Abstract:
The New Zealand Meteorological Service has archived AVHRR (Advanced Very High Resolution Radiometer) data from the NOAA meteorological satellites over a period of seven years. This data covers a 128 x 192 km region of the southern part of the North Island, New Zealand. The size of the archive (over 10 000 satellite passes), makes it ideal for the study of longterm properties of the satellite and sensing equipment, as well as determining a climatology of the region. The orbital characteristics of the NOAA satellites and the calibration of the AVHRR instrument are examined in this thesis, and a new algorithm for climatological cloud analysis is developed.
The NOAA satellites are shown to drift from sun-synchronous orbits. Although this drift is caused by the interaction of many complex forces upon the satellite, it can be closely approximated with a simple quadratic curve. Once the nature of the satellite drift is known, its effect on climatological studies is easy to assess.
The calibration constants of the AVHRR change over the course of the archive. This is investigated to determine whether these changes are to compensate for varying instrument response, or due to problems with instrument operation. Although some of these changes are due to deteriorating satellite performance, the variation in calibration constants appears to adequately compensate for differences in instrument response.
Existing cloud detection algorithms, developed primarily for operational purposes, derive parameters from single images. In a climatological study, this method may be inappropriate for the analysis of the vast quantity of data. Once the implications of the above features of the archive were considered, a new algorithm was developed to analyse climatological cloud information. This method interprets characteristics of frequency distributions of sensed temperature for 4 x 4 groups of pixels collected during each month of the year, over the entire archive by the NOAA 7 and 9 satellites. These characteristics are described by the sum of four independent Normal Gaussian distributions, which are considered to represent the properties of the ground and three different levels within the atmosphere. The parameters of the gaussian fits immediately provide information on the frequency of different cloud types, and their mean sensed temperature and variance. The results of the algorithm were verified by comparing its output with surface-based data collected in the region.
