The Dvorak technique was developed in the 1970s and early 1980s. At that time, most satellite images were viewed on paper using black and white printers. To accommodate this medium, Dvorak developed the “BD Curve”. This curve assigned each satellite brightness temperature value to a specific shade of black, white, or gray.
The Dvorak technique relies on the analyst’s ability to identify each of these shades. Trained experts can usually do this relatively quickly. The BD Curve can be confusing, however, especially to newer analysts. Some colors are repeated, and it can be difficult to discern one shade of gray from another. We have developed a new full-color satellite enhancement for the Dvorak technique to address these issues. In addition to using this new color scheme for Cyclone Center, we plan to share it with tropical analysts around the globe.
The image above compares our color scheme with the BD Curve. Both schemes use gray shading to highlight clouds warmer than 9°C (48°F). The BD Curve then uses a second series of grays, while we give it a pink tint to help differentiate it from the warmer values.
Both color schemes use solid shades at varying intervals for temperatures colder than -30°C (-22°F). In our scheme, this begins with a dark red (which flows naturally from the pink). The colors become progressively less warm (orange, yellow, then shades of blue). Where the BD Curve is forced to repeat Medium Gray and Dark Gray shades, our colorized scheme is able to use unique colors throughout.
Note that the BD Curve uses black for temperatures from -63°C (-81°F) to -69°C (-92°F). This bold color marks a transition from moderate to tall clouds. This same transition is marked by the change from warm to cool colors in our scheme.
We have also included an additional color (white) for temperatures colder than -85°C (-121°F). This color is never used by the Dvorak technique, but it provides us additional information about the coldest clouds.
The images above show two views of Super Typhoon Gay (1992). The one on the left uses the BD curve; the one on the right is our color scheme. All features are identical in both color schemes, but we believe the colorized scheme makes them easier to identify.
We also wanted to ensure that our imagery could be easily interpreted by everyone, including people with color vision deficiencies. We were guided by the principles laid out by Light and Bartlein (2004). Specifically, we avoided any color scale that included both red and green. We also sought a scheme that varied in both hue and intensity. Our ultimate selection was inspired by the “RdYlBu” scheme from colorbrewer2.org.
The images above simulate how Super Typhoon Gay would appear to these users. These simulations are performed using vischeck.com. The one on the left simulates Deutarnopia, and the middle simulates Protanopia. These are both common forms of red/green deficiency. The image on the right simulates Tritanopia, a rare form of blue/yellow deficiency. These simulations suggest that any analyst, regardless of color deficiencies, would be able to identify the same features in our imagery.
In Cyclone Center, one of your first tasks is to: “Pick the cyclone type, then choose the closest match.” You may be wondering how we found the images that you’re matching against.
One of the first steps in the Dvorak technique is to determine the storm’s “Pattern” strength. It’s an initial estimate of the storm’s strength based on how the clouds are organized. Here are Dvorak’s original patterns:
Each of these patterns gets stronger as we move from left to right, similar to in Cyclone Center. We could have used these patterns in Cyclone Center. However, the strengths are irregularly spaced, and there are only two levels of strength for Eye storms. We chose instead to use real satellite images to identify each pattern.
Some of the most highly trained Dvorak analysts in the world work in the Tropical Analysis and Forecast Branch (TAFB) at the National Hurricane Center. To take advantage of this expertise, we sorted the satellite imagery from the Atlantic in 2003–2006 by the strengths and cyclone types that they assigned. We then chose representatives from each category based on these criteria:
- Image quality
- Similarity to the original Dvorak patterns
- Representativeness of that image compared with others of the same strength and cyclone type
- Continuum of strengths for a given cyclone type
The last criteria was particularly important since we wanted to show a clear progression from weakest to strongest in each cyclone type. So if you are ever debating between two images to select, remember that they go from weakest to strongest and see if that helps.
The World Meteorological Organization (WMO) has assigned the task of forecasting tropical cyclones to different agencies in different regions. For instance, NOAA’s National Hurricane Center in Miami, FL, USA provides forecasts for the North Atlantic and Eastern Pacific. The Central Pacific Hurricane Center in Honolulu, Hawaii provides forecast for the Central Pacific. Several other countries have responsibility for providing forecasts in other regions (such as Japan, Australia, India, etc.). These same agencies that produce forecasts of tropical storms, also produce post season analysis of each storm’s position and intensity – which is called best track data.
In addition to the WMO agencies, numerous other entities forecast and provide best track data. For instance, countries often provide the capability in their nation’s interest, such as China forecasting storms in the Western Pacific. The U.S. Joint Typhoon Warning Center (JTWC) provides forecasts for U.S. interests around the world. These agencies also provide best track data which overlap best track data from other agencies. However, there is not always complete agreement between organization on the strength of a given tropical cyclone. This is often due to different data available to each agency, different procedures in place for forecasting systems, personnel, etc.
