Satellite-based estimates of land surface temperature (LST) are widely applied as an input to models. A model output is often very sensitive to error in the input data, and high-quality inputs are therefore essential. One of the main sources of errors in LST estimates is the dependence on vegetation structure and viewing and illumination geometry. Despite this, these effects are not considered in current operational LST products from neither polar-orbiting nor geostationary satellites. In this paper, we simulate the angular dependence that can be expected when estimating LST with the viewing geometry of the geostationary Meteosat Second Generation Spinning Enhanced Visible and Infrared Imager sensor across the African continent and compare it to a normalized view geometry. We use the modified geometric projection model that estimates the scene thermal infrared radiance from a surface covered by different land covers. The results show that the sun-target-sensor geometry plays a significant role in the estimated temperature, with variations strictly due to the angular configuration of more than ±3°C in some cases. On the continental scale, the average error is small except in hot-spot conditions, but large variations occur both geographically and temporally. The sun zenith angle, the amount of vegetation, and the vegetation structure are all shown to affect the magnitude of the errors. The findings highlight the need for taking the angular effects into account when applying LST estimates in models and when comparing LST estimates from different sensors or from different times, both on the daily and seasonal scale.