The fate of mercury (Hg) content of four poor quality coals was investigated under various pyrolysis conditions (temperature, time, pressure), using helium flow. The resulting chars were analyzed for Hg and compared with the parent coal's content. Hg removal efficiency was found to be affected by both pyrolysis temperature and residence time. Mercury removal continuously increased for pyrolysis temperatures up to 600 degrees C, and was sharply decelerated for higher ones, resulting in a rather sigmoid shape of the Hg removal curve. Pyrolysis time up to 20 min showed significant impact on Hg removal, negligibly affected Hg release. Thus, pyrolysis at temperatures between 500 and 600 degrees C for similar to 20 min could be considered sufficient enough for "Hg-free" coal production. The role of pyrolysis pressure on Hg removal was found to be limited, and its impact (if any) was observed at short pyrolysis times (<= 10 min), being diminished when the residence time is prolonged (up to 20 min). The results revealed three characteristic temperature regimes; a low temperature (200 degrees C-300 degrees C) physical desorption mechanism with low apparent activation energy, a mild temperature regime (300 degrees C-600 degrees C) with mixed physico-chemical desorption and sublimation mechanism with higher activation energy, and a high temperature regime (700 degrees C-900 degrees C) where Hg evolution dramatically decreased. Mercury coal removal during pyrolysis was simulated using a simple first order volume reaction model for the three temperature regimes identified. The apparent activation energy was initially found similar to 3.1 kJ/mol at the low temperature regime, increased to similar to 6.1 kcal/mol at the mild temperature range, and almost fourfold to similar to 22.4 kJ/mol at the high temperature regime. As the apparent activation energy increased, the pre-exponential factor (k(0)) also increased linearly, tending to compensate the increase of the activation energy. However, the conventional compensation theory does not offer a robust explanation of this behavior and further investigation is required.