Spatial-Temporal Population Dataset of Hainan Island (2013-2018)

Lai, X. L.  Han, N. L.^*  Huang, P. J.

Department of Land Resources Management, School of Public Administration, Hainan University, Haikou 570228, China

Abstract: The spatial-temporal population dataset of Hainan Island (2013-2018) is based on the 2013-2018 NPP/VIIRS night light data, and performs Image synthesis, unstable light sources removal, and year-by-year corrections to eliminate the effects of transient light sources and background noise. Based on the correlation analysis between the processed night lighting data and population statistics data, a regression model is established, and the population spatial modeling is carried out according to the error classification, so as to eliminate the influence of mountain topography, small population and other factors on the simulation accuracy. Finally, the spatial distribution dataset of population in Hainan Island is generated. The dataset is archived in .tif format with a spatial resolution of 500 m and consists of 26 data files, 11.0 MB (compressed into one file, 168 KB), reflecting the characteristics of temporal and spatial distribution of population in Hainan Island from 2013 to 2018.

Keywords: Hainan Island; population; NPP/VIIRS night light data; regression model; population spatialization

DOI: https://doi.org/10.3974/geodp.2022.01.08

CSTR: https://cstr.escience.org.cn/CSTR:20146.14.2022.01.08

Dataset Availability Statement:

The dataset supporting this paper was published and is accessible through the Digital Journal of Global Change Data Repository at: https://doi.org/10.3974/geodb.2021.08.02.V1 or https://cstr.escience.org.cn/CSTR:20146.11.2021.08.02.V1.

1 Introduction

Nightlight remote sensing images can provide important basis for the estimation of socio-economic parameters such as GDP, population, electricity consumption, greenhouse gas emissions, poverty index and Gini coefficient and so on^[1]. At present, the commonly used night light data are DMSP/OLS night light data and NPP/VIIRS night light data, both of which come from the National Oceanic and Atmospheric Administration (NOAA). The difference between them is that the time series of DMSP/OLS night light remote sensing image is from 1992 to 2013, and the spatial resolution is 1 km, while the time series of NPP/VIIRS night light remote sensing image is from 2012 to now, and the spatial resolution is 500 m. At the same time, NPP/VIIRS night light data were captured by Suomi-NPP sate­llite using VIIRS, using polar orbit, and obtained by stitching together multiple cloud-free images^[2]. The NPP/VIIRS sensor has 22 bands, with a wavelength range of 0.4-12μm, covering the visible and infrared spectrum. Spectral resolution is 16 bit, and radiation detector with wider band and on-orbit radiometric correction technology effectively improve the quality of light image^[3]. Thanks to its stronger light capture sensitivity, higher spatial and temporal resolution, NPP/VIIRS data is more suitable for small and medium-scale human activities research^[4], and the spatial distribution of population information extracted from this can be intuitive reflects the characteristics of the temporal and spatial evolution of the population. Therefore, this dataset is processed based on NPP/VIIRS night lighting data, and the spatial distribution information of population in Hainan Island is obtained by correlation analysis and regression modeling with the demographic data of 18 cities and counties of Hainan Island, in order to provide data basis and decision-making basis for Hainan’s territorial spatial planning or scientific research.

2 Metadata of the Dataset

The metadata of Night lights and statistical data fusion of Hainan Island population 500-m grid dataset (2013-2018)^[5]dataset is showed in Table 1.

Table 1  Metadata summary of the Night lighting and census integrated 500 m raster population dataset of Hainan Island (2013-2018)

3 Methodology

3.1 Study Area and Data Sources

The study area of ​​this paper is the 18 cities and counties in the land area of ​​Hainan Island (excluding Sansha city), which are Haikou city, Sanya city and Danzhou city, and 15 province-level counties. Hainan Island is located between 108°37¢E-111°03¢E and 18°10¢N-20°10¢N. The terrain of the whole island is low and flat all around, with a high tower in the middle, presenting a dome-long mountain topography. The original night light data of this dataset are mainly the NPP/VIIRS night light remote sensing monthly composite mean image during 2013-2018, which comes from the NPP/VIIRS sensor dataset in the National Oceanic and Atmospheric Administration (NOAA)^[1], with a resolution of 500 m. Hainan Island’s administrative division data comes from the China Natural Resources Database^[2]; Hainan Island’s demographic data comes from the Hainan Provincial Statistical Yearbook^[7].

3.2 Technical Route

The technical route of this study includes three parts: NPP/VIIRS night light data acquisition, data processing and population spatialization modeling (Figure 1).

