IX-B8, 2012
PAR vCf (eq. 2)
GI vCf, while the
h higher. This is
ce of the rainfall
losses are mostly
ons with high vCf
hest predicted soil
for the study area
flow and changed
(such as increased
0mparison to an
nts of cumulative
imflow at Myuna
and TSS measures
nset of events and
nconsistencies are
8 s 3$
8
8
FparErosion {t/10 000)
Streamflow (cumec/s)
* s: %
Fparkrosion {t/10 000)
Steeamflow (tumec/s)
8
e
Streamflow (cumoec/4)
FparBrusion (t/10 000)
POUSSE ES
e) and streamflow
nd predicted soil
) 000 for scaling
(a) 2003/2004, (b)
| soil loss was
| Figure lor eq. 2
GC. Data sources:
v from QDERM.
b, and c.
characteristically
. The wet season
between all three
Sons.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
Similarities between time series of TSS measures and soil loss
predictions for certain parts of events are strong. À linear
regression TSS (mg/l) and FparEros (t) suggests that 74% to
96% of the variability in TSS measures of certain sub-event
over the wet season 2004/2005 could be explained by the
MODIS FPAR based soil loss predictions, e.g. for the period
between 9.12.2004 and 01.02.2005 (results not shown).
Integrating high temporal frequency vCf predictions into USLE
is suggested to reduce the dominant effect the only other high-
temporal frequency factor (R-factor) had on the soil loss
predictions. A polynomial equation fitted to the relationship
between daily soil loss predictions made using the high
temporal frequency vCf estimates from eq. 2 using the MODIS
FPAR time series and average daily rainfall has an R? of 0.74
(Figure 5).
350000
- . y= 26.731x? + 1674.3x + 95.632
Es: 300000 R^- 0.9313
wn U
2 £ & 250000
= oo >
0 E =
^ 5 9 200000
vo 50
9$ ug
© 7 3 150000
"C ©
02€
RE © 100000
= 0
co 50000
©
0
0 20 40 60 80
average daily rainfall (mm)
Figure 4. The relationship between daily soil loss predictions
made using the classical, annual vCf estimates and average daily
rainfall.
2500000
eo y = 212.09x? + 7710.6x + 957.61
4 8 2000000 A R2= 0.7825 E
o
=z ;
5 "T e
a £ {1500000
95c + 7
5 2 & 1000000 7
Ron i fii
EZ e e tnt
> -
z E 500000 "
i
+ ide
0 4 ^ : :
0 20 40 60 80
average daily rainfall (mm)
Figure 5. The relationship between daily soil loss predictions
made using the high temporal frequency vCf estimates from eq.
2 using the MODIS FPAR time series and average daily
rainfall.
In comparison, the relationship between soil loss predictions
using the formerly used, annual vCf estimates and daily rainfall
lies at R° of 0.93 (Figure ) (p < 0.001) (Figure 4 and 5). This is
taken as an indication of the reduction of a strong temporal
dependency of the USLE-based soil loss predictions on the
daily rainfall/-erosivity factor when integrating high temporal
frequency vCf estimates.
Limitations of this feasibility study are predominantly related to
the use of the empirical USLE (e.g. development and validation
and thus validity of the (R)USLE, no account for sediment
transport or storage, sensitivity to variations and scale of input
factors, no account for streambank or gully erosion, tendency to
overestimate soil losses) (Kinnell 2005). It also has to be
acknowledged that the interpretation of TSS concentrations is
not the only factor to consider when interpreting the soil loss
predictions. Also, the vCf equations were not developed for
FPAR measures. Nevertheless, Searle and Ellis (2009) suggest
that their R/USLE variable cover model as applied in this study
made sensible erosion estimates in semi-arid savannas in
Australia and the MODIS FPAR has been shown to be sensitive
to relevant vegetation properties (Schoettker, Scarth et al.
2010).
Whether the observed relationship between remotely sensed vCf
based soil loss predictions and in stream TSS measures
represents event-typical behaviour, such as supply limited or
transport limited events, cannot be clearly identified at this
stage without further field based data.
4. CONCLUSION
This study has provided the first suitability study of MODIS
FPAR as an additional input parameter for estimating vCf in
combination with information from ICESat and Landsat based
VSC to improve existing erosion modelling studies and
applications in a tropical semi-arid savanna ecosystem.
Integrating those dynamic vCf into a modified version of the
USLE, we presented the first high temporal frequency time
series of soil loss predictions for the study area.
The high-temporal frequency vCf predictions of this thesis
might be regarded as a new and promising approximation of the
antecedent catchment conditions. We propose we have provided
valuable results to show steps towards required improvement of
existing erosion modelling approach in the study area, and
possibly elsewhere. Yet, the soil loss predictions of this study
have to be interpreted with care until a future study can validate
the predictions.
Future research aims to identify drivers of observed temporal
and spatial variations in soil loss predictions (e.g. by using
physical based erosion models, multivariate analysis, including
more recent discharge events and using the new collection of
MODIS FPAR data). Further research also intends to validate
the dynamic, remotely sensed vCf predictions with existing field
data of ground cover and foliage projective cover, compare to
vCf predictions based on a Landsat fractional ground cover
product, a predicted Landsat FPAR product (Schoettker, Scarth
et al. 2010). Developing FPAR based vCf equations and
improved CanCov calculations is suggested as a target for
future remote sensing studies that could combine optical, radar,
and laser remote sensing techniques (Lucas, Lee et al. 2010).
To finally quantify the linkages between the spatially and
temporally variable vegetative cover, rainfall and erosion
processes and their impact on the adjacent riverine and coastal
environments is a continuing task for inter-disciplinary
research.
5. ACKNOWLEDGEMENTS
This study was partly funded through a University of
Queensland Research Scholarship and was supported by CSIRO
Land and Water, Canberra (A/Prof Arnold Dekker). The authors
would like to thank Colin Rosewell for his advice, Peter Scarth,
and Robert Denham for their involvement in determining the
potential usability of the MODIS FPAR in the study area.
A similar Figure to Figures 1 has been published in Schoettker,
Scarth et al. (2010).
6. REFERENCES
Asner, G. P., 1998. Biophysical and biochemical sources of
variability in canopy reflectance. Remote Sensing of
Environment 64(3) pp. 234-253.