One of the most important performance indicators of a fiber optic distributed temperature sensing (DTS) system is its operational stability, which depends on the performance of each device making up the system. Such individual device performance itself is affected by temperature and other environmental conditions. As a result, ambient fluctuations invariably lead to measurement uncertainties and drifts, seriously degrading the system performance. To ameliorate such adverse effects of the environment, we propose a dynamic sampling-correction scheme, which we have implemented on our fiber optic Raman DTS system, resulting in an average temperature error within 1°C when the main chassis undergoes a temperature variation over the range -25°C-45°C. The drastically improved system stability foreshadows better performance and reliability in engineering applications.