Key Assumptions and Application to SASS Data

In order to apply our resolution enhancement method to SASS data, conditions must be imposed on the data and various assumptions made. As an example of the former, in order to estimate both and for the resolution elements, each element must be observed with a diversity of incidence angles. Our assumptions in applying this method to SASS data for this region are listed below.

1. The instrument calibration remains stable over the data acquisition interval. Limited tests of this assumption over the Amazon rainforest have indicated the calibration stability to be adequate [4].
2. Over the study regions, there is no dependence of or on the azimuth angle of the measurement, i.e., there is no azimuthal modulation of . The validity of this assumption for Amazon data is borne out by previous investigators [6][1]. We plan further study of this issue for this and other regions in later papers.
3. The effects of surface topography are small and can be neglected. Surface slope affects the local incidence angle which, in turn, may affect the estimates of and . However, , which describes the incidence angle dependence of , is relatively small (globally, dB/deg). Given the small value of , the relatively flat topography of the Amazon basin, and the relatively large size of the resolution elements, this assumption is reasonable.
4. The dependence of the time of day on is small or known and correctable. Previous researchers [4][1] have noted some time-of-day variation in over the Amazon region. In the early morning hours dropped approximately 0.5 dB, possibly due to the accumulation of dew [4]. However, only a limited number of scatterometer measurements were made during this time with most occurring later in the day. Uncorrected diurnal variations is treated as ``noise'' in the imaging algorithm. This variation appears to have very little impact on our results and has, therefore, been ignored (see the discussion below). Further studies of the diurnal variability are being conducted.
5. The and parameters over the target region remain constant over the imaging time interval. For a sufficiently short imaging time interval this assumption is justified. However, in applying the method to SASS data, the imaging time interval must be very long (days to weeks) to obtain sufficient measurement overlap in order to achieve the desired resolution.

   Kennett and Li [5] specifically examined the time dependency of and at resolution in their search for extended scatterometer calibration regions. During the 3 month SASS mission in the late summer of 1978 they observed time-dependent changes in the parameter in some areas. However, over the extended Amazon basin, the only variation occurred during the first 2-3 weeks of the mission and was limited to less than 1 dB (compare Fig. 3). We have considered imaging time intervals which include and exclude these first weeks. In this paper, the full data set was used. In a later paper we will report on seasonal variations.