Part of the work that I do involves observations of solar low wavelength UV observations, specifically UV-B and UV-A II (up to 340nm). I have noticed that when I observe responses on a CCD or CMOS based sensor, that the lower the wavelength observed the greater the random fluctuations in the pixel responses. This is also noted in the paper Evaluating UVA Aerosol Optical Depth using a Smartphone Camera (abstract only), that suggests reason as being
The greater relative error occurs at lower wavelengths (340 nm) due to increased atmospheric scattering effects, particularly at higher air masses and due to lower signal to noise ratio in the image sensor.
So, the question is how is it possible to experimentally diminish the random noise from low wavelength UV observations, while preserving as much of the signal as possible?