During the last decade, serious accidents have continued to occur in the process industry. Apparently the scenarios of various undesired events leading to those accidents are still not sufficiently controlled. The key question is how potentially hazardous situations develop, what processes form the basis for this development, and how to control them? Safety level is not static but depends on many risk factors that change in presence and intensity over location and time. Safety level is dependent not only on technical process parameters that have immediate effects on the 'frequency' or probability of catastrophic consequences, but also depends on equipment integrity degradation, operational and management quality, attitudes, and cultural processes which may change over a prolonged time. The time and human interaction aspects make dynamic risk assessment complex. This paper will outline a conceptual approach using in addition to the regular process parameter signals received, also weak and slowly changing signals from various safety indicators, enabling to keep track of the risk factors. In theory this could lead to obtaining an instantaneous safety level 'measure' making possible forecast alarming for an imminent event to occur. Such concept could be regarded as a 'writing' safety barometer, or barograph. However, there are quite a number of problems to be solved which in the paper will be discussed. (C) 2013 Elsevier Ltd. All rights reserved.