A trend towards earlier sea-ice melt is detected in many ice-covered regions in the Arctic. The timing of the melt onset has a strong impact on the sea-ice energy budget. Melt onset changes the radiative properties of the ice due to increasing snow wetness and meltwater. So far, satellite passive microwave data are used to detect the melt onset. We analyzed transmitted radiation spectra as collected underneath drifting sea-ice using a remotely operated vehicle during the ARTofMELT expedition in the Fram Strait in spring 2023. We colocated those spectra with measurements of snow depth, sea ice and surface topography, chlorophyll-a concentration in the water column, and with aerial images. This combined dataset enables us to track down possible subsurface pathways and accumulation pools of meltwater. Areas of low snow load and depressed surface topography are characterized by higher transmitted radiation compared to areas with a thick snow cover. Those areas overlapped with areas that showed the first signs of surface melt. Chlorophyll-a concentrations varied only slightly in magnitude and did not match with the heterogeneous pattern of snow depth and ice topography. Here we discuss how to disentangle the influences of chlorophyll a and the subsurface meltwater on the spectral shape of transmitted radiation. We propose that upon successful disentanglement, the spectra can be used as an indicator for subsurface melting. Our study suggests that sea-ice melting starts subsurface and that measurements of transmitted solar radiation spectra could be used to identify the melt onset prior to surface melting. This can provide an interesting complementary information on melt occurrence and on the location of the water in the snowpack in addition to satellite passive microwave data.