Global Warming: A study of erratic fluctuations in land and oceanic temperature as influenced by heightening industrial activity and population growth

Abstract

Tackling the global climate crisis is a daunting task. Excessive greenhouse gas emissions lead to rising temperatures and sea levels, stratospheric ozone depletion and a greater occurrence of intense, sporadic natural disasters. In order to commence the development of remedial technologies to subdue the adverse effects of global warming, it is essential to understand the range of anthropogenic and industrial activities that cause global land and oceanic temperatures to fluctuate beyond their expected statistical range. The following paper employs a myriad of machine learning algorithms in the development of deep neural networks (DNNs) that are used to forecast land and oceanic temperatures. Data input during the model’s training process, (covering an overall sample period of 28 years from 1990 - 2018), consists of sectoral carbon emission values of the top 20 contributors to the climate crisis worldwide—as well as additional correlated factors, such as population growth rate and gross domestic product (GDP) growth per capita, incorporated to create a comprehensive and versatile predictive tool. Results indicate high predictive accuracy of the model, with a low mean square error (MSE) of 0.208 and a high coefficient of determination (r²) of 98.7%, implying a high degree of explainability regarding variations in predictions.