Accurately classifying laser-scanned point cloud data remains a critical challenge in geospatial analysis, particularly due to the complexity and volume of the data. This thesis presents a novel, confidence-aware deep learning framework designed to improve the classification accuracy of point cloud data, specifically focusing on the Actueel Hoogtebestand Nederland (AHN) dataset. The framework integrates geospatial knowledge into the deep learning process, enabling the model not only to refine its predictions through iterative learning but also to enhance the training data along the way via iterative online learning, ensuring continuous improvement in both training data quality and model performance.

The preprocessing phase assigns confidence scores to each point in the point cloud based on local neighborhood properties, with additional input from multispectral imagery (MSI) to further enhance the confidence estimation. These confidence scores are central to the online learning process, where the model prioritizes high-confidence points for training while progressively updating lower-confidence points to improve accuracy. To test the hypothesis that confidence-aware learning can enhance point cloud classification, we selected the KPConv network due to its suitability for handling unstructured data and capturing complex geometric features.

Extensive experiments demonstrate that the proposed framework, particularly with the Online strategy, enables deep learning models to perform better when trained solely on native point cloud attributes (elevation and intensity) compared to models without this strategy. Importantly, the Online strategy qualitatively enhances the training data by refining labels and reducing noise, thereby supporting more robust model performance. While incorporating additional features from aerial imagery showed no overall improvement, specific classes, like High tension and others did see performance gains.

The reconstruction of 3D city models has garnered significant interest in recent years. However, the majority of existing reconstruction methods primarily focus on LOD2 models, while LOD3 model reconstruction often relies on manual labor, and the primary data sources are street view images. This research aims to advance this field by reconstructing LOD3 models through the addition of windows and doors to existing LOD2 models, thereby maximizing the utility of available 3D building models, as well as the accurate addition of windows and doors. This research innovatively utilizes aerial oblique images as the data source for extracting building openings and employs 3D BAG LOD2.2 models as the basic 3D building structures. The 3D facades are projected onto the 2D aerial image space using perspective projection and registration is employed on the projection facade and oblique aerial images. Subsequently, Mask R-CNN is employed to detect and extract the building openings from these projections. Following the extraction, the layout of the openings within the same facade is optimized in terms of both size and position. Lastly, the relative positions of the openings on the facade images are combined with the 3D coordinates of the corresponding facade to calculate the positions of the openings in 3D space. This information is then integrated into the LOD3 model, resulting in a more detailed and accurate representation of the buildings.
This approach successfully reconstructs the final LOD3 model in CityJSON format, which passes the val3dity validation. By effectively utilizing existing 3D building models, this approach conserves a considerable amount of computational resources required for reconstruction. The simplicity and high level of automation of this approach make it a promising solution for reconstructing large-scale LOD3 buildings, leading to more accurate and detailed large 3D urban models.

In a rapidly evolving digital landscape, 3D city models have become more accurate and complex. Despite their widespread availability of open-source 3D city model datasets, these invaluable resources remain underutilized. Our primary goal centers on the classification of building and roof types. For our client, Spotr, our work directly impacts on their current project of house value estimations for insurance companies. Beyond insurance, it can be used in energy level rating, sustainability assessments, and so on.
In classification schema design, we designed 15 building types and 7 roof types that are compiled with cities in Europe. We labeled 6599 buildings and 2551 roofs in total using CityJSON files of Delft, Rotterdam, Den Haag, Munich, and Berlin. The accuracy for building classification is 70% and for roof classification is 74%.
For future work, we can increase the labeling amount and include more data from different regions. What's more, it can be extended to 2D roof classification: With the help of 3D roof surfaces, we can easily have the bounding box of the roof and additional 3D features can also be used.

