MICS-Asia II: Model intercomparison and evaluation of particulate sulfate, nitrate and ammonium

Eight chemical transport models participate in a model intercomparison study for East Asia, MICS-Asia II. This paper analyzes calculated results for particulate matter of sulfate, nitrate and ammonium through comparisons with each other and with monthly measurements at EANET (the acid deposition monitoring network in East Asia) and daily measurements at Fukue, Japan. To the EANET measurements, model ensemble means better agree with model individual results for sulfate and total ammonium, although total nitrate is consistently and considerably underestimated. To measurements at Fukue, the models show better agreement than for the EANET measurements. This is likely because Fukue is centered in many of the model domains, whereas the EANET stations are mostly in Southeast Asia and Russia. Moreover, it would be important that Fukue is in Northeast Asia, where the emission inventory is more reliable than Southeast Asia. The model-model comparisons are made in view of the total amount in the atmosphere, vertical profile, coefficient of variation in surface concentrations, and transformation changes with distance. All the models show reasonable tendencies in vertical profiles and composition ratios. However, total amounts in the atmosphere are discrepant among the models. The consistency of the total amount in the atmosphere would influence source-receptor analysis. It seems that model results would be consistent, if the models take into account the primitive processes like emission, advection/diffusion, chemical transformation and dry/wet deposition, no matter the processes are modeled simply or comprehensively. Through the comparison study, we learned that it would be difficult to find any problems from one comparison (model-observation comparison with one data or many but at one station or in a short period). Modelers tend to examine model performances only from model-observation comparisons. However, taking budget in a certain or whole model domain would be important, before the models are applied to source-receptor analysis.