Whelen uses plain ol' DTMF, but in a precise timing sequence. If the timing is off, they won't decode. I had about 150 tornado sirens in a system I used to manage before I vested out. The sequence also allows for polling and diagnostics, so if you have the E969 or E2010 units, you can collect maintenance data. But they're footprint hogs being large hunks of mostly vacant plastic with a circuit board, display, and a keypad membrane. So, most consoles can emulate the sequences if the timing can be set. The SCADA boxes could be placed "somewhere else" out of the way, and maybe put a printer with a Centronics interface on them.
Some old tornado and fire siren systems used QC1 to activate, but I haven't seen anything recent to decode them, and if they break, I'm guessing whomever maintains them needs a "plan B."
Motorola and Zetron supported the 2+2 format. I have to plead guilty to unnecessarily putting a 2+2 sequence into my former department's CentraCom II alerting stack, but I grew out of it, and so did they. It was kinda cool, but a few LA area comm guys have told me it was notorious for falsing. Funny thing is that aircraft use them on HF for SELCAL, but their decoders are software driven now, and falsing resistant.
Knox has proprietary tone frequencies for their sequences, and are time coded.