The impact of the February 6, 2023 earthquakes on natural gas consumption: Case of Malatya

The earthquakes that struck Kahramanmaraş on 6 February 2023 not only caused severe physical destruction but also reshaped the energy consumption patterns of the affected cities. This thesis examines the impact of the disaster on natural gas consumption in Malatya and estimates the demand level that would have occurred in the absence of the earthquake, using an LSTM-based time series forecasting approach. The analysis relies on monthly data from 2019 to 2023, including gas consumption, number of subscribers, mean temperature, HDD/CDD indices, public holiday counts, and lagged consumption variables. Seasonal patterns were captured through trigonometric transformations, moving averages, and lag features. The model was trained with modern optimization techniques such as EarlyStopping and ReduceLROnPlateau, achieving a strong performance with R² = 0.96. Counterfactual results indicate that gas consumption for the February-December 2023 period would have reached 329.5 million m³ under normal conditions, whereas the actual consumption amounted to 246.6 million m³ which is a substantial 25% decline. The magnitude of this reduction varied across customer groups: industrial zones experienced a relatively modest decrease, while residential and service-sector subscribers faced much sharper drops. Spatial analyses further reveal that the recovery of consumption patterns did not progress uniformly across neighborhoods; in several areas, the post-disaster normalization took significantly longer. Overall, the findings demonstrate that large-scale disasters disrupt energy demand both temporally and spatially, while the LSTM framework provides a robust tool for understanding such structural breaks and supporting post-disaster tariff planning, budgeting, supply security, and risk management.