For the National Weather Service, what are they looking for in terms of credentials and experience for entry-level meteorologists?

You’ve probably heard it’s very,very,very competitive, even more so than when I got in. I worked in the Milwaukee NWS office during grad school and then I was placed into an office when I graduated. You can get in with a bachelor’s, but you have to have a bachelor’s and some other experience. Most people now, just because of the sheer number of applicants (thousands of people are applying), have Master’s, and some people have private-sector experience when they come in. It’s really important to set yourself apart from the crowd; have something that makes you different from everyone else past your schooling. If you can even volunteer one day a week at the weather service office or some other entity that’s weather related just to get your foot in the door. Conferences or local meetings, or that sort of thing where you can meet people outside of your field will be really helpful because sometimes all it takes is someone remembering your name and having a conversation with you to set you apart from everyone else. Make sure you get good grades and do your job in school but try to have some other things that sat you apart from every other person that’s applying!

How exactly are forecasts made at NWS? Do you really rely on the models or do you have a lot of human presence in making the forecast, or does that vary from office to office?

The way the weather service is headed is less time working on the forecast and more time taking it and communicating it and interacting with your core users. We still look at all the forecast models and go through that process, but we actually make the forecast on a computer. It’s kind of a graphical editor, and we can put any kind of model we want into that. The forecast is less time-intensive than it was before but the weather service has this national blend of models that’s a starting point for the forecast. We then tweak that blend. The blend often times does better than humans can do, which is really hard for a lot of people to accept.

Are you hiring any people specifically for communication or social media outreach, or is the NWS mostly self-taught for people who would be forecasters and then take on this new responsibility?

What we do now is we hire a bunch of people who are scientists and then teach them how to be communicators, how to make graphics, how to be graphic designers. What, in my opinion, they really should be doing is hiring people who maybe don’t have a meteorology degree but are really good at communicating and can take our forecast and make really good graphics for it. I think there’s some push to in the future to start going that way but the government is very slow to start to change. A lot of us just learn on the fly.

You mentioned it before, but when you said you were part of this hydrology team, could you elaborate on what you do and what happens at these emergency meetings?

As I said in California we get our water from the snowpack in the mountains in the summer, so whatever melts and goes into the reservoirs, that is then let out and controlled. Farmers have water rights and crazy things I never knew existed. The hydrologist at our office plays a huge role even during a drought because water is such a big deal so I’ve kind of been working with her. Did you guys hear about the Oroville Dam situation last winter? It’s one of the biggest reservoirs in the state. It holds 3 million acre-feet of water, which is a lot of water! The spillway, where they release water from the dam, was damaged so they couldn’t really send any water so the lake filled up and then they thought the spillway itself was going to collapse. Basically, all that water would go all the way down the valley all the way to Sacramento, and that would have been really catastrophic, so now there’s a big focus in California on dam safety and water and reservoir operations, so we go out and meet with the state of California and the USGS and the US Bureau of Reclamation to make sure we’re all on the same page and having the face time with the people so when it does hit the fan, we already have those relationships established with everybody.

Would getting a Ph. D make you more valuable to the weather service?

Honestly, I didn’t even need my Master’s. I just got it because I needed it to get in and I had that internship. A lot of your basic forecasting knowledge from your undergrad, you will learn when you’re on the job. Everything I needed to know besides the basics (why does it rain/storm), I learned on the job. We do have a few Ph. D students, but you really don’t need it and sometimes I think they’re looked at like they’re a Ph. D. You don’t really need it and I don’t think it would help you.

Describe your dream job

Well I love what I’m doing now but I wish I didn’t have to do shift work. That’s kind of the other part of the big equation is I just finished midnight shift and it takes so long to adjust to it and adjust from it that I never know when I’m sleeping or what day it is. I love what I’m doing so I put up with it for now but I wouldn’t want to do it for the next 20 years!

Thanks Courtney! Return to Top