Fitting a Nonlinear Curve to Temperature Data and the Accuracy Estimation of the Curve

Since June 2000, we have been observing the temperature of some points continuously to estimate the regional temperature change and also the paddy field effect to the boundry temperature. The temperatures were evaluated every 10 minutes, so that the data size is 144 per day and 4320 (=144x30) per month. From these data, we derived a maximum temperature series, a minimum temperature series and a mean temperature series of the target month and year. For each series of the month, the size is 144, representing the maximun, the minimum and the mean temperature changes of the month respectively. We have proposed a nonlinear model for these data and give a practical method to evaluate the coefficients of the model. A fitness criterion between the model and the obtained nonlinear curve is presented. The efficiency of the criterion is discussed comparing it with the multiple correlation coefficient R(2).