*inter alia*) the Queanbeyan and Molonglo River basins. Each month I send off a report of the daily flow ratings and rainfall and they get added to the information bank.

However, until now I have never tried analysing the data myself. Given the difficulties I have found in compiling this post that is possibly not surprising!

Cutting to the chase, what the data suggests is that if there is:

- a high amount of rain there will be a high number of days with higher flows; or
- a lot of days of light rain gives a low number of days with higher flows but a high number of days with lower flows

The monitoring site is where our driveway crosses the Creek by means of 2, 1m diameter, pipes. One of the pipes takes the primary flow while the other only gets active in heavier flows.

The flow codes are fairly subjective and the words chosen are such that they sort alphabetically in order of strength of flow:

- Zero; the pipe under the driveway is dry.
- Trace; the pipe isn't dry but the sound of running water isn't audible unless close to the outlet;
- Light: one pipe running noisily;
- Heavy: both pipes running;
- Flood: water is across the drive.

Rainfall has been recorded one way or another whether we have been present or not. I have looked at 5 measures of rainfall - all are quite obvious in their definition.

- Annual rainfall in mm;
- Number of days with a fall >5mm;
- Number of days with any rain recorded;
- Number of days with 0.2 -5mm of rain;
- Proportion of rain-days with 0.2 -5mm of rain.

My first way of looking at the data was to chart the % of days in a year for each of the 5 flow codes.

That chart is a bit too complicated to make much sense of, beyond illustrating that in most years the incidence of codes usually follows the alphabetic order (ie most zero, least flood etc).

To try to reduce the complexity I combined the three lowest values and the two highest. I then looked at the correlations between those two series and for measures of rainfall. The correlation with heavy-flood has the same number as for the zero-light range , but with reversed sign (ie if one is -0.87 the other is +0.870. That has logic, but I can't prove why it's so: trust me, this is on the internet so must be true. (I suspect it is something like a zero-sum game - if one item is big the other must be small: this applies to both the flow codes and measures of rainfall.)

Since the combination of flow codes above was more or less chosen arbitrarily I also looked at a few others: combing zero and trace (and thus light - flood); and just looking at zero . None of them had significant (using the word loosely) values of the correlation coefficient. The largest measures of correlation were between the 3+2 flood codes with mm of rain (+/- 0.87) and proportion of days with 0.2 -5mm of rain (+/-0.78) so I will concentrate of them.

**correlation of r=0.87 between amount of rain and heavy - flood creek flow**

__positive__rather than the equal but negative relationship with lower flows.

Note that in the first chart the lines move, more or less in the same direction while in the second they move in the opposite direction.

The conclusion is that with a lot of rain I observe more days with heavy flows.

When thinking about correlations with number of days with some, but not much rainfall, I will only show the comparison of days with light rainfall and zero-light flow. (As demonstrated above the chart for heavier flows would be the mirror of this one.)

Again the two series move in the same direction in most cases. My conclusion here is that with a high proportion of days with low rainfall there are not enough days of heavier rain to generate heavy or flood flows. This is the zero-sum situation referred to above.