One of the most significant challenges faced by disaster preparedness and response is creating efficient strategies to mitigate the impact of catastrophic events. The phases of disaster planning and strategy development usually necessitate extensive cooperation among communities and individuals. As such, both online and offline social network dynamics play a vital role in the formation of effective disaster management plans. In this paper, we focus on efficient community detection in disaster networks. The input network graph, which is usually a dense graph, is initially sparsified through a spectral-based edge sampling technique. This sparsification procedure is applied to remove less important edges while preserving those critical for maintaining a modular structure, aiming at realizing a faster, greener and reliable community detection. Next, we run a genetic algorithm-based community detection method that maximizes modularity as the objective function on the sparsified graph which is a proxy of the initial graph. Experimental results on synthetic networks demonstrate the effectiveness of our approach compared to other benchmark methods.