Efficient GPU-Based Parallel Kriging Algorithm for Predicting the Air Quality Index

Air quality index (AQI) is an evaluation standard of air quality and a major concern in the daily lives of people. However, predicting the AQI is difficult; existing prediction methods are few and they cannot satisfy the requirements of real-time prediction. This paper proposes an efficient GPU-based parallel Kriging algorithm (GP-Kriging) for real-time AQI prediction. Parallel computing strategy is used in the sub-steps of the serial Kriging algorithm, and some steps are designed and implemented in GPU to accelerate the computing speed. The data sets of the experiments are collected from Beijing Environmental Monitoring Center. Experimental results show that the time requirement of the GP-Kriging algorithm is approximately 20 times faster than that of serial Kriging. The successful application of the GP-Kriging algorithm in this study suggests that the method can be used to predict AQI.