Analysis of Grid Distribution Characteristics of Agricultural Population in China Based on Land Use Data

MI Ruihua, NI Shilong, LIU Shumin
(School of Economics and Management, Yan'an University, Yan'an 716000, Shaanxi, China)

Abstract

High-precision spatial distribution research on agricultural population constitutes foundational work for constructing modern agricultural industrial systems and holds significant decision-making value for fostering new-quality productive forces in agriculture. Based on county-level data from the Seventh National Population Census and the China Land Use Remote Sensing Monitoring Dataset (CNLUCC2020), this study explores a gridding method for agricultural population and achieves a visual representation of China's agricultural population density at the grid scale. Validation indicators demonstrate that the results exhibit good accuracy. The findings reveal: (1) Agricultural population distribution along the Hu Huanyong Line shows significant differentiation, with the southeastern half (30.57 persons·km⁻²) being 15.9 times denser than the northwestern half (1.92 persons·km⁻²). (2) Agricultural population distribution increases gradiently along the three-tiered terrain ladder, with densities of 0.98 persons·km⁻², 11.27 persons·km⁻², and 30.76 persons·km⁻² respectively. (3) Terrain and climate in each agricultural region substantially influence agricultural population distribution, with dense concentrations in warm, humid, low-altitude agricultural zones and relatively sparse populations in cold-arid, high-altitude plateau and hilly agricultural areas. The study recommends implementing differentiated digital agriculture promotion strategies, prioritizing agricultural digital transformation in densely populated regions such as the Huang-Huai Plain, promoting integrated agritourism development in ecologically fragile areas, and accelerating the cultivation of new-type professional farmers.

Keywords: agricultural population; land use type; gridding method; Hu Huanyong Line; three-tiered terrain ladder; accuracy verification

1. Study Area Overview

China's topography is complex and diverse, with terrain descending from west to east across three distinct steps. The Altun, Qilian, and Hengduan Mountains in the west form the boundary between the first and second terrain steps, while the Greater Khingan, Taihang, Wushan, and Xuefeng Mountains in the east separate the second and third steps. Average elevation decreases progressively from west to east across these three tiers (Fig. [FIGURE:1]). Temperature gradually declines from south to north, forming five temperature zones: tropical, subtropical, warm temperate, temperate, and cold temperate, plus the Qinghai-Tibet Plateau vertical zone. Precipitation decreases from southeast to northwest, creating four moisture regimes: humid, semi-humid, arid, and semi-arid. Modern agricultural zoning (first-level division) reflects regional heterogeneity in agricultural cultivation conditions under the influence of natural geography and socio-economic factors \cite{18}. Due to data limitations, this study focuses solely on China's mainland administrative regions (excluding Hong Kong, Macao, and Taiwan).

2. Data and Methods

2.1 Data Sources

The primary datasets employed in this study include: agricultural population data, administrative boundary data, land use remote sensing monitoring data, three-tiered terrain ladder data, and national modern agricultural zoning data. Agricultural population data were extracted from the "Population by Industry Category (Long Form)" in the Seventh National Population Census County-Level Data (2020), which sampled 7,135,407 agricultural workers nationwide. Based on the 1.35% sampling ratio, we reasonably estimated agricultural population sizes at the county level. Vector data for China's 2020 administrative divisions were obtained from the National Natural Resources Department's Standard Map Service System (http://bzdt.ch.mnr.gov.cn). The China Land Use Remote Sensing Monitoring Dataset (CNLUCC2020) was sourced from the Resource and Environmental Science Data Registration and Publication System \cite{19}. This dataset, compiled at a 1:100,000 national scale, includes land use categories such as cropland, forest, grassland, water bodies, urban/rural/industrial/mining/residential land, and unused land. Cropland is further subdivided into paddy fields and dry farmland, while urban/rural/industrial/mining/residential land comprises urban land, rural settlements, and other construction land. Data for China's modern agricultural zoning (first-level division) were obtained through application to the research team of Agricultural Regional Differentiation and Modern Agricultural Zoning Scheme in China \cite{18}, while three-tiered terrain ladder data were compiled based on relevant research findings \cite{20}.

