There's also the potential of co-locating non-evaporative cooling with CO2-scrubbing using a strong KOH solution in the big cooling towers and regenerating the hydroxide with calcium hydroxide (regular process done in kraft process paper mills, though they use NaOH with slightly favoured properties over KOH).
KOH is very hygroscopic; it does not have to be humid for a barely-still-liquid solution to actually dehumidify in the process.
Or, if that's too much effort for them, use e.g. LiBr or other such atmospherically-stable salt that can maintain humdity equilibrium with dry ambient air.
The reason for even involving a water-based solution at all is because you can spray it or at least run it over dense corrugation (sheets alternating orientation, but overall with the channels roughly pointed upwards) "packed beds" with free contact between the coolant and the air, instead of having to maintain a barrier layer between the two (typical car/computer radiators, but also AC coils), which notably saves you from even a potential for a there-required barrier layer to leak and from having to clean such a barrier layer.
If you make the corrugations large enough and have some simple mesh filters in the intake path that you just roughly shake/rinse/blow/brush off every once in a while, you can prevent particles larger than a rice grain from getting to the coolant in the first place and wash/rinse all the sand grain and smaller dust particles down into the coolant sump where their densities are far better matched (than air vs. dust) and volume flow is much easier to handle/filter.
The big part of these is still that they don't require active fan ventilation to cool radiators, which would be a substantial increase in critical electrical power needed after a SCRAM to keep the core from melting down; vs. the passive evaporative cooling towers and the KOH/LiBr non-humidifying variant I mentioned.