Title: SEDeblend: A new method for deblending spectral energy distributions in confused imaging

Abstract: For high-redshift submillimetre or millimetre sources detected with single dish telescopes, interferometric follow-up has shown that many are multiple submm galaxies blended together. Confusion-limited Herschel observations of such targets are also available, and these sample the peak of their spectral energy distribution in the far-infrared. Many methods for analysing these data have been adopted, but most follow the traditional approach of extracting fluxes before model spectral energy distributions are fit, which has the potential to erase important information on degeneracies among fitting parameters and glosses over the intricacies of confusion noise. Here, we adapt the forward-modelling method that we originally developed to disentangle a high-redshift strongly-lensed galaxy group, in order to tackle this problem in a more statistically rigorous way, by combining source deblending and SED fitting into the same procedure. We call this method "SEDeblend." As an application, we derive constraints on far-infrared luminosities and dust temperatures for sources within the ALMA follow-up of the LABOCA Extended Chandra Deep Field South Submillimetre Survey. We find an average dust temperature for an 870 micron-selected sample of (33.9+-2.4) K for the full survey. When selection effects of the sample are considered, we find no evidence that the average dust temperature evolves with redshift.