Modern navigation heavily relies on Global Navigation Satellite Systems (GNSS) and digitized road network databases, but faces limitations in GNSS-denied areas and complex 2D road netowrks. This project addresses these challenges by developing a methodology to create and store a comprehensive 3D road and terrain dataset for enhanced navigation. In collaboration with TomTom, a company that aims to fulfill software requirements, making significant advancements in geolocation technology and societal contributions. The main research question of the project is: ”How can we create a 3D map of roads using information about the center of the road and elevation data?”. The approach to answer this question involves extracting 2D road polygons from centerline data based on width of the roads, the direction and the amount of lanes of them. These 2D polygons undergo enrichment with elevation data, with techniques like filtering, segmentation, and primitive extraction ensuring alignment with the digital terrain model. The methodology encompasses data acquistion, creation of polygons using the centerlines dataset, 2D-to-3D polygon conversion, elevation integration and data storage in CityJSON format.

The building permit process in the Netherlands is mostly digitalized, however, there are still some issues. A downsize of information, the manual checking from the municipality, and the duration of the process are some of these issues. To overcome these issues information between 3D building design models, so-called Building Information Modeling (BIM)s, and 3D city models must be exchanged. To exchange interoperable information between BIMs and 3D city models is called integration. In this research, automatic rule checking is performed after the BIM encoded in Industry Foundation Classes (IFC) is converted to a 3D city model encoded in CityJSON. Integration, however, is not as straightforward as it seems. Other researches have been carried out to perform a full integration from 3D city models encoded in CityGML to BIMs encoded in IFC. This is rather complex and so use cases are utilized. In most other researches using the building permit process as a use case, the automatic rule checking is performed in the BIM domain. In this research, a conversion is performed from a BIM encoded in IFC to a 3D city model encoded in CityJSON. The first step in this research is to analyze land use plans and select the most often used rules. These rules are further analyzed on required information to check the rules. For both IFC and CityJSON, the required information for the rules representing the same information as entities in the standards are selected. In the next step, the input models are analyzed on the presence of the entities from the standards. Before the conversion is performed, it is determined which entity will be converted from which input model and whether or not additional information is needed. Finally, the conversion is performed. The 3D city model can be used for rule checking and satisfies the selected rules. The implications of this research are described for the digital building permit process as well as for the integration of the two domains. Guidelines to model correct BIMs for the digital building permit process and further integration are drafted. In conclusion, the tool created in this research works successfully. Automatic rule checking on all the rules in land use plans is technically possible. In practice, automatic rule checking will most likely not take over soon, since rules are still written ambiguously, builders work with 2D drawings mostly, and the Environmental and Planning Act is soon to be established.

A digital reconstruction of real-life trees could provide many benefits in fields such as botany, forestry management, biology, and urban planning. Plant growth modelling in particular would enable the analysis of plant structure and behaviour in a customizable, widely applicable and non-destructive manner. Although many data-driven plant reconstruction methods exist, it remains a complex problem due to the intricate branching systems of trees and the need for balancing model soundness with adherence to the often incomplete input data. Modelling plant growth currently proves difficult as well due to the large number of factors involved in the growth process and the level of prior botanical knowledge and/or detailed field data that is often required.

This work uses an automatic MST (Minimum Spanning Tree)-based reconstruction method to obtain tree skeleton models from LiDAR input data. Multiple scans of the same tree, gathered at different years, are related to each other to improve and expand upon the reconstruction. A procedural model is used to simulate the growth in the tree tips using a lobe-based approach and a region-growing algorithm. The growth-based models provide a temporally informed reconstruction that is visually, geometrically and topologically sound. Establishing correspondences in the main structure between timestamps can assist the reconstruction of the tree at a time for which the input data was noisier or incomplete, as well as provide an estimate of the tree's structure in between known data points. This type of reconstruction can be used to both model and study the growth behaviour of trees, for multi-temporal visualisations, and to provide more informed tree models for reconstruction purposes.

