Temperature prediction based on a space-time regression-kriging model.

Area of Science:

• Environmental Science
• Geospatial Analysis
• Statistical Modeling

Background:

• Spatio-temporal phenomena often suffer from low data sampling rates and sparse observation networks.
• Accurate interpolation is crucial for understanding and predicting these phenomena.

Purpose of the Study:

• To introduce and apply a novel space-time regression-kriging model for accurate spatio-temporal interpolation.
• To evaluate the model's performance against traditional time forecasting methods.

Main Methods:

• Applied time series decomposition and multiple linear regression for space-time trend fitting.
• Utilized a nonseparable spatio-temporal variogram function to model residual similarities.
• Implemented space-time kriging for monthly air temperature prediction, validated with jackknife techniques.

Main Results:

• Achieved high correlation coefficients (close to 1) between predicted and observed monthly temperatures.
• Demonstrated significantly lower Mean Absolute Error (MAE) and Root-Mean-Square Error (RMSE) compared to Autoregressive Integrated Moving Average (ARIMA) models.
• Observed conspicuous improvements in interpolation accuracy using the space-time kriging approach.

Conclusions:

• The developed space-time regression-kriging model offers superior accuracy for spatio-temporal interpolation of environmental data.
• This method provides a robust framework for analyzing and predicting phenomena with sparse spatio-temporal observations.
• The findings highlight the limitations of pure time forecasting for complex spatio-temporal datasets.