2.2 Methodology

Population gridding methods typically employ census data \cite{21} or statistical demographic data \cite{22}, combined with land use \cite{23}, nighttime lighting \cite{24}, points of interest \cite{25}, location-based data \cite{26}, residential attributes \cite{27}, or other social sensing data. These approaches utilize kernel density estimation \cite{28}, random forest models \cite{29}, similarity matching models \cite{30}, or other techniques to allocate population data to grid units \cite{31}, thereby mitigating the Modifiable Areal Unit Problem (MAUP) while significantly enhancing spatial detail and producing finer-scale population distribution outputs \cite{32}. Drawing on existing research, this study converts agricultural population data from administrative units to grid units based on land use type data. The gridding process involves four key steps: selecting an appropriate grid scale, determining land use type weights, calculating grid density, and visualizing the gridded data.

2.2.1 Selecting an Appropriate Grid Scale

An optimal grid scale enhances spatial representation without compromising computational efficiency. Excessively large grids reduce spatial expressiveness \cite{33}, while overly small grids fail to improve spatial relationship representation \cite{34} and create data redundancy. Previous studies have selected various grid scales—such as 10 km \cite{35}, 1.5 km \cite{36}, and 100 m \cite{37}—depending on study area and data purpose. Mi et al. \cite{38} calculated appropriate grid dimensions based on minimum administrative area to avoid both oversized grids encompassing multiple small administrative units and undersized grids reducing spatial expressiveness. The formula for determining appropriate grid side length is:

$$
g = \sqrt{S_{\min}}
$$

where $g$ represents the appropriate grid side length and $S_{\min}$ is the minimum administrative area. Calculations show $S_{\min} \approx 10$ km², thus $g$ should be less than 1.78 km. To ensure horizontal comparability with existing research \cite{22,24}, we selected grid side lengths of 10 km and 50 km for comparative validation. Using the Create Fishnet tool, we generated grids, trimmed them using national boundaries, and calculated each grid's area $a_n$ (where $n$ is the grid ID).

2.2.2 Determining Land Use Weights

The key to agricultural population gridding lies in assigning weights to land use types. Rural settlements include living areas of agricultural, forestry, and pastoral farms, as well as all villages, tents in pastoral areas, and isolated houses in sparsely populated regions \cite{39}, serving as fundamental spaces for agricultural production and living. Paddy fields and dry farmland concentrate agricultural activities, while urban land, other construction land, forest, and grassland host relatively fewer agricultural activities \cite{40}. Agricultural activities on water bodies and unused land are even scarcer but not nonexistent. Using stepwise regression to model county-level agricultural population size against land use types, we found rural settlements exhibit the strongest positive explanatory power, followed by paddy fields, dry farmland, forest, urban land, and water bodies. Other construction land, grassland, and unused land showed no statistically significant explanatory power. Combining literature review and regression results, we employed the expert scoring method to assign weights (Table [TABLE:1]). Comparative simulations revealed that weight scheme 2 best satisfies three principles: (1) reflecting heterogeneity in agricultural population distribution across land use types consistent with empirical reality; (2) ensuring natural transitions across county boundaries under total population constraints; and (3) maintaining similar density values for identical land use types across counties with comparable agricultural conditions. Therefore, weight scheme 2 was selected as the optimal weighting scheme.

2.2.3 Density Calculation Formula

Overlaying land use data, county boundary data, and 1 km grid data through map algebra yields minimum polygons with unique administrative attributes ($i$), land use type attributes ($j$), and grid attributes ($n$). Calculating minimum polygon area ($a_m$, where $m$ is the polygon ID) and summarizing yields county $i$'s area of land use type $j$ ($a_{ij}$). The agricultural population density for county $i$ and land use type $j$ ($D_{ij}$) is calculated as:

$$
D_{ij} = \frac{p_i}{\sum_j a_{ij} \times f_j} \times f_j
$$

where $p_i$ is the agricultural population of county $i$ from census data; $f_j$ is the weight of land use type $j$; $a_{ij} \times f_j$ represents the weighted area; and $p_i / \sum_j a_{ij} \times f_j$ serves as the baseline density for uniform distribution.

2.2.4 Gridded Data Visualization

Assigning corresponding $D_{ij}$ values to each minimum polygon based on administrative and land use attributes, we multiply $D_{ij}$ by $a_m$ to obtain agricultural population counts per minimum polygon ($p_m$). Summing $p_m$ values by grid attribute ($n$) yields total agricultural population per grid ($p_n$). Dividing $p_n$ by grid area ($a_n$) produces final agricultural population density per grid ($D_n$). Using natural breaks classification with manual boundary adjustments, we generated China's agricultural population density distribution map, achieving spatial visualization of the gridded dataset.

