Flow Summary

Flow data for the stream from July and August 2004 are shown in Table 1. July was a particularly dry month with below average rainfall. Records from the Musselburgh weather station show this July to have recorded the second lowest total rainfall since records began: the Dunedin monthly average was 26 mm compared to the historic average of 61 mm.

Rainfall statistics for August show that the month was considerably wetter than July: the monthly average was 150 mm compared to the historic average of 56 mm. The peak in August flow occurred between the 6th to 7th due to a significant storm rainfall event (45 mm of rain fell over the 48 hours to 8 pm on 7/8/04 - Musselburgh weather station).


Flow (L s-1)
July (n=20)
August (n=31)

Table 1. Showing flow data for the three months of monitoring starting June 2004. The symbol 'n' denotes the number of days included in the statistics.


Base flow is defined as those that which occurs under steady state fair weather conditions and do not include natural or man made event flows such as those described below. They therefore correspond to flow rates attributed to the slow infiltration of drainage and surface and sub-surface runoff. Base flow conditions are expected to fluctuate on a weekly, monthly and seasonal basis; an example for July is shown in Figure 1 below.

Figure 1. Graph showing the base flow condition before and after a change in flow due to a storm rainfall event in Kaikorai Stream on the 15th of July. Base flow conditions are taken at about 100 L s-1 following the yellow. The Deep Stream overflow from Mt Grand during this time is considered to be zero - see Figure 5 14/07/04.


Storm flow

Storm events produced a characteristic change in flow rate. With the onset of rainfall, flow increases rapidly and then tails off more gradually - producing a flow curve which is asymmetrical. The rapid increase in flow is considered a characteristic of urban streams where stream channelling, road and paved surfaces and the storm water drainage system help to funnel water quickly into the stream. The duration of the rainfall events determine how long the event lasts and the potential for overlap with other subsequent rainfall events (Figure 2).

Figure 2. Graph showing the change in flow due to a storm rainfall event on the 13th of August in Kaikorai Stream. Note the rapid increase and more gradual decrease in flow for the event.


Melt flow

Dunedin commonly experiences at least 1 snow falls to near sea level each winter. Two large falls occurred on the night of the 15th and early 16 th of August 2004 and on the 23rd and 24th of the same month. The subsequent thaw and melt contribution to the stream appears to have produced a different flow profile from storm rainfall events. The profile shows pronounced doming involving a gradual increase and decrease in flow (Figure 3); the occurrence of this doming in flow in the afternoon, perhaps aided by solar melting, may also be significant.

Figure 3. Graph of stage and flow for the 25 th of August in Kaikorai Stream. Snow fell in Dunedin to sea level on the 23rd and in the morning of the 24th.


Man-made event flow

The input of water from the Mt Grand water treatment station gives the hydrology in Kaikorai Stream a special character. Water is piped from the Deep Stream catchment in the coastal Otago hinterland some 40 km to help supply Dunedin with water.

Ordinarily this flow enhances Kaikorai Stream by supplying water of good quality and low temperature. In addition, this helps to dilute any other pollution that comes from human activity further down the stream course and flushes such water towards the mouth. Without this flow in Kaikorai Stream, we might expect the flow and level of dissolved oxygen to decrease, and turbidity, temperature, and conductivity to rise.

The flow in the pipeline is at a maximum of about 525 L s-1 or 46000 m3 per day. Water not needed to supply Dunedin at any particular time is allowed to flow (directly without passing into the station itself) into a tributary of Kaikorai Stream in the upper catchment. Depending on the demand for water in Dunedin or periods of shut down at the treatment station the amount of water allowed to flow into that tributary changes.

These changes are observed downstream at the monitoring station in Kaikorai Valley College by sharp changes in base flow (see Figure 4). These alternating base flow conditions produce characteristic profiles that are special to Kaikorai Stream. On the 4th of July for example (Figure 4), a rapid rise in flow occurs around 10.30 am, followed by a steady elevated flow until about 10.40 pm, and a drop so that by midnight the flow is similar to pre event levels.

Figure 4. Line graph showing changes in flow due to increase in volume of water coming from the Deep Stream pipeline as determined from the Mt Grand Water Treatment Station (for explanation see text).


Larger and smaller increases and decreases in flow have been recorded that follow the shape of flow curve in Figure 4. A maximum value under fine weather conditions appears to be about 700 L s-1 - suggesting a maximum contribution from the Deep Stream pipeline source.

