Determining dust temperatures and masses in the Herschel era: The importance of observations longward of 200 micron

Context. The properties of the dust grains (e.g., temperature and mass) can be derived from fitting far-IR SEDs (>= 100 mu m). Only with SPIRE on Herschel has it been possible to get high spatial resolution at 200 to 500 mu m that is beyond the peak (similar to 160 mu m) of dust emission in most galaxies. Aims. We investigate the differences in the fitted dust temperatures and masses determined using only <200 mu m data and then also including >200 mu m data (new SPIRE observations) to determine how important having >200 mu m data is for deriving these dust properties. Methods. We fit the 100 to 350 mu m observations of the Large Magellanic Cloud (LMC) point-by-point with a model that consists of a single temperature and fixed emissivity law. The data used are existing observations at 100 and 160 mu m (from IRAS and Spitzer) and new SPIRE observations of 1/4 of the LMC observed for the HERITAGE key project as part of the Herschel science demonstration phase. Results. The dust temperatures and masses computed using only 100 and 160 mu m data can differ by up to 10% and 36%, respectively, from those that also include the SPIRE 250 & 350 mu m data. We find that an emissivity law proportional to lambda(-1.5) minimizes the 100-350 mu m fractional residuals. We find that the emission at 500 mu m is similar to 10% higher than expected from extrapolating the fits made at shorter wavelengths. We find the fractional 500 mu m excess is weakly anti-correlated with MIPS 24 mu m flux and the total gas surface density. This argues against a flux calibration error as the origin of the 500 mu m excess. Our results do not allow us to distinguish between a systematic variation in the wavelength dependent emissivity law or a population of very cold dust only detectable at lambda >= 500 mu m for the origin of the 500 mu m excess.