3. Results

3.1 Agricultural Population Distribution Across the Hu Huanyong Line

Using the Hu Huanyong Line as a boundary, we analyzed spatial differentiation patterns of China's agricultural population (Fig. [FIGURE:2]). Under the influence of natural factors including topography, altitude, monsoon climate, temperature, and precipitation, the southeastern half exhibits relatively higher agricultural productivity and population carrying capacity, resulting in greater agricultural population density. Conversely, the northwestern half, characterized by high altitude, mountainous terrain, arid/semi-arid conditions, and cold climate, primarily practices animal husbandry and specialty crop cultivation. Agricultural population density remains low overall, concentrating mainly in locally favorable areas such as plains, river valleys, terraces, dam areas, and oases, while being extremely sparse across the Qinghai-Tibet Plateau, desert zones, ecological reserves, and water source conservation areas. Overall, the Hu Huanyong Line serves not only as a crucial demographic boundary but also as a significant dividing line for agricultural population distribution in China, manifesting a clear "high east, low west" pattern.

Statistical analysis of land area, per capita cropland, and agricultural population density across both sides (Table [TABLE:2]) reveals that the southeastern half, comprising 42.65% of national territory with 98.14% of paddy fields and 92.22% of dry farmland, concentrates 77.86% of agricultural population at a density of 30.57 persons·km⁻², with per capita cropland of 1.18 hm². The northwestern half, accounting for 57.35% of land area, holds only 1.86% of paddy fields and 7.78% of dry farmland. Due to poor hydrothermal conditions unfavorable for cultivation, it hosts merely 22.14% of agricultural population at a density of 1.92 persons·km⁻², with per capita cropland of approximately 2.86 hm².

3.2 Agricultural Population Distribution Across Terrain Steps

To analyze distribution characteristics across the three terrain steps, we generated contour maps based on 1 km gridded agricultural population density data (Fig. [FIGURE:3]). Contour mapping, adapted from geology, meteorology, and hydrology \cite{41}, connects areas with equal population density to reveal distribution patterns by omitting unnecessary details \cite{42}. Contour density indicates population concentration gradients, while closed contours signify high- or low-density centers.

The results show marked differences across terrain steps. The first step exhibits sparse, few contours, indicating low and relatively uniform density. The second step shows variable contour density, reflecting significantly increased density with substantial spatial variation. Northern areas of the second step, with insufficient hydrothermal conditions, display sparse contours; central-southern areas with favorable conditions show dense contours and numerous closed centers, indicating dramatic density fluctuations and several high-density zones. The third step features the densest contours with pronounced regional differences: northern areas exhibit dramatic density variations due to cold climate and complex terrain; central plain zones contain multiple peak density areas; and southern hilly regions, despite favorable hydrothermal conditions, show alternating peaks and valleys due to topography. Overall, altitude significantly influences spatial distribution, with agricultural population density increasing progressively down the three terrain steps, making these boundaries important dividing lines for agricultural population distribution.

Statistical analysis across terrain steps (Table [TABLE:3]) shows the first step, comprising 27.71% of national territory but only 0.20% of paddy fields and 2.35% of dry farmland, hosts merely 1.92% of agricultural population at 0.98 persons·km⁻². The second step, with 43.13% of land area and 34.47% of cropland, accommodates 29.16% of agricultural population at 11.27 persons·km⁻². The third step, though only 29.16% of territory, contains 63.61% of cropland and, benefiting from superior hydrothermal conditions, supports 55.06% of agricultural population at 30.76 persons·km⁻². In summary, agricultural population differentiation across terrain steps is primarily influenced by altitude, cropland proportion, and cultivation conditions.

