Low Flow Modelling to detect rating PERTURBATIONs G.H. Macky

advertisement
LOW FLOW MODELLING TO DETECT RATING PERTURBATIONS
G.H. Macky,1 E.J. Brown,2 and J. C. Smith3
1
DHI New Zealand
Waikato Regional Council
3
Water Resource Science Ltd, formerly Waikato Regional Council
2
Aims
There is now increasing pressure on Councils to have accurate flow data available in real time for
managing water allocation, in particular when restrictions should occur during low flow conditions.
Monitoring low flow conditions in rural rivers can be difficult because of the lack of a reliable and
stable stage-flow rating. Significant weed growth during such periods may cause unstable ratings
and inaccurate estimates of flows. During low flow conditions in the Piako catchment of the
Hauraki Plains near Hamilton, more than 50% error can occur in the rated flow within a week of
the rating being corrected. This is largely due to either weed growth or sudden die off.
In the Piako catchment, calibrated hydrological models have been developed for two gauging
sites, to see whether the models can determine when rating curves become invalid. The ultimate
aim is to deploy such a model in real time, so that the Regional Council can identify at an early
stage when the rating is changing and arrange fresh gaugings or simply adjust the rating.
Method
The prime component of these models is a lumped, conceptual rainfall-runoff model, NedbørAfstrømnings-Model (NAM) (Nielsen & Hansen 1973). It simulates overland flow, interflow
(horizontal leakage), and base-flow components of runoff as a function of the moisture contents in
four interconnected storages, but in this application only the baseflow component is of direct
interest. NAM accounts first for surface storage, which loses water to evaporation, interflow and
infiltration into the lower zone and groundwater storage. Evapotranspiration demands are first met
at the potential rate from the surface storage, then at a reduced rate by root activity from the
lower zone storage. Once maximum surface storage is reached, some of the excess water
enters the streams as overland flow, and some goes to groundwater storage. Baseflow is
computed from groundwater storage and a user-specified time constant.
There are up to 11 parameters in NAM that the user can adjust during model calibration, and an
auto-calibration function that speeds the process. For this study, auto-calibration was used as a
first step in calibration, but refinement of the calibration was completed manually to match
modelled and rated river flow best during the baseflow recession which typically occurs from
about November to mid-April.
To convert the NAM runoff to stream flow, groundwater flow out of the catchment should be
subtracted. A desktop study (White et al. 2014) had quantified long-term groundwater flow rates
from the catchments. This flow has been assumed constant except for a sinusoidal seasonal
variation, with the magnitude of the seasonal variation used as a further calibration parameter.
The catchments of two long-established flow gauging sites were modelled: Piako @ PaeroaTahuna Rd (537 km2) and Piako @ Kiwitahi (104 km2). Flow-stage gauging data pairs from
recent years were accessed to confirm the rated flows and conversely to identify periods when
the rating might have drifted. Typically, more than 10 gaugings per year have been carried out.
Rainfall and potential evapotranspiration data were obtained from virtual climate station data and
averaged over the catchment.
Results
Modelled, rated and gauged flows at the Paeroa-Tahuna Road are compared in Figure 2 for
2012-13, showing baseflow recession in early 2013. The recession in the rated flow is interrupted
only briefly by minor rainfall events. This period is typical of the results obtained: modelled flows
are slightly higher than measured in some years and slightly lower in others, but the calibration
process has replicated the trend of the recession.
This result means that this graphical comparison of modelled and measured flows should clearly
indicate when rated flow data become inaccurate due to weed growth, weed die-off or bed
erosion or scour. The model for the Piako @ Paeroa-Tahuna Rd is now being set up to run in
real time, linked to the same real-time rainfall and flow information as the flood forecasting system
for the Hauraki basin.
Some adjustment of model parameters may be desirable because the nearest rainfall data
available in real time are at sites outside the catchment. Regardless, the real-time model will
provide a useful early indicator that a stage-flow rating has significantly changed. With
experience, it should be practicable to apply an adjusted rating prior to obtaining and analysing a
new gauging. Based on past flow records and past gauging programme, this would allow a
management response to decreasing flows 2-4 weeks earlier than has previously been possible.
Deploying the model is therefore expected to help the Regional Council understand low-flow
patterns and manage abstractions within the catchment.
References
Nielsen, S.A. & Hansen, E. (1973) Numerical simulation of the rainfall runoff process on a daily basis. Nordic
Hydrology, 4, 171-190.
White, P.A., Tschritter C. & Rawlinson Z. (2014) Water budgets in the Piako catchment, Hauraki Plains.
Letter Report, GNS.
Figure 1
Figure 2
Piako River at Paeroa-Tahuna Rd, looking downstream
Piako River at Paeroa-Tahuna Rd, modelled, gauged and rated flows, July 2012 – June 2013.
Download