The temperature of the sea surface (SST) is important for studies of the oceans to be able to
indicate the status of the meteorological weather and climate global. Knowing this, the SST
has been included in numerical models of prevision weather, climate and ocean circulation as
input to obtain a result closer to the real, however, that the predictions are successful, these
models require that data have high accuracy. Thus, the study aimed to validate the SST
extracted by the algorithms multichannel MCSST satellite NOAA-14, obtained by the sensor
AVHRR/2, through measurements of in situ SST collected by ship in May 1999 the Program
REVIZEE in the Continental Shelf Amazon. Was used 37 points and compared with 17
satellite images obtained with time interval maximum time range of ± 12 hours from the
purchase of those, 4km resolution were processed with extraction of SST, georeferencing and
interpolation getting 19 SST maps. Statistical analysis was calculated values of MBE, MAE
and RMSE, anomaly standard percentage and rate of agreement. The differences maximum
and minimum between values of SST measured in situ and satellite estimates were 1.4 and
0.01 °C, respectively. The SST obtained by ship, in general, presented more value than the
SST estimated by satellite. However, despite this, the data showed a good significance
because the magnitude of error was less than 6% (in module), and the values of MBE, MAE
and RMSE were small (respectively, -0.47, 0.60 and 0.72 °C). The concordance index was
0.51 showing that the data compared were not even close nor distant. Thus, statistical analysis
showed divergent, the correspondence between the data may be due mainly to clouds that lie
in the area and period of data collection in situ, which reduced the number of observed data
with the images of satellite and influenced the calculation of obtaining the SSTs and other
errors.
|