3.3 Agricultural Population Distribution by Agricultural Region

Analyzing spatial differentiation using China's modern agricultural first-level zoning reveals significant inter-regional variations and clear spatial clustering characteristics (Fig. [FIGURE:4]). The Huang-Huai Plain and Beijing-Tianjin-Hebei-Shandong Plain-Hill regions exhibit the densest agricultural populations due to flat terrain and fertile soil. The Sichuan Basin, South China Tropical Crop Region, and Middle-Lower Yangtze Plain show high concentrations due to low average elevation, favorable hydrothermal conditions, and developed irrigation agriculture. The Southeast Coastal Hill region, with undulating terrain and good water-heat resources, supports relatively dense populations engaged in tea, fruit, and rice cultivation. The Yunnan-Guizhou Plateau, despite favorable hydrothermal conditions, exhibits alternating dense and sparse patterns due to complex mountainous terrain. The Northeast Plain, with lower temperatures but fertile land, concentrates populations in the Songnen and Liaohe Plains. The Loess Plateau, characterized by arid climate and complex topography, shows overall low density with a narrow concentration zone in the Guanzhong Plain. The Northeast Mountain-Hill region, Inner Mongolia Plateau, Gansu-Xinjiang Desert Plateau, and Qinghai-Tibet Plateau exhibit the lowest densities due to cold climate, high altitude, or arid conditions.

Statistical analysis by agricultural region (Table [TABLE:4]) shows the Huang-Huai Plain, with 7.47% of territory and 11.98% of cropland, hosts 13.97% of agricultural population at the highest density (63.61 persons·km⁻²). The Beijing-Tianjin-Hebei-Shandong Plain-Hill region, with 3.41% of land area and 12.70% of cropland, accommodates 12.87% of agricultural population at the second-highest density (60.54 persons·km⁻²). The Sichuan Basin, South China Tropical Crop, and Middle-Lower Yangtze Plain regions also show high densities (>30 persons·km⁻²). In contrast, the Northeast Mountain-Hill, Inner Mongolia Plateau, Gansu-Xinjiang Desert Plateau, and Qinghai-Tibet Plateau regions all have densities below 10 persons·km⁻², with the Qinghai-Tibet Plateau recording the lowest density (0.43 persons·km⁻²) despite comprising 24.07% of national territory. Overall, agricultural population distribution correlates closely with cropland area, average elevation, topography, and hydrothermal conditions.

3.4 Validation

Given the complexity and mobility of population distribution plus inherent methodological and data limitations, simulated gridded datasets inevitably contain errors \cite{43}. We assessed dataset quality through weight scheme validation and grid scale validation.

3.4.1 Weight Scheme Validation

To examine weight assignment impacts, we generated density maps under different weight schemes (Fig. [FIGURE:5]). While subtle differences exist in detail representation, all schemes capture major spatial patterns without significant divergence. Following Xiao et al. \cite{44}, we aggregated gridded data to county-level estimates and compared them against census values, calculating absolute and relative estimation errors:

$$
REE_i = \frac{PE_i - P_i}{P_i} \times 100\%
$$

$$
AEE_i = PE_i - P_i
$$

where $PE_i$ is the estimated agricultural population for county $i$ from gridded data, $P_i$ is the census value, $REE_i$ is relative error, and $AEE_i$ is absolute error. We defined critical thresholds: $REE_i \leq -10\%$ indicates severe underestimation; $(-10\%, -5\%]$ indicates moderate underestimation; $(-5\%, 5\%]$ indicates accurate estimation; $(5\%, 10\%]$ indicates moderate overestimation; and $REE_i > 10\%$ indicates severe overestimation.

Spatial error analysis (Fig. [FIGURE:6]) shows weight scheme 2 correctly estimates 90.16% of counties, with severe under/over-estimation in only 4.80% of cases—primarily urban districts with small agricultural populations where minor absolute errors create large relative errors. Accuracy assessment (Table [TABLE:5]) demonstrates strong performance across all metrics, with weight scheme 2 achieving the highest correlation coefficient ($r = 0.9988$), indicating excellent consistency with census data and minimal systematic bias.

3.4.2 Grid Scale Validation

To validate grid scale effectiveness, we generated agricultural population maps at 10 km and 50 km resolutions (Fig. [FIGURE:7]). Both scales capture major distribution patterns, but 1 km grids provide superior textural detail and granularity. As grid size increases, spatial variability diminishes and fine-scale information is lost. The 1 km grid achieves peak density values up to 500 persons·km⁻², with statistical means and standard deviations decreasing as grid size expands, confirming that larger grids substantially reduce spatial detail.

