An Analysis of Multi-Source Temperature Datasets using Statistical Techniques

Abstract

Temperature is considered as one of the important weather parameters so a qualified study for multi-source dataset is carried out by inspecting different parameters. The maximum temperature (T_max) and minimum temperature (T_min) for the Haridwar district are taken from two distinct sources, including data from the Indian Meteorology Department (IMD) and the National Aeronautics and Space Administration (NASA). Initially, NASA Power Larc and IMD datasets are compared using Standard Anomaly graphical representation. Afterwards, different indices were evaluated for Maximum and Minimum Temperature such as Bias, mean absolute error, mean square error, root mean square error. Further, correlation coefficient analysis & Wilcoxon–Mann–Whiney test is carried out to test the equivalence of both datasets. The outcome shows that the yearly standard anomalies of the Power Larc data and the annual standard anomalies of the IMD data both follow the same trend. According to the various indices assessed, MAE, MSE, and RMSE all remain in one standard deviation of the data being observed. Moreover, Power Larc data is well correlated and shows equivalence with IMD dataset. It is observed that both datasets show very strong (0.87) to weak correlation (0.37) for minimum temperature on seasonal scale and strong correlation (0.71) on annual scale. For maximum temperature both datasets show a correlation range of 0.41–0.83 on seasonal scale and weak (0.37) correlation on annual scale. From the correlation values it is observed that both datasets are identical on monthly and seasonal scale. Therefore, it may be concluded that Power Larc dataset is reliable dataset and may be used in place of IMD dataset.