From the course: Data Visualization: A Lesson and Listen Series

Listen with Rob Simmon

From the course: Data Visualization: A Lesson and Listen Series

Start my 1-month free trial

Listen with Rob Simmon

“

(bright upbeat music) - I'm very happy to welcome today Rob Simmon, senior data visualization engineer at Planet Labs, to talk about color. Rob, thank you so much for being here with me today. - Hi, it's happy to be here, I'm happy to be here. - Great. So, we're here to talk about color and that is an impossibly broad topic I realized to cover it in a 15 to 20 minute interview. But when I say to you, what is the most important thing that you want to tell people working in data visualization specifically, when it comes to color, what is it and why? - I really think that it's good to remember that color's extremely difficult, it's a very complex topic and so don't get discouraged if when you first start trying to make decisions about color, that you might not get results that you really like immediately. And so like be patient with yourself and that's really for a number of reasons. Mostly because there's so many interdependent factors. So you have to deal with the light source itself, you have to deal with the surfaces that the light's falling on, you have to deal with the medium that the light's traveling through, in most cases, of course that's air and then you have to deal with the complexities of the human eyes and brain, which are basically taking all of the stimulus and then reconstructing color, you know. Color is really not an objective thing, it's very much something that's within us and that varies from person to person and varies from circumstance to circumstance. And so like keep all that in mind and just don't get discouraged when you're trying to do interesting things with color. - Yeah. It is a very difficult subject, it covers so many fields, there's so many issues involved, you know. Designers learn how to work at color, they study it, they learn a lot of those issues, you know, a lot of those things you were talking about and I see not necessarily all of them but I guess one question for you, is, are you a trained designer? Did you learn color theory and all those intricate details that you need to really become an expert in working with color? Or did you sort of pick up what you know about it just by working and doing and putting in your 10,000 hours? (laughs) - It was very much the 10,000 hours. I actually have degrees in material science, which is a type of engineering/physics that doesn't really deal with color design at all. And had the opportunity when I began working at NASA to learn by doing and to learn by reading and basically studying what at the time was a little bit of a new field as far as like how to use color in data visualization and scientific visualizations. It was sort of the transition period from sort of very simple computers that had a limited color palette. And so when scientists were doing visualization, they were constrained by those colors. But then 24 bit displays started being on the scene and we had full control over what we were displaying and sort of bridging that transition. And so, people like Rogowitz and Treinish who had written a paper, I think it's called, 'Why Should Engineers and Scientists "be Concerned About Color?" Which sort of went into the basics of color theory and how to present data with color. And then when all the way back to, you know, Burton who had done some of this work in the '60s. I think I was first inspired actually by a lecture by Tufty where he talked about how cartographers used color and how that differed from sort of your typical scientific visualization in the mid '90s and the rainbow palette. So in some ways you could consider myself self-taught, except it was really, you know, absorbing what other people had learned and they're writing about color and then trying to adapt that to all of the work that I was doing at NASA, where, you know, we were visualizing literally dozens, maybe hundreds of different scientific data sets and trying to make them all meaningful, yet all distinct. - Yeah. Well and you brought up names of academics and researchers, so you're building on best practices, you're not just making it up, you're not just trying to make it pretty, you're actually using color theory and all these things you're learning along the way. And I think that was a great point about this idea that, yeah computers used to be black and white. And then eventually they had 8-bit color, you know, 256 colors and then eventually more and then eventually more and became more and more important that we learn how to actually make all these colors work in a way that makes sense. So you mentioned something earlier, you phrased it, you didn't use these exact words but you said color is difficult maybe. And you know, I always say my pat answer when people ask me about color is that, the phrase I use always, yeah color is hard, it's really hard to do. And to be honest, I can't think of another subject that is as hard in design except maybe typography, just because there's a lot of nuanced typography. But even typography, I don't think has quite as many things going on. Why do you think color is so difficult to master beyond what you've already said, that it, you know, involves, you know, the physical, you know, the physicality of the eye and the light and the spectrum and all these other things. What makes us so hard to really master even once you understand those things? - Okay, well I think the real fundamental challenge is how variable it is. And so that's both, it varies based on your environment and