The flexibility of the Industry Foundation Classes (IFC) format makes it challenging to blindly rely on the data it stores. This is because, to allow its flexibility, the stored data in an IFC file can be missing, incorrect, redundant, or stored in vague ways while the file can still be considered format compliant. This missing reliability is an issue for a format that is so central to the Architecture, Engineering and Construction (AEC) industry. This also severely hampers the chances for automated processes to be applied in the AEC industry. Chances that could reduce the time and cost spend on a construction while enlarging its quality. Prior done research on this subject is sparse. The small subset of this research that covers the ”repair” or reliable extraction of IFC data primarily focuses on idealized situations that do not often occur in practice. Due to the combination of these factors, this master thesis attempts to discover if it is possible to extract reliable data from the unreliable IFC files with an automated method. Because of the large flexibility and broadness of the IFC format, the scope of the thesis had to be limited. This was done by focusing on the extraction/reconstruction of storeys (and their elevations), rooms, and apartments. For these three subjects not only extraction/reconstruction methods have been created, but they have also been implemented in an open source software tool1. To develop reliable processes, the processes have to be based upon reliable data. From the data available in the IFC format, the tangible geometry is considered to be the most reliable. However, not all the other data can be discarded. Only the data that is implicitly stored in the tangible geometry can be ignored without losing information. In practice this means that most, if not all, topologic data and a subset of the semantic data can be ignored. It was found that for the extraction/reconstruction of storeys (and their elevations), a z-value extraction based on created IfcSlab/IfcRoof groups was fairly effective. These groups were made based on their size, height, and location. This process does show a lowered reliability in buildings with complex storey structures. It was found that for the extraction/reconstruction of rooms, a voxelization approach functioned very well. The boolean refinement that followed, which took the shape from a rough voxelization to the precise room shape, showed some reliability issues. Finally, it was found that for the extraction/reconstruction of apartments that rules applied to a created space syntax graph showed promise. The rules that were used were simple, resulting in lower reliability in structures that had more complex room relationships. In conclusion, this research has been unable to give a definite answer to the issues at hand. However, all three the subjects that have been covered showed promising results and can be considered steps towards a solution.

In recent years, there has been an exponential growth in the need for 3D spatial information, particularly 3D models representing the geographic information of the urban environment. The existence of a comprehensive integrated 3D model representing the condition of the underground utility networks in accordance with the above-ground city objects is crucial in the ever-increasing infrastructure demands. To acquire such models the related spatial information (geo-information) is required. This information and its quality determine the quality of the models since it describes the functionality, physical properties, semantic details of the urban features as well as the between them relationships (if exist) and interconnections. Currently, there are available various 3D models that, however, are limited in the representation of objects at a city level, with the corresponding underground information about the utility networks supporting cities’ functionality being restricted and/or underdeveloped. Although the modeling of underground utility networks is under development, there are available models that allow for the mapping of the existing information to predefined schema. However, these models have some limitations that restrain the display of the integrated condition with the above-ground city objects as well as they do not support (completely) the interdependencies between networks of different uses.
This thesis focuses on how to develop a three-dimensional model of underground utility networks integrated with above-ground objects, in order to utilize the model in real-world scenarios (disaster management, cost-effective
routes). These networks concern sewage networks (and the stored sub-networks), while the objects are related to the buildings of the study area. For the research, vector datasets were used, related to the existing underground utility networks of a limited area of the Delft University of Technology campus, as well as network elements related to them. For the examination of the proposed model creation, a methodology was developed that could constitute a general approach for relevant applications, considering the condition of the available vector datasets of the similar and/or relevant content.
This methodology is divided into seven stages that concern the preparation of the research and data collection process, the statistical and spatial analysis of their
content, the integration and cleaning approaches to reconstruct invalid information, and the creation of a relational database to store them. Finally, based on the results from these steps model, routing analysis and functionality were implemented to examine the effectiveness of such model in real-world scenarios. According to the results and limitations that arose, suggestions for the harmonization of the data are addressed as well as future application proposals.
Experimentation demonstrates that the developed methodology lead to the creation of a model that, although it represents a simplification of the existing geographical information, it can successfully be used for the implementation of the proposed case studies (disaster management, cost-effective routes). However, taking into account the poor quality of the input data, the model necessitates improvements in order to be used out of the box for other applications, as well as to verify its compliance with available standards supporting the mapping of the including information.

