Research On Land Surface Temperature Time Series Analysis Methods

LST as an important indicator of evaluation and assessment of the surface balance,its analysis allows people to better understand the variation of surface temperatures.So the surface temperature trends and studies law has very important significance.However, in the application of exponential smoothing, ARMA and ARIMA models and other methods of surface temperature time series analyzes of inadequate research,and there is a lag in the statistical analysis and forecasting results forecast, the larger multi-step prediction error and poor sequence analysis of the effect of non-stationary problems. To solve these problems, the paper ARIMA model and single exponential smoothing method is improved, the specific research content and research work includes the following aspects:(1) Application of ARIMA model of surface temperature time series analysis and forecasting. First check whether the acquired data points have large error exists, then white noise test, research and analysis by selecting a suitable model from the AR(p)model, MA(q) model and the ARMA(p, q) model and then determine the order elected suitable model, least squares method of surface temperature time series data parameter estimation. In this paper, to overcome the difficulties in surface temperature time series model pricing model to estimate other aspects of the experimental results show that the model can predict the surface temperature time series, Theil coefficient ranging from two experiments were 0.47 and 0.449,respectively, co-variance ratio 0.76 and 0.711, indicating that the model has satisfactory predictions.(2) of the ARIMA model is improved. First time series at multiple time points were analyzed, and then apply formula equidistant nodes to improve ARIMA(p, d, q)model. Experimental results show that the improved model Theil coefficient of0.264582 than the traditional model decreased 55.64%, indicating that the unit root mean square error is smaller, closer to the actual value of the predicted value; thecovariance ratio of 0.854773 than the traditional model the 0.763663 increased by11.93%, indicating that non-systematic error is larger, ARIMA model improved with better prediction results. It said improvements ARIMA model to predict surface temperature time series analysis provides a new way of thinking.(3) will apply exponential smoothing short-term analysis and forecast surface temperature time series data, the Holt-Winters no seasonal model analysis and prediction experiments through continuous tests to determine the damping factor,predictive formula. Two experiments results Holt-Winters seasonal model no residual sum of squares were 13.78303,12.23737, respectively, lower than the first exponential smoothing 15.05871,16.97791 secondary exponential smoothing 14.68269,18.75255;Holt-Winters no seasonal root mean square error of prediction models were0.677816,0.638680, respectively, lower than the first exponential smoothing0.708489,0.752283 secondary exponential smoothing 0.699588,0.790623. Studies have shown that Holt-Winters no seasonal model accuracy of analysis and forecasting short-term time-series data of the surface temperature is higher.(4) single exponential smoothing method is improved. For residual traditional single exponential smoothing variation contains only information historical data values, without taking into account the impact of the back of already acquired data values this limitation, it has been improved, single exponential smoothing proved the improved method Ward decreased 22.73% than the traditional single exponential smoothing, root mean square error of prediction than the traditional single exponential smoothing reduces 15.49%. Studies have shown that improved prediction of single exponential smoothing method is more accurate.(5) improved by ARIMA model and Holt-Winters seasonal models predict no comparative analysis of the results, the improved single exponential smoothing and Holt-Winters seasonal models predict no comparative analysis of the results and the ARIMA model improved ARIMA model and the improved single exponential smoothing forecast results of comparative analysis found that the ARIMA model improved and improved single exponential smoothing method is higher than the conventional exponential smoothing prediction accuracy of the other models, andimproved predictive value and predictive value than the improved single exponential smoothing method is closer to the real value, more suitable for surface temperature time series analysis and forecasting.