Spatial error analysis at different scales (Fig. [FIGURE:8]) shows 1 km grids correctly estimate 90.55% of counties, while 50 km grids correctly estimate only 73.38%, with severe errors increasing from 4.80% to 13.97%. Accuracy metrics (Table [TABLE:6]) confirm that 1 km grids achieve the highest $r$ value (0.9988) and lowest errors, while 50 km grids show significantly degraded performance. This demonstrates that appropriate grid scale selection is critical for dataset precision.

4. Conclusions and Recommendations

4.1 Conclusions

(1) China's agricultural population distribution exhibits pronounced spatial heterogeneity, with the Hu Huanyong Line serving as a crucial dividing line. Agricultural population density shows a clear east-west gradient, with the southeastern half—rich in paddy fields and dry farmland—hosting 77.86% of agricultural population, while the northwestern half, with limited cropland and poor hydrothermal conditions, accounts for only 22.14%.

(2) Average elevation significantly influences distribution, causing agricultural population density to increase progressively down the three terrain steps. Within each step, density is further shaped by topography and climate, with higher concentrations in plains and basins versus sparse distribution in mountains and hills, and greater density in warm, water-rich areas versus cold-arid regions.

(3) Under the combined influence of natural geography, socio-economics, and location, agricultural population varies dramatically across agricultural regions. High-density regions include the Huang-Huai Plain and Beijing-Tianjin-Hebei-Shandong Plain-Hill areas, while low-density regions comprise the Northeast Mountain-Hill, Inner Mongolia Plateau, Gansu-Xinjiang Desert Plateau, and Qinghai-Tibet Plateau. Industrialization and urbanization have reduced agricultural population densities in major city centers.

4.2 Recommendations

(1) West of the Hu Line, where ecosystems are fragile and agricultural populations are sparse, we recommend constructing water conservancy facilities where feasible, rehabilitating deserts and swamps, enhancing ecological functions, and promoting integrated agritourism tailored to local conditions. East of the Hu Line, with abundant cropland and high agricultural population density, we advise accelerating agricultural digitalization, developing socialized agricultural services, and supporting smallholders' integration into modern agricultural cooperative systems.

(2) On the first terrain step, characterized by extreme ecological fragility and minimal agricultural population density, priority should be given to ecological conservation, developing characteristic ecological agriculture, and promoting agritourism integration. The second step, with better agricultural resources and moderate density, should focus on soil improvement, farmland transformation, smart agriculture development, and mechanization enhancement. The third step, with optimal cultivation conditions and highest density, should lead in constructing high-standard farmland and promoting primary-secondary-tertiary industry integration.

(3) Given substantial differences in agricultural resource endowments across regions, we recommend formulating differentiated and flexible agricultural policies tailored to each zone. This includes promoting agricultural digitalization adapted to local terrain and climate, developing digital farming tools for diverse conditions, training farmers in digital literacy, cultivating new-type professional farmers, and prioritizing digital transformation in high-density regions such as the Huang-Huai Plain and Beijing-Tianjin-Hebei-Shandong Plain-Hill areas.

References

\cite{1} Hu Huanyong, Zhang Shanyu. Population Geography of China (Volume 1) [M]. Shanghai: East China Normal University Press, 1984.

\cite{2} Wang Kedong, Zeng Weizhong, Wu Peipei. Leading agricultural and rural modernization with new-quality productive forces: Intrinsic logic, opportunities, challenges, and implementation pathways [J]. Journal of Sichuan Agricultural University, 2024, 42(6): 1159-1164, 1178.

\cite{3} Dai Rong, Guo Luo, Xue Dayuan, et al. Spatio-temporal characteristics of agricultural population spatial distribution in China [J]. Journal of China Population, Resources and Environment, 2010, 20(Suppl. 2): 186-189.

\cite{4} Chen Yang. A preliminary study on distribution characteristics and driving factors of agricultural population at county scale: A case study of Sichuan Province [C]// Proceedings of 2019 China Urban Planning Annual Conference. Beijing: China Architecture & Building Press, 2019: 11.

\cite{5} Wang Haitao, Wu Tongyi, Bai Zhengbing, et al. Research on spatial distribution and spatio-temporal evolution characteristics of agricultural population in Shaanxi Province [J]. Geospatial Information, 2020, 18(10): 106-109, 6.

\cite{6} Jiang Jie, Yang Xin. Quantitative segmentation of China's three gradient terrain based on EDM [J]. Journal of Spatio-temporal Information, 2009, 7(1): 8-13.