Although the monthly data of NPP/VIIRS night light data have eliminated the effects of lightning, moonlight irradiation and cloud coverage, there are still transient light sources such as aurora, fire, and ship lights and background noise. Therefore, it is necessary to process the original data to obtain stable annual average data. Therefore, NPP/VIIRS night light data processing includes the process of annual mean image synthesis, elimination of negative and extremely high values, removal of unstable light sources, and continuous correction^[8,9].

Population spatialization modeling includes correlation analysis, regression modeling and correction. We toke the city and county to analyze the correlation between the night light brightness from 2013 to 2017 and the corresponding demographic data. The regression modeling analysis of demographic data and night light brightness values ​​shows that the R² of the cubic linear function model is the highest, so the population spatialization model uses a cubic linear function. The 2018 demographic data was used to verify the 2018 population simulation results of the model. The results show that the average error of the overall simulated population of Hainan Island is 36.55%, and the error of individual cities and counties is much larger than the average error.

Figure 1  Data development technology roadmap

4 Data Results and Validation

4.1 Data Composition

The population spatial dataset of Hainan Island from 2013 to 2018 is composed of 26 data files, and the data is named in the form of “subject + time”.

4.2 Data Products

After processing the original data through the above technical route, the Night lighting and census integrated 500 m raster population dataset of Hainan Island (2013-2018) with a resolution of 500 m´500 m is obtained, in which the data unit is the total population on the 500m grid. The characteristics of the spatial and temporal distribution of Hainan Island’s population from 2013 to 2018 from the dataset show that the population distribution of Hainan Island is scattered on the whole. The population is mainly distributed in coastal cities and counties, and mainly concentrated in the two cities of Haikou and Sanya. The population of central cities and counties is relatively sparse, and formed the spatial polar distribution pattern of north Haikou and south Sanya (Figure 2, 3). From 2013 to 2018, the population of Hainan Island has a relatively obvious growthtrend. The population increment is mainly distributed in the suburbs of cities and counties, and spreads from the urban center to the periphery, such as Haikou and Sanya are particularly obvious. On the whole, the growth population is mainly distributed in the urban areas of coastal cities and counties, the areas along the high-speed railways and the high-speed roads around the island (Figure 4).

Figure 2  Population spatialization of Hainan Island in 2013

4.3 Data Validation

The NPP/VIIRS night light data from 2013 to 2018 is used to retrieve the population of the past years through the model, and the errors are verified according to the statistical data, and the large errors are grouped and then modeled again. Based on the secondary modeling after error grouping, the accuracy of population simulation is significantly improved compared with the accuracy of the first simulation. The average relative error of the model inversion

Figure 3  Population spatialization of Hainan Island in 2018

Figure 4  Population growth of Hainan Island from 2013‒2018

results is 18.19%, and the regression coefficient R² is above 0.9. The regression coefficient R² results are close to the results of Li et al.^[10] and Cao et al.^[11]. The model error is similar to that of Zhao et al.^[12], and higher than the average relative error of Wang et al.^[13] and Chen et al.^[14].

At the level of cities and counties, Wuzhishan city has a large error in the retrieve results . The reason is that Wuzhishan city is located in the hinterland of the central mountainous area of Hainan Island, and its urban topography is undulating and its population is sparse, so it is difficult to accurately retrieve its population spatial distribution only by night light data. According to the administrative divisions of Hainan province, the cities and counties are divided into prefecture-level cities, county-level cities, ordinary counties and ethnic minority autonomous counties. It can be found that the population simulation error of prefecture-level cities is the smallest, followed by ethnic minority autonomous counties. The error of county-level cities is slightly higher than that of minority autonomous counties, and the error of ordinary counties is the largest. There is no regular difference in population simulation accuracy between cities and counties with different administrative levels.

At the same time, Fu et al. conducted population regionalization based on the spatial distribution characteristics and regional differences of the population in China, and screened out land use types with strong correlation with population distribution as model parameters, established a multiple regression model to achieve population spatialization, and four factors of urban population density, traffic conditions, DEM and total amount control are used to calibrate the model^[15], and finally obtained the national population distribution dataset. Among them, the average relative error of the population spatialization data of Hainan Island in 2010 was only 0.92%^[16]. Therefore, these methods can be referenced in the future to combine night light data for population spatialization to improve accuracy.

Table 2  Model simulation results

5 Discussion and Conclusion

NPP/VIIRS night lighting data population spatial modeling can basically reflect the spatial distribution of population, but the simulation accuracy of Wuzhishan city is poor. The reason may be due to the fact that Wuzhishan city is located in the central mountainous area, with large terrain undulations, which greatly affects the night light data, and the population of Wuzhishan city is small, which is far from other cities and counties^[14].

