Scholarlink

<p>Global gridded population (GGP) datasets provide estimates of a population within a grid and are used across various disciplines. However, their accuracy, particularly at the spatially fine grid-cell level, is poorly understood. Therefore, in this study, we empirically evaluated the accuracy of four common GGP datasets (GHS-POP, GPWv4, LandScan, and WorldPop) for Japan and several EU countries (France, Germany, Italy, and Sweden) at approximately 1-km<sup>2</sup> resolution. In addition, we examined an ensemble of GGPs and smart selection of GGPs using a random forest model. The results indicated that (1) GHS-POP achieved at least second-best accuracy in most cases for both Japan and the EU; (2) simple averaging of GHS-POP, GPWv4, and WorldPop may improve the accuracy, although the sole use of GHS-POP was generally associated with better accuracy; and (3) dataset selection using a random forest model with the GGP datasets and infrastructure as explanatory variables did not improve accuracy outside the country for which the model was trained.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Accuracy Assessment of Global Gridded Population Datasets at Approximately 1-km2 Grid Level

  • 期刊论文

  • 发表日期：

    2025年10月31日

查看全文
  • Masashi Tomari,
  • Hajime Seya,
  • Ryo Saito

摘要

Global gridded population (GGP) datasets provide estimates of a population within a grid and are used across various disciplines. However, their accuracy, particularly at the spatially fine grid-cell level, is poorly understood. Therefore, in this study, we empirically evaluated the accuracy of four common GGP datasets (GHS-POP, GPWv4, LandScan, and WorldPop) for Japan and several EU countries (France, Germany, Italy, and Sweden) at approximately 1-km2 resolution. In addition, we examined an ensemble of GGPs and smart selection of GGPs using a random forest model. The results indicated that (1) GHS-POP achieved at least second-best accuracy in most cases for both Japan and the EU; (2) simple averaging of GHS-POP, GPWv4, and WorldPop may improve the accuracy, although the sole use of GHS-POP was generally associated with better accuracy; and (3) dataset selection using a random forest model with the GGP datasets and infrastructure as explanatory variables did not improve accuracy outside the country for which the model was trained.