You'll need to measure the torque yourself - then allow a heathy safety margin (factor of two or three perhaps?). Measure the force needed to pull the arm, multiply the force by perp. distance from pivot. Piece of string and a weights pan might be the simplest.
For motorising the globe valve things will be more complex as the end-stops can only be determined by measuring the torque or stall condition. You have to avoid over-tightening yet be able to supply a larger torque to reverse from the end-stops (such valves bind when closed, more torque is needed to un-stick them due to static friction).
Offhand I'd say you could use something like a motor with a series resistor. You normally power via the resistor and monitor the current by measuring voltage across the resistor - this allows you to detect when the motor stalls (endstop). To then reverse back you briefly bypass the resistor to get more current/torque to start the motion again before going back to current-monitoring mode. So having an H-bridge supplied via a load resistor with a high-side switch in parallel to the resistor might be one way.
As for the mechanical issue, the motor would need to either be in a frame than can slide up and down with the screw-thread motion, or (perhaps better) done via a reduction gear with the pinion long enough to cope with the along-axis motion.
Another method is to have the motor drive a set of prongs that go into the holes in the hand-wheel - easily removed and allows manual override.
Note that if you have two much reduction-gearage it will be hard to measure / limit the torque to valve, too little and the motor will need be over-powered.