The data available to study, forecast and understand tropical cyclones has changed significantly through time and varies from agency to agency. Prior to the 1940s, the primary observations came from ships and land-based weather stations. Beginning in the 1940s, the U.S. military began testing – and later made operational – flights into Typhoons (in the West Pacific) and Hurricanes (in the North Atlantic). They found that these reconnaissance flights could be conducted safely and that they provided a wealth of information on the storm’s structure, intensity and environment. Routine aircraft reconnaissance in the Western Pacific ended in 1987 but still continues today in the North Atlantic.
The satellite era was ushered in during the 1960s, providing more information on tropical cyclones. Numerous studies began relating cloud forms to intensity, which culminated in the Dvorak Technique in 1984. However, the availability and quality of satellite data varied, with some agencies still receiving imagery by fax in the 1990s. Similarly, newer satellites provide a wealth of information beyond imagery – microwave satellites provide information on storm structure, radar satellites observe precipitation and other satellites measure wind speed at the ocean’s surface. Again, different agencies have different levels of access to this data.
The result is 1) best track data has changed in time as availability of data changes and 2) best track data varies between agency, due in part to access to different data and routine procedures. This means that differences occur in the best track record.
More information on the WMO agencies is available here.
More information on IBTrACS is available here.
Image provided by MeteoFrance
When Dvorak developed his method, he knew that you could tell something about a storm’s strength by looking at its lifecycle. If a storm looks stronger than it did yesterday, odds are that it probably is! That’s why the first step of most classifications is to ask which of two images look stronger. These are actually two images of the same storm taken within 24 hours. If the image you see is from the first 24 hours of the storm (or the image 24 hours prior is missing) then you’ll skip this step.
We’ll use your answer to calculate something called the Model Expected strength. It starts with the storm’s strength from 24 hours ago. If you say the newer one looks stronger, then we’ll bump it up a notch. If the older one looks stronger, then we’ll bump it down. And if they’re about the same, then we just hold it constant. This isn’t as sophisticated as some of the other ways we estimate strength (see the upcoming posts on the Detailed Classifications), but it gives us a good first guess.
A number of characteristics determine whether a storm is stronger, weaker or about the same. There are two main measures of strength to look for:
1. How cold are the clouds?
Colder colors in infrared imagery indicate taller clouds that release more energy into a storm. Stronger tropical cyclones tend to have taller clouds and more of them. For example:
- The presence of more colder colored clouds in an embedded center suggest a stronger storm.
- Colder clouds surrounding an eye suggest a stronger storm.
2. How organized are the clouds?
This question is a bit more subjective, so just give it your best shot. Some features that might indicate which storm image is stronger:
- Stronger storms have spirals that wrap farther around the storm.
- The cold clouds near the center become more circular as a storm strengthens.
- Typically Shear and Curved Band storms are weaker than those with an Embedded Center.
- Storms with an eye are almost always stronger than storms without one.
- For storms with an eye, consider the shape, size and color of the eye. Eyes that are more circular, smaller and/or warmer tend to be associated with stronger tropical cyclones.
In some cases, the storm on the left may appear to have some of these characteristics, while the storm on the right may appear to have others. If this is the case, they can actually cancel out, in which case we would say that they are about the same. For example: If the storm on the left appears better organized and more tightly-wrapped, but the storm on the right has more cold colors, you would say that they are about the same.
You can use the images below to help you gauge a storm’s relative strength.
The satellite imagery in CycloneCenter were created from HURSAT data. These data are snapshots focused on tropical cyclones from around the world. The location of the snapshot moves with the storm, generally keeping the storm centered. Of course knowing where cyclones have occurred requires a separate data source (IBTrACS).
While both images from visible and infrared (and other channels) are available from HURSAT, the Cyclone Center uses the infrared data because its observations are available both day and night (since visible images are only available during the sunlit hours).
The image shows how storm data is remapped and gridded over each individual storm. A large enough region is included in each grid to allow studies of the cyclone’s environment. Sometimes this means that some grids will have another cyclone on its outer edge when cyclones form or traverse near enough.
The satellite data that make up HURSAT are from international geostationary satellites from the past 30+ years. The data – collected at NCDC – have been processed in a manner to minimize inter-satellite differences in an effort to create a seamless climate record of hurricane observations.
The images you see on Cyclone Center were observed by infrared sensors on weather satellites. These sensors provide an estimate of the temperature at the tops of clouds. Cloud top temperatures are very important because they give us an idea of how tall the clouds are. Temperature decreases with height in the lower atmosphere (up to 10 miles), so cold clouds are taller than warm clouds. Taller clouds are responsible for the heavy rain and thunderstorms that drive tropical cyclones.
In the Cyclone Center images, the cloud top temperatures are represented by a range of colors. The scale on the image above shows the temperatures in degrees Celsius that correspond with each color.
Black and gray are the warmest, indicating temperatures from 9°C (48°F) to 30°C (86°F). Often these will be the temperatures we experience at the land or ocean surface, but they can also be associated with very low clouds. Shades of pink go down to -30°C (-22°F). In our images, these are almost always associated with low clouds. Red, orange, and yellow come next, and they indicate medium-level clouds.