By 2050 the municipality of Amsterdam plans to become climate neutral. The historic city center of Amsterdam poses a challenge in this ambition, as it is home to numerous monumental buildings. Conservation is the starting point for the renovation of monumental buildings and consequently additional regulations are in place. Furthermore, each building is unique, thus renovation of monuments requires customization, making it more expensive and unattractive. Research has shown that due to restricted renovation possibilities, the indoor comfort and energy performance of monumental buildings are often below standard.
This research aims to create an energy-neutral renovation for a monumental building in the oldest neighborhood of Amsterdam: the Burgwallen-Oude Zijde. Resulting in the following research question: “How can monumental buildings be renovated to become energy neutral and comfortable for the users, whilst preserving the monumental status?” A renovation design is made for a mixed-use monumental building and elaborated into a generic approach.
Through careful analysis of the Oudezijds Voorburgwal 30, the chosen case study building, it was determined that a significant part of the building is not monumental. Furthermore, it was confirmed that the energy usage of the building is relatively high, and that the indoor comfort is currently not up to standard. To improve the energy performance and indoor comfort, a design is proposed in which different design measures are suggested. The design makes use of the potential of the neighborhood as it uses aquathermal heat (LT) from the canal in combination with heat pumps as basis for the heating of the building. The building is insulated using a combination of exterior and interior insulation, crack & seam filling, and double-glazed windows. For the monumental parts of the building a wall heating system is proposed, in which the walls are slightly heated to minimize energy losses to the exterior (‘WarmBouwen’ concept’). Energy is produced on site using PV roof tiles on the South facing roofs and several other small design interventions are proposed. Evaluation of the proposed design reveals that none of the identified monumental elements are altered, the energy demand of the building is lowered by 73% and that the indoor comfort increases significantly.
By following the proposed generic approach of (1) identifying the heritage values, (2) assessing the current performance of the building, (3) creating concept designs, (4) finalizing the design, and (5) evaluating the design, the remaining energy demand of the proposed design for the Oudezijds Voorburgwal 30 has not been reduced to zero and is thus not energy neutral. This is mostly due to restricted renovation possibilities and limited roof area for on-site energy production. The proposed design does make the Oudezijds Voorburgwal 30 BENG compliant: a step in the right direction, but for monumental buildings to become energy neutral a fairer middle ground between conservation and sustainability should be found. We should ask ourselves: how far should one go to preserve the past if it limits or damages our future?

The growing complexity of many projects and applications require methods that help integrate vasts amounts of data coming from different sources. In all cases, semantic interoperability plays an important role - the meaning of content should be organized logically, as to allow machines to interpret it. The Semantic Web vision has been developing standards to support this idea, with the goal of creating a 'web of data', were pieces information can be accessed and linked to one another consistently. The ideas behind Linked Data have been explored in different areas of study. The geospatial domain presents an interesting challenge, as geospatial data offers many different views on the same physical universe. This is especially the case with geo-registries - collections of official geospatial data used by governments. This data already contains links in the form of spatial relations. However, there is no consensus on how these links should take form. Thus, this research will explore how linkable geospatial data from registries actually is by answering the following question:"To what extent can ontology-based solutions using semantic web technologies contribute to the integration and use of data from geo-spatial registries?". This question is broken down in three sub-questions that address ontology-based integration techniques (including tools and languages), the impact of dataset characteristics and the added value of semantic relations. The research includes a literature study to provide a theoretical background, a case study of two Dutch geo-registries (Basisregistratie Grootschalige Topografie and Basisregistrie Topografie) and a conceptual framework that is applied to the case study as a way to explore the main question. The results of this framework reinforce the importance of using instance-level data to understand the connections between different conceptual models and show the unavoidable subjectivity that is involved in the process - from the conceptualization of the data (model) to the creation of query rules. The findings indicate that the capabilities of ontology languages (such as OWL) are not necessarily practical for data from registries, and that custom processes might be necessary depending on the application envisioned by data users. Country-wide registries managed by different data owners and different interpretations of data acquisition rules will lead to inevitable variations in the registration of objects in datasets. And more efforts could be invested in providing quality indicators for alignments.

