I have streamflow data (daily average discharge) from 1980 to 2006 at one site. In other words, for each day there is one value that represents the average streamflow for that day. I can assume that streamflows at this site are affected only by precipitation and irrigation return flows. In other words, if there is no rain for a few days the only water in the stream is from irrigation return flows.

Also, I have precipitation data for the same period, but it is monthly totals.

In 2005 and 2006, irrigation return flows were theoretically reduced due to changes in the canal system operation. I need to find out if irrigation return flows really were reduced.

First, I am trying to subtract out the precipitation part of the streamflow. What is left should be irrigation return flows. I have added daily streamflows to get monthly totals, then I regressed monthly streamflow on precipitation to see if there was any correlation (using Excel). R-squared was only about 0.5. So, I don't think I can subtract out precipitation.

What can I do? Are there any procedures you would recommend to test if irrigation return flows were reduced over the 2005-2006 period? Can I use a non-parametric test to compare means and/or medians of flow data from 1980-2004 to those of 2005 & 2006?

Finally, the average-daily flow data is compiled from instantaneous measurements collected every 15-minutes. So can I assume that the central-limit theorem applies here and thus average daily flows are normally distributed?