Wavelet analyses and comparative denoised signals of meteorological factors of the namibian atmosphere

Abstract With population growth and climate change increasing the exposure of communities and assets to extreme hydrological events, such as floods and droughts, it is crucial to make accurate and timely early warning, climate adaptation and water management information available that can help minimize the loss of the limited quantities of water in arid regions. Trend estimation using wavelets has continued to attract ubiquitous applications in various fields including atmospheric and water studies. While Fourier analysis employs big waves, wavelet analysis uses small waves. Since wavelets localise features in a signal data to different scales, we can therefore preserve vital signal features while removing noise in the signal. Hence the basic notion of wavelet denoising or wavelet thresholding is that the wavelet transform results in a sparse representation for many real-world signals, meaning that the wavelet transform concentrates signal features in a few large-magnitude wavelet coefficients. Since small wavelet coefficients are typically noise, we can “shrink” or remove them without affecting the signal quality. This paper employs wavelet analyses of some selected meteorological factors of the Namibian atmosphere to investigate the noise in the data collected from a number of Southern African Science Service Centre for Climate Change and Adaptive Land Management (SASSCAL) stations, for the years 2012 to 2015. The results show some significant levels of noise around the thresholds of various optimal time series peaks of the data. Hence the denoised data serve to represent fairly refined representations of data for the given period. Moreover, the results of comparing the optimal game model solutions for the original and denoised data specifically identified some weather stations that are “humidity noise stable” and “temperature noise stable”. Thus, these suggest that despite the noise in the respective meteorological data for these stations, the effects of the noise in the data appear to be manageable or optimally controllable, within the given period.