If you know me chances are you’ve heard me rant about wind roses, and how they are misguiding architectural design. Chances are however, that you’ve also seen me use them to inform project design. How can that be?
In this post I’ll guide you through why wind roses are so popular, why most designers are using them wrong, and how, despite all, if properly used I believe they can be a powerful design tool.
What is a Wind Rose?
Wind rose diagrams are one of the staples of climate analysis: colorful, simple, and seemingly intuitive. They speak on their own. They are often used to drive massing decisions, inform natural ventilation strategies, and qualitatively assess pedestrian comfort.
Appearing in the 14th century, compass roses were conceived to convey the directions of the incoming winds, including eight major winds, the eight half-winds, and the sixteen quarter-winds. While direction was the main variable conveyed in these compasses, you’ll notice that the Tabla Anemographica below already conveys a notion of either wind speed or frequency, represented by the amount of air blown by each character.
As data availability and visualization evolved over time, the graphic representation of wind compasses evolved to pack as much information as possible into a single graphic. The modern wind rose shows:
- Wind direction: concentric wedges, following the cardinal directions
- Frequency of winds blowing from particular directions: concentric circles, emanating from zero at the center to increasing frequencies at the outer circles.
- Wind speed (or other): color of each wedge segments. (I’ve seen this variable replaced by Universal Thermal Comfort Index in outdoor comfort studies.)
Do Wind Roses Lie?
This wind rose is, however, both flawed by design and improperly used.
1. Frequency is inadequately represented.
Edward Tufte, the father of data visualization, states that a one-dimensional variable should only be represented one-dimensionally. Without bending your brain too much, what this means is that no one-dimensional variable (say, wind frequency) should be represented by elements that grow/shrink in two dimensions (say, a wedge). In other words, you should never visually represent your kids’ weight, say, 20 and 40 lb, by two bubbles of radius 20 and 40, respectively. Why? Because your eyes will read “area” and not “radius,” and will therefore interpret that your older kid is four times heavier than your youngest, instead of two times.
Similarly, when using a cropped wedge to represent wind frequency, your eyes are giving much more weight to the largest wedge, instead of the smallest one, despite the fact that they should be considered identical in value.
If you’d like to learn more, we highly encourage you to read Tufte’s chapter called, fittingly, Graphics That Lie, in his book The Visual Display of Quantitative Information.
2. Calm wind data is almost always left out.
Take a look at the most recent wind rose you created or used to design a building: does it tell you how frequently calm conditions occur? The answer is: most likely not. While a few climate analysis tools allow you to account for calm conditions, most will simply leave this very important piece of information out.
The two wind roses below represent the same set of data, with one key difference: the one of the left does not include calm wind conditions, which amount to a whopping 33% of the year for this location.
When using wind roses to design for natural ventilation, ignoring third of the data as the upper left rose does could mislead us into thinking that a building can be fully ventilated with wind-driven strategies.
3. Frequency is just hard to grasp numerically.
Sure, one can easily tell qualitatively which is the most frequent wind direction. But how many times have you actually counted the concentric circles of a wind rose to quantify frequency? Can you easily tell how many hours of the year the wind blows in a particular direction? In fact, most wind roses don’t even label the circles… and nobody seems to notice! This is a clear sign that modern wind roses aren’t actually very helpful quantitatively!
Take a look at the wind rose below and indicate from where the wind blows most of the time. Odds are, you’ll say south/southwest.
And you’ll be wrong.
Why? The wind only blows from south/southwest 25% of the time, which means the wind does not come from that direction 75% of the time! How can that be possible? First of all, because wind roses create an optical illusion that distorts the reader’s perception of frequency. Second, because it takes adding up all those donut slices of 1% frequency to estimate a total frequency for a particular direction, which is not something our eyes are competent at.
Wrongfully designing a natural ventilation strategy around the “dominant” 25% will yield very different outcomes to occupants than correctly designing to all possible wind directions.
Creating wind roses requires making sure that the data is appropriate for the site being analyzed.
