As outlined in the section of Methodology, climate analysis and projections as well as hazard modeling and mapping are conducted for the HCR pilot site to analyze the potential hazards that have been already identified.
Climate Analysis and Projections
The climate change signal was determined by comparing future climatic conditions (based on climate projections for the 30-year period 2036–2065) to recent climatic conditions (based on the 30-year period 1971–2000). As climate model forcing scenario the RCP8.5 (Representative Concentration Pathway) was selected, which describes a future in which greenhouse gas emissions continue to increase (Riahi et al. 2011). This scenario only describes one of many possible “climate futures,” and the actual change in climate will depend on how greenhouse gas emissions, among other things, will develop. RCP8.5 is the highest emission scenario available, and was chosen to provide an upper bound for the risk assessment.
As knowledge of the local climate conditions for the pilot site region is needed, the resolution of available global climate model results is too coarse for the analysis. Therefore, two different types of downscaling were considered to obtain results specifically for the immediate study area. A statistical downscaling was performed (Benestad et al. 2008). Global model results provided by the CMIP5 initiative (Coupled Model Intercomparison Project Phase 5) were downscaled using the local monitoring station data for Heraklion obtained through ECAD to optimize the coarse global model results for the pilot site region. In addition, regional climate model results provided by the EURO-CORDEX (Coordinated Downscaling Experiment—European Domain) project (Jacob et al. 2013) were used.
The selected climate indices were then calculated based on the local observations, statistical downscaling results, and regional climate model data. Average (and where appropriate extreme) values were determined for the historical and future reference periods. Prior to further analysis, multi-model ensembles for the statistically downscaled results and regional climate models were created. An ensemble refers to the averaging of the results of several models, and provides an improved “best estimate” projection, as the mean of the ensemble can be expected to outperform individual ensemble members under the assumption that simulation errors in different models are independent (IPCC 2007).
In the next step, the analysis results were represented graphically and the different indices were checked for consistency. After this quality check, the results were summarized in tabular form for all relevant STORM indices. The concept of these tables is illustrated in Table 4, for the heat wave and intense rainfall hazards. In the left column, the indices assigned to the hazards are listed. In the second column, the 1971–2000 baseline based on observations as well as the historical climate model runs is shown. Note that for the climate model results, the ensemble average and spread (standard deviation) are determined after calculating the 30-year baseline for each model individually. In the next two columns, the change for the time period 2036–2065 with respect to the 1971–2000 baseline (derived from the historical run) is shown for the statistical downscaling and regional climate model results. Here, the change projected by the individual models was derived first, and based on this the ensemble mean and spread (standard deviation) were determined as reported here. In the “comments” column, remarkable features are listed and an estimate of the projected climate change signal is given. Finally, these projected changes are summarized in a qualitative fashion to aid the integration in the consecutive risk assessment steps. This qualitative classification consists of five levels, and ranges from very low (indicated by a green color coding), low, medium, high, to very high (red). The assessment is performed for all indices separately, and an example of the assignment of the projected signals to the classification of change is shown for the heat wave indices in Table 5. It should be noted that this classification is solely based on climatological considerations. The level of risk will further be influenced by the exposure and vulnerability of the individual heritage cases to the hazard in the next steps of the risk assessment to determine the final risk level.
Hazard Modeling and Mapping
The hazard (GIS) modeling and mapping process of a series of natural hazards that were identified as the potential threats to the HCR pilot site area is provided in this section.
Earthquakes constitute one of the natural hazards worldwide that cause extremely severe damages to cultural heritage sites, especially in seismically active regions, such as the southeast Mediterranean basin (Papazachos and Comninakis 1978; Shaw et al. 2008). In order to assess earthquake hazards, several factors associated with earthquake events need to be acknowledged such as earthquake epicenters, proximity to active faults, and the type of geological formations (Sarris et al. 2010). This study considers events for shallow depth (< 70 km) earthquakes around Rethymno City (within a buffer zone of 100 km) for the period 1900–2006. There were 1167 earthquakes considered in total and the calculation of their density map (Mw magnitude) was generated. Moreover, the fact that the seismic intensity decreases (attenuates) with distance to faults, a cost distance parameter from the pilot site to the recorded faults was acknowledged (Cooke 1997). The geological formations were also considered as another parameter in this hazard map development process, since deep, weak soils tend to amplify and prolong the seismic waves shaking more than the stronger rock bed (Argyriou et al. 2016). All the datasets were implemented into a GIS environment following appropriate standardization, rating, and ranking of the datasets (Argyriou et al. 2016). The reason is that in order to combine them in a single analysis, each cell for each factor needed to be reclassified into a common hazard assessment scale such as 1 to 10, with 10 being a location with the highest likelihood or severity of the hazard occurrence. A weighted overlay procedure with an equal weighting for the sum of the diverse three parameters was performed in order to derive the final earthquake hazard map (Fig. 5).
