Climate change poses increasing risks to what is valuable to humans around the globe. Changing, often intensifying, weather and climate extremes can increasingly be attributed to anthropogenic climate change, and change is projected to accelerate throughout the 21st century. Against this backdrop, risk awareness is growing across sectors and so is the demand for research and tools supporting efforts to mitigate climate change and adapt to its adverse consequences. Over the past decades, more and more global-scale climate impact data and models have become available, including archives of reported disaster impacts, storm track datasets, and global gridded crop models simulating yield responses to climate variables. The risk modeling platform CLIMate ADAptation (CLIMADA), implemented in the programming language Python, provides a modular open-source and -access platform for the probabilistic, event-based assessment of climate-related impact and risk. The underlying conceptual framework describes risk as a function of hazard, exposure, and vulnerability. Hazard is represented by the intensity and frequency of weather and climate events. Exposures constitute the presence of people, ecosystems, or assets that can be affected by a hazard. Vulnerability, implemented in CLIMADA in the form of ‘impact functions’, relates hazard intensity to the degrees of damage experienced by the respective type of exposure. CLIMADA is a growing modeling platform, and the three studies in this thesis aim to bring it to global consistency. The studies constituting this thesis were conducted within a joint research and development project with an implementing partner in the financial sector applying the results directly for forward-looking asset valuation. The main objective of this collaboration was to develop, evaluate, and implement climate risk modeling configurations with a global scope for the assessment of physical climate impacts and related economic risk. The implementing partner has not only integrated the model components as developed in the present thesis, engagement with academia also helped him to design and implement a more consistent risk assessment framework well beyond the scope if this collaboration. In Chapter One of this thesis, both the applied and scientific context of the project are introduced, providing insight in the conceptual framework and its application. Chapter Two enables the spatially explicit modeling of direct impacts to economic assets by providing a globally consistent asset value exposure layer. The method proposed and implemented makes use of the spatial correlation of economic activity and asset values in a country with both nightlight intensity and population density. The so-called LitPop (‘[night] Li[gh]t Population’) method combines satellite-based nightlight data and population data to disaggregate country-level asset value estimates to a sub-national high-resolution grid. The disaggregation skill is evaluated both quantitatively and qualitatively, comparing varying weights for nightlight intensity and population count. A global gridded data set of disaggregated asset values in US dollars for the year 2014 is provided alongside the paper at a resolution of 30 arcsec (approx. 1 km globally). In Chapter Three, LitPop exposure data is combined with a hazard set based on track data of hundreds of historical tropical cyclones (TCs) that made landfall in 53 countries between 1980 and 2017. The objective of the study is the calibration of the vulnerability component for TC risk modeling. For this purpose, wind speed footprints are modeled for each TC event. This allows the fitting of regional impact functions by comparing simulated with reported damage values for 473 reported events matched to individual TC tracks. Chapter Three is concluded with an explorative case study of TC damage in the Philippines, where the calibration comes with a large spread in fitted impact function parameters. With Chapter Four, attention is shifted from TC impacts to a sectoral risk perspective, for a global, country-level assessment of historical and twenty-first century risk to crop production. This study is based on global gridded crop yield simulations for maize, rice, soybean, and wheat. It uses an unprecedented ensemble of transient yield simulation output from eight global gridded crop models driven by bias-corrected output from five global climate models, as facilitated by the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP). Applying two complementary risk metrics, crop yield simulations are used to calculate the annual probability of projected crop production falling short of a given threshold, by country. Country-specific 21st century crop production risk is assessed by comparing these probabilities for historical and future levels of global warming, considering model agreement as a measure of robustness. The three main chapters are followed by a summary of key findings: quantitative estimates of exposure value distribution, TC vulnerability, and crop production risk per country. This is followed by discussions of the cascading uncertainties intrinsic to complex risk modeling chains, and practical implications of the thesis within the context of the joint research and development project as well as beyond. To make the resulting data and tools available for research and application – also beyond the scope of this project – the work in this thesis pays special attention to using scientific data and tools licensed for both academic and commercial use. At the same time, methods developed here are published open-source and -access, both as part of the CLIMADA repository and as peer-reviewed research papers. As for an outlook, future research is proposed with the potential to mitigate some of the entailed uncertainties, expand TC risk modeling from historical to future risk, and build on the findings presented here, further integrating output from climate and impact models in the probabilistic risk modeling framework of CLIMADA for globally consistent multi-hazard climate risk modeling.