If the maximum flow that the pipeline can deliver to the station is in the order of 525 L s-1 then, provided filters are not back flushing at the time (see below) this should leave about 175 L s-1 that is residual flow for this particular day - mostly it is assumed is from natural sources.
That notwithstanding, a case of variance occurred on the 13-14th of July 2004 that suggests the supply may on occasion be greater than 525 L s-1. What appears to be residual flow is only of the order of 100 L s-1 see figure 5. This level of flow was maintained as a base flow for the subsequent days of the 15th and 16th of July.
We are currently trying to assess the error ranges associated with the flow measurements to see if they might explain discrepancies in residual flow values.

Figure 5. Line graph showing changes in flow due to decrease in volume of water coming from the Deep Stream pipeline as determined from the Mt Grand Water Treatment Station (for explanation see text).


Repeated episodic changes are also apparent in the flow data for Kaikorai Stream (see Figure 6). These events have a duration that is in the order of 40-50 minutes and typically involves flow changes that, although variable, are between 80 and 120 L s-1. The flow peaks always show a sharp rise, but more gradual fall producing slightly asymmetrical shape. These changes are most clearly evident at times of lower base flows on fine weather days.

The Mt Grand Water Treatment Station has a resource consent to periodically discharge to waste the contents of 6 large chambers that filter the water before it leaves the plant to supply the households of Dunedin city.
These chambers discharge when the water in them is judged, electronically, to exceed minimum values for turbidity that are set to maintain high quality water is received by ratepayers. We are currently studying the link between these discharges and the periodic changes in flow recorded at the Kaikorai Valley College monitoring station.

Figure 6. Line graph showing changes in flow possibly due to periodic discharge to waste of water in filter chambers at the Mt Grand Water Treatment Station (for explanation see text).


Artefacts and errors

Occasionally artefacts appear in the recorded values for some of the water quality parameters. At present these artefacts occur for reasons that are unclear. They are probably due to some electronic irregularity or an error in data transfer. This means that you need to be careful in interpreting some of the data from the monitoring station.

One example of an artefact is the rare occurrence of a spike in dissolved oxygen that then gradually corrects itself over time (see shark-fin shape in Figure 7). Another case in point are turbidity spikes. These are most easily seen when they occur over say one 5 minute record on days with no change in flow (see Figure 8). Be aware also that the turbidity sensor is cleaned each week to prevent build up of algae and clay on its window surface. This would otherwise cause a gradual increase in turbidity values over time. A correction is applied to the turbidity data if required.

Figure 7. Graph of Water Quality Parameters in Kaikorai Stream for the 12/9004 showing an artefact dissolved oxygen peak (shark-fin shape). -flow during this time was unchanging. Note also the diurnal fluctuations in temperature and pH.

Figure 8. Graph of Water Quality Parameters in Kaikorai Stream for 10/09/04 showing an artefact turbidity spike and the effect of manual cleaning of the turbidity sensor. Note also that flow was constant for the day and the fluctuations in temperature, dissolved oxygen, conductivity and pH (see water quality parameters section).

Any scientific work should have an appreciation of the errors involved in measurement. The calibration of the sensors has shown that each of the water quality parameters had the following overall errors associated with them when they were calibrated: conductivity +/- 1.4% (of full scale 0-1000 µS cm-1), temperature +/- 0.2 ºC (of full scale 0-50), dissolved oxygen +/- 0.2 mg L-1 (of full scale 0-20), pH +/- 1.4% (of full scale 0-14) and turbidity +/- 2% (of full scale 500 NTU).

Finally, note that the flow data is also subject to error. The flow data calculated manually by gauging have errors that were all under +/- 6%. The rating curve developed from this data is a line of best fit and so also has errors associated with it. This is particularly so where the line is extrapolated beyond the controls of actual gauging points: very high flows and very low flows by their nature may be one-off events or seasonally influenced and therefore cannot be added to the rating curve until they occur. The actual flow value should fall within +/- 8% of the rating curve calibration to be considered acceptable.

An example of such an estimate of accuracy can be made where later manual gauging is done between the points that make up this curve. A case in point occurred on the 6th of August when the monitoring station recorded an average flow of 1931 L s-1 for a period of 15 minutes in which a manual gauging showed a value of 2053 +/- 5%. Since 1931 is a value that is within 6-7% of 2053 then the value (1931) is an acceptable one for estimating flow; bear-in-mind also that the figure of 1931 L s-1 was based on an extrapolation (flows as high as this had not been recorded before).