Designing urban CO2 sensor networks under realistic uncertainties and biases: Results from synthetic studies in Berlin and Munich

Accurate knowledge about urban carbon dioxide (CO2) emissions is essential to support effective climate change mitigation. Yet large discrepancies among emission inventories within cities highlight the need for well-designed atmospheric CO2 sensor networks that can deliver independent CO2 emission estimates. Most cities still lack observation systems capable of meeting this challenge. We conducted Observing System Simulation Experiments (OSSEs) for 2 German cities, Berlin and Munich, to assess how the number, quality, and placement of in situ CO2 sensors affect the ability to quantify city-wide emissions under various error conditions. The experiments included random errors from transport and limited precision of the sensors, as well as systematic biases arising from background concentration errors and limited sensor accuracy. Networks were evaluated by their ability to recover assumed true emissions from incorrect prior assumptions using simulated measurements in a Bayesian inversion framework. The results show that thoroughly investigating network designs using OSSEs prior to physical deployment enables identification of optimal sensor locations that substantially improve emission estimates compared to random placement. Across all tested error cases, we found mid- and high-cost sensors to provide reliable constraints on total city emissions, but mid-cost sensors can introduce biases on sub-urban scales. Sensors with accuracies of 5 ppm or worse produced considerably smaller improvements or even degraded prior emission estimates. Spatially uniform background errors could generally be corrected through an appropriate choice of state vector, whereas spatially varying biases propagated into the inferred emission patterns. These findings were consistent across both cities, suggesting their applicability to similar urban areas. This study thereby provides practical guidance for designing cost-effective urban emission monitoring networks and highlights the importance of assessing diverse error contributions within a network’s planning phase.