On December 13th the core team from the Netherlands went to visit the partners in Italy for another support and advice meeting. Since the previous SAM meeting, Forlì has reached agreement on the final definition of the 8 impact criteria, as described in the document “Metodologia di calcolo della matrice multirischio”. The first 6 criteria are defined in accordance with the discussion in the previous meeting and the methodological advice that was provided by Ruud. The five impact levels of criteria 5.1 ((violation of public order and security) and 5.2 (sociological and psychological impact) have been defined by the local working group in accordance with the example methodology from Germany: a qualitative description of the levels, rather than the use of specific indicators.
Forlì has experimented with different “relative weights” of the impact criteria, as well as with the different functions for assigning a numerical value to the impact labels (linear, exponential). It seems that the relative weight makes little difference (most scenarios remain the same, only in some very limited instances the impact level can be one more or less), but the function makes more difference. Ruud explains it is a good thing that the relative weight makes little difference. This means that the impact scores are robust and not heavily depending on one or the other criterion. The experiment that has been done by Forlì can be referred to as a “sensitivity analysis”. The goal of this sensitivity analysis is to establish whether the impact scores are robust, to make sure that the whole risk assessment has a sound and stable basis. Now that the sensitivity analysis has been done, it is best to use the original weights of the criteria: all the criteria have the same weight. In this way there will be no useless discussion amongst stakeholders or political decision makers on the relative importance of different criteria or societal functions. At the same time, a footnote can be placed by those specific risks that have some (limited) sensitivity for the relative weights (just one impact level more or less), to explain this sensitivity and that the actual score for that scenario point has a “bandwidth”.
With regards to the used function, Ruud explains that the exponential function (base number 3 or 10) is mostly used to have more “contrast”. It favours the extreme “disaster scenarios”: the scenarios with one or several D or E impact scores are higher in the matrix, while the scenarios with no D or E remain low. In case of base number 10 this is the most extreme. An exponential function is mainly used at national level, because it fits with the purpose of a national risk assessment: identifying those risks that are important for national safety/security and filter out the less important risks, that are dealt with on regional or local level. Because this analysis is done for local level (municipality of Forlì), it makes more sense to use the linear function. In future it might be interesting to experiment using the 3 different functions for the national, regional and local levels in Italy, to distinguish risks according to their different responsibilities.
The working group presents the risk curves that they have prepared. The study of is Forlì focuses on floods, earthquake, landslides and Seveso. For the first 3 risks the curves have been prepared, using the sectorial prescribed “time of return” scenarios (3 for each). For Seveso the curve still has to be prepared. This subject meets with another kind of discussion. On the one hand the scenarios could be selected from any number of different toxic, flammable and/or explosive substances. On the other hand, Forlì only has one Seveso site, with only one main scenario: propane explosion. It seems impossible to define 3 different kinds of scenarios in this case. An additional problem is the estimation of probability: should this only be done on the basis of internal failure or also external scenarios, like extreme weather or flooding? Ruud suggests not to limit the view on hazardous substances to this one Seveso site, but also to include rail and road transport (many different substances, also toxic and flammable scenarios), as well as industries and storages that do not have to comply to Seveso, but have hazardous substances nonetheless. The best is to include at least explosions, pool fires, toxic liquids and toxic gasses. As the methodology and impact criteria now are set, it is easier to quickly experiment with different kinds of scenarios. These scenarios do not have to be that much in detail or elaborated: the main goal is to make a quick assessment of their potential impact on the 8 criteria. What would make the scenario score much higher on the different criteria? Ruud also suggests to add specific scenarios related to environmental pollution and to social unrest (no fatalities, but still social consequences). With regards to the question of internal versus external triggers. Industrial sites can always be a subject to a domino effect of other incidents, mainly extreme weather, earthquake and floods. This specific issue will be dealt with in Seveso IV, but can also be taken into account in the current analysis. First, it is important to consider for which external scenarios an industrial sub-scenario might be likely or credible. For those external scenarios an additional “worst case scenario” can be added to their respective curves. So for example: adding a worst case flood with additional environmental catastrophe due to chemical spills. Secondly, the industrial scenarios themselves have to include the external trigger in their probability analysis: what is the total likelihood of a specific scenario, including external triggers. As for example floods might have a much higher likelihood than the internal safety of chemical industry, those external triggers can be very relevant. However, it has to be considered that those external triggers do not necessarily result in the same (worst case) industrial scenario, so for example no BLEVE but just an environmental pollution. In any case, it is recommended to do some quick and dirty experiments on extra scenarios, rather than try to be too detailed and scientific. This is also the case for other subjects besides Seveso: Ruud suggest to try and experiment with adding other risks besides the 4 selected ones.
Once the risk curves for each specific risk have been prepared, the next step is to combine them in one all-hazard risk matrix. The risk analysis module provides the option to combine all scenarios into one matrix, but that might result in an incomprehensible overview: too many different scenarios and curve lines. The adding of curves instead of single scenario points is a very important improvement of the methodology, compared to existing examples from the Netherlands, Norway and all across the EU. But it requires one additional step: first interpret each of the single risk curves and secondly combining the “priority parts” of those curves into one all-hazard matrix. This kind of prioritization of curves has been discussed in SAM 3 for the example of earthquakes. The earthquake curve has two specific characteristics. Firstly, earthquakes cannot be prevented, so reduction of likelihood (i.e. focus on minimizing likelihood of the more frequent scenarios) is not a factor of consideration in the prioritization. Secondly, the curve is very steep: the more likely the scenario, the much lower the impact. Together this means it is most logical to prioritize the worst case scenario as a basis for building codes (vulnerability reduction) and preparedness. A similar interpretation of the other risk curves will have to be made. This is all about understanding the curve: if it is more or less horizontal or vertical, the scenario with respectively the highest probability and the highest impact should be prioritized. Vertical is more or less the case with the example of earthquakes, whilst horizontal or even from down left to top right could be argued for flu pandemics (the higher impact is also the more likely one, as it is more likely to spread). If it is bent upwards (concave), it could be argued that the top point (with highest combined impact and likelihood), should be prioritized. If it is bent downwards (convex), it could be argued that the both ends of the curve should be prioritized: the worst case (high impact, low probability), as well as the most frequent (low impact, high probability). It will require some practice and comparison to really understand the curves. As said before: only the priority points/scenarios of the curves will be presented together in the all-hazard matrix, for the overall prioritization.
Preview of the process of making the risk management strategy
Ruud presents the MiSRaR thoughts on risk management. After the risk assessment, the next step is a so-called capability assessment. Like the risk assessment, this step includes identification, analysis and evaluation. The nature of this process is brainstorming and networking: trying to find smart solutions to improve safety. The key is trying to find win-win situations. In the MiSRaR project Forlì itself has presented many good examples, like using abandoned carries as water retention basins. The working group discusses a very important window of opportunity to find win-win: the municipality of Forlì is at the very beginning of the development of a new spatial master plan. The political assignment for this process still has to be set. This is a great opportunity to make safety a part of this assignment. CRISMAS could offer to deliver all-hazard spatial principles for increasing safety. This fits perfectly within the main goal of the whole CRISMAS project. It could also help to clarify the mission and advice role of the new regional Civil Protection organization towards the municipalities: besides advice on preparedness this could include advice on spatial planning.