Incorporate some programmatic checks into the program that allow for qa/qc of the final data, prior to it being entered into the database.
significant change in rolling variance in the time series - we have been having a lot of issues with null units in many of our transducer downloads; if the processing misinterprets these units, it can cause a huge difference in the observable variance of the time series.
water level above the casing top elevation - keep in mind that at a few locations, this is possible (ex. transducer in a piezometer that can become completely submerged in a pool)
a check for negative, null, or unrealistic water elevation values
I found that temperature can be used to eliminate spikes caused by freezing conditions or situations where the instrument is out of water
Incorporate some programmatic checks into the program that allow for qa/qc of the final data, prior to it being entered into the database.