I am designing an aquatic center on the top floor of a multi-story building. The critical beams and columns to be designed are supporting a 25 m lap pool and I am wondering whether to apply a live-load or a dead-load safety factor to the water load acting on the structural elements.
Given that the water level will probably remain almost entirely stable, dead loading is the way to go.
In fact, the only time I imagine the water level would change is if the pool is emptied for cleaning, but that is quite a rare event and, in fact, would reduce the applied load, so no worry. Sure, when the pool is then being refilled there would be some dynamic action, but filling up a pool takes a lot of time, so the "dynamic factor" should be negligible.
Also, the reason safety factors are greater for live loads is due to their inherent variability. You may design a bridge for a 45-ton truck, but who knows how heavy trucks will be in 40 years' time? Meanwhile, dead loads are far more consistent: in 40 years' time, the bridge will weigh the same as it does now. And unless your material is defective or your construction is stupendously shoddy, the self-weight you use in your analysis will probably be within a few percentage points of the true weight. You can therefore place a lower safety factor for dead loads, while adopting large ones to cover the uncertainty in live loads.
As @AndyT mentioned in a comment, there may be elements that are relieved by the pool. An empty pool in such cases would not have this relieving effect. Relieving dead loads are usually fully considered (maybe slightly reduced). Relieving live loads are ignored, causing no relief. For these elements, one may decide to consider the empty-pool case as well.
As @Ethan48 mentioned in a comment, it may be relevant to calculate the structure with the pool both empty and full. The empty pool could be relevant for the columns, when checking the "maximum bending moment (due to wind, for example) and minimal axial compression" case.
So, perhaps the most reasonable approach is to consider the load as a hybrid: with a dead load factor when unfavorable but a live load factor (0.0) when favorable.
This mostly applies to somewhat "simple" structures. If your building requires more complex calculations, then things might get messy.
For instance, I'd guess this may have an important effect in seismic calculations, since the water (or lack thereof) would represent a large mass (or lack thereof) at the very top of the structure. As well, as mentioned by @CarlWitthoft in a comment below, if the building is slender and therefore vulnerable to wind-induced dynamics, the dynamic motion of the water could come into play and therefore act more like a live load.
You need to check with your local building codes. California has special codes for treating elevated water tanks and pools and containers. Not only they are treated with dynamic load factor but other considerations in structural assembly, detailing and use of facility apply.
Intuitively in case of an earthquake after part of pool water is spilled out, after a few seconds the left over will start to swing back and forth and collide to the walls of the pool and break on them in an explosive impact, much like the shore line retaining walls and piers.
So you need to be mindful of local seismic spectra and avoid resonance and in the absence of dynamic lateral analysis follow the prescribes scheme advised by you local building department.