Field observation of H2O2 was conducted at an urban site in Beijing from Sep.23rd to Nov.4th of 2018. The daily peak concentrations of H2O2 and organic peroxides (OPs) during the measurement were 0.57 ppbv (Range: 0.07–1.89 ppbv) and 0.16 ppbv (Range 0.06–0.35 ppbv). Using an observation based model, the modeled HO2 level reproduced a much lower H2O2 production rate than the observation, especially in the morning under the weak wind, non-turbulent surface winds conditions. With the modeled-HO2 concentration corrected based on the previously reported observed/modeled HO2 ratio in this city, the daily integrated H2O2 production rate increased by 144%–164%, which is important but still not enough to explain the observed high H2O2 in the lightly polluted and polluted days. Heterogeneous production of H2O2 on aerosols and the overestimation of H2O2 removal rates were examined not the underlying causes of this additional H2O2 missing source. H2O2 concentrations in residual layers were evaluated. The residual layer was revealed a stronger reservoir of H2O2 than its reservoir effect on O3. In the morning of non-turbulent conditions at the urban site, the residual layers H2O2 were estimated about 20–80 times of the ground surface H2O2 before the air layers vertical mixing. The vertical mixing significantly affects gaseous H2O2 concentration on the ground in the morning. This work highlights the importance to consider the HO2 underestimation and air mass vertical mixing in atmospheric H2O2 studies.