Modeling Population Density Using a New Index Derived from Multi-Sensor Image Data

Abstract

The detailed information about the spatial distribution of the population is crucial for analyzing economic growth, environmental change, and natural disaster damage. Using the nighttime light (NTL) imagery for population estimation has been a topic of interest in recent decades. However, the effectiveness of NTL data in population estimation has been impeded by some limitations such as the blooming effect and underestimation in rural regions. To overcome these limitations, we combine the NPP-VIIRS day/night band (DNB) data with normalized difference vegetation index (NDVI) and land surface temperature (LST) data derived from the moderate resolution imaging spectroradiometer (MODIS) onboard the Terra satellite, to create a new vegetation temperature light population index (VTLPI). A statistical model is developed to predict 250m grid-level population density based on the proposed VTLPI and the least square regression approach. After that, a case study is implemented using the data of Sichuan Province, China in 2015, and the results indicates that the VTLPI-estimated population density outperformed the results from other two methods based on nighttime light imagery or human settlement index, and the three publicized population products, LandScan, WorldPop, and GPW. When using the census data as reference, the mean relative error and median absolute relative error on a township level are 0.29 and 0.12, respectively, and the root-mean-square error is 212 persons/km2. The results show that our VTLPI-based model can achieve a better estimation of population density in rural areas and urban suburbs and characterize more spatial variations at 250m grid level both in both urban and rural areas. The resultant population density offers better population exposure data for assessing natural disaster risk and loss as well as other related applications.

The effect of population distribution measures on evaluating spatial accessibility of primary health-care institutions: A case study from China

“…Gao et al (2019) proposed that more accurate night-time light data simulates a more detailed population distribution, and these data are real-time and easier to obtain (Huang et al, 2016;Li and Zhou, 2018;Song et al, 2019). However, the various population distribution measures produce highly variable accuracies depending on which regions they are applied Luo et al, 2019;Wang et al, 2019). For instance, PoiPop shows a higher accuracy than WorldPop in Beijing, Shanghai, Guangzhou and Chongqing (Ye et al, 2019), while LandScan tends to underestimate people counts in Poland (Calka and Bielecka, 2019), and the LuoJia1-01 simulation data had a higher accuracy than that of Landscan and European Commission's Global Human Settlement population (Gao et al, 2019).…”

Mapping Multi-Temporal Population Distribution in China from 1985 to 2010 Using Landsat Images via Deep Learning

“…Several approaches have been developed to produce fine-scale gridded population data in the past few decades, such as areal weighting [20], spatial interpolation [21][22][23], and dasymetric mapping [24][25][26][27][28][29][30][31][32]. Among them, dasymetric mapping technology [33], which uses fine-scale auxiliary variables and specific weighting schemes to re-allocate census counts to grid cells, is the most widely used and effective one [19].…”

Optimization of Modelling Population Density Estimation Based on Impervious Surfaces

“…Unlike the national census, remote sensing data are not subject to time restrictions [6][7][8][9]. Owing to the rapid development of remote sensing and computer technologies, some Land 2021, 10, 791 2 of 17 scholars have applied these to estimate the population [10][11][12], enabling the possibility of mapping populations at the pixel scale. The commonly used remote sensing images are as follows: high-resolution images, such as IKONOS, QuickBird, and Worldview images, and moderate-resolution images, such as the Landsat series, hyperspectral, and radar images [2,13,14].…”