Due to the geomorphology of the terrain, landslides have been considered as a hazard causing damages to cultural heritage assets (Agapiou et al. 2015). The occurrence of landslides can be a result of human interventions in the landscape and/or due to geomorphological and climatological factors. In order to calculate the landslide hazard map, various factors were considered such as hydrolithology, geomorphometry (slope gradient), and climatic attributes (a complete time series of rainfall data of the 1990–2000 period) (Alexakis and Sarris 2010; Kouli et al. 2010). The spatial distribution of the mean rainfall for each month of the year was derived by applying the inverse distance weighting (IDW) interpolation method, as collected from various allocated meteorological stations and provided by the Hellenic National Meteorological Service (HNMS). The overall sum of these spatially distributed months provided the mean annual rainfall distribution. The hydrolithological formations (IGME 1971) were categorized in relation to the mean annual rainfall datasets, with higher values capable of triggering landslide events in relation to lower values (Alexakis and Sarris 2010). Similarly, the slope gradient was derived, via the TanDEM-X elevation model, and then categorized in relation to the mean annual rainfall datasets. All these factors were imported as spatial layers into a GIS environment and their range of values was reclassified with various ratings according to their association and linkage to possible landslide phenomena. The final landslide hazard map was based on the combination of the various ranked datasets through the application of a simplified weighted-factors model to provide an outcome advantageous to regions exposed to potential landslide phenomena (Fig. 5).
Because the Historic Centre of Rethymno (HCR) is located on the northern coast of the island of Crete, exposure to the strong north winds may cause serious damages to the historic structures. The wind hazard map was developed considering geomorphological characteristics such as aspect, with respect to the dominant strong north wind direction as recorded by the available climatological data information (HNMS 2017). The north-facing facades of the historical monuments in the HCR are exposed to the dominant north wind. There were five classes derived according to the facing aspect categories: (1) very high for north facing; (2) high for northwest and northeast facing; (3) moderate for east and west facing; (4) low for southeast and southwest facing; and (5) very low for south facing.
Flash and Coastal Flooding Hazard
With respect to rainfall-induced flooding in the HCR, it occurs only in the case of intense and abrupt rainfall, lasting only a few hours at maximum. Coastal flooding (for example, storm surge) seems to be a significant threat to the historical buildings and monuments of the town of Rethymno, especially for those structures that are facing towards the north, as is the case for the Lighthouse in the Venetian harbor. The flash flooding hazard map was developed considering as an input dataset the TanDEM-X elevation model, while the spatial modeling process was conducted with various spatial analysis algorithms such as flow accumulation, flow direction, cost allocation, and cost distance (Fig. 6). Based on the combination of the derived information, through the various spatial analysis algorithms, the least accumulative cost distance and its least cost source were determined to define the flood-prone areas at the HCR. The zones that resulted were the ones more exposed to potential flash flood events for the HCR site (Fig. 6). Similarly, the coastal flooding hazard map was derived highlighting the zones exposed to high waves or tsunamis. This was achieved by acknowledging the distance to the coastline (20 m inundation) with respect to elevation (up to 3 m) within a weighted overlay model.
Salinization is a slow-onset threat, in particular for monuments located close to the coastline (Robinson et al. 2010; Agapiou et al. 2015). In the case of the HCR, the north section of the Fortezza castle and the Egyptian Lighthouse in the Venetian harbor are the most exposed to the sea. A salinization hazard map can be developed by acknowledging both the distance to the coastline with respect to elevation and the derived aspect related to the dominant recorded wind direction. Specifically, the derived geomorphological information (elevation and aspect) can ensure the determination of the areas exposed to higher degrees of salinization, that is, more prone to salt-induced decay, with regard to their lower elevation and their facing aspect (northwards) regarding the dominant wind direction (north). The spatial datasets were then implemented within the GIS environment, and by using a weighted overlay procedure for all datasets, the areas with the higher susceptibility to salinization were derived. There were five classes produced to categorize the salinization susceptibility in relation to distance from the coastline (Fig. 6).
