On June 10, 2024, Sugandha Doda defended her thesis entitled Population Estimation Utilizing Earth Observation Data, supervised by Prof. Dr. Xiaoxiang Zhu. The defense committee was chaired by Prof. Dr. Martin Werner. We particularly appreciate the engagement of the examiners Guest Prof. Dr. Yuanyuan Wang and Prof. Dr. Monika Kuffer from the University of Twente!
Congratulations, Dr. Doda!
The rapid population growth demands additional infrastructure, food, water, healthcare, and schools, putting pressure on the environment. Accurate population data is essential for effective urban planning and policy-making. Traditional census methods are costly and offer low spatial resolution. Modern statistical and machine-learning methods provide better population estimates but often depend on existing census data, limiting their transferability and accuracy.
This Ph.D. thesis addresses these challenges by creating a large-scale benchmark data set using publicly available data, facilitating more accurate and transferable population estimation. An end-to-end deep learning framework is developed to predict populations without relying on census data, ensuring easy transferability and improved accuracy. The model's performance is compared with standard products for cities in Europe and the US, and an explainable AI technique is used to understand the model's decisions, enhancing trustworthiness.
The thesis also focuses on improving data resolution by using a hybrid deep learning approach that generates gridded population maps and disaggregates counts to buildings. This method integrates building data to enhance accuracy and resolution. The study includes a qualitative comparison of building data sources and examines the impact of data quality on fine-scale population estimates.
References:
Doda, Sugandha, et al. "So2sat pop-a curated benchmark data set for population estimation from space on a continental scale." Scientific data 9.1 (2022): 715.
Doda, Sugandha, et al., "Interpretable deep learning for consistent large-scale urban population estimation using Earth observation data", International Journal of Applied Earth Observation and Geoinformation 128, 103731, 2024.