Spatio-temporal variability of the winter surface mass balance is a major uncertainty in the modelling of annual surface mass balance. Moreover, its measurement at high spatio-temporal resolution (sub-200 m) is very useful to force, calibrate or validate models. This study presents the results of year-round field campaigns to study the evolution of the surface mass balance in a ~2 km2 portion of the accumulation zone of the Mer de Glace (France). It is based on repeated LiDAR acquisitions, submergence-velocity measurements and meteorological records. The two methods used to quantify submergence velocities show good agreement. They present a linear temporal evolution without significant seasonal changes but display significant spatial variability. We conclude that a dense network of submergence velocity measurements is required to reduce the uncertainties when computing winter and annual surface mass balance from digital elevation model differencing. Finally, a hight spatio-temporal variability of the winter surface mass balance is highlighted (e.g., a std dev. of 0.92 m in April) even though the topography is homogeneous (std dev. of 25 m). Attempts to relate this variability to different morpho-topographic variables and wind-related indexes show the need for studies conducted at the snowfall event scale to obtain a better understanding of the variability in mass balance at the glacier scale.