Executive summary

For this project 485 flow records containing at least five years of mean daily flow data were assembled. These records were used to create regression (random forest) models to calculate several hydrological indices. Calculated values of the same hydrological indices were also extracted from NIWA’s national TopNet model and databases available from the previously completed Hydrology of Ungauged Catchment (HUC) projects. Calculated values for each method were then compared with observed values using scatterplots and by calculating root-mean-square deviance. 

Results indicated that:

• the HUC method is the best currently available method for calculating mean flow for application to Auckland region;
• the Random Forest method is the best currently available for calculating 7-day mean annual low flow (MALF) for application to the Auckland region;
• the Random Forest method for calculating 7-day one-in-five low flow (Q5) is the best currently available method for application to the Auckland region;
• the Random Forest method for calculating the proportion of flow in February is the best currently available method for application to the Auckland region.  
• the Random Forest method for calculating FDCs over all-time and for each month of the year is the best currently available method for application to the Auckland region.

These data were used to demonstrate how calculated hydrological values at all rivers in a region can be used to inform river management decisions. For example, maps were used to illustrate that the proportion of time in February that is lower than the 7-day MALF varies across a region. This has implications for both reliability of supply to water users and for ecological effects when setting minimum flow and total allocations as a proportion of 7-day MALF.  

Auckland Council technical report, TR2012/042

June 2012