Türkiye’s GHG Emissions Towards the 2030 Target: MDAM and LSTM-Based Analysis with Key Energy Factors

Abstract

Greenhouse gas (GHG) emission, caused by heat-trapping gases in the atmosphere, is a major driver of global warming. The World Economic Forum highlights rising emissions, ecosystem degradation, and climate-related disasters as long-term threats to global stability. The accurate modeling and prediction of GHG emissions are crucial for evidence-based climate governance. This study investigates Türkiye’s GHG emissions by comparing two advanced time-series models: a deep learning-based Long Short-Term Memory (LSTM) network and the Multi Deep Assessment Model (MDAM), which incorporates Caputo fractional derivatives. Data from the Turkish Statistical Institute (TurkStat) and the Turkish Electricity Transmission Corporation (TEIAŞ) were used covering the period between 1993 and 2022. These datasets include information on GHG emissions, fossil-based generation, renewable energy, and electricity demand. Both models were trained to predict 2030 emissions and assess the contributing factors. Results show that MDAM achieved superior accuracy compared to LSTM. Both models projected 2030 emissions above Türkiye’s 695 Mt target, with 721.87 Mt (MDAM) and 709.49 Mt (LSTM), highlighting the need for stronger policy action. A no-COVID scenario yielded higher forecasts, confirming the pandemic’s suppressive effect. Impact factor analysis revealed that domestic electricity demand and fossil-based generation are the strongest drivers, while renewables mitigate emission.