I am comparing some data, it is daily water usage (megalitres) over the same time period from one year to the next. I have constructed some histograms of the data. It follows a log-normal distribution.

To compare say August 2007 to August 2008 I will have 31 pieces of data in each. I'm thinking a paired T-test.

And then if I want to do it on a yearly basis 2007 to 2008 i'll have 365 days of data in one & 366 days of data in the other.