The paper titled ''Global gridded population datasets systematically underrepresent rural population'' by Josias Láng-Ritter et al. provides a valuable contribution to the discourse on the accuracy of global population datasets, particularly in rural areas. We recognize the efforts put into this research and appreciate its contribution to the field. However, we feel that key claims in the study are overly bold, not properly backed by evidence and lack a cautious and nuanced discussion. We hope these points will be taken into account in future discussions and refinements of population estimation methodologies. We argue that the reported bias figures are less caused by actual undercounting of rural populations, but more so by contestable methodological decisions and the historic misallocation of (gridded) population estimates on the local level.
The paper titled "Global gridded population datasets systematically underrepresent rural population" by Josias Làng-Ritter et al. provides a valuable contribution to the discourse on the accuracy of global population datasets, particularly in rural areas. We recognize the importance of the work, as it adds significant validation efforts and draws further attention to rural population estimation challenges.This focus is valuable for improving datasets and fostering discussions on uncertainty communication. However, we caution on the conclusion drawn in the paper that the result "implies that rural population is, even in the most accurate dataset, underestimated by half compared to reported figures". Taking this statement seriously, it would constitute an undercounting of the world's population by 1-2.6 billion (by applying the study's reported bias of -80% to -32% for 2010 to the 2010 rural population of 3.24 billion according to the World Bank 1 ). What we claim, here, is that researchers should exert greater caution in their claims to avoid ambiguity and non-academic press coverage with titles such as, "There could be billions more people on Earth than previously thought" 9 . We caution against such bold claims, if not properly backed by evidence, not only because of the methodological limitations outlined below, but also due to their potential to fuel harmful socio-political and environmental narratives as they could amplify climate anxiety, xenophobic or alarmist discourses and skepticism toward science and expert institutions.
We argue that the reported bias figures are less caused by actual undercounting of rural populations, but more so by contestable methodological decisions and the historic misallocation of (gridded) population estimates on the local level. Furthermore, we are emphasising that it is not possible to draw from the reported bias a coherent, global picture of the undercount of rural populations. We would like to highlight several critical points regarding the methodology and conclusions drawn in the study:
The study compares gridded population estimates from approximately ten years before dam completion with reported resettlement figures. However, project inception, usually entailing land acquisition combined with resettlement incentives often begin much earlier-sometimes decades before dam completion. Using two of the three German dam projects included in the study as examples, it took almost 30 years from inception to completion for the Brombachsee project and the settlements in the flooded area were purchased soon after project inception 8 . Thus, ten years before completion, around the time of the reference year, the affected population has left, while still being counted as resettled, which would inflate the bias estimates.
In the case of the second German dam, the Rothsee dam, roughly 13 people still lived in the area at the time of resettlement 7 . While a gridded population estimate of e.g. 7 or 8 for that area is not necessarily a “big miss”, a percentage error (such as the sMAPE used in the study) looks large and may not be appropriate for extrapolating to country-level population numbers. Furthermore, extrapolating from three dam projects (affecting roughly 100 people) built on a methodology that does not properly seem fit-for-purpose to state that over 50% of the German rural population goes uncounted as alluded to in Fig. 7 and Fig. 8 of the study might risk overgeneralization, which also lacks backing from other data collection efforts such as recent census rounds in Germany. We expect that both the temporal misalignment and the metric argument is not limited to the German example, but holds true for other projects under study as well.
gridded population counts it is key to have a reliable link between population count and spatial extent of the enumerated location. The authors noticed that the polygons they used, which are from the GeoDar dataset, have a systematically smaller surface than the reservoir area reported by ICOLD. In order to circumvent the issue, the authors apply a fixed factor of 1.23 (that is an additional 23%) to the population number extracted from the gridded population. This procedure questions the reliability of the ground truth regarding fine-scale population counts.
The study acknowledges, and we further emphasize it, that historical gridded population data are inherently limited due to the lack of detailed settlement maps before 2010. As described in the study, the bias in population counts decreased from -80% to -32% over the period of 2000 to 2010, which contradicts the study’s claim of a consistent underrepresentation of the rural population. Since gridded population layers in this study usually disaggregate census counts with the help of satellite imagery-derived pixel-level maps of human presence, inaccurate spatial patterns of detected human presence are, in our view, an important source of potential bias in hyper-localized population es
This content is AI-processed based on open access ArXiv data.