Remote sensing techniques are applied to sodar data for providing information regarding the planetary boundary layer. The temperature inhomogeneity that results from convective mixing in lower atmosphere is responsible for the backscattering of sound wave transmitted by the sodar. The intensity of echo returns fluctuates with time in the mixing height zone due to convection. In this work, fractal dimension (FD) of the data is considered as a measure for classifying the mixing height zone. A method has been proposed to tackle the noisy sodar data without losing relevant meteorological information. A comparison of the computed FD for noise-prone sodar data sets of tropical urban settings with those of relatively noise-free Antarctic settings is reported in this context.