中文摘要：

基于广东省环境信息综合发布平台发布的珠三角9城市空气质量日报数据构建VAR模型，通过格兰杰检验和脉冲响应分析，探讨2002—2012年珠三角9城市空气质量的区域影响特征。结果显示：珠三角9城市间空气质量的区域影响包括自我影响和跨域影响。且城市个体的自我影响是负效应，其强度远远超过跨域影响。在跨域影响中单向性影响占比84％，双向性影响占比16％，邻域影响占42％，非邻域影响占比58％。南部沿海城市影响范围普遍大于北部城市的影响范围。具有因果关系的城市间的跨域影响中，65％表现为正效应，22％为负效应，有13％显示为正负效应转换。而正负效应转换的情况说明存在两种不同的影响模式。

英文摘要：

The Interaction of air quality across the cities exists objectively as a consequence of air pollution spread around. It is meaningful to do some research on regional effects in which both cross -border and intra -city effect involved at the level of administrative units for government making the environmental governance policy and build- ing the mechanism of ecological compensation between the different administrative districts. The article builds the VAR model composed of nine cities in the Pearl River Delta, based on the daily data of API of the cities from 2002 to 2012 year, which is from the Guangdong Province Environmental Information Integrated publishing plat- form. The results of the Granger Causality test and the impulse response analysis show there exist interaction of air quality intra - city and inter - city. The interaction degree of intra - city is generally stronger than that of inter - cit- y. It is seen that unidirectional influence is high up to 84% and the bi -directional effect only accounts for 16% ~ The cross -border effects of air quality are happened between the two cities no matter whether they share the common geographical boundary or not. However, such effects occur much more in non - adjacent domain than in adjacent domain. The proportion of the former is 58% while of the later is 42%. Much more cities are influenced by the cities on the south of Pearl River Delta, rather than by those on the north of the region~ The regional effects of those cities which satisfied the test of Granger Causality are identified to three kinds which positive, negative or transforming between positive and negative effect, and their proportions are 65%, 22% 13% in proper sequence. Specially, the transforming between positive and negative effect reveals the two different interaction processes.