Noise power spectral density estimation is an important componentof speech enhancement systems due to its considerable effect onthe quality and the intelligibility of the enhanced speech. Recently,many new algorithms have been proposed and significant progressin noise tracking has been made.In this paper, we present an evaluation framework for measuringthe performance of some recently proposed and some well-knownnoise power spectral density estimators and compare their performancein adverse acoustic environments. In this investigation we donot only consider the performance in the mean of a spectral distancemeasure but also evaluate the variance of the estimators as the latteris related to undesirable fluctuations also known as musical noise.By providing a variety of different non-stationary noises, the robustnessof noise estimators in adverse environments is examined.