THE NETHERLANDS ROVE PROGRAM - RADAR REFLECTION DATA 10 GHZ 
1039 
L but also on 
Efects the radar 
for all inci— 
. This soil 
ated by the ex- 
is are caused by 
iscription of 
Eluence could be 
}] (7) 
this assumption, 
equation 
. The model des- 
/ (Attema e.a. 
i case of small 
scatterometers 
low the develop- 
the growing sea- 
3 we will see, 
sets alone also 
Eerence purposes, 
ites a measure- 
soil by vege- 
le underlying 
>eak on May 7 
! caused by a 
;urement. 
;e later in time 
in soil moisture 
are soil scat- 
lue to rainfall 
ises through 
icur due to the 
ion. In this 
'Othed by the 
een the total 
dd-season val- 
rimination for 
has to take 
on and stabil- 
as used to 
79). From this 
e differences 
cation of crop 
APA MAY JUN JUl RUG SEP 
1990 
LESENOt © SUGAA8EETS » POTATOES + S.VMEAT X OATS OB. SOIL 2 
Figure 13. Back scattering coefficient y versus time for various test 
plots. Incidence angle 60°, HH-polarization, season 1980. 
By taking into account that a water cloud is a 
volume scatterer we may write, after the introduction 
of the usual approximations, that 
Y =35- { 1-exp (- 
veg 2Q 
2NQh1 
COS0 
(8) 
where : 
0 is the radar cross section of one droplet 
Q is the so-called attenuation cross section of 
one droplet 
N is the number of the droplets per unit volume 
h, 0 are as indicated in figure 14 
Figure 14. Geometry of the cloud model. 
types should be possible. One of the more important 
results being that the success percentages could be 
greatly improved by the introduction of multitemporal 
analysis. Although basically each observation will 
add some information and therefore should improve 
the classification results a number of three obser 
vations can be considered as an optimum. During the 
growing season of 1980 an X-band SLR campaign was 
imitated to verify the conclusions of the simulation 
study (Hoogeboom 1983). 
Multitemporal applications can only be utilized to 
their full extent when radar systems are calibrated 
with sufficient accuracy. Although such calibrations 
ask for a raise of standards in radar technology they 
offer at the same time the possibility to add a 
memory function to the observation system. This mem 
ory function will probably turn out to be the most 
important element of operational radar-based remote 
sensing systems. It not only opens perspectives with 
respect to monitoring but it of fers also a possibility 
for meaningful comparisons between observations made 
in different years and over large arease.g. on a Euro 
pean or even on a worldwide scale. 
An important step towards the quantification of 
the relative importance of the soil and vegetation 
contributions in fig. 13 was made by the introduction 
of the, so-called, cloud model (Attema e.a. 1978). 
The underlying ideas are that the microwave dielec 
tric constant of dry vegetative matter is much small 
er than the dielectric constant of water; a vegetation 
canopy is usually composed of more than 99% air by 
volume. Therefore, the canopy can be modelled as a 
water cloud, the droplets of which are held in place 
by the vegetative matter. As a first step, it is 
assumed that this cloud consists of small spherical 
droplets with the same radius and with a uniform 
random distribution (fig. 14) . 
It is convenient to simplify this formulation a bit 
further. Since all water particles are assumed to be 
identical in shape and size we may replace the ratio 
0/2Q by a parameter C. If we define W as the water- 
content of the cloud per unit volume (kg/nP), N is 
proportional to W and therefore 2NQ can be replaced 
by DW, where D is the second modelparameter; eq.(8) 
becomes: 
Yveg = c[ 1 - exp(-DWh/cos0)] (9) 
In eq. (9) there is one single crop parameter Wh re 
presenting the amount of water per unit surface. This 
quantity Wh is equal to the biomass per unit area 
times the volumetric moisture content of the'plant. 
Since the equivalent dropsize is unknown the model- 
parameters C and D must be determined for each crop 
by non-linear regression analysis. 
Because the vegetation layer is partially trans 
parent for microwave radiation the return from the 
underlying soil must be accounted for. Assuming an 
incoherent addition of the soil and vegetation con 
tributions we may simply add Y so ii to eq. (9) taking 
into account the attenuation by the vegetation layer. 
In this way we arrive at the cloud model equation 
Y=c[ 1-exp(-DWh/cos0)]+Y so ^^exp(-DWh/cos0). (10) 
In the development of the model described by eq.(10) 
using radar backscattering measurements in X-band, 
throughout the growing season, of 8 different crops 
it turned out that the attenuation parameter D is 
rather insensitive to the incidence angle. For crops 
with relatively large leaves (sugarbeet, potatoes 
and peas) the scattering parameter C is dependent on 
the incidence angle. 
The analysis showed further that the appearance of 
so-called ears in the cereals has a dramatic effect 
on the geometry and consequently on y. After this