In order to limit the global warming to well below 2 degrees Celsius, all sectors have to reduce their greenhouse gas emissions and become more sustainable. This also includes the building sector, which is in Europe responsible for 40% of the total energy consumption (European Commission, 2020). A way to work towards this goal is by retrofitting the existing building stock to become more energy efficient. Urban Building Energy Modelling (UBEM) can help in this endeavour by identifying energy-saving potentials and thus to effectively allocate the required resources (Horak et al., 2022). Yet, UBEM involves many stakeholders which is why standards are crucial to facilitate data exchange and interoperability among them. In this context, the Energy ADE v1.0 was developed as an extension for the semantic 3D city model standard CityGML 2.0. It serves two purposes, first by storing energy related information on the individual building level, and second by providing the necessary input data for UBEM simulations (Agugiaro et al., 2018). However, in September 2021 CityGML 3.0 was released. The introduced changes directly affect the structure of the Energy ADE, which is why it cannot fully function on it anymore. This thesis therefore answers the question, how and to what extent the Energy ADE for CityGML 2.0 needs to be adapted to be conformant with the new CityGML 3.0 standard. It is accomplished by following a model-driven approach, where the UML class diagrams for the mapped Energy ADE are created first, before automatically deriving the corresponding XSD schema file. Through the lossless mapping itself, the Energy ADE is integrated as much as possible into CityGML 3.0, while also maintaining a logical symmetry. As such it accounts for the introduced changes of CityGML 3.0, by making use of the space and geometry concept, the versioning possibilities as well as the provided structures to model time-dependent data. The result is eventually tested and verified by converting a sample dataset to the Energy ADE for CityGML 3.0. This work provides an example on how other ADEs can be adapted to fit the new CityGML 3.0 standard and thus hopefully to the further establishment of it.

In the Global South, large urban spaces resulted in the duality between the so-called ‘formal’ and ‘informal’ cities. It is the case of São Paulo, a twenty-two-million people metropolis and a financial hub in Latin America. Albeit a vast literature addresses the social-spatial segregation emerging from this dual built environment, the scarcity of spatial datasets regarding informal settlements also enforces a geo-information segregation, resulting in a terra incognita. This is exemplar in favelas, defined as precarious, spontaneous, and unorganised land occupation built on third-party property, most of which lack cadastral data. Since favelas are often not mapped, assessing urban phenomena becomes a technical challenge for several application domains, e.g., the energy one. A recent public initiative in Brazil estimates solar irradiation and photovoltaic potential for buildings at city scale, but favelas are intentionally excluded from the resulting web-based solar maps. Technicians believe that the absence of a spatial pattern in favelas calls for investigation on how to refine a roof mapping methodology. The main research question becomes: “How far is it possible to perform solar analysis on buildings of favelas in São Paulo, with the goal of estimating PV Potential?”. The research is structured into two topics: 1) Roof Mapping, which investigates the data pipeline that leads to a digital reconstruction of favelas; 2) Solar Irradiation, which investigates how existing solar irradiation modules – GRASS GIS, ArcGIS, CitySim, SimStadt, Ladybug and the one developed by Virtual City Systems – perform when assessing buildings of favelas. From a roof mapping perspective, the experiments reveal that the absence of cadastral datasets represents a complex technical challenge. Nevertheless, the reconstructed and post-processed building footprints cover the extension of the cadastral footprints that are available for the study area, with building shapes that are satisfactory as a first approximation. Regarding the solar irradiation perspective, qualitative and quantitative analyses are carried out to compare the results coming from the six solar modules. The qualitative analysis indicates that each solar module offers potentialities but also limitations. Therefore, a straightforward choice is not possible, since the optimal solution will be derived from a data-driven approach that considers, among other factors: the scale of the favela, its topographical characteristics, the presence/absence of urban features other than buildings (such as vegetation), a possible pre-selection of buildings of interest, etc.; The quantitative analysis reveals that ArcGIS outputs an annual summation of irradiation values that is the closest to the one offered by the meteorological station of São Paulo, adopted as ground truth. Nevertheless, from an accuracy perspective, CitySim outputs a daily curve that best corresponds to the ground truth one. In conclusion, based on the geometrical model and the weather dataset criteria, the author expresses his preference for a raster-based solar module – GRASS GIS or ArcGIS – if, on the one hand, the reconstructed building footprints result in an unrealistic or excessively complex vector-based model. On the other hand, if the resulting vector-based model is simple enough and representative of the built environment of the favela, the author suggests the adoption of CitySim.