the products by the Moderate Resolution Imaging 
Spectroradiometer (MODIS) and the Geoscience Laser 
Altimeter System on the current NASA Ice, Cloud, and land 
Elevation Satellite (ICESat) (Scarth, Armston et al. 2010). 
Regionally developed Landsat Thematic Mapper (TM) and 
Enhanced Thematic Mapper (ETM+) products in Queensland 
also provide estimates of properties of the TVC that are highly 
relevant for supporting national and state NRM - specifically as 
these products are validated for the unique conditions in 
Australian savanna ecosystems (Danaher, Scarth et al. 2010; 
Scarth, Röder et al. 2010). The collective, remotely sensed 
information on certain vegetation properties could be valuable 
for erosion modelling studies in the tropical semi-arid savannas 
in Queensland. 
The sensitivity of a time series of the global, biophysical 
Moderate Resolution Imaging Spectroradiometer (MODIS) 
Fraction of Photosynthetically Active Radiation absorbed by a 
canopy (FPAR) (Knyazikhin, Glassy et al. 1999) to a time series 
of regionally developed Landsat TM and ETM- based green 
and non-green fractions of ground cover and vegetation 
structural categories (VSC) has been shown for a tropical semi- 
arid catchment in Queensland, Australia, in an earlier study 
(Schoettker, Phinn et al. 2010; Schoettker, Scarth et al. 2010). 
In a multiple regression analysis (including interaction terms) 
75% of the variability in dry season MODIS FPAR was 
explained by the Landsat datasets in a catchment of 9500 kn? 
that lies adjacent to the GBR (the Bowen/Broken subcatchment) 
(Schoettker, Scarth et al. 2010). The catchment has been 
considered an important contributor to terrestrial discharges 
into the GBR lagoon (Lewis, Sherman et al. 2009). 
Which potential global and high temporal frequency 
biophysical products, such as the MODIS FPAR and the ICESat 
might have to complement regional remote sensing products for 
the mapping and monitoring of TVC properties relevant to 
erosion modelling has not been identified to date — specifically 
not in Australian savannas. The main aim of this research was 
thus to determine the global MODIS FPAR's potential 
suitability to improve erosion modelling via an integrated 
approach combining global and regional vegetation remote 
sensing products in the same study area as Schoettker, Phinn et 
al. (2010) and Schoettker, Scarth et al. (2010), a tropical semi- 
arid catchment in Queensland. 
1.2 Overview and references 
In erosion modelling a so called C-factor measures the 
combined effect of all the interrelated vegetative cover and crop 
management variables (Rosewell 1997). This definition has 
been used in empirical soil erosion models such as the 
Universal Soil Loss Equation (USLE) (Wischmeier and Smith 
1978) and its subsequent Revised Universal Soil Loss Equation 
(RUSLE) (Renard, Smith et al. 1997). The C-factor is also 
commonly applied in varying forms in most other erosion 
models worldwide. Above ground vegetative C-factor (vCf) 
estimates for non-cropping areas under Australian conditions 
were published by Rosewell (1997). 
Many water driven erosion models worldwide have been 
applied in recent decades and some have integrated remote 
sensing information (Vrieling 2006; USDA 2008; Searle and 
Ellis 2009). In Australia, however, to date most of the water 
driven erosion models applied still use the basic concept of the 
empirical USLE model, e.g. the SedNet whole-of catchment 
modelling (Lewis, Sherman et al. 2009). Aside from a number 
of major considerations which limit the utility of models, such 
as the USLE for, recent model applications by Searle and Ellis 
(2009) have improved the USLE utility in the semi-arid tropics 
of Queensland. 
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 
Despite these substantial research efforts, it is surprising that no 
study - to the knowledge of the author until this date - has 
integrated high temporal resolution remote sensing imagery to 
derive vCf estimates for use in erosion modelling other than by 
using classical vegetation indices (de Jong 1994; Lu, Prosser et 
al. 2003; Symeonakis and Drake 2004). However, in open plant 
communities, such the tropical semi-arid savannas of the study 
area, classical vegetation indices have been shown to perform 
less reliably for quantifying temporally variable TVC and its 
components and hence erosion or biomass modelling (van 
Leeuwen and Huete 1996). The derivation for or inclusion of 
remotely sensed structural characteristics of the TVC in erosion 
modelling is to date also very limited (Lu, Prosser et al. 2003; 
de Jong and Jetten 2007). 
2. METHODS 
2.1 Deriving high temporal frequency vegetative cover 
factor estimates 
To account for the role vegetation plays in impeding soil loss, 
erosion models classically include so called cover subfactors 
that separate the total vegetation cover into two major vertical 
components: Canopy cover and surface cover (as for example 
described in USDA (2008) and Rosewell (1997). As most 
Australian plant communities feature a distinctive upper and a 
ground or lower stratum (Specht 1981), the separation of the 
TVC into a canopy and a surface cover is considered to 
adequately represent open plant communities that cover most of 
the Australian continent and the study area. 
The equations typically used to derive vCf and subfactor 
estimates based on the earlier work by Wischmeier and Smith 
(1978) take the following form as published in Rosewell 
(1997): 
C=CanCov*SurfCov (1) 
, Where CanCov is the canopy cover subfactor and SurfCov 
is the surface cover subfactor. The concept of vegetative cover 
subfactors applied here was taken from the Revised Universal 
Soil Loss Equation USDA (2008) and Rosewell (1997) for 
Australian conditions. The relevant equations to determine the 
vegetative cover's subfactors are commonly given as follows: 
SurfCov= AHH GCHGC +d GC) o 
CanCov 21- (CC/100)* g COS = 
with, h, =h, +a,*a,(h, —h,) (3a) 
1 
h, =—*CH 
or 3 es 
, Where a, b, c, d are coefficients given in Rosewell (1997), CC 
is Canopy Cover (?6), h, is effective drop height, h, is height to 
the bottom of the canopy, 4, is height to the top of the canopy, 
a, is a coefficient for canopy shape, a, for concentration of 
surface area within canopy given in (USDA 2008), and is CH is 
canopy height. The cover factor has commonly been determined 
simply as in eq. (2) for low wFPC areas (Searle and Ellis 2009) 
or as the product of eq. (2) and (3a or b) (USDA (2008) and 
Rosewell (1997), respectively). 
The calculation of the dynamic vCf in this study were designed 
to advance and yet replicate useful aspects of more recent 
applications by USDA (2008) (RUSLE2 model) or 
conventional approaches by Rosewell (1997), to date used in 
  
  
     
    
  
   
     
   
   
   
   
   
   
   
   
    
    
   
   
   
   
    
   
    
    
    
   
  
     
  
   
    
  
   
    
    
   
    
   
  
   
   
    
    
    
    
   
    
   
   
   
  
   
    
   
    
    
   
Int 
Australi 
method 
calculate 
based oi 
develope 
product 
Queens: 
Manage: 
into VS( 
The rele 
follows: 
variable 
modellir 
more gre 
a, b, Ce 
CC per‘ 
by Scart 
was deri 
been pro 
and A, w 
the uppe 
and a, i 
CH was 
canopy | 
al. (201 
each of 
eq. 1 w 
represen 
the time 
distribut 
and 10 k 
2.2 Mo 
The (R) 
soil loss 
Smith et 
, where 
(MJ mm 
S 1s slo 
factor (h 
of data ı 
A temp 
modifiec 
written | 
The vari 
the spe 
represen 
(Renard 
made fo 
had bee 
techniqu 
Due to t 
soil loss 
the stud 
of the s 
measure 
streamfl 
To eval 
frequenc 
compari 
The rel: 
date cor 
imagery 
tempora