it varies from person to person. And that's on top of all of the physics that go into the transmission and reflection of light. And to reiterate sort of what I earlier, like, it's an artifact of our brains as much as it is a reflection of the physical world. And so you have a light source which can vary in color. So if you think of the sun, it's, you know, it's white light, quote unquote, but it's also, it's really more yellow than that and then it's affected by the atmosphere, so there's direct light and indirect light. Then there's the surface which is absorbing and reflecting different wavelengths. And then that's traveling again through the atmosphere to your eyes. And then your eyes are taking that electromagnetic energy of the light and turning that into electrical impulses and chemical impulses within the brain and then the brain is trynna construct a 3D reality out of that. And color is one of the things that we use to sort of interpret our world. And so the colors that we see are our brain's way of helping us to interact with the real world. And they're not, it's not a constant thing, it really changes based on our environment. It can change based on our mood, it can change based on the brightness. And of course you need to account for the fact that about 5% of the population has some sort of color deficiency. Which means that they can't distinguish some hues from other hues. So very commonly it would be reds and greens predominantly in men, interestingly enough. And so it's really important to make sure that you're considering both people with full color vision and people who might have a little bit of a challenge perceiving some subtle hue changes and things like that. - Yeah. Add on top of those issues the fact that, you know, different printing types, you know, CMYK versus RGB, different paper quality, Macs versus PCs, laptops versus, yeah it's just an endless challenge. (laughs) So given that, you know, it's hard and we have to be very thoughtful about the strategies we use to effectively use color. We also have to think about those default settings in our software programs because of that variability. And, you know, by the way, here we are shooting during the pandemic in my house, which has lots of windows, I have a lot of trees around so it's very green in this room sometimes. And the sun goes behind clouds, we're experiencing the variability throughout this recording. So coming back to software, we have these default settings, default colors, Excel, you know, for instance, had really terrible default colors back in the day, they've gotten better over time. Tableau has some, it seems more adherent to best practices, default colors these days. Default color swatches in Adobe Illustrator are I think much more interesting than they used to be but it's still this, you know, very set palette. So our software seems to be catching up to best practices and bringing more interesting options available to us, I would say. And I just wanted to ask you about that. So are there any sort of specific examples of, you know, kudos, you want to praise any software that's doing it particularly well and, or any challenges you see with default colors in the software programs that we tend to use? - Yeah that's a really good question. And it's probably worth pointing out that there's sort of two different uses of color in visualization. There's color as a design element, where you're using color to direct the viewer's attention. Think like a big red button for, you know, performing an important function or stopping something that, before it goes runaway in and causes a crisis. And so the palettes in Excel, illustrator, Photoshop are more oriented around that and so like constructing a visual hierarchy with color. There's also using color and data visualization to denote quantity. And so where you're actually like, okay, this color represents a temperature or this color represents an amount of charge or things like that. And there, you really want to make sure that the colors are accurately representing the differences and the proportionality of the numbers that you're trying to get across. It's, because color is so difficult, it's really hard to like get an absolute quantity out of a color palette. And so they're really more useful for relative. And that means that you just need to be very very careful about that relative relationship. So, you know, in most cases you want the difference between zero and one to be the same difference between 10 and 11 and 100 and 101. And so you want to make sure that there's no like breaks or anything on the palette. And that's a really important difference in use of color and in defaults. And as you said, like for those design type palettes, Excel is definitely getting better. Like they used to have sort of like just a pure RGB CNYK colors and now it's a more sort of harmonious color palette. And those are good for design and also good for qualitative. So if you're doing like a map of land cover classification and you want to distinguish forest from water from urban from agriculture, you know, you would use different colors that all work together but are also distinct. Versus, you know, vegetation health where it's a quantitative palette and you want to say like, you know, this vegetation is, you know, a certain amount of biomass per square kilometer and this other vegetation is another and really maintain those relationships. And so that's more the realm of things like the Viridis palettes, which are coming out. And before that Color Brewer, which was this extraordinary tool, it's at this point 15 years old or so that really helped mapmakers to map