\cite{7} Mi Ruihua, Yang Xin, Feng Fei. Population distribution projection in urban development zones and its planning value: Taking Xi'an High-tech Zone as an example [J]. Journal of Northwest University (Natural Science Edition), 2019, 49(5): 801-807.

\cite{8} Zheng Chuan, Jia Zhaoliang, Xu Ruijie, et al. Refinement study of population distribution in Yingjiang County based on risk census building hazard data and nighttime lighting data [J]. Journal of Seismological Research, 2023, 46(3): 403-414.

\cite{9} Yang Yanxin, Cao Yi, Wang Wenbin. Fractal study of agricultural population distribution in Jinzhong City [J]. Journal of Shanxi Agricultural University (Natural Science Edition), 2011, 31(6): 573-576.

\cite{10} Mi Ruihua, Liu Shumin, Ni Shilong. Research on grid distribution characteristics and spatio-temporal evolution of agricultural population in Shaanxi Province [J]. Journal of Shaanxi Normal University (Natural Science Edition), 2025, 53(2): 37-47.

\cite{11} Chun Jin, Zhang Xinchang, Huang Jianfeng, et al. A gridding method for redistributing population based on POIs [J]. Geography and Geo-information Science, 2018, 34(4): 83-89, 124, 2.

\cite{12} Zhang Huize, Gao Qijie. The phenomenon of new farmers and construction of rural talent revitalization mechanism: Based on social-industrial dual network perspective [J]. Modern Economic Research, 2021(2): 121-125.

\cite{13} Javanmard R, Lee J, Kim J, et al. The impacts of the modifiable areal unit problem (MAUP) on social equity analysis of public transit reliability [J]. Journal of Transport Geography, 2023, 106: 103500.

\cite{14} Mei Y, Gui Z, Wu J, et al. Population spatialization with pixel-level attribute grading by considering scale mismatch issue in regression modeling [J]. Geo-spatial Information Science, 2022, 25(3): 365-375.

\cite{15} Yang X C, Ye T T, Zhao N Z, et al. Population mapping with multi-sensor remote sensing images and point-of-interest data [J]. Remote Sensing, 2019, 11(5): 574.

\cite{16} Li Shanshan, Lin Wentan. Spatial distribution and characteristics of population in Guangdong-Hong Kong-Macao Greater Bay Area based on nighttime light remote sensing [J]. Tropical Geography, 2023, 43(3): 384-394.

\cite{17} Gui Zhipeng, Mei Yu'ao, Wu Huayi, et al. Urban population spatialization by considering heterogeneity in local resident attraction force of POIs [J]. Journal of Geo-information Science, 2022, 24(10): 1883-1897.

\cite{18} Liu Yansui, Zhang Ziwei, Dou Shiqing. Regional differentiation and comprehensive regionalization scheme of modern agriculture in China [J]. Acta Geographica Sinica, 2018, 73(2): 203-218.

\cite{19} Xu Xinliang, Liu Jiyuan, Zhang Shuwen, et al. China multi-period land use remote sensing monitoring dataset (CNLUCC) [DB/OL]. Resource and Environmental Science Data Registration and Publication System (http://www.resdc.cn/DOI), 2018.

\cite{20} Jiang Jie, Yang Xin. Quantitative segmentation of China's three gradient terrain based on EDM [J]. Journal of Spatio-temporal Information, 2009, 7(1): 8-13.

\cite{21} Wang L, Fan H, Wang Y. Improving population mapping using Luojia 1-01 nighttime light image and location-based social media data [J]. Science of the Total Environment, 2020, 730: 139148.

\cite{22} Sinha P, Gaughan A E, Stevens F R, et al. Assessing the spatial sensitivity of a random forest model: Application in gridded population modeling [J]. Computers, Environment and Urban Systems, 2019, 75: 132-145.

\cite{23} Ye T, Zhao N, Yang X, et al. Improved population mapping for China using remotely sensed and point-of-interest data within a random forests model [J]. Science of the Total Environment, 2019, 658: 936-946.

\cite{24} Yu B, Lian T, Huang Y, et al. Integration of nighttime light remote sensing images and taxi GPS tracking data for population surface enhancement [J]. International Journal of Geographical Information Science, 2019, 33(4): 687-706.

\cite{25} Han D R, Yang X H, Cai H Y, et al. Modelling spatial distribution of fine-scale populations based on residential properties [J]. International Journal of Remote Sensing, 2019, 40(14): 5287-5300.

