Identification and Quantification of Synoptic Structure in Simulated Tiros N Radiance Soundings.
There has been continual improvement in the accuracy of retrieved atmospheric soundings from satellite radiance measurements; however, satellite data still provide an overly smooth product which is of limited use for synoptic scale analysis. Considerable smoothing occurs when information present in satellite radiance measurements is lost in the retrieval process. This research describes and develops techniques which (1) identify synoptic features and (2) quantify their fine structures directly from TIROS N satellite radiance data. A series of nine, overlapping, synthetic temperature and moisture soundings is constructed which contains synoptic features representative of the tropical eastern Pacific Ocean. Each of the soundings in the series is perturbed randomly, within limits, to create a statistically meaningful training data set. The training data set is processed through a radiative transfer algorithm to generate equivalent TIROS N satellite observations. The variance of the training set is partitioned into a portion accounting for synoptic class and a portion describing synoptic feature variation within a class. The variance structure of feature quantification differs strongly from that associated with classification. In a two-step process, the training data observations first are separated according to synoptic feature, and, second, the fine structure of the synoptic feature is quantified. Three classification techniques are developed to partition the training data observations into synoptic groups and to classify new soundings: a subjective, graphical, interpretative procedure; canonical discriminant analysis; and, discriminant analysis. Accurate classification was accomplished until atmospheric sounding perturbations exceeded 12% of the observed temperatures of the synoptic class means. Two experiments of feature qualification are described: a prediction of trade wind and frontal inversion height and strength using both synthetic and observed soundings; and, a prediction of frontal structure based on discriminant analysis of a training data set containing frontal inversion soundings. Positive skill, with some ambiguity, was demonstrated in both experiments.
- Pub Date:
- Physics: Atmospheric Science; Remote Sensing