Membrane oxygenators are medical devices used to support or take over the gas exchange of the natural lung. In modern oxygenators, the gas exchange surface is provided by a hollow fiber membrane packing. While blood is pumped through the shell side of the hollow fiber packing, O2 is used to sweep the fiber lumen. CO2 and O2 are exchanged through the membrane following the partial pressure gradient. Consequently, blood is enriched with O2 and purged from CO2. Initially, membrane oxygenators were developed to supplement the natural lung during cardiopulmonary bypass. Here, the oxygenator has to take over the total metabolically required O2 and CO2 transfer. With continuous development, oxygenators were applied as lung support to manage acute respiratory distress syndrome (ARDS). Patients suffering from ARDS are often treated with lung protective ventilation (LPV). While LPV allows sufficient O2 transfer, the CO2 removal is limited. The limited CO2 removal evokes serious side effects such as upcoming hypercapnia and hypercapnic acidosis. Consequently, oxygenators are increasingly used to provide additional CO2 removal during LPV to circumvent the mentioned side effects. This doctoral research aims to improve the development process of oxygenator-based CO2 removal. A particular focus is placed on the initial development phase, where the work is strongly characterized by engineering challenges such as design, assembly, and first basic performance tests. The development process shall be improved in two ways. First, by designing experimental campaigns that are simple, inexpensive, and reliable. Second, by developing numerically inexpensive and accurate computational fluid dynamic (CFD) methods for in-depth insights into the CO2 separation process. The fundament for efficient development of oxygenator-based CO2 removal is an accurate measurement of the CO2 removal rate. We compared the two available CO2 removal rate determination methods, i.e., the determination based on CO2 concentration decrease in the blood (blood-based) and the determination based on CO2 concentration increase in the sweep flow (sweep flow-based). Our study shows that the sweep flow-based method performed superior with a CO2 removal measurement error of 3 % of reading. This error lies significantly (p < 0.05) under the CO2 removal measurement error of the blood-based method (16 % of reading). Furthermore, blood mimicking fluids for the determination of the CO2 removal rate of oxygenators were evaluated. While water tests are a common method to reduce experimental effort and avoid animal tests, its application limits and reliability have never been analyzed systematically in the literature. Consequently, we compared the CO2 removal rate of blood and water at three pathological elevated CO2 partial pressures (50, 70, 100 mmHg) and three blood flow rates commonly applied in blood oxygenation (1000, 1300, 1600 mL/min). Our experimental data shows an average 10 % deviation between the CO2 removal rate of blood and water. The low deviation can be attributed to the opposing influences of the material properties of the two liquids. Using CFD simulations, we could quantify the contributions of the different material properties. Compared to water, the higher CO2 solubility of blood and the accompanied increased CO2 removal rate (+ 125 %) is in most parts compensated by the lower CO2 diffusion rate of blood (- 53 %), followed by the lower CO2 permeance available with blood (- 18 %) and the higher viscosity of blood (- 10 %). While this leads to comparable macroscopic CO2 removal rates, we could elaborate that the boundary layer built up – the main CO2 transport resistance in oxygenators – is fundamentally different between blood and water, i.e., the two liquids do not follow the same dimensionless mass transport analogy. Hence the use of water as a blood model should be limited to the macroscopic determination of the CO2 removal rate and not be used in studies of the boundary layer. In addition to water as a CO2 transport model, this work investigated aqueous and animal blood models as rheological models for blood. While experimental data is most reliable, it is mainly limited to pointwise data at easily accessible locations. Computational fluid dynamic (CFD) simulations can extend this data. However, they are numerically expensive due to the highly refined computational mesh required to resolve the diffusive CO2 transport in the membrane packing. Consequently, in the current literature, a gap between the geometric size scales of mass transfer and hydrodynamic simulations of oxygenator membrane packings can be recorded. In order to bridge this gap, an up-scaling method was developed. It allows scaling the transmembrane transport predicted in species transport simulations of a reduced geometry on the geometrical scales of flow simulations. This is done by calculating velocity inlet conditions of the reduced geometry based on the average velocity within the complete packing determined via CFD flow simulations. By doing so, the flow distribution in the reduced geometry is representative of the flow regime within the complete packing. This was proven by comparing experimental and numerical results. The deviation between the experimentally determined and numerically predicted CO2 removal rate of a prototype oxygenator amounts on average to 6 % for blood and 3 % for water. As a further novelty, our CFD model for the species transport in blood oxygenators can resolve the CO2 transport in the membrane wall and the fiber lumen. While the membrane wall resistance is often considered negligible in literature, we could show that membrane permeance can reduce to 22 % of its original value due to plasma leakage or pervaporation. CFD simulations show that this would result in a proportional decrease of the CO2 removal rate. This is important since numerical overprediction of the CO2 removal rate poses a risk of incorrect validation of CFD models. In order to guarantee high accuracy of a CFD model, suitable models for the complex material properties of blood are required. In the scope of this work, models for two of the most relevant material properties in oxygenator-based CO2 removal, CO2 solubility of blood and viscosity of blood, were investigated. Based on our data, we can recommend a simple and reliable CO2 solubility model proposed in the literature. Furthermore, viscosity models for ovine, bovine, equine, and porcine blood are presented as a function of shear rate, hematocrit, and temperature. Finally, a less complex CFD model based on the calculation of local Sherwood numbers was successfully tested. It allows to qualitatively assess different microstructures of hollow fiber membranes regarding their CO2 transport resistance in blood. To conclude, the research conducted in this doctoral thesis offers a solid foundation for designing reliable experimental and numerical investigations of oxygenator-based CO2 removal. Due to the experimental findings and advances in CFD modeling, oxygenator-based CO2 removal can be efficiently developed in the future.