To solve this problem, more detailed population zoning of the study area will be considered in the future. At the same time, combined with land use data, POI and other multi-source data^[17], and multi geographical factor weighting method, spatial weighted regression or neural network models will be adopted to improve the inversion accuracy. This dataset is the 500-m resolution data of Hainan Island’s population based on NPP/VIIRS night light data retrieve. This dataset can reflect the spatial and temporal distribution of Hainan Island’s population from 2013 to 2018, and can provide an auxiliary basis for the current Hainan territorial spatial planning and related research.

Author Contributions

Han, N. L. designed the algorithms of dataset. Lai, X. L. collected and processed the night light data and designed the model.  Huang,P. J. made data verification.  Lai, X. L. wrote the data paper.

Conflicts of Interest

The authors declare no conflicts of interest.

References

[1]      Li, D. R., Li, X. On noctilucent remote sensing data mining [J]. Journal of Surveying and mapping, 2015, 44(6): 591–601.

[2]      Li, F., Mi, X. N., Liu, J., et al. Spatialization of GDP in Beijing based on NPP-VIIRS nighttime light data [J]. Remote sensing for land and resources, 2016, 28(3): 19–24.

[3]      Li, F., Sun, G. T., Wang, Q. L., et al. Spatialization of power consumption at county level based on NPP-VIIRS nighttime light data [J]. Geomatics & Spatial Information, 2018, 41(4): 8–11, 15, 18.

[4]      Gao, Y., Wang, H., Wang, P. T. Spatial analysis of coastal population based on population census and multi-source nighttime light data [J]. Resources science, 2013, 35(12): 2517–2523.

[5]      Lai, X. L., Han, N. L., Huang, P. J. Night lighting and census integrated 500 m raster population dataset of Hainan Island (2013-2018) [J]. Digital Journal of Global Change Data Repository, 2021. https://doi.org/10.3974/geodb.2021.08.02.V1. https://cstr.escience.org.cn/CSTR:20146.11.2021.08.02.V1.

[6]      GCdataPR Editorial Office. GCdataPR data sharing policy [OL]. https://doi.org/10.3974/dp.policy.2014.05 (Updated 2017).

[7]      Hainan Bureau of Statistics, National Bureau of Statistics Hainan Survey Team. Hainan statistical yearbook [OL]. 2014-2019. http://www.hainan.gov.cn/hainan/tjnj/list3.shtml.

[8]      Hu, W. A., Liu, C. L., Zhan, Q. W. Synthesis method and comparative verification of nPP-VIIRS annual night light data in China [J/OL]. Journal of Guilin University of Technology, 2020: 1–10 [2020-12-23].

[9]      Zhou, Y., Chen, Y., Liu, Y., et al. Generation and verification of NPP-VIIRS annual nighttime light data [J]. Remote sensing information, 2019, 34(2): 62–68.

[10]    Li, X., Chen, Z. J., Wu, J. X., et al. Research on grid method of urban resident population based on night light data and spatial regression model [J]. Journal of geo-information science, 2017, 19(10): 1298–1305.

[11]    Cao, L. Q., Li, P. X., Zhang, L. P. Urban population estimation based on DMSP/OLS nighttime light data: a case study of counties in Hubei province [J]. Remote Sensing Information, 2009(1): 83–87.

[12]    Zhao, Y., Wang, L., Li, X. X., et al. Spatialization of population in Dalian based on NPP/VI RS night light and land use [J]. Resources development and market, 2018, 34(10): 1372–1376.

[13]    Wang, X. X., Zhao, M. S, Xie, Y., et al. Spatialization of population data in Jiangsu province based on NPP- VIIRS night light remote sensing [J]. Journal of heilongjiang institute of technology, 2020, 34(3): 31–35.

[14]    Chen, Q., Hou, X. Y., Wu, L. Comparative analysis of spatialization model of population based on land use data and night light data: a case study of high efficiency ecological economic zone in the Yellow River Delta [J]. Human geography, 2014, 29(5): 94–100.

[15]    Fu, J. Y., Jiang, D., Huang, Y. H. Dataset of kilometer-grid population distribution in China [J]. Acta Geographica Sinica, 2014, 69(Sup.): 41–44.

[16]    Fu, J. Y., Jiang, D., Huang, Y. 1 km grid population dataset of China [DB/OL]. Global Change Data Repository, 2014. https://doi.org/10.3974/geodb.2014.01.06.V1.

[17]    Zhao, X., Song, Y. Q., Liu, Y. L., et al. Spatialization of population based on satellite remote sensing and POI data: a case study of Guangzhou [J]. Tropical geography, 2020, 40(1): 101–109.