AbstractVolatile organic compounds (VOCs) have crucial influences on atmospheric chemistry. Accurate quantification of the VOC emissions is critical for air pollution research, especially when applying to air quality models. However, current bottom‐up emission inventories have biases, making observational constraints of VOC emissions necessary. We conducted concurrent VOC measurements in the Pearl River Delta (PRD) region during the summer of 2018 and found large discrepancies in the spatiotemporal variations of VOCs between observations and model simulations when using the priori VOC emission inventory (Multi‐resolution Emission Inventory for China). The normalized biases of total VOC concentrations ranged from −55% to 85% across the PRD cities in the study period. To improve the simulations, we constrained the anthropogenic VOC emissions based on their measured concentrations. The observation‐constrained VOC emissions showed clear diurnal variations and resolved the spatially‐concentrated priori emissions by reducing the high emissions by 15%–36% in the central PRD cities while elevating the sparse emissions in other cities. After employing the observation‐constrained VOC emissions, the model better reproduced the spatiotemporal variations of VOCs in the PRD region, alleviating the biases to −13%–13%. Furthermore, simulations of peak ozone (O₃) concentrations were amended to reduce the mean normalized bias by 5%–12% on high O₃ days. Our work has effectively combined VOC field measurements with air quality modeling to achieve better simulations of VOCs and O₃. Besides, the observational‐constrained emissions are flexible for studying short‐term emission changes and their subsequent impacts on air quality.