International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
  
apud Lagouarde and Brunet, 1993). In contrast, the surface 
temperature can rise rapidly (Goetz, 1997). 
The daily amplitude of Tskin is strongly related to the intensity 
of soil cover and evapotranspiration. Under clear-sky conditions 
Tskin generally poses the same type of evolution (Lagouarde 
and Brunet, 1993). All over this behaviour the loss energy 
evolution of Tskin may be affected by fluctuations due to 
instantaneous variations of local weather conditons associated 
for instance with cloud passes or wind speed changes 
(Lagouarde and Brunet, 1993). 
2.2 Emissivity 
The potential of obtaining information about energy budget and 
water status of a surface through the relation between remotely 
sensed surface temperature and vegetation index (NDVI) has 
been investigated by several authors (Goetz, 1997; Sandholt er 
al. 2002: Ouaidrari er al. 2002). The LST and NDVI can 
provide informations of soil covering conditions because the 
amount of vegetation is important to LST estimations. 
In keeping with Kerr et al. (1992) for no corrections on the 
emissivity factor the split window technic will give good results 
over water, slightly less over fully vegetated areas, and poor 
results on dry bare soil. In very low fractional vegetation 
covering situation, the energy interactions with the surface is 
controlled by soil type. The differences between LST and Tair 
are also highly related to temperature magnitude and 
predominant soil type (Ouaidrari er al, 2002). Many 
researchers (Kerr et al. 1992; Van de Griend & Owe 1993; 
Andersen 1997 and Kerényi & Putsay 2000), have applied 
successfully the accurate calculation of NDVI for emissivity 
estimation associated to LST as a tracer of soil conditions. 
For the LST estimation we applied the Sobrino et al. (1993) 
method that use a combination of the channels 4 and 5 of the 
AVHRR sensor. Gusso (2003) evaluated the LST results the 
method proposed by Sobrino et al. (1993) over this same 
region. This algorithm, called Weak Split-window (WSW), they 
consider four atmospheric models: Tropical (T), Midlatitude 
summer (MLW), 1976 standard USA (US) e Midlatitude winter 
(MLW) as shown at equation 1: 
T, 2T, * [053 - 0.02(T, -T,)]J(T, -T,) *64(1-e) D 
Ts = land surface temperature; 
T, and Ts = brightness temperature for channels 4 and 
5 of AVHRR; 
£= mean emissivity for channels 4 and 5, (g4 + €5)/2. 
where: 
The Sobrino et al. (1993) method (eq. 1) use the retrieved 
emissivity computed by Valor & Caselles (1996) (eq. 2, 3 and 
4) algorithm by means of the NDVI as representative of the 
surface. The 
emissive characteristics of the irradiating 
emissivity factor was retrieved as follows: 
¢=0985P. +096(1-P, )+0,06P.(1-P,) (2) 
thus: 
(1- i ) 
7 ig 
p=s————— 
| (1- i )-kli- i) (3) 
/ig / iv 
and k is: 
y. 2n 
k = P» f (4) 
Pag = Pig 
158 
where: 
surface; 
i. ig. iv = NDVI, NDVI bare soil, NDVI vegetated 
pi and p, = channels 1 and 2 reflectances of AVHRR 
v and g — indexes for vegetation and bare soil. 
The monthly (July and September) mean values of reflectances 
of channels 1 and 2, NDVI for bare soil and vegetated areas of 
AVHRR are presented in the Table 3. The NDVI data computed 
(Rouse et al., 1974) were extracted from monthly composites 
produced using the maximum NDVI procedure (Holben 1986), 
The later LST is subtracted from the early LST so that (final 
LST — initial LST = LST var). In the same way to the Tair. 
Table 3. NDVI and channels 1 and 2 reflectances values applied 
corresponding for vegetation and bare soil. 
  
  
Sample £i £2 NDVI 
Bare soil 0.18 0.2 0.1 
Vegetation 0.12 0.48 0.6 
  
2.3 Analysis 
Figure 1 shows the scatterplot graphic of the variation in the 
LST and Tair (in situ) measurements between two overpasses 
during night-time. À slightly direct trend is observed. 
  
  
  
| July/15 y = 1.0956x + 0.0703 | 
| x 40 e | 
i S | 
| 3 3.0 ® . | 
| € 20 9 6. 
| S 140 $ 9.9.9 
e 
0.0 + . : | 
0.00 1.00 2.00 | 
LST var 
September/04 y = 1.5127x + 0.5188 
4.0 - 0 
® 
520 
© 
> 
3 0.0 
= -20 
-4.0 @ : 
-2.00 -1.00 0.00 1.00 2.00 
LST var 
  
Figure 1. Scatterplot of LST and Tair (in situ) variation 
between two overpasses during night-time. 
It is worthy to note the variation range of LST is closely to 2K 
and Tair variation is around 4K in the both dates. Small changes 
in the LST are followed by severely changes in the Tair. 
3. CONCLUSIONS 
The general patterns of decreasing night-time temperature 
conditions for which minimum air and surface temperatures 
appear correlated, can improve the understanding of the 
mechanisms between their day-time maximum values. 
The modest differenciating to the levels of the reflectances for 
different types of bare soil and vegetation, the performance of 
the radiometric resolution available in the AVHRR (1024 
digital counts), has the potential to improve data on the quality 
Internationd 
ur Cid 
of land surf 
before any 
layers dynat 
November ( 
into account 
the produce 
theoreticaly 
variable sta 
one day to à 
This type o 
but will be 
features. 
Preliminary 
there is no 
associated t 
3.1 Select: 
Andersen, 
NOAA-AV! 
Journal of 1 
Becker, F. : 
method ow 
Sensing, 1l 
Goetz, S. . 
temperature 
Internation: 
Gusso, A. ( 
superficie t 
do sensor o 
8lp. Disse 
Centro est: 
Meteorolog 
Porto Alegr 
Gusso, A. | 
métodos d 
terrestre no 
NOAA. Xl 
p.1185-119 
Holben, B 
composite 
Journal of 
Kerényi, J. 
temperature 
Advances ii 
Kem, Y. H 
land surfac 
of an im 
Environme 
Ll 7-5 
temperatur 
Remote Ser 
Lagouarde, 
estimating 
from NOA 
Sensing, 14