Advancing COVID-19 poverty estimation with satellite imagery-based deep learning techniques: a systematic review

In today’s world, where the global population is expanding at an unprecedented rate, addressing the challenge of poverty has become more critical than ever before. In the wake of the COVID-19 pandemic, the issue of poverty has taken on renewed urgency as communities worldwide grapple with the socioeconomic fallout. Amidst the difficulties posed by the COVID-19 crisis, innovative approaches are required to address the evolving nature of poverty in the context of a pandemic-stricken world. This study systematically reviews the usage of machine learning (ML) and deep learning (DL) methods for COVID-19 poverty estimation using satellite imagery, emphasizing the need for innovative approaches due to the growing global population and poverty levels. It assesses how ML and DL leverage diverse data sources, such as mobile phone records, satellite imagery, and household surveys, to identify poverty indicators and enhance analysis precision. This study identifies challenges, including data availability and model biases, and suggests future directions focusing on dynamic models and multidimensional COVID-19 poverty assessment. It highlights the implications for spatial information science, advocating for improved data integration and model transparency to support effective COVID-19 poverty alleviation policies.