Noise simulations require finding the paths between multiple receiver and source points. In the current approach, only 3D polylines can be used as input to describe the terrain. These 3D polylines are semi-automatically generated, based on the principle of describing the terrain profile with as few height lines as possible. In order to propose a more efficient, standardised and economic modelling approach, a partnership between RIVM/RWS and the 3D Geoinformation Group at TU Delft was launched in 2017, aiming to generate these height lines automatically from the available datasets, namely AHN3, BAG, and BGT, which are publicly available via PDOK for free. However, it was then proposed to prove that the paths between receiver and source points can be directly generated from a TIN without creating the height lines. The following report provides proof of concept to the hypothesis: ‘Using a TIN directly allows automated 3D noise modelling according to the guidelines of CNOSSOS-EU’. A code was written to generate the paths between receiver and source points using an LoD2 TIN. The paths were then checked visually and were fed to test_Cnossos software to prove their validity. Finally, noise maps were generated and compared to noise maps generated with the current method.

Our linear way of producing has to come to an end. More and more companies and government institutions see the value of circularity of raw materials. The municipal- ity of Amsterdam published several reports to clarify the directions towards circu- larity. A report on waste-chains highlights the importance of organic residual flows for circular economy transition. This thesis focuses on organic residual streams, the lack of understanding of these streams has been mentioned as an obstacle to plan- ning by many of regional stakeholders. The topic of this thesis is formulated within the context of the REPAiR project by the AMS Institute. The starting point of this thesis is the waste collection location in public space and follows the waste back to the retailers of food. The households are identified, after which the used trashcan and the retailers of food are identified.
To create the geographical connection between the three nodes in this thesis, a related works research is conducted. The main source of food waste is food retail and the main disposal is collection of residual waste. Since the municipality of Amsterdam is not collection organic waste separately, residual waste is considered. Multiple researches added spatial dimensions to their material flows, trying to cre- ate a better insight in sinks and flows of the chains. Household level is not reached. This leads to the research question: How can household waste flows be mapped in reverse based on their environment?
To create the mentioned connections, two main methods are developed. First, the method household-to-trashcan and second, the method household-to-retail. The first method is simulating the pedestrian path from a household towards their used trashcan. A least cost raster is created from the BGT, the direct environment is translated into walkability values representing the willingness to walk on a certain object. By comparing the paths to different trashcans, the trashcan reachable with the smallest least cost value is picked. Another factor is inserted, considering that walking to the trashcan is mandatory trip, executed in as less time as possible, city hotspots are inserted. The hypothesis suggests that people are willing to walk further, if the residents can combine their trip to the trashcan with their trip to, for example, the supermarket. By changing the values in the least cost raster, this direction is made preferable. After the validation of this method is concluded, that a stronger influence of the city hotspot (among the tested values), results in a higher positive correlation of the number of households per trashcan and the mass of waste in the trashcan in kg.
The second method household-to-retail, is a demonstration, due to the uncer- tainly and the many involved personal factors. Food retailers in different isochrone areas are given importance values based on their distance towards the considered household and the floor size of the retailer.
After establishing the connections from trashcan to household and from house- hold to food retail, reverse distribution of the waste measure in the trashcan can be executed. In this way is known which household contributed how many kg per year to the trash composition of the used trashcan. When adding the other step, the contribution by the food retailers can be given.
This thesis generates the first step in mapping residual waste flows touching household level. Promising positive correlations between the number of households using a trashcan and the kg measured inside this trashcan are shown. The method household-to-retail cannot be validated and is therefore only a demonstration. The outcome of this thesis is a step in the right direction to be able to follow waste through the waste chain, by analyzing results in trash composition and knowing which households and retailer contribute to this composition.

