Nightlight as a Proxy of Economic Indicators: Fine-Grained GDP Inference around Chinese Mainland via Attention-Augmented CNN from Daytime Satellite Imagery

Abstract

The official method of collecting county-level GDP values in the Chinese Mainland relies mainly on administrative reporting data and suffers from high costs of time, money, and human labor. To date, a series of studies have been conducted to generate fine-grained maps of socioeconomic indicators from the easily accessed remote sensing data and achieved satisfactory results. This paper proposes a transfer learning framework that regards nightlight intensities as a proxy of economic activity degrees to estimate county-level GDP around the Chinese Mainland. In the framework, paired daytime satellite images and nightlight intensity levels were applied to train a VGG-16 architecture, and the output features at a specific layer, after dimensional reduction and statistics calculation, were fed into a simple regressor to estimate county-level GDP. We trained the model with data of 2017 and utilized it to predict county-level GDP of 2018, achieving an R-squared of 0.71. Furthermore, the results of gradient visualization confirmed the validity of the proposed framework qualitatively. To the best of our knowledge, this is the first time that county-level GDP values around the Chinese Mainland have been estimated from both daytime and nighttime remote sensing data relying on attention-augmented CNN. We believe that our work will shed light on both the evolution of fine-grained socioeconomic surveys and the application of remote sensing data in economic research.

Using Multi-Source Nighttime Lights Data to Proxy for County-Level Economic Activity in China from 2012 to 2019

“…The use of nighttime lights (NTL) data to proxy for economic activity is well-established in remote sensing and other disciplines [1][2][3][4][5][6][7][8][9][10]. This proxy enables research when traditional economic activity data, such as Gross Domestic Product (GDP), are either absent or are not trusted because of concerns about either measurement error or manipulation [11][12][13].…”

Global GDP Prediction With Night-Lights and Transfer Learning

“…This is a textbook use case for the application of machine learning, particularly deep learning (DL) [18] [19]. The relevance of daytime satellite sensor imagery to the task at hand has been confirmed by Liu et al [20] who used a convolutional neural network CNN trained on such imagery to predict GDP at the county level for China (R 2 = 0.71) and Khachiyan et al [21] who paired a similar CNN with socioeconomic indicators to predict total personal income at fine spatial resolutions in the USA (R 2 = 0.90), though neither of these examples took advantage of nightlights data.…”

Integrating Street Views, Satellite Imageries and Remote Sensing Data Into Economics and the Social Sciences

“…For example, researchers have used satellite imagery to examine the relationship between nighttime lights and economic growth, finding that the percentage of nighttime light-covered areas strongly correlates with economic growth. Bickenbach et al, 2016;Bluhm & Krause, 2022;Cauwels et al, 2014;Chanda & Cook, 2022;Chen et al, 2022;Elvidge et al, 2012;Henderson et al, 2012;Hu & Yao, 2022;Liu et al, 2021;McCallum et al, 2022;McCord & Rodriguez-Heredia, 2022;Pérez-Sindín et al, 2021;Weidmann & Schutte, 2017;Xu et al, 2021;Yeh et al, 2020 used nighttime light estimations to construct a wealth index to measure economic well-being and provide a decision-making basis for areas facing impoverishment. Similarly, Gibson et al (2020), Gibson et al (2021), Goldblatt et al (2020), Luu et al (2019), Maldonado (2022), Mellander et al (2015) and Proville et al (2017) explored the relationship between nighttime lights and economic activity, finding that differences in lights intensity relate to variations in economic activity.…”