The determination of atmospheric aerosol layers from lidar returns is possible through automated algorithms. This product is useful, for example, in monitoring the Boundary Layer Height (BLH) as well as volcanic plumes. Aerosol layers are usually detected using the gradient method, i.e. by finding the inflection points of the range-corrected backscattered signal. These points can be either calculated as the minima of the numerical derivative of the signal, or by zero-order Digital Wavelet Transforms. Since for low signal-to-noise ratios the numerical derivative is very prone to noise-induced fluctuations, a moving average is often performed. We demonstrate why this procedure should be avoided in the gradient calculation. Finally, an alternative approach to the Digital Wavelet Transform is proposed, giving the same results but lowering the computational times by about one order of magnitude.