Diagnostics for Model Structure:

1
Diagnostics for Model Structure Improving
Hydrological Models using Data from Experimental
Basins
Recession Analysis
Analysis of streamflow recessions can be used to
give insight into the storagedischarge behaviour
of a watershed, and hence to appropriate
lower-zone representations. We study the
relationship between flow (Q) and its derivative
dQ/dt.
A weir used to measure flow in Satellite Catchment
Hilary McMillan1, Martyn Clark1, Guillermo
Martinez2, Dave Goodrich3, MS Srinivasan1,
Maurice Duncan1, Ross Woods1, Andrew Western4
1 National Institute of Water and Atmospheric
Research, Christchurch, New Zealand 2 Department
of Hydrology and Water Resources, University of
Arizona, USA 3 US Department of Agriculture,
Agricultural Research Service, Tucson, Arizona,
USA 4 Department of Civil Environmental
Engineering, University of Melbourne, Australia
h.mcmillan_at_niwa.co.nz

• Conclusions
• No single Q vs dQ/dt relationship exists, so a
  single baseflow reservoir formulation (e.g.
  TopModel) is rejected
• Season controls the proportions of storage in
  each reservoir, by influencing the recharge
  history of the catchment
• Synthetic experiments show that multiple linear
  perennial reservoirs or a non-linear baseflow
  reservoir are required to reproduce low-flow
  behaviour

Mahurangi Experimental Catchment, Northland, New
Zealand
Figure Recession Analysis by season at Satellite
Catchment
Satellite Sub-catchment - Intensive monitoring
of soil moisture
Soil Moisture Analysis
In the satellite sub-catchment we have direct
measurements of soil moisture. Although there is
a clear relationship between soil moisture and
flow (figure below), surprisingly there is no
relationship between time-derivative of soil
moisture and flow. This is consistent with tracer
studies at the catchment which suggest that flow
is controlled by a deeper reservoir with
residence times of months-years, and suggests
interflow is less important in this catchment.
Aerial view of Satellite Sub-catchment
29 flow gauges (circles) 13 rain gauges
(triangles)
Our aim Using data to inform modelling decisions
Our research aims to use experimental data
collected at Mahurangi to inform the structure(s)
of a national hydrological model for New Zealand.
We tested many different model structures using a
modular modelling framework. All gave good
Nash-Sutcliffe fits to the hydrograph. We want to
identify structures which give the right answers
for the right reasons.
Examples of modelling decisions
ET-available water
Upper Zone Architecture
Average soil moisture over storm events reveals
strong connectivity between upper and lower soil
layers. The need for multiple reservoirs (see
recession analysis) probably reflects deeper
aquifer systems.
Lower Zone Architecture
Productions mechanisms for saturation excess /
interflow