LAUSR

Abstract In the present study, MOZART-4 (Model for Ozone and Related chemical Tracers-Version-4) simulation has been made from 2003 to 2007 and compared with satellite and in-situ observations with a specific focus on Indian subcontinent to illustrate the capabilities of MOZART-4 model. The model simulated CO have been compared with latest version (version-6) of MOPITT (Measurement Of Pollution In The Troposphere) carbon monoxide (CO) retrievals at 900, 800 and 700 hPa. Model reproduces major features present in satellite observations. However model significantly overestimates CO over the entire Indian region at 900 hPa and moderately overestimates at 800 hPa and 700 hPa. The frequency distribution of all simulated data points with respect to MOZART error shows maximum in the error range of 10–20% at all pressure levels. Over total Indian landmass, the percentage of gridded CO data that are being overestimated in the range of 0–30% at 900 hPa, 800 hPa and 700 hPa are 58%, 62% and 66% respectively. The study reflects very good correlation between two datasets over Central India (CI) and Southern India (SI). The coefficient of determination (r 2 ) is found to be 0.68–0.78 and 0.70–0.78 over the CI and SI respectively. The weak correlation is evident over Northern India (NI) with r 2 values of 0.1–0.3. Over Eastern India (EI), Good correlation at 800 hPa (r 2  = 0.72) and 700 hPa (r 2  = 0.66) whereas moderately weak correlation at 900 hPa (r 2  = 0.48) has been observed. In contrast, Over Western India (WI), strong correlation is evident at 900 hPa (r 2  = 0.64) and moderately weak association is found to be present at 800 hPa and 700 hPa. Model fairly reproduces seasonal cycle of CO in the lower troposphere over most of the Indian regions. However, during June to December, model shows overestimation over NI. The magnitude of overestimation is increasing linearly from 900 hPa to 700 hPa level. During April–June months, model results are coinciding with observed CO concentrations over SI region at 900 hPa. Model simulation has been compared with surface in-situ observations over ten Indian locations. Model performance is found to be moderate to good over various observational locations. However, over highly polluted megacities, model underestimates observed CO concentration by up to 3500 ppbv. A case study over the forest fire prone area reveals the clear increase of modeled and retrieved CO in February–March and a decrease in May which is coinciding with biomass burning emissions and fire counts. Model performance is found to be relatively poor over this region with r 2 of 0.29 and slope of 0.56.