Retrieving low LWP values using the microwave spectral regime reveals large relative errors, whereas the potential for infrared methods is high. Therefore robust and computationally low demanding synergistic retrievals based on a multivariate regression and a neural network are derived to estimate LWP and cloud effective radius. While the regression-type synergy retrievals are strongly influenced by the nonlinearities of saturating signals in the infrared regime for higher LWP, the neural network retrieval is able to retrieve LWP and cloud effective radius with a higher accuracy than the single instrument retrievals. This is achieved by examining synthetic observations in the low LWP range. Furthermore, the performance of the retrievals is assessed in a radiative closure study for the downwelling shortwave flux, using measurements of a microwave radiometer and a spectrally highly resolved Atmospheric Emitted Radiance Interferometer (AERI).