Menü überspringen

Design Practices

Energy retrofit contributing to
climate-proof built environment

A further tool to climate-proof the built environment, is the newly developed method of resilience-enabling energy retrofit interventions.  The method, which is further described in the public report D8.4, consists of an area-specific building typology that classifies buildings according to their structural and non-structural features. 

Understanding and classifying building typologies is essential for developing targeted climate-resilient retrofit strategies. A building typology refers to the systematic classification of buildings based on shared architectural, structural, and functional characteristics—such as construction period, number of storeys, structural system, façade type, and layout. These typological attributes enable tailored interventions for improving energy efficiency, enhancing seismic resilience, and ensuring thermal comfort across different contexts. 

Building typologies support a range of sectors and objectives: 

  • Targeted retrofitting: Retrofit needs and solutions differ significantly across building types. For instance, pre-1945 masonry multifamily buildings often require a different approach than post-2000 reinforced-concrete structures. Understanding typology allows interventions to be fine-tuned to each building’s vulnerability and energy behavior. 
  • Refined risk assessment: Typological data improves hazard modelling by enhancing the precision of risk assessments.  

Policy and planning guidance: Building typologies are embedded in national energy strategies, influencing energy performance certificate (EPC) systems, retrofit incentive schemes, and zoning laws. The JRC’s 2014 overview of building typologies in Europe highlights their central role in implementing long-term renovation strategies. 

The format of this method is a python code, exploiting machine learning clustering algorithms for the grouping of the buildings. The functions of this code are:  

  • Data Loading & Preprocessing: The first step involved importing and loading the building dataset from Excel, which contained all relevant features for analysis. Where necessary, derived variables were computed, such as composite damage scores for structural components. The dataset was then cleaned to remove inconsistencies and handle missing values, ensuring that only reliable information was included in the clustering process. 
  • Feature Definition & Selection: Candidate feature sets were defined separately for each hazard (heatwaves and earthquakes) to capture the specific characteristics most relevant to vulnerability. A correlation analysis was conducted to explore relationships between variables and identify redundancy. Based on this, we selected a subset of non-redundant and physically meaningful features, ensuring that the clustering would be driven by variables that truly reflect building characteristics and hazard exposure. 
  • Data Normalisation environment – environment – & Transformation: To prepare the selected variables for analysis, continuous and ordinal features were normalised using MinMax scaling, placing them on a comparable range between 0 and 1. Nominal and categorical features were encoded appropriately so that they could be integrated alongside numerical data. These transformations ensured that all variables were represented in a consistent format suitable for clustering and distance computation. 
  • Distance Matrix Computation: Because the dataset included a mix of numerical, ordinal, and nominal variables, we employed the Gower distance metric. This method allows the calculation of pairwise dissimilarities across mixed data types, ensuring that each type of variable contributes appropriately to the clustering. 
  • Cluster Analysis & Validation: Clustering was then performed using algorithms capable of handling mixed-type data, specifically Hierarchical (Agglomerative) clustering and K-Medoids, both applied to the precomputed Gower distance matrices. To evaluate the quality of the clustering, internal validation indices such as the Silhouette score and the Dunn index were calculated. These measures guided the determination of the optimal number of clusters (k), balancing compactness and separation of the resulting groups. 
  • Cluster Assignment: Based on the chosen method and optimal cluster number, each building in the dataset was assigned to a cluster. These cluster labels were integrated back into the main dataset, allowing further interpretation and comparison of results across different building groups. 
  • Statistical & Spatial Analysis: For each cluster, summary statistics (mean, median, minimum, maximum, and standard deviation) were computed to describe the distribution of features. Spatial analysis was also performed, mapping the distribution of clusters and key variables across Camerino’s urban fabric. For earthquake-related clustering, the analysis was extended to include cluster-specific distributions of structural damage types and levels, offering a more detailed understanding of vulnerability patterns. 
  • Visualisation: Finally, a range of visualisations supported both validation and interpretation. These included feature distributions, correlation matrices, silhouette and Dunn indices, dendrograms, and spatial cluster maps. Maps were generated not only for clusters but also for each key feature, enabling the identification of distinct geographic patterns and their relation to hazard vulnerability. 

No, it is not accessible. The code can be accessible upon request. Currently, is available as a service offered by the C-EREL (Climate- Environmental REsearch Laboratory) of NCSR Demokritos. 

Currently, there is no user group for this application. It can be applied only by the scientific staff of the C-EREL. The area of interest that has been applied and is planned again to be applied in the foreseeable future are EU Municipalities. 

The potential clients of this service (e.g. municipalities) can benefit from the application of building typologies against the negative impacts of climate change and natural hazards. From a methodological perspective, building typology supports both top-down (desk-based) and bottom-up (data-driven) approaches. The former relies on existing classifications from literature and institutional sources, while the latter employs computational techniques—such as clustering algorithms—to extract structural patterns from large datasets. The resulting typology enables: 

  • Scenario modelling for energy and water resilience 
  • Prioritized strengthening of seismically vulnerable typologies 
  • Development of hazard-specific and scalable retrofit guidelines 
  • Enhanced communication and engagement with local stakeholders 

In practical terms, building typologies serve across various domains: 

  • Energy retrofits: Envelope upgrades, insulation, and passive cooling strategies are matched to typological features to optimize energy performance without compromising heritage value. 
  • Seismic resilience: Typologies allow estimation of expected structural failure patterns under seismic stress, aiding prioritization of reinforcement efforts. 
  • Urban planning: Planners use typologies to tailor zoning policies, emergency response plans, and funding allocation for renovation. 

Monitoring and modelling tools: Typologies are integrated into models and databases—such as EPC registries or GIS-based hazard simulations—to project impacts and test adaptation scenarios. 

Building typology methodology was developed for and implemented in the Camerino demo case, specifically for its historical center. The study was limited in the characterization of the buildings and divide them in groups, matching the proposed energy solutions, originated from D2.4, to each group, with a focus on nature-based solutions, and examining potential restrictions on the implementations of those solutions due to cultural heritage regulations.  Thus, the scope of this research was on a theoretical basis and the potential implementation is intended in Task 11.1. 

Contact for more Information