Geoscience Reference
In-Depth Information
T4 and T5 are the brightness temperatures recorded in channels 4 and 5 of the
AVHRR, and Ts is the temperature of the surface. The NOAA-AVHRR infrared
data have to be considered as one of the variables that predicts air temperature. The
two temperatures (air and surface temperature) should not be confused in any way.
The other geographical variables used are latitude, longitude, distance from the
coast, and altitude.
The results in Table 3.3 show that latitude is the variable that has a significant
influence on daytime air temperatures. This is not particularly surprising as the
study was carried out on a sunny day and the main variable that increases air
temperature during the day is the amount of solar radiation emitted in relation to
latitude. The fifth column of Table 3.3 shows the correlation after carrying out a
multiple regression between the four above mentioned geographical variables as
well as air temperature (Ta): the coefficient is very important and confirms that the
temperature of the air does depend on latitude and distance from the coast. The
second to last column shows that the correlation between NOAA surface
temperatures and the Ta is identical to the correlation that exists for the four
geographical variables and the Ta. This means that surface temperature recorded by
satellites is a better indicator of air temperature than any of the four chosen
geographical variables.
Altitude
(alt)
Distance from
the coast (dist)
Longitude
(long)
Latitude
(lat)
Satellite
(Ts)
04-10-97
Four variables
Total
Correlation
0.26
0.58
0.36
-0.68
0.81
0.81
0.92
Table 3.3. Correlations between air temperature (Ta)
and geographical and satellite variables
With the aim of developing an interpolation formula for air temperature, a
multiple linear regression was carried out. It is therefore possible to express air
temperature as the result of a linear combination with several variables. The formula
which was developed for the 10 th April 1997 is as follows:
Ta = 80.75 + 0.27Ts + 0.0028alt + 0.0303dist - 0.028lat - 0.0024long
[3.2]
Once the difference between the estimated temperatures (calculated by the
model) and the actual temperatures that were measured on the 45 stations was
calculated, differences of more than 2° accounted for only one case in 10 (9.6%).
For two out of every three cases (between 62 and 71%) the estimation error that was
generated by the model was less than 1°. The study of residues in relation to this
model means that most of the 45 stations chosen as part of our research were located
along the coast: the model seems to be better at producing spatial variations in
temperature further inland, and this is due to the fact that there is a lot of mixed
pixels near the coast. It is also possible to create an air temperature map from a
Search WWH ::




Custom Search