1. A comparative assessment of AWBM and SimHyd for forested watersheds
Bofu Yu1, Zhongli Zhu2
1School of Engineering, Griffith University, Australia
2 School of Geography, Beijing Normal University, China

2. Background
Workshop objective: to apply and evaluate model performance in simulating ‘non-stationary’ hydrologic time series;
Climate and streamflow data were provided for 14 watersheds/catchments around the world;
Protocol was stipulated for consistency for all modellers;
AWBM and SimHyd are by far the most commonly used models for streamflow estimation in Australia;
Two smallest of the 14 watersheds/catchments were selected for comparison purposes.
Reasons for selecting small watersheds were is that rainfall and climate data are usually much better for smaller watersheds, and the effect of hydrologic lag is minimal.

3. Two models AWBM SimHyd Model type Conceptual Developer Walter Boughton
Francis Chiew
Where model developed
Brisbane, Australia
Melbourne, Australia
Latitude (degree)
-27.50
-37.78
Time step
Daily
No. of parameters 8 9
Input data requirements
Rainfall and potential evaporation

4. AWBM

5. SimHyd

6. Model comparison With default parameter values for
AWBM and SimHyd, and
max. retention value of 105mm
for the SCS Curve Number method,
all models show a non-linear
relationship between rain excess
and runoff.
NB: All the ‘stores’ in AWBM and
SimHyd were assumed to be half
full initially.

7. Two watersheds Rimbaud Fernow Country France The U.S. Watershed
The Ruisseau du Rimbaud at Collobrieres
The Fernow EF River at Watershed 6
Closest major city
Nice
Pittsburgh
Latitude (degree)
44.24
39.07
Area (km2)
1.4
0.2
Period
Land use
Forest
Cause for non-stationarity
Fire in 1990
Clearing , re-forestation in 1973
Gross runoff coefficient
60%
41%
Solid prec./total prec.
<1%
14%

8. - models developed
- watersheds applied

9. Working Hypotheses None of the 2 models is going to work;
Conceptual models would perform better for watersheds with a higher runoff coefficient and (for these 2) with less snow fall

10. Two watersheds Rimbaud Fernow Country France The U.S. Watershed
The Ruisseau du Rimbaud at Collobrieres
The Fernow EF River at Watershed 6
Closest major city
Nice
Pittsburgh
Latitude (degree)
44.24
39.07
Area (km2)
1.4
0.2
Period
Land use
Forest
Cause for non-stationarity
Fire in 1990
Clearing , re-forestation in 1973
Gross runoff coefficient
60%
41%
Solid prec./total prec.
<1%
14%

11. Model comparison in terms of the N-S coefficient of efficiency for 30 calibration-validation combinations
AWBM
SimHyd
Fernow, the US
0.43±0.11
0.40±0.12
Rimbaud, France
0.79±0.07
0.77±0.09
There is a much larger difference
between watersheds than between models!!

12. Model performance in terms of NSE between 2 watersheds
Histograms of the Nash-Sutcliffe
coefficient of efficiency for 30
calibration-validation combinations
(level 1 & 2 requirements)
For the Rimabaud, no NSE is less than 0.6, for Fernow none is above 0.6.

13. Rimbaud

14. Rimbaud

15. Fernow

16. Fernow

17. Observed hydrologic changes at Fernow
Variable
Period
No. of years
Rate
(mm.yr-1)
p-value
Precipitation 28
+6.7±3.1
0.04
Streamflow
+11.4±2.5
<0.01 24
-2.2±5.5
0.69
-10.1±4.6 51
1.3±1.5
0.40

18. Fernow

19. Observed hydrologic changes at Rimbaud
Variable
Period
No. of years
Rate (mm.yr-1)
p-value
Precipitation 39
-11.3±4.3
0.01
Streamflow
-14.8±4.2
<0.01
Both precipitation and streamflow have
significantly decreased over the 39 year period.

20. Rimbaud

21. Conclusions (what we have learnt from this exercise)
Difference in model performance is small if calibration method is used consistently;
Source of non-stationarity is important;
Conceptual models tested are adequate in describing the effect of changes in precipitation, not the effect of changes in vegetation;
Changes to streamflow are greater than those in precipitation for these watersheds.

22. Conclusions .. (what we have learnt from this exercise)
The effect of the fire in the Rimbaud watershed is secondary in the context of the overwhelming decreasing trend in precipitation and streamflow over the 39 years;
The 2 conceptual models tested did not perform well for the Fernow (W6) watershed because of the significant increase in streamflow early in the study period, and the subsequent significant decrease in streamflow from And it snows a lot over there.

23. Thank You!