Cloud-Drift Winds

Read...

As you've already seen, remote sensing from satellites can be used to estimate the intensity of a tropical cyclone via the Dvorak Technique. Indeed, several other satellite-based remote sensors help forecasters observe various aspects of a tropical cyclone's structure and intensity as well. We'll cover several more of these sensors as we continue through the lesson. In this section, I'm going to focus on a remote-sensing technique that has broader applications than just tropical cyclone analysis.

The same geostationary satellites used to execute the Dvorak Technique can also be used to remotely retrieve tropospheric wind information by calculating what are formally called atmospheric motion vectors (AMVs). You'll also sometimes hear AMVs referred to as cloud-drift winds (CDWs) or cloud-tracked winds. On the Web, you may encounter any of these phrases, so just realize that they refer to the same thing. For simplicity, however, I'm going to stick with the term "cloud-drift winds" (CDWs) because it intuitively describes how this product is created. In a nutshell, this technique retrieves estimates of wind speeds and directions at various altitudes by tracking the movement of clouds on satellite loops. The process sounds pretty simple, but it can actually be quite challenging.

Before we really explore cloud-drift winds, I should point out that they rarely get a mention in any NHC discussions. So, what's the role of CDWs in tropical weather forecasting? Well, as you know, there's a serious dearth of routine radiosonde observations over the oceans, where the lack of data introduces errors into numerical simulations of the atmosphere. Thus, the capability of getting a proxy for winds by measuring how fast clouds drift over open seas is invaluable. With numerical weather prediction in mind, it should come as no surprise to you that CDWs are assimilated into computer models. 

While cloud-drift winds often play somewhat of a "behind-the-scenes" role in weather forecasting, we can still use available CDW data to infer important things about tropical weather. For example, check out the CDW data in the image below, which shows wind barbs annotated on infrared satellite imagery over the Southeast Indian Ocean. The various colors are described by the key in the upper-right corner of the image (green wind barbs are in the layer between 800 mb and 950mb; yellow = 600 - 799 mb; blue = 400 - 599 mb). For example, look off the west coast of Australia and note the closed, cyclonic circulation. Remember that this image is from the Southern Hemisphere, so "cyclonic" refers to a clockwise circulation. Using infrared imagery alone, the system would only appear as an innocuous blob of relatively low clouds that are hard to pick out, but the yellow barbs derived from cloud-drift winds showed that a closed circulation existed between 799 mb and 600 mb. These CDW observations helped forecasters not be fooled by the unremarkable blob of low clouds and realize that there was actually a tropical depression present.

Credit: CIMSS

Now that you have a little background on CDWs and their applications, let's look at the technique of retrieving winds from various types of satellite imagery (they're not all based on IR imagery). As its name suggests, cloud-drift winds are derived from a sequence of satellite images. In the simplest sense, you could spot a cloud and watch it move with time; but, as you can imagine, in reality, it's not that straightforward. As an example, I'll describe the technique for retrieving winds in the lower to middle troposphere from infrared satellite data. First, the technique requires three successive images. Next, "target clouds" are selected according to brightness gradients (large gradients in brightness, for example, typically mark cloud edges). The pressure altitudes of the cloud targets are then estimated from the intensity of infrared radiation detected by the satellite.

Credit: CIMSS

An important caveat here is that the brightness gradients associated with candidate targets must remain relatively consistent in time, which means that not all clouds provide viable targets. To get a feel for what a sample of suitable cloud targets might look like, check out the dots on the infrared satellite image on the right. Indeed, multilayered clouds (decks of low, middle, and/or high clouds lying over the same geographical area) are eliminated as potential targets because trying to assign altitudes to multilayered clouds poses nightmarish challenges. So, ultimately, we can't accurately determine CDWs everywhere using infrared imagery because of these challenges, and the fact that some areas have no cloud cover at all.