quantities under their maps. So overall, there's a trend towards better use color and better palettes. But I think there's a little bit of stickiness and a little bit of resistance. So, people especially in scientific visualization are kind of sticking to the older ones, but there's a lot of progress being made that's really encouraging. - Yeah. I had no idea that Color Brewer was that old, 15 years. So you bring up a good point. You know, we think about, you know, the color palettes like the rainbow palette, which still seems to persist in meteorology and other places. And even though more and more people understand that this is a bad idea for, you know, uses like that. You know, you brought up this idea of continuous scales versus, we didn't talk about it in these words, but versus threshold scales. And, you know, so, you know, the idea that you can use gradients of color to, you could go down to increments of one and of zero to one versus five to six versus 10 to 11 versus 100 to 101. Or maybe we should just break things into bins of 50 and just say zero to 50 is this color and 50 to 100 is that color. Any advice you have there on when, with color in particular, because it's so much harder to differentiate hue variance. When to make the decision to go with the threshold scale versus at that continuous scale. - Yeah, ironically, that's something that I've never really come up with a good answer to and I definitely struggle with. And I think if it's critical for the viewer to be able to distinguish like groups of colors, like if 10 and 20, like there's some life and death decision that you need to make based on that difference, absolutely use a segmented color scale. And you really can't go above seven or eight segments, you know, more than that and they become indistinguishable in certain circumstances. So, if the quantitative information is the important part, I think it's okay to go or appropriate to go with a discreet scale. If you're more going for trynna show relationships and patterns, like say, ocean currents. Like, you don't necessarily need to break it up into those discreet bands, you can leave it as a more continuous gradient. And then, you know, we can see it as these flows and you can see the gradients and really like distinguish more, realistic is not quite the right word, but like it's, they're softer, they're more bigger picture as opposed to like trying to be precise. - Yeah. That's actually a perfect segue into my next question, because, you know, for the audience here, these are people who, you know, some of them are data visualization practitioners, experts, all this, you know, that sort of category. Others are, you know, people in IT or HR who are just making charts and graphs for their PowerPoints and they work with data all the time but they're not really thinking about data viz to that sort of level. For them arguably, oftentimes it's going to be more appropriate to use threshold scales. But for you, you use, you work with satellite imagery, right? And so you're thinking big picture, literally the globe or beyond. (laughs) And so you're telling a more nuanced story and that nuance and then it's okay to see that the entire ocean looks like this and it sort of goes from deep blue over here, there's something going on over here and it sort of gradually gets lighter blue over there cause it's different over here. And so, you know, you have this interesting collection of challenges on your side of that fence with satellite imagery, so you have the challenge of trying to visualize things that we expect to have certain colors. Like if you make the land mass purple and then the water orange, your audience might get a little bit confused, right, trees should be green, et cetera. So what other challenges do you face that might give us some insights, even on the more of the standard data visualization side. Any sort of things that pop out of you from those challenges that might help us on the other side of the fence? - Yeah, that's again, another excellent question. And I think you hit the nail on the head when you said meeting expectations. So, for satellite imagery in particular, like we have a frame of reference, especially for places that we live or places that we frequent. Like, I know what a tree looks like, I know what a road looks like, I know what water looks like although you would be surprised at how different those things might look from directly above. And so when I'm creating satellite images for a broad audience, I need to make sure that that's as interpretable as possible. And so, when people are viewing the image, it matches their everyday experience. The downside of that is sometimes, like you can't show certain things with that true color view that sort of match, again, matches what we see from the ground level. And you need to move to other false colors like other band combinations or, you know, showing a quantity and not a photograph like image. And there, I think you need to make sure that you frame what you're doing. And so you show this is what it would look like to you normally and then this is what it looks like in false color infrared and show both and make sure that you explain the two. And that way people have a bridge to go from one to the other. And then if I'm doing something that's more quantitative, if it's something that people are familiar with, again, like vegetation, that's a good, or clouds, you know, I want heavy vegetation, dense vegetation should be dark green. Clouds, dense collage should be bright white. That way, people don't have to try to retrain their brains to understand what they're seeing. It gets a little bit