Descriptive and Diagnostic Analysis of NASA and NiMet Big Weather Data

Abstract

In this paper, descriptive and diagnostic analysis of the National Aeronautics and Space Administration (NASA) weather data predictions and the Nigerian Meteorological Agency (NiMet) ground truth weather data is presented. Also, the prediction accuracy of the NASA weather data with respect to the NiMet weather data is evaluated. This research was based on ninety-two months of daily weather data from NiMet (January 1, 2010 to August 14, 2017) alongside corresponding NASA data, with specific consideration on minimum and maximum temperature meteorological parameters. Using time series plots, descriptive analysis was implemented to identify patterns and trends, while the diagnostic analysis shows the basis for the patterns and trends. The time series plots show that the patterns and trends can be attributed to seasonal weather changes in the study area. Furthermore, statistical evaluation of the NASA predictions with respect to the NiMet ground truth data resulted in a root mean square error (RMSE) of 1.2, and correlation coefficient value of 0.6, for minimum temperature, then, RMSE of 2.6, and correlation coefficient value of 0.7, for maximum temperature. Evidently, the NASA prediction was able to better predict the study area minimum temperature compared to the maximum temperature. However, the correlation analysis clearly establishes a good correlation between the NASA and the NiMet weather data, which indicates that a model can be developed to predict more accurate ground truth weather data, using the NASA data as input feature.