Gradient Factors

Two numbers, one decision

Open the settings on a technical dive computer and you'll find a pair like 40/85 or 30/70, written GF Lo first, then GF Hi.

Those are your gradient factors, and they quietly shape every decompression stop you do.

They don't change the physics of your tissues. They change how close to the edge your computer is willing to let you ride.

Erik Baker wrote them up in 1998 as a way to put a conservatism dial on top of an existing model: the dissolved-gas Bühlmann ZHL-16 algorithm. They are not a new theory of bubbles. They are a margin you choose against an old theory you already trust.

This article is about what that margin actually does, what the two numbers control, and why the diving science community has quietly moved away from the "lower is safer" instinct most of us started with.

If you haven't met the underlying model yet, start with M-values, half-times, and tissue compartments. Gradient factors only make sense once you know what an M-value is.

What a gradient factor is a fraction of

Bühlmann's model splits your body into tissue compartments and tracks the inert gas dissolved in each one.

For every compartment, at every depth, there's an M-value: the most supersaturation (the excess of dissolved-gas pressure over the surrounding pressure, the thing that makes a tissue want to fizz) that compartment is allowed to carry before the model says you're risking bubbles.

Picture two lines on a graph. The bottom line is ambient pressure, where the gas pressure in the tissue exactly matches the water around you, no supersaturation at all. The top line is the raw Bühlmann M-value, the model's hard ceiling.

A gradient factor is just where you sit between those two lines, written as a fraction.

• GF 0 is the bottom line. No supersaturation allowed at all, which would pin you to the surrounding pressure and make any ascent impossibly slow. Nobody dives this; it's just the floor of the scale.
• GF 100 is the top line. The full, unmodified Bühlmann M-value.
• GF 50 sits halfway between ambient and the M-value.

So a gradient factor doesn't invent a new limit. It tells your computer to use a chosen slice of the gradient Bühlmann already permits.

What each dial actually does

Lower the GF Lo and your first stop gets deeper.

A deep first stop caps how much your fast tissues are allowed to supersaturate early in the ascent: it slams the brakes on while you're still deep.

Lower the GF Hi and your shallow stops get longer, and you arrive at the surface carrying less supersaturation.

The part that surprises people: GF Hi is the dial that matters most for your decompression sickness risk. The shallow stops, governed by GF Hi, are where the bulk of off-gassing (shedding the dissolved gas back out) happens and where your surfacing margin is actually set. GF Lo gets all the attention, but it's the junior partner.

That single fact is behind most of the modern advice that follows.

Have a play with it. The dive below is fixed at 45 m for 25 minutes on air; only the two dials move. Drag GF Lo down and watch the first stop sink deeper. Drag GF Hi down and watch the shallow stops stretch out and the total deco grow. It's the same Bühlmann model the whole time. You're only choosing how much of its margin to spend, and where.

The numbers people actually use

Shearwater ships three presets, and they're a useful map of the landscape:

Note the direction: "Low conservatism" has the higher numbers, because higher GF means closer to the raw M-value, which means less margin.

Then there's a rule of thumb that caught on as the diving-research community moved away from deep stops:

GF Lo ≈ 0.83 × GF Hi

Pick your surfacing margin first (GF Hi), then set GF Lo to about 0.83 of it. So a GF Hi of 85 gives a GF Lo near 70, the familiar 70/85 pair. It is one popular way to land a shallow-biased profile, not a law.

The reason isn't arbitrary. Setting the two numbers in that ratio roughly flattens supersaturation across all your stops, instead of piling extra stress deep and racing through the shallows. You're spreading the load evenly rather than betting it on the deepest stop.

A dial, not a bubble model

Baker built gradient factors for one job: give a diver a single honest knob to back away from the raw Bühlmann limit. You pick how close to the M-value you're willing to ride, and the computer interpolates the rest. That's all it was meant to be: a conservatism dial on a dissolved-gas model from the 1980s.

Then the deep-stops era arrived, and the dial got borrowed for something it was never built to do. Bubble models like VPM were pushing deep, early stops, and divers wanted those same deep stops out of the Bühlmann computers they already owned. The lever that produces them is GF Lo: crank it down and your first stop sinks deeper. So a very low GF Lo, 20 or sometimes less, quietly became the field hack for "give me a bubble-model-shaped ascent on a dissolved-gas computer." Try it in the widget above: pull GF Lo to 20 and watch the first stop plunge. That deep, front-loaded profile is exactly the shape the bubble-model crowd was chasing.

It's worth being precise about what that does and doesn't do. Dropping GF Lo really does change the shape of your ascent to look more like a bubble model's. But the computer underneath is not modelling bubbles, and it isn't quietly switching theories on you. It never tracks a single nucleus. A low GF Lo just spends more of your existing dissolved-gas margin deep instead of shallow: same model, same M-values, a different slice of them used earlier in the climb.

