Economy Estimation of Mainland China at County-
Level Based on Landsat Images and Multi-Task
Deep Learning Framework
Bo Yu, Ying Dong, Fang Chen, Yu Wang
Abstract
The social-economic statistics collected from local govern-
ments are the main access for the central government to
achieve national economic circumstance, especially for
China. However, the statistics of almost 10% of national
counties are missing or inconsistent due to the statistical
caliber change in the wave of urbanization during economic
development. Some researchers proposed to apply a night
luminosity product to solve such issue. However, it lacks the
ability to distinguish between the wealthy populations with
a dense distribution and the less developed places. In this
paper, the publicly available daytime Landsat images are
used to estimate economic statistics. An end-to-end multi-
task deep learning framework is constructed to estimate the
county-level economy of Mainland China and the overall
accuracy of this model achieves higher than 85%. The experi-
ments show that our model provides a potential strategy to
make up for the missing statistics and examines the reli-
ability of the statistics collected for the central government.
Introduction
Data, which have penetrated into all walks of life, have be-
come an important factor that drives production in our study,
work, life, and social development. Socioeconomic data,
especially in economy, education, and agriculture, are the key
factors in measuring development level (Cheng
et al.
2018).
With well-documented socioeconomic data, governments can
make accurate evaluations and predictions, and put forward
reasonable policies to ensure rational allocation of public
resources to realize the more balanced and full development
goal especially under the unstable social and economic situa-
tion. Moreover, detailed data record supplements internation-
al researchers or institutions to understand economic status
of different places from different aspects (Puzo, Mehlum, and
Qin 2018). Despite the significant work in collecting data,
improving data quality, and distributing data, actions mo-
bilized by global governments, companies, and researchers,
there are still many places suffering from a lack of socioeco-
nomic data or the lack of accessibility to the data (Ieag 2014).
Such inaccessibility hinders objective economy analysis for
the developing or undeveloped places, which should have
needed more resources or pertinent strategies to grow.
China has spent billions of U.S. dollars to collect and share
data collected by different departments of the government (Li
and James 2015). The quantity and quality of economic data
available in China have improved steadily, but continuous
and complete data on key measurements of economic situa-
tions are still missing to some degree (Lo 2016). In addition,
there is a data inconsistency problem due to statistical caliber
and standard change during economic development (Liu
et
al.
2016). The data gaps are hampering efforts to identify and
understand variation in these outcomes and target interven-
tion effectively to areas of greatest need (Wei and Zhang 2013;
Yu and Jin 2016).
Moreover, the data gaps in China’s county economic data
are particularly constraining (Tao
et al.
2015). The Chinese
County Statistical Yearbook covers annual detailed macro-
economic statistics in comprehensive economy, agriculture,
industry, and education for more than 2000 counties of
mainland China except for prefecture-level cities (Dong
et al.
2016). However, the macro-economic statistics do not involve
all the counties. Almost 10% of the national counties do not
have statistics. On the other hand, due to the issue of statis-
tical caliber, some statistics deviate from the reality (Wang
2015). The collection of county economic data in China relies
on the statistical demand of the research hotspot and urban-
ization degree. The statistical caliber of county economy
indicators is inconsistent because of the frequent data update
and modification of urban and rural population classifica-
tion standards in the census (Wang 2015). In addition, it is
especially difficult for the current statistical caliber to take
China’s “semi-urbanization” phenomenon into account,
which is mainly occurring in Central China. Central China
is a typical place where villages are being withdrawn and
the corresponding towns are being established with differ-
ent versions of statistical calibers (Pan, Wei, and Wang 2015).
However, consistent county-level economic data are crucial
in economic prediction and policy making process under the
large-scale population flow as well as the urban-rural integra-
tion development (Friedmann 2005; Wang 2006; Zhang 2011).
Therefore, it is urgent to find an objective way to fill in the
missing data and correct the inconsistent statistics collected.
Bo Yu and Fang Chen are with Key Laboratory of Digital
Earth Science, Institute of Remote Sensing and Digital
Earth, Chinese Academy of Sciences, Beijing 100094, China
(
).
Fang Chen is also with Hainan Key Laboratory of Earth
Observation, Institute of Remote Sensing and Digital Earth,
Chinese Academy of Sciences, Sanya 572029, China.
Fang Chen is also with the University of Chinese Academy of
Sciences, Beijing 100049, China.
Ying Dong is with the College of Economics and Management,
South China Agricultural University, Guangzhou 510642,
China.Yu Wang is with the Beijing Twenty-First Century
Science & Technology Development Co., Ltd., Beijing 100096,
China.Yu Wang is also with the School of Electronics and
Information Engineering, Harbin Institute of Technology,
Harbin 150001, China.
Photogrammetric Engineering & Remote Sensing
Vol. 86, No. 2, February 2020, pp. 99–105.
0099-1112/20/99–105
© 2020 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.86.2.99
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
February 2020
99
