Using satellite imagery to understand and promote sustainable development

Area of Science:

• Environmental science
• Data science
• Sustainable development studies

Background:

• Accurate measurement of sustainable development outcomes is crucial for research and policy.
• Satellite imagery is increasingly used to monitor human-related development indicators.
• There is a growing need to integrate satellite data with advanced analytical techniques.

Purpose of the Study:

• To synthesize literature on using satellite imagery and machine learning for sustainable development outcomes.
• To assess the impact of data scarcity and quality on machine learning model performance.
• To highlight applications, constraints, and future research directions in this interdisciplinary field.

Main Methods:

• Systematic literature review of studies combining satellite imagery and machine learning.
• Analysis of trends in satellite data resolution (spatial, temporal, spectral).
• Evaluation of machine learning approaches for handling scarce and noisy training data.

Main Results:

• Quantified the limited availability of ground data for human-related outcomes.
• Documented the increasing abundance and improving resolution of satellite imagery.
• Demonstrated how data noise can lead to inaccurate assessments of machine learning model performance.
• Quantified recent model performance across various sustainable development domains.

Conclusions:

• Satellite imagery and machine learning offer significant potential for monitoring sustainable development.
• Addressing data scarcity and noise is critical for reliable assessments.
• Further research is needed to refine models and expand applications in policy and practice.