Acoustic travel-time tomography of the atmosphere

Knowledge of temperature and wind velocity fields in the near-ground atmospheric layer is important for many areas, e.g. boundary layer meteorology, theories of turbulence, and study of sound and electromagnetic wave propagation. The conventional way to measure temperature and wind velocity uses in-situ sensors (e.g., meteorological thermometer-anemometers). However, such measurements provide the values of temperature and wind velocity only at widely separated points. More sensors are required to measure the fields at other points. However, a large number of sensors will significantly distort the original fields. In contrast, acoustic travel-tune tomography allows remote sensing of temperature and wind velocity fields, and requires far fewer transducers for the same amount of data than the conventional in-situ measurements. This dissertation studies several different tomographic approaches, which can be divided into three classes: (i) algebraic techniques that divide the fields into constant-valued cells; (ii) a basis function approach that represents the fields as the sum of known basis functions with unknown coefficients; and (iii) stochastic approaches that take into account spatial or spatial-temporal correlations of the fields. First, it is shown that the conventional stochastic inversion approach, which utilizes spatial correlations, yields more detailed and accurate reconstruction of temperature and wind velocity fields in the atmosphere than traditional algebraic techniques and a basis function approach. Second, it is shown that a more general time-dependent stochastic approach, which utilizes both spatial and temporal correlations, allows more accurate reconstruction than any of the other techniques mentioned above.