Fortunately, we're not limited to infrared imagery for CDW observations. As you already know, loops of water-vapor images can be used to assess winds in the upper troposphere (forecasters infer these winds even without a formal CDW technique). Indeed, in areas without high clouds to track the upper-level winds, water-vapor imagery becomes indispensable because it allows us to follow "vapor targets (opens in a new window)" (like cloud targets on infrared imagery) in time. Following "vapor targets" allows us to deduce the speed and direction of upper-level winds over tropical seas quantitatively (as opposed to the qualitative approach that forecasters often use when looking at water vapor loops).

The schemes used to generate middle- and upper-tropospheric winds from water-vapor loops are similar to the technique used to generate lower-altitude winds from infrared-satellite loops. For starters, three successive water-vapor images are required, and horizontal gradients in water-vapor (or high cloud tops) that remain coherent in time serve as potential vapor targets. Even though water vapor imagery provides more target options, many of the vapor targets end up being removed because of various quality control issues. After applying the same principles involved in retrieving CDWs from infrared imagery, out pops water-vapor images with middle- to upper-tropospheric winds -- just like the one below showing CDWs over parts of North America, Central America, and South America (and surrounding oceans).

Credit: CIMSS

The different colors of the wind barbs correspond to their layers as described by the key at the top right of the image. Note however, that color schemes for plotting various layers of cloud-drift winds can vary from Web site to Web site, so it's wise to not get locked in on one particular color scheme. At the time of this image, a tropical depression had just formed off the coast of Central America, but a corridor of relatively fast north-northeasterly flow in the upper troposphere over the storm (note the wind barbs in the 250-100 mb layer indicating winds as fast as 40-50 knots or more near the storm), with a few in the 350-251 mb layer, too) was helping to create strong vertical wind shear over the storm. Persistent vertical shear over the storm helped prevent it from becoming anything more than a minimal tropical storm during its lifetime. 

Cloud-drift winds aren't limited to just traditional infrared and water-vapor imagery, though. CDW data can also be produced from visible imagery, though it has an inherent limitation of only being available during local daytime when there's enough reflected visible sunlight to produce a visible image. Shortwave infrared imagery is often used to supplement visible imagery so that there's no data void at night. If you're interested in learning more about CDWs from visible imagery, check out the Explore Further section below.  

Finally, I should point out that CDWs have long been used to assess large-scale wind patterns, but they often aren't the best tool for detecting the details of wind patterns around tropical cyclones. Remember that tropical cyclones are rather small features in the scheme of things, and available CDW targets are often too few in number or too far apart to capture the finer details. However, improvements in satellite technology, including the introduction of very high-resolution sector scans focused on important weather features, allows for better tracking of wind fields around tropical cyclones. These high-resolution satellite scans with their more frequent updates make it possible to track a higher number of cloud targets (potentially thousands within the region of a tropical cyclone), allowing for a higher-level of detail in CDWs.

To see what I mean, check out the image of these "High-Resolution CDWs" below, showing CDWs in the vicinity of Hurricane Melissa (2025) as it was beginning a period of rapid intensification. The blue and cyan wind barbs represent winds above 250 mb, and they show a clear pattern of upper-level divergence over the storm (note how the wind barbs blow outward from the center of the storm). This upper-level divergence, of course, would help air columns near the center of the storm lose weight and lower surface pressures.

Credit: CIMSS

You might have noticed in the image above that most of the wind barbs were located in the upper troposphere, which isn't surprising, given the plethora of high-topped clouds in and around a hurricane. However, in some storms that don't have well-developed central dense overcast regions or expansive high-level cirrus flowing outward at the top of the storm, high-resolution CDWs can track lower level winds around the storm. But, in Melissa's case above, CDWs below 500 mb were limited to the bottom left corner of the image (purple and red wind barbs blowing from the west-southwest -- in toward the storm).

If you want to learn a bit more about how cloud-drift winds are produced, or want to view real-time CDW data, you may be interested in the resources in the Explore Further section below. In the final analysis, assessing CDWs at a variety of wavelengths (corresponding to those that create water-vapor and infrared imagery, as well as shortwave infrared and visible imagery if available) gives forecasters a more complete picture of winds throughout the depth of the troposphere. Such a "multi-channel" approach was essential for assigning heights to potential cloud targets. But, the utility of the multi-channel approach in satellite remote sensing has broader applications, which we're about to investigate.