Finding optimal meteorological observation locations by multi-source urban big data analysis

In this paper, we try to solve site selection problem for building meteorological observation stations by recommending some locations. The functions of these stations are meteorological observation and prediction in regions without these. Thus in this paper two specific problems are solved. One is how to predict the meteorology in the regions without stations by using known meteorological data of other regions. The other is how to select the best locations to set up new observation stations. We design an extensible two-stage framework for the station placing including prediction model and selection model. It is very convenient for executives to add more real-life factors into our model. We consider not only selecting the locations that can provide the most accuracy predicted data but also how to minimize the cost of building new observation stations. We evaluate the proposed approach using the real meteorological data of Shaanxi province. Experiment results show the better performance of our model than existing commonly used methods.