A mixed strategy for estimating pollution series at non-monitored sites

Air quality is a core topic, because air pollution is one of the most important pollution problems in the world, especially in large cities. In particular, predicting or detecting a future extreme air pollution episode or predicting the violation of an air quality standard, is of particular interest in the field of pollution control.

This is why in this article we propose a new strategy, a threshold autoregressive asymmetric stochastic volatility strategy (TA-ARSV), for alerting immediate violations of the quality standards. It takes into account the different answer of the volatility to a positive or negative, but equal in magnitude, relative variation of the level of the pollutant in the previous period.  But, given that the TA-ARSV strategy only provide alerts at the locations where the monitoring stations are sited, it is combined with a double kriging procedure carried out over the area under study. On the one hand, the map of coefficients of asymmetric answer of volatility is obtained by using a univariate punctual kriging; on the other hand, a functional kriging strategy will provide the map of pollutant curves. Then, provided that both curves and coefficients of asymmetry have been estimated all over the area under study, TA-ARSV models will be able to easily and accurately predict violations of the standard.This novel approach has been applied in the city of Madrid (Spain) to particulate matter (PM10) levels, the reason being that PM10 is one of the most dangerous pollutants and Madrid is a large city with a high ratios monitoring stations per inhabitant and square kilometer