In 2021, noise pollution monitoring will be mandatory in the Netherlands, which requires data on traffic that can be re-used for air quality estimation models. One of the important input parameters for the latter is the street type, which is required by the dilution parametrisation used within the air quality model.
The goal of this project is to show whether automatic street classification for air quality estimation is feasible and reliable, considering the geo-spatial data currently available in The Netherlands. The motivation for this project originates from the common data used in noise and air quality monitoring tools by the Dutch National Institute for Public Health and the Environment, (RIVM).
Currently, street classification is performed manually by many municipalities. The larger municipalities are legally obliged to monitor air quality levels, which makes use of the street types. Automating the process by using existing datasets can save a lot of time, costs, and resources, while providing standardised results in comparison to manual classification. In addition, our method is extendable to the whole of the Netherlands. Consequently, our method can have a large societal impact, since it allows the provision of air quality estimations for all municipalities; even those that are not yet required to do so. To our knowledge, no similar work has been conducted in this field, which made it even a bigger challenge.
The implementation of the automatic classification algorithm, which is thoroughly explained in this re- port, shows very promising results. We first tested the approaches in a small area, the Weesperstraat in Amsterdam, where we have success rates from 76.7% to 83.3% for the different classification methods when compared to the NSL classification. After evaluating the performance of each of the methods, the optimal approach has been tested on larger areas where visual inspection shows a priori promising results as well.
In addition to the automatic classification algorithm, air quality measurements with new Flow sensors from Plume Labs were performed in the city of Amsterdam. The goal was to investigate whether different street types can be identified through the use of small air quality sensors. The limited measurements did not provide distinct patterns for the different street types, and therefore identification based on pollutant concentrations was not possible within the project.
We hope that the results of this project will motivate public bodies and agencies in the Netherlands to invest in automated workflows using currently available and high accuracy geo-spatial data. This can potentially improve their efficiency, while creating a more standardised and scalable framework.

This study presents a methodology to convert from the most dominant 3D city model standard in 3D GIS to BIM. Namely, CityGML to IFC. IFC is chosen because it is the common open standard to exchange data in the BIM world. For the aim of this study, the two standards are divided into 5 comparable sub-parts; Semantics, Geometry, Geographical coordinates, Topology, and Encoding. The characteristics of each of these sub-parts are studied and a theoretical conversion method is proposed for it from the first standard to the other. This is done by performing a semantic and geometrical mapping between the two standards, converting the georeferencing from Global to local, converting the encoding that the two standards use from XML to STEP, deciding which topological relations are to be retained, and providing a basic implementation that is created using Python to combine the above tasks. The work presented in this thesis can provide a foundation for future work in converting CityGML to IFC. It provides an insight into the relationship between the two standards and a methodology for the conversion from one to the other, and the process of developing software to perform such conversion. This is done in a way that can be extended for future specific needs.

3D city models are now common planning and analysis tools. Urban vegetation as a feature in these models, however, is neglected overshadowed by the focus on buildings, so its inclusion in 3D city models is often symbolic. On the other hand, urban vegetation improves the comfort and social well being of a city’s inhabitants and is a resource for sustainable urban growth and an environmentally friendly resource for mitigating the negative effects of climate change, e.g., frequent heat waves, floods from storm downpours and extended dry periods. Trees also mitigate Urban Heat Island (UHI) effects. Urban vegetation’s ecosystem services (ecoservices) as mitigation functions of pollution and negative effects of climate change have propelled research, studies, applications, and simulations that need data and 3D models of existing vegetation e.g., for spatial simulations, to assess the canopy cooling impact to surroundings or to identify areas prone to UHI. In view of these needs, urban vegetation in 3D city models is underrepresented. Guidelines for modelling vegetation are already provided in CityGML, the 3D city modelling standard, but they are insufficient for today’s needs because it has vagueness and focuses mostly on built infrastructures.

