jwbernin:
I'm going to write the software such that readings are taken when d1 is high, and not when d2 is high.
I did something similar and found it was quite tricky to get meaningful readings out of the sensor. In my case, the best correlation with moisture seemed to occur when I took a reading about 5ms after applying power. Like you, I reversed the sensor polarity after taking the reading to counteract any ionization effects.