Maps with depth use every pixel to describe something about that particular part of the Earth's surface. There was certainly a time when this kind of data simply didn't exist. Charted territory was restricted to towns and rivers and shorelines. The first real data set to be created that (at least by inference or extrapolation) covered every square inch of land on continental scales was contour line data representing elevation. This occurred in the early 1900s thanks to air-photo interpretation techniques developed over the course of WWI. If you look at atlases from the 1920s forward, you'll begin to see the proliferation of this first class of “deep” maps. They used “hypsometric tinting” to describe elevation at every point on the land area of a map; hypsometric tinting simply being continuous or discrete colouration based on height above sea level.
But there's a potential pitfall with this generalization. Do we find greenness in deserts that occur at low elevations? Do we find greyness in a high alpine meadow? Maybe, but in these instances - probably not. So while an interesting relational story can be told with hypsometric-tinted maps (this road goes over a high mountain, or this land with few lakes is very flat), it is not necessarily a story that relates most closely to our human experience of landscape, and not necessarily the best choice as a background for most maps.
Even if we consider elevation on it's own merits (regardless of how we colour it), unless we're in the mountains, we don't really define our experience of place in terms of elevation. Our experience of place is far more defined by the environments we find at different elevations. Prairie vs. forest vs. city. These definitions (or to use a map term “land cover data classes”) are how we experience the world. So how to represent these environments on a map? The default approach these days is to use a satellite image. After all – this is a photo of what is actually there at any given place on the planet.
But as my teacher told me in my first cartography class “a satellite image is not a map”. This point can certainly be argued either way – but the point he was making was that a satellite image has not undergone the important cartographic process of classification and symbolization. And I agree that maps are best when all elements shown have been classed into categories that can be described clearly and understandably. Though there are unquestionably challenges in applying this process at a continental landscape: where exactly does the city end and the farmland begin? is this area a wetland or a forest? - we get a much clearer graphic communication when the reader of our map knows what every pixel on the map is meant to represent.
As for the ocean (generally under-attended-to on maps), elevation (or in this case depth (also know as bathymetry)), actually is the most intuitive way to understand the marine environment. Ocean depth is the best single indicator of the kind of environment one might find in a given part of the ocean, and so seems a good analogue for land cover on a deep (pardon the pun) map.
For the purposes of this post – I have shown only the backgrounds (imagery, land cover, etc.) and have stripped off the roads, rivers and town names that would typically be nestled into these contexts to create the relational “ah-ha” moments. As I said to begin this post – maps are by definition relational, and a truly deep map is one that shows a river passing not through an infinite sea of yellow background, or an undefined series of textures and colours, but instead through clearly defined mosaic of cities, forests, deserts, etc. And while I continue to see far more maps with hypsometric tinting or satellite imagery as their contexts, my hope is to see the rise of landcover as the new background of choice for mapmakers.