Either way I don’t feel like you could get this off the ground and have ‘reliable’ data to gather, without some serious experimentation to find a ballpark resistance, based on the distance between the probes(handwidth is fine but is better defined in inches or centimeters), the composition of the soil being sampled, and of course the amount of water in the composition. But you’d also have to take into consideration the possibility of water pooling and skewing the readings, and many other possible and unforseeable variables that could make getting ‘accurate’ data very difficult.
I would try and use this method as a way of ‘thresholding’ the reading and finding the right value that resembles a relative ‘dry’ or relative ‘moist’ condition. After all, this is just for watering plants, correct?
Find the ballpark value that is indicative of ‘moist’, and a ballpark value that is indicative of ‘dry’ by experimentation, and in code use ‘greater than’ or ‘lesser than’ comparisons to decide whether to water, or not to water; and when watering, when to stop watering. Leave adjustment points (potentiometers) in your design and tweak it in. If you find that adjusting is no longer needed, measure the pot values and replace with discrete resistors.
Once you have the circuit tweaked to where you can accurately determine when your plants need watering and when they don’t, some interesting data to gather would be how often it waters and how much water it dispenses per watering.
But to find the exact voltage that means ‘dry’ or ‘moist’ must be found experimentally.