And we now know that deep redistribution isn't the free safety it felt like. The NEDU trial and the work that followed found that front-loading stops deep, whether a bubble model told you to or a low GF Lo did, leaves the slow tissues quietly loading while you hang there, and does not cut the bends (cases of decompression sickness). The dial can imitate a bubble model's ascent; it cannot give you a bubble model's reasoning, and on the evidence the imitation isn't the upgrade it was sold as. The whole story is in bubble trouble.

Three things people get wrong

"Lower GF Lo is always safer." This is the big one, and it's false.

A lower GF Lo means a deeper first stop. While you sit deep, your fast tissues do off-gas, but your slow tissues keep on-gassing, because at that depth the surrounding pressure is still loading them up. You've traded a manageable shallow obligation for a heavier slow-tissue one you now have to clear later.

The Belgian military study (De Ridder, 2023) pointed the same way: its more conservative, deep-biased gradient-factor settings were not safer than shallower-biased ones, and a deep-stop-style profile showed no advantage. Deeper wasn't safer.

"Gradient factors add bubble modelling." They don't, as the section above lays out. It's still pure dissolved-gas Bühlmann underneath, with no bubble ever tracked. A low GF Lo only changes the shape of the ascent, not the model doing the maths.

"GF Lo is the important number." As above, it isn't. GF Hi and the shallow stops carry your real risk.

What's settled, and what isn't

Settled: a shallow-biased gradient-factor schedule is defensible and at least as safe as a deep-biased one. Four independent lines of evidence converge on that, and they don't usually all agree on anything. The instinct to stop deep and stop early, the thing bubble models pushed for years, did not hold up when it was actually tested on divers.

This is the same lesson as safe ascents: the shallow zone is where decompression is won, and rushing it is what hurts you.

Contested: the best setting for helium-heavy trimix is genuinely still open. Helium off-gasses differently from nitrogen, and deep mixes bring in complications (isobaric counterdiffusion, where swapping gases lets one diffuse in while another diffuses out, and inner-ear decompression sickness) that muddy the simple "as shallow as possible" advice. The flat-supersaturation logic is a strong starting point for trimix, not a closed case. This is where good instruction and real planning software earn their keep.

How to read it on the computer

You don't have to do any of this maths in your head underwater. Two live numbers do it for you.

GF99 is your current supersaturation, right now, as a percentage of the M-value at your present depth. If it reads 92, your leading tissue is at 92% of its limit. If it climbs past 100, you're over the M-value, so hold or drop down slightly.

Surface GF answers a different question: if I bolted to the surface this instant, what would my supersaturation be up there? You hold your shallow stops until Surface GF drops to your GF Hi setting. When it does, you've cleared your obligation and may surface.

Together they turn gradient factors from a planning abstraction into something you can simply watch tick down.

That's the short version. Actually reading these two numbers in the water, knowing what to do when GF99 creeps up and how to bank a little extra conservatism with Surface GF after a cold or hard dive, is its own skill. Safe ascents covers flying the stop on a real computer.

The honest summary

Gradient factors are a margin, not a model. They tell a decades-old dissolved-gas algorithm how much of its own permitted supersaturation to actually use.

GF Lo sets your deep stop; GF Hi sets your surfacing, and GF Hi is the one that matters for risk. Most computers default to 40/85, and that's a defensible place to start; if you want to flatten the stress across the ascent, the 0.83 rule nudges GF Lo higher, toward something like 70/85.

The old instinct to stop deep and stop hard didn't survive contact with the evidence. Lower is not automatically safer, and a GF Lo in the teens has nothing behind it. Pick a sensible pair, watch GF99 and Surface GF on the way up, and be more conservative when the dive was long, cold, or hard.

References

1. Baker EC. Clearing Up the Confusion About "Deep Stops." Immersed. 1998;3(4).
2. Doolette DJ. Gradient Factors in a Post-Deep Stops World. InDEPTH. 2019.
3. Doolette DJ, et al. NEDU TR 11-06: An Evaluation of Decompression Algorithms for Use in a Fleet Dive Computer. US Navy Experimental Diving Unit, 2011.
4. De Ridder S, Pattyn N, Neyt X, Germonpré P. Selecting optimal air diving gradient factors for Belgian military divers: more conservative settings are not necessarily safer. Diving Hyperb Med. 2023.
5. Doolette DJ, Mitchell SJ. Recreational technical diving part 2: decompression from deep air dives. Diving Hyperb Med. 2013;43(2).

Gradient factors and ascent strategy are core to every technical course I teach. Enquire about training →