Forecasting of Meteorological variables using statistical methods and tools

TL;DR

This paper reviews statistical methods for meteorological forecasting, applies them to Nigerian data, and demonstrates significant temperature increases and the effectiveness of regression models in predicting refractivity.

Contribution

It provides an in-depth review of statistical techniques like Mann-Kendall tests and regression models, and applies them to real meteorological data with practical results.

Findings

Maximum ambient temperature in Calabar is increasing significantly.

Both dry and wet season trends show increasing meteorological parameters.

Regression models effectively predict refractivity with minimal residual error.

Abstract

The need to understand the role of statistical methods for the forecasting of climatological parameters cannot be trivialized. This study gives an in depth review on the different variations of the Mann-Kendall (M-K) trend test and how they can be applied, regression techniques (Simple and Multiple), the Angstrom-Prescott model for solar radiation, etc. The study then goes ahead to apply some of them with data obtained from the Nigerian Meteorological Agency (NiMet), and applying tools like the python programming language and Wolfram Mathematica. Results show that the maximum ambient temperature for Calabar is increasing (Z=2.52) significantly after the calculated p-value < 0.05 (significant level). The seasonal M-K test was also applied for the dry and wet seasons and both were found to be increasing (Z=3.23 and Z=4.04 respectively) after their calculated p-values < 0.05. The relationship between refractivity and other meteorological parameters relating to it was discerned using partial differential equations giving the gradient of each with refractivity; this was compared with results from the correlation matrix to show that the water vapour contents of the atmosphere contributes significantly to the variation of refractivity. Multiple linear regression has also been adopted to give an accurate model for the prediction of refractivity in the region after the residual error between the calculated refractivity and predicted refractivity was minimal.