4. The data is often inadequate.
Leaving wind rose design aside for a minute, it is critical to understand that, like with any other analysis, a wind rose follows the “garbage in, garbage out” principle. Have you made sure that the weather file you use aligns with the geographical location you’re analyzing? Great! But that’s just step one.
The wind speed must match the height of the analysis plane. Most weather stations (though not all) measure wind speed 30 ft (10 m) above ground. If you are studying the impact of wind speed at a different height than that (say, pedestrian level or the top of a skyscraper), you must modify the data to match the height, according to the wind profile. Climate analysis tools like LadybugTools have this capability.
The wind speed must match the terrain type of the site. Imagine a city like Boston, whose airport is by the sea. The airport wind data will have the profile of unobstructed terrain. However, if you want to use that data for a project downtown, you must adjust the wind speed to align with an urban terrain. Again, you must find a climate analysis tool that offers this capability.
The wind rose should come with a disclaimer: “This graphic lies. Use at your own discretion.”
5. Wind roses are rarely answering a question.
Perhaps the greatest reason why wind roses mislead architectural design is that they are typically generated without an analytical question in mind: they are created as a generic climate graphic waiting to incite a design reaction. Given that wind roses are optical illusions very prone to displaying the wrong data set, if one is going to use them it must be done as part of an analysis led by a driving question. Think about it: an annual wind rose will barely answer whether you can use natural ventilation or not and will not answer when you can use it unless you break it up into smaller seasonal roses. Smaller seasonal roses will not tell you when in the day you may use natural ventilation either unless you break them further into daytime/nighttime.
If you think about it, there are few analytical questions that a wind rose can answer. They exist to give a high-level overview of wind direction and wind speed, neither of which should be influencing design without further digging into the data. In fact, the wind rose should come with a disclaimer: “This graphic lies. Use at your own discretion.”
If wind roses are so terrible, should we still use them?
We are still looking for a wind data visualization format that fulfills the needs that we think the wind rose does.
Our favorite alternative so far was introduced by a client looking to understand the impact of daily and seasonal wind, temperature and humidity trends on an island project we were working on. The diagram summarizes wind direction and intensity on an annual and daily basis, layered by outdoor thermal comfort conditions. It identifies the ocurrence of strong winds. In this graphic wind intensity is qualitatively represented through arrow thickness, and only the dominant wind direction is highlighted. This graphic was created with a specific question in mind, and was much more successful at highlighting specific climate properties (strong winds, daily and seasonal trends) than a wind rose would have.
That said, most designers are so used to wind roses and no matter how much some of us rant about them, they are likely to extend their thousand-year presence among us. We therefore recommend being very careful with your use of wind roses, and only showing them when they are answering a specific performance question, as opposed to showing them as an informative graphic in a climate analysis.
Tips to create an acceptable wind rose
If, having learnt the good, the bad and the ugly about wind roses, you will think they are the best way to display the wind data for your project, be sure to follow the following steps.
Begin by getting the right data, by making sure it:
- Corresponds to the location you’re studying.
- Has been modified to match the terrain height that your design question is focused on.
- Has been modified to the terrain type you’re studying.
In terms of composition, make sure that the rose:
- Answers a specific analysis question.
- Isolates data for the season or time-of-day of interest.
- Takes advantage of the color legend for an extra layer of information.
Finally, be sure that you are including critical information such as:
- Indicate the frequency of calm conditions.
- Indicate the frequency associated with each concentric circle. (We like to use hours as opposed to % of time.
Here are a couple of examples we’ve created to guide the design team we’re working with. The first set was created to address how to implement a natural ventilation strategy in the shoulder season, considering both daytime and nighttime ventilation options.
With a very different question in mind, the second example below was created to evaluate the risk of exceeding extreme wind speeds on an annual basis.
More information
If you’re interested in hearing other experts’ opinions on the influence wind roses have on architectural design, we invite you to watch the Wind Rose Debate, organized by Project StaSIO in January 2023.