In most images, the coldest clouds you see will be shades of blue. Sometimes you’ll even see a cloud that’s so cold it shows up as white. These clouds are colder than -85°C (-121°F). Coastlines and political borders are also drawn in white, so make sure the white clouds are surrounded by dark blue. Otherwise, you might just be looking at a small island.
Sometimes there is a problem with parts of the satellite data. These missing data will show up as black lines in the images. Just ignore them and carry on with the analysis when you see them.
‘Detailed classification’ allows you to delve deeper into the imagery. That helps us better understand the storm’s characteristics and more precisely determine its intensity.
By performing the basic classification, you have provided us with what we refer to as the “Pattern” strength. It’s an initial estimate of the storm’s strength based on the how the clouds are organized (the pattern they make).
The detailed classification takes you further, asking more precise questions about the cloud structure itself. To opt in for the detailed classification, just click the checkbox below the image. You’ll be asked 3-5 questions that depend on the storm type you have chosen. For example, if you have chosen a curved band, we’ll be asking you what color the band is. If you’ve chosen an eye storm, you’ll be measuring the size of the eye and answering some questions about the clouds that surround it. The answers to these more detailed questions help us to really pin down the storm’s strength.
Each question that you are asked will have an accompanying “?” bubble to explain how that step is performed. It takes a little longer than the basic classification, but allows you to really get into the heart of the technique that we use. Plus it’s a lot of fun! It likely won’t take you more than a couple of minutes to go through the whole classification, and you’ll probably get quicker over time.
The more of these detailed classifications we can acquire, the more information we’ll be able to gather about the cyclones themselves. As with the basic version, there are no “right” answers, so give it your best shot! If you decide it’s not for you, you can always opt-out at any time by simply unchecking the box.
Tropical cyclones have enormous impact on life and property around the world. These storms can destroy cities, economies and even armies. However, our ability to understand the strengths or even just to count the numbers of storms that have occurred globally or even in each ocean basin is limited by deficits in the historical data. For instance, there are studies in published literature that suggest that typhoon activity is both increasing and decreasing in the western Pacific Ocean. Clearly both cannot be true!
Changing personnel, forecasting procedures, technological innovation and the introduction of new types of satellite data have created a record of past tropical storm information that is heterogeneous – that is, non-uniform. So looking at how the climatology or history of storms has changed over time is quite difficult and riddled with assumptions. For example, researchers interested in the number of tropical cyclones that have occurred each year throughout history have to make decisions about what years to begin or end their studies – the “period of record.” This choice can influence the results substantially. One possible choice is to choose a starting date as of when all recordkeeping began, say, 1851, ending in the present. This very long record would have a greater number of storms included but the data quality may not be as good as if the researcher made a different choice and started in the 1980’s satellite era. This alternative choice, limiting their years of interest to the time period when the entire globe has satellite observations would influence the number of storms included in the study. Before satellites, how would anyone know if a storm existed if it did not, in some way, impact humanity? Neither choice is “incorrect” but the assumption that is made clearly affects the results.
Looking at the history of one storm at a time, even fairly recent storms have discrepancies in their historical records. For example, four forecast agencies were responsible for tracking Typhoon Yvette in 1992 – the Joint Typhoon Warning Center (US Navy), the Japan Meteorological Agency, the China Meteorological Administration, and the Hong Kong Observatory. These agencies have different methods and procedures for estimating the wind speed in tropical cyclones. The differences can be quite large. As shown in the graph below, on October 13th, 1992, the Joint Typhoon Warning Center estimated Yvette’s winds to be 155 knots while The Hong Kong Observatory estimated only 90 knots. This would be the difference between a Category 2 and Category 5 tropical cyclone on the Saffir-Simpson Hurricane Scale used in the United States. It isn’t clear what the best answer is for Yvette.
Using group consensus, citizen scientists making storm strength estimates in CycloneCenter.org can help clear up this confusion. By participating in this endeavor, you and other citizen scientists are analyzing the satellite images through a process similar to the Dvorak Technique used by meteorologists, thereby providing a more consistent record of tropical cyclone strength. Your choices as you classify storms – your very mouse clicks – will lead to a better understanding of tropical cyclones. You never know, the statistics from the decision process you use to classify storms could even lead to further refinement of the intensity estimation process meteorologists use as well!
Greetings, and welcome to the official blog of Cyclone Center! We are very excited to give users the opportunity to answer questions about tropical cyclones using the same images seen by meteorologists around the globe!
Over the past few decades, technological advances in satellites have given us the opportunity to observe tropical cyclones, even if they never make landfall. NOAA’s National Climatic Data Center (NCDC), one of the world’s leading organizations in the collection of weather and climate information, has archived a dataset of nearly 300,000 cyclone images dating back to 1978. We are asking you, the public, to help us gather information about these images, in order to gain a better understanding about these fascinating weather phenomena.
In this blog, we will keep you updated on how the project is progressing, discuss current happenings in the tropics, and hopefully answer a lot of questions you may have about the weather and tropical cyclones in general!
Watch this space.