\cite{26} Wang Mingming, Wang Juanle. Gridded data of town-level population distribution in Shandong Province (2000 & 2010) [J]. China Scientific Data, 2019, 4(3): 182-192.

\cite{27} Leyk S, Gaughan A E, Adamo S B, et al. The spatial allocation of population: A review of large-scale gridded population data products and their fitness for use [J]. Earth System Science Data, 2019, 11(3): 1385-1409.

\cite{28} Fang Y, Jawitz J W. High resolution reconstruction of the United States human population distribution, 1790 to 2010 [J]. Scientific Data, 2018, 5: 180067.

\cite{29} Wang F, Lu W, Zheng J, et al. Spatially explicit mapping of historical population density with random forest regression: A case study of Gansu Province, China, in 1820 and 2000 [J]. Sustainability, 2020, 12(3): 1231.

\cite{30} Wang L Y, Fan H, Wang Y K. Fine resolution population mapping from international space station nighttime photography and multi-source social sensing data based on similarity matching [J]. Remote Sensing, 2019, 11(16): 1900.

\cite{31} Zhang Xuezhen, Wang Fahui, Lu Weidong, et al. Grid reconstruction of Chinese traditional agricultural area population from 1776 to 1953 [J]. Scientia Sinica (Terrae), 2022, 52(3): 528-539.

\cite{32} Dong Nan, Yang Xiaohuan, Cai Hongyan, et al. Suitability evaluation of gridded population distribution: A case study in rural area of Xuanzhou district, China [J]. Acta Geographica Sinica, 2017, 72(12): 2310-2324.

\cite{33} Bai Z Q, Wang J L, Wang M M, et al. Accuracy assessment of multi-source gridded population distribution datasets in China [J]. Sustainability, 2018, 10(5): 1363.

\cite{34} Luo Yongzhen, Dong Chun, Zhang Yu. Study on the method of evaluating the suitable grid for population spatialization [J]. Journal of Geo-information Science, 2023, 25(5): 896-908.

\cite{35} Xiao Dongsheng, Wang Ning, Liu Zhicheng. Accuracy of "Representative Population Grid Dataset" in arid areas: A case of Gansu-Ningxia-Qinghai region [J]. Arid Land Geography, 2023, 46(3): 505-514.

\cite{36} Wu Jinghang, Gui Zhipeng, Shen Li, et al. Population spatialization by considering pixel-level attribution grading and spatial association [J]. Geomatics and Information Science of Wuhan University, 2022, 47(9): 1364-1375.

\cite{37} Guo Zihan, Chen Fei, Liu Xiaoli, et al. A method to improve the spatial accuracy of population gridded data based on land use [J]. Bulletin of Surveying and Mapping, 2021(12): 66-70.

\cite{38} Zhou Zugen, Wang Jinglong. Application research on contour map of population density [J]. Chinese Journal of Population Science, 1992(5): 28-33.

\cite{39} Wu Li. Introduction to the newly compiled 1:6,000,000 China population distribution map and contour density map [J]. Population Research, 1987(4): 59-60, 58.

\cite{40} Li Huimin, Luo Dawen, Wang Jieyong. Regional population estimation on multi-spatial scale using Tencent location big data [J]. Bulletin of Surveying and Mapping, 2022(6): 93-97.

\cite{41} Mi Ruihua, Shi Ying. Identifying method of CBD and sub-CBD based on resident population distribution: A case study of Xi'an City [J]. Journal of Shaanxi Normal University (Natural Science Edition), 2014, 42(3): 97-102.

\cite{42} Javanmard R, Lee J, Kim J, et al. The impacts of the modifiable areal unit problem (MAUP) on social equity analysis of public transit reliability [J]. Journal of Transport Geography, 2023, 106: 103500.

\cite{43} Mei Y, Gui Z, Wu J, et al. Population spatialization with pixel-level attribute grading by considering scale mismatch issue in regression modeling [J]. Geo-spatial Information Science, 2022, 25(3): 365-375.

\cite{44} Xiao Dongsheng, Wang Ning, Liu Zhicheng. Accuracy of "Representative Population Grid Dataset" in arid areas: A case of Gansu-Ningxia-Qinghai region [J]. Arid Land Geography, 2023, 46(3): 505-514.