Data-Based Models for Global Temperature Variations

This paper presents two data-based models for the measured time series of global annual average tropospheric temperatures. One model is a smoothing spline fit designed to give an optimal separation of signal from noise. The other combines an optimal spline fit to the measured record of CO2 concentration in the atmosphere with a previously reported 70 year cycle in the temperatures. It assumes a simple linear relation between changes in temperature and changes in the CO2 concentration. When the cycle is added to the model, its fit to the temperature data is very similar to the optimal spline fit. The differences between the two fits are smaller in magnitude than the residuals for either one of them.