For all intents and purposes, I think only counting "surface" water is more useful and intuitive. It's essentially any water that can participate in the hydrologic cycle on Earth, and that water locked beneath the crust doesn't really "matter" for what I think the intended purpose of this graphic is.
If we want to talk about the total amount of H2O around, on, or in the Earth then inclusion makes sense.
If we want to talk about water interacting with the surface environment (atmospheric, sea, ice cap, fresh, and subsurface aquifers and tectonic water), then splitting those into distinct categories probably also makes sense. In which case we can also show the subsurface water.
How much mantle water does make it to the surface over time is a good question. I've no idea though I'd suspect that some does through geothermal and tectonic activity. The more interesting question might be how we'd determine this (all but certainly through isotopic composition), and if a net flux could be determined.
Over geological time, additional reservoirs of water are significant simply because surface water boils off into space over time, with estimates I've seen of up to 25% of Earth's original allotment having done so over 4.5 billion years or so. As the Sun eventually grows warmer, this rate will increase. At the same time, tectonic activity will slow.
Note that there's a fair bit of water transport through the lower crust / upper mantle as oceanic plates subduct under continental plates, with the water absorbed into the oceanic plates playing a major role in volcanism at those plate boundaries, e.g., along the "Rim of Fire" surrounding the Pacific basin.
I simply presumed that water that's pumped from layers 100m or lower below the surface, water that's sometimes 10.000 years "old", wasn't surface water. But it makes sense to lump it in there too.
Is the water pumped from beneath the crust in the Netherlands poisoned too or have you guys got the last of the good stuff?
The breakout is for comparison: surface freshwater (total) and surface freshwater (lakes and rivers) relatively.
That one actually has percentages on the subsets too for interesting differences like glacier and icecap volume vs ground water (which I think still excludes the kind of deep mantle water mentioned up thread because it's not usably extractable).
Even if it was accessible water to accessible non-water I don't really see how the metric is relevant in any decision making. Is it warning against a half-baked plan to mix water with every available cubic meter of soil or rock? Because there wouldn't be enough water to do that crazy thing? Thanks, I'll bear that in mind.
Given how many miles I can travel over land or through the air, 2.5 miles _into_ the earth is amazingly shallow in similar ways to how the article already anticipated my amazement of how small the spheres of water were.