Inconsistencies in air quality metrics:: 'Blue Sky' days and PM10 concentrations in Beijing

International attention is focused on Beijing's efforts to improve air quality. The number of days reported as attaining the daily Chinese National Ambient Air Quality Standard for cities, called 'Blue Sky' days, has increased yearly from 100 in 1998 to 246 in 2007. However, analysis of publicly reported daily air pollution index (API) values for fine particulate matter (diameter <= 10 mu m, PM10), indicates a discrepancy between the reported 'Blue Sky' days (defined as API <= 100, PM10 <= 150 mu g m(-3)) and published monitoring station data. Here I show that reported improvements in air quality for 2006-2007 over 2002 levels can be attributed to (a) a shift in reported daily PM10 concentrations from just above to just below the national standard, and (b) a shift of monitoring stations in 2006 to less polluted areas. I found that calculating daily Beijing API for 2006 and 2007 using data from the original monitoring stations eliminates a bias in reported PM10 concentrations near the 'Blue Sky' boundary, and results in a number of 'Blue Sky' days and annual PM10 concentration near 2002 levels in 2006 and 2007 (203 days and similar to 167 mu g m(-3) calculated for 2006-38 days fewer and a PM10 concentration similar to 6 mu g m(-3) higher than reported; 191 'Blue Sky' days and similar to 161 mu g m(-3) calculated for 2007-55 days fewer and a PM10 concentration similar to 12 mu g m(-3) higher than reported; 203 days and 166 mu g m(-3) were reported in 2002). Furthermore, although different pollutants were monitored before daily reporting began and less stringent standards were implemented in June 2000, reported annual average concentrations of particulate (diameter <= 100 mu m, TSP) and nitrogen dioxide (NO2) indicate no improvement between 1998 and 2002. This analysis highlights the sensitivity of monitoring data in the evaluation of air quality trends, and the potential for the misinterpretation or manipulation of these trends on the basis of inconsistent metrics.