more complicated when you're doing something like electric charge, you know, which doesn't have a color and there you have a lot more freedom for choice but I think you still need to keep in mind your audience. Like the core tenant of data visualization and really all types of communication is understand who you're talking to. Because if you're say, doing something for a chemist, they're used to these ball and stick models that have set colors for set types of atoms, like oxygen and carbon and chlorine. And if you vary from those colors, you're going to confuse the acanemics, even if you're making something that's more appropriate for a general audience. - Right. - So you really, sometimes there's a negotiation between the source of a dataset and the audience for a dataset. But really, you know, just try to balance, you know, what people's preconceived notions versus their learned experience, you know and things like that, if that makes any sense. - Yeah it makes perfect sense. And so it sort of circles back to the theme that I come back to every single time I teach anything to anybody, which is, it's always all about the audience, right. It's, you have to understand your audience, you have to understand, as you said, what their expectations are, what their norms are, their standards and do your best to match those. And then that includes with color for sure. We're running low on time. I did want to thank you very much for being here but before we go, you know, I want to acknowledge that we're, I said at the beginning, we're covering an impossibly broad topic to be sure. But do you have any final thoughts you'd like to share with this audience? And again, keep in mind, very broad audience here. This is not just data viz practitioners day-to-day people, this is business people who work with charts and graphs here in Excel and PowerPoint, et cetera. It's both. So any final thoughts on advice around color for those folks? - Yeah I think it's really important to keep your eyes open and sort of be aware of your environment and try to, if, especially if you're interested in color and want to use color in communication, notice, you know, what's the color of the light of the room that you're in? How does the light change from morning to noon to afternoon, from winter to summer, from a cloudy day to a sunny day? You know, you were mentioning that yeah, the sky is changing because you're in a room with a lot of windows. We're going from, - We're experiencing that - the marine layer. - all the time. - A nice gray layer and that's starting to break up a little bit. And so I'm getting sort of yellower, brighter, crisper light coming through the window in front of me. So, you know that's changing and that's going to affect how I perceive my environment here, how the computer's camera is recording things. And, you know, even the light inside the room, because as the light bounces around different surfaces are absorbing and reflecting the light differently. And so, like I have pretty neutral gray walls and like everything around it's a little bit yellow. And so you can really, you know, sort of pay attention to how mutable everything is. Like how bright summer day, you're in a sort of dark inside and you walk outside all of a sudden you can't see anything at all because it's so blindingly bright. And sort of how shadows change. Like people think, oh, a shadow, it's black. Well, shadows are not, they're rarely if ever black, they're actually the opposite color of the light that's falling on them. And so, if you keep that in mind and like use palettes that are the same palettes that we're used to in nature, I think they tend to work a little bit better. - Yeah. That was actually one of my favorite tools. I can't remember who did it, but essentially you could take a picture of it with your iPhone and it would generate like a five color palette based on that natural palette. And yeah, it was always the, they always generated the most beautiful, perfect, wonderful palettes that-- - I think there's an Adobe tool that does that. - Yeah, maybe, that sounds about right. So yes listen, Rob, thank you so much for joining me here today. I really appreciate all your insights into color and otherwise. And, I know that you write, you have a blog about the work that you do and maybe you can just tell us a little bit about it and so people can know where to find it and how to check it out. - Sure. I have a, I blog on Medium, I also will blog on the Planet website. Things range from color obviously, GTL, which is programming, sort of quasi programming language for manipulating special datasets like satellite imagery, like maps, things like that, all on the command line, which basically I'm writing about as I learn. And, sometimes about things more, like that transition from very restricted computers to computers that are much more powerful and how to navigate like sort of the change from handcrafted data viz and handcrafted visualization before computers to automated. And sort of like the joy and the skill of having things that are still made in a handcrafted way. And so I've got a ton of blog ideas. I've several unfinished series that I need to, I'll follow up on and hopefully in the coming months we'll get an opportunity to flush out. - Great. Yeah and so you mentioned, you know, you're writing as you're learning essentially and there's no better way to learn than to document, write it down, try to teach others. And I'm learning every day as I'm doing this course and I've learned a lot from you. So again, thank you so much for being here, really appreciate it. - You're welcome, it was fantastic.

Contents