Turbulent sensible heat flux is a measure of heat transfer between the surface and the atmosphere. In stably stratified conditions, such as found during night-time and over the glaciers, the sensible heat flux is warming the surface since the heat is transferred from the atmosphere towards the ground. On a sloping surface at night, the surface tends to cool down by radiative cooling so the buoyancy drives the air downslope. This downslope flow is called katabatic flow and it affects the energy flux exchanges. The influence of katabatic flows on turbulent fluxes is still not perfectly understood since the interaction between boundary layer, mesoscale flow and the terrain give rise to complex flows. During HEFEX campaign, five meteorological stations and eight turbulent sensors were set up on Hintereisferner for 3 weeks in August 2018. The aim of this campaign was to study the interaction between the surface, the atmosphere and the topography along and across the glacier. The main objective of this thesis is to understand the main drivers of the sensible heat flux over a sloped glacier surface, and the variability of the flux in the horizontal and vertical direction. Furthermore, a set of bulk methods commonly adopted to simulate turbulent sensible heat flux over flat, sloping and glacier surfaces is evaluated. During the three weeks of measurements, a stronger and more turbulent flow is found in the center of the glacier. At higher altitudes and close to the margins the flow is weaker. The increase of the sensible heat flux is associated with an increase of temperature and wind speed. For a weak wind the dependence of the kinematic heat flux on temperature is relatively small but when the wind strengthens, the temperature highly influences the flux. Close to the katabatic jet, the kinematic heat flux has a low absolute value. However, the lowest absolute value of this flux is found above the jet height. In this thesis the katabatic jet was classified into three regimes. The first is dominated by weak winds (up to 3.2 m/s), in this regime the jet is shallow. The second regime occurs when the jet deepens and has a wind speed between 3.2 m/s and 4.5 m/s. Finally, in the third regime the jet is deep and has a wind speed higher than 4.5 m/s. The divergence of the kinematic heat flux in the vertical direction becomes significant only in the second and third regime. When the divergence becomes significant, the two levels of measurements differ on the dissipative timescale. The dependence of the studied heat flux on the stability of the atmosphere, was also diagnosed. Stable conditions did not lead to a collapse of turbulence. The bulk methods tested in this thesis are taken from Mahrt (2017), Grisogono et al. (2001), Schaefer et al. (2020), Radić et al. (2017). The first model was developed for flat terrains, the second model was developed for sloping terrains, and the other models were developed for glaciers. The comparison between the models was done by studying the Pearson correlation coefficient, the slope, the intercept and the bias of the linear fit. Each parametrizations was tested considering every day of the campaign, and considering only katabatic periods. Every model and every tower (except CSAT02 located at the margin of the glacier), showed higher correlation coefficients in katabatic conditions. The highest correlation coefficients were given by C-SR, C-log methods (by Radić et al. 2017), RF method (by Schaefer et al. 2020) without the stability correction, and Mahrt's method. The biases of C-SR, RF (without the stability correction) and C-log method were the highest. Mahrt's model had a low bias but slope values far from 1. The tuning of the C-log model was performed, and it led to a decrease of the bias. The correlation coefficients remained the same. Therefore, the best performance was given by the models with the highest correlation coefficients that were tuned. The choice of the roughness length for momentum highly affected the performance of the fits. Moreover, the approximation of a melting surface was effective. Higher correlation coefficients and lower biases were found for a longer time average interval up to 120/240 minute (depending on the model). Further investigations are required from different glaciers in order to prove the universality of these findings.