Abstract. Urban areas are major contributors to anthropogenic CO2 emissions, yet detailed monitoring remains a challenge due to the cost and operational constraints of traditional sensor networks. As a scalable alternative, we established the ACROPOLIS (Autonomous and Calibrated Rooftop Observatory for MetroPOLItan Sensing) network in the Munich metropolitan area, using mid-cost sensors to enable dense, city-scale observation. This work outlines the development of the hardware and software of the system, its performance and the first 1.5 years of operation, during which more than 90 million CO2 measurements were collected in urban, suburban and rural environments. The primary goal was to evaluate whether mid-cost Vaisala GMP343 sensors, when combined with manufacturer internal corrections and environmental stabilization, can reliably measure CO2 concentrations with sufficient accuracy to resolve urban gradients. We implemented a fully automated 2-point calibration procedure using synthetic dry reference gases and conducted a multi-week side-by-side comparison with a high-precision Picarro reference instrument to assess sensor performance. Our results show that, despite inter-sensor variability in temperature sensitivity, the hourly aggregated mean root mean square error (RMSE) of all sensors is 1.16 ppm with a range of 0.57 to 2.58 ppm. For the specific sensor housed in our second-generation enclosure with PID-controlled heating, the performance improved from 0.9 to 0.6 ppm RMSE. Analysis of spatial and temporal patterns reveal distinct seasonal cycles, urban–rural concentration gradients, and nighttime accumulation events, consistent with expected biogenic and anthropogenic activity, and atmospheric transport mechanisms. We conclude that mid-cost urban networks can provide scientifically valuable, spatially highly resolved greenhouse gas observations when supported by appropriate calibration and stabilization techniques. The open-source design and demonstrated performance of the ACROPOLIS network establish a blueprint for future deployments in other cities seeking to advance emissions monitoring and urban climate policy.
Aigner et al. (2026) studied this question.