Analysis of Average Air Temperature in New York based on SARIMA Model

Abstract

Temperature prediction is one of the most vital part of weather forecasting. The temperature change has a great impact on agricultural, environmental evolution, human activities etc. The monthly average temperature of the Central Park Meteorological Observatory from 1869 to 2020 is the data to get sorted and analyzed. The SARIMA model is selected to forecast and analyze the monthly mean temperature of New York City. The first step is using the ADF test to check whether the original time series are basically stable, and through the test, the period was obviously 12. Then, by comparing with the data of first-order difference and first-order seasonal difference, the AIC value of the original time series was the smallest, it is no need to carry out the difference. The function in R is used to determine the order, and it is found that SARIMA (1,0,2) (2,1,0,12) model has the best fitting effect. Meanwhile, the residual is no autocorrelation of white noise. Finally, the average monthly temperature from 2010 to 2020 is predicted based on the original data. The difference between the predicted temperature and the actual temperature in most months is less than 5%, and the model is basically consistent with the reality. However, during the counting of the average annual temperature, the prediction results are almost unchanged after using the monthly data. This situation is quite different from the current state.