In this research, 14 Single Vegetation Object (SVO) Levels of Details (LOD) and four root LODs are proposed. They target to meet different adherence requirements and scales. Their formulation is based on LOD specification approaches, and on a needs analysis that identified the vegetation models and data most commonly required in applications in the urban environment. Vegetation LOD description approaches include semantic 3D modelling standards, industry, and common practices of municipality users which are also GIS data providers.

Current vegetation LOD descriptions approaches fall into two different groups based on the geometry they adopted for their specifications: implicit or explicit, and no one approach fulfills identified needs. Acquisition techniques and demand in IT resources have influenced the definitions of vegetation LODs, the adoption of one geometry type or the other, and the wide use (or not) of certain LODs.

Refined SVO LODs specifications of this research combine the strengths of each group with descriptions that cover beyond geometric specifications. Refined LODs incorporate implicit components, underground representations and reconstruction LODs not defined by any approach. With them, most datasets can be represented by at least one LOD, and modelers can tell what LOD is possible to implement based on the data they already have. For acquisition, it is possible to tell what data is required for a particular LOD, and which LOD can be used to obtain data needed for an application. The broad spectrum of refined LOD allows them to meet different requirements.

A shadow analysis case study was done with implementations of refined SVO LOD specifications. Acquisition from aerial and mobile LiDAR data was done in a workflow that brought the 0D tree inventory of the municipality of Rotterdam to 3D models using mainstream and open source tools. The case study confirmed a quantitative impact in shadow duration and extent by each LOD indicating that each is independently differentiated. Volumetric and non-volumetric models had different shadow over and underestimation impacts. The study further highlighted the properties of the crown of the real-world object that help in choosing a LOD and gave insights offered by lower LOD models.

The implementation of the assorted LODs revealed that while much research has been done in acquiring vegetation parameters from LiDAR data, the many options, methods and algorithms are scattered necessitating a unifying process or tool.

To cut down immense greenhouse gases emission and energy consumption in the rapidly urbanizing world, a holistic understanding and rethinking of our dynamic urban energy system are inevitable. Performing bottom-up building energy modelings at urban scale based on Geographic Information System (GIS) and semantic 3D city models could be a promising option to provide quantitative and integrated energy solutions.

Nevertheless, input uncertainties either caused by limited data accessibility in most cities or parameters with stochastic variability (e.g. house occupancy profile) become one of the biggest obstacles to produce reliable and acceptable building energy modeling results. This study aims to address the heating demand simulation performance gap caused by input uncertainties. In this case study based on Amsterdam residential building stock, parameter importance ranking of the 14 simulation inputs are first derived according to the sensitivity analysis. The selected key uncertain parameters are then modeled in a probabilistic distribution way at postcode 6 level (approximately or slightly more than 10 buildings). Model calibration is based on the Bayesian approach and given six years (2010-2015) of gas consumption data to infer parameter posterior distributions. After the training phase, the calibrated annual heating demand simulation results of the validation years show significant improvement in modeling accuracy. Comparing the baseline and the calibrated simulation results, the averaged absolute percentage errors of energy use intensity (EUI) among at least 84 valid postcodes have decreased from 24.96% to 8.31% in 2016 and from 19.93% to 7.70% in 2017 respectively.

The calibrated urban building energy model would be most interested by municipalities, urban planners, and engineering consultancies. It can be used to evaluate long-term energy supply and demand strategies, identify building renovation saving potential, perform large-scale building performance mapping, and carry out retrofit measures assessment.