Beyond global metrics: A geographically weighted framework for exