Haze pollution is a serious interregional problem for many countries and wind direction plays a crucial role in the process of pollution transport. Upwind cities have obviously greater impacts on downwind cities due to atmospheric transport. Hence the directions of pollution spillover effects should be emphasized in empirical study to avoid model bias. With the panel data of 44 cities in North China Plain from 2013 to 2017, we constructed a novel wind direction weight matrix to analyze the spatial variations of PM_2.5 concentration. The matrix was then incorporated into a spatial panel model to quantitatively evaluate the impacts of socioeconomic and natural factors on PM_2.5 concentration. The results show that wind direction dominates the distribution of PM_2.5 concentration. The growth of per capita GDP facilitates the reduction of PM_2.5 pollution while the increase of the other socioeconomic factors aggravates haze pollution. Besides, natural factors directly or indirectly affect PM_2.5 concentration. Particularly, the spillover effects of socioeconomic factors are greater than their local effects. Based on the results, we suggested that investment of air pollution control in a neighboring area may be more effective than in the local city itself. The mechanism of pollutant transport should be fully considered in the fields such as the construction of urban air ducts, industrial layout, and eco-compensation.

烟霾污染是许多国家的严重区域间问题，风向在污染运输过程中起着至关重要的作用。由于大气运输，上风城市对下风城市的影响显然更大。因此，在经验研究中应强调污染溢出效应的方向，以避免模型偏差。利用2013年至2017年华北平原44个城市的面板数据，我们构建了一个新颖的风向权重矩阵来分析PM _2.5浓度的空间变化。然后将矩阵合并到空间面板模型中，以定量评估社会经济因素和自然因素对PM _2.5浓度的影响。结果表明，风向主导了PM _2.5的分布浓度。人均国内生产总值的增长有助于减少PM _2.5污染，而其他社会经济因素的增加则加剧了霾污染。此外，自然因素直接或间接影响PM _2.5浓度。特别是，社会经济因素的溢出效应大于其局部效应。根据结果​​，我们建议在邻近地区进行空气污染控制投资可能比在当地城市中更为有效。在城市风管建设，产业布局，生态补偿等领域应充分考虑污染物的传播机理。