Hazard Analysis and Evaluation
As described in the Methodology section, a semiquantitative method was applied to incorporate the ranking factors of likelihood, severity, relevance of hazards to the sites, and expected intensity of impacts into the hazard analysis procedure. Accordingly, two main factors of “event parameter” and “expected intensity of impact” were derived to be incorporated into a hazard analysis matrix (Table 6). The criterion of “event parameters” was calculated by summing up hazard likelihood, severity, and relevance to the site. Then it was reclassified in order to allow its integration into the hazard matrix. Applying the equal interval method, the event parameter scores were reclassified to divide the range of values into five equal-sized classes of very low (< 5.4), Low (5.4–7.8), Medium (7.8–10.2), High (10.2–12.6), and Very high (> 12.6) (Table 6). The factor of “expected intensity of impact” is a rapid estimation of impact, according to the existing data and expert opinions to facilitate the identification of significant hazards. In the further steps of risk assessment, however, a detailed analysis of potential impacts needs to be carried out while applying exposure and vulnerability analysis methods.
According to the ranking factors, the identified hazards and threats may fall into one of the five zones of Very low (dark green), Low (light green), Medium (yellow), High (orange), and Very high (red) (Table 6). The qualitative interpretation of the hazards in the hazard analysis matrix shares a better understanding of the situation in a multi-hazard context with the site managers and stakeholders engaged in the safeguarding of the heritage site. Subsequently, a hazard evaluation needs to be conducted based on the above hazard matrix in order to determine those hazards that need to be incorporated into the next steps of the risk assessment procedure. The ALARP (As Low As Reasonably Practicable) principle (adapted from AEMC 2010) was applied to prioritize the hazards and threats for the further steps of ongoing risk assessment.
Figure 7 shows how the results of the hazard assessment contribute to the risk assessment procedure. According to the proposed procedure (Fig. 7) and the results of the hazard analysis matrix for the HCR (Table 6), those hazards with a High or Very high significance (for example, earthquakes and strong winds) fall into the red zone and those with a Medium significance (for example, coastal floods and heat waves) fall into the yellow zone. Within the overall framework, it means that both zones need to be incorporated into the further steps of risk assessment, in particular vulnerability assessment, to adequately analyze the structural sensitivity of the heritage assets to the corresponding hazards. Hazards in the green zone (for example, surface runoff) are not necessarily subject to the further risk assessment procedure; however, they should be considered within the heritage conservation system through the regular monitoring of the structural elements and climate parameters, for instance.
In the further steps of the risk assessment procedure, risks need to be identified and analyzed according to the elements of risk defined in the methodologies, depending on the objectives of risk management for a heritage site. For instance, in the exposure assessment, the value of movable and immovable heritage assets and their associated intangible elements can be analyzed. Vulnerability assessment may focus on the structural susceptibility of the heritage elements to the hazards of interest as well as on the nonstructural factors that influence the adaptive and coping capacity of the institutional, conservation, and management system. The results of the risk assessment will provide risk reduction strategies with tolerable and intolerable risks that need risk treatment options. The linkage between the risk management, whether at the individual site or overall urban level, and the heritage management system is essential to ensure the efficiency and effectiveness of the planning and implementation of strategies.
Apart from the overall significance of the hazards at the site level, the GIS hazard maps allow for a common understanding of the hazards at the single monument level. The Fortezza Fortress (HCR-1), for example, is exposed to earthquake and landslides (Fig. 4), and its fortification walls in the northern part are exposed to salinization (Fig. 5). A detailed exposure and vulnerability assessment needs to be conducted for the case of the Fortezza Fortress to address the potential impacts of the corresponding hazards. This is the case for the Lighthouse in the Venetian harbor (HCR-4) regarding salinization, and for the Soap Factory (HCR-3) and the Rimondi Fountain (HCR-5) regarding earthquakes. In general, the larger the structural detail that is mapped within the GIS, the better resolution of the definition of the specific hazards is achieved. The same holds true with respect to the input information that defines the hazards themselves, either coming from a dense network of close-by sensors or other measurements and analyses. Looking at the climate projections (Table 4), there are some slow-onset hazards (for example, heat waves) that have not been considered as serious threats, but are expected to increase dramatically over the next 50 years. In this case, even though there might not be any past damages or deterioration patterns due to such hazards, their potential impacts on historic structures need to be addressed within the risk assessment and management procedure.