Most anthropogenic carbon dioxide (CO₂) emissions originate from urban areas. To improve understandings of urban and regional emissions, we design and construct a low-cost UAV coordinated carbon observation network (LUCCN) which uses mid-accuracy (±1 ppm) CO₂ sensors. In this paper, we introduce our multi-variable non-linear regression method for calibrating the non-dispersive infrared (NDIR) CO₂ sensors for LUCCN’s ground stations. We tested our calibration method with concentration data collected at the Xinglong Atmospheric Background Observatory. With comparison against data simultaneously collected by a high-accuracy cavity ring-down spectrometer, we found the maximum standard deviation of LUCCN’s sensors to be 0.782 ppm in a controlled laboratory environment with a 1-second window size and 0.53 ppm in an outdoor environment with a 1-hour running average window size. As validation of LUCCN’s ground measurements, we identify and present consistent trends between local CO₂ concentration variations and aerosol pollution events captured by the space-based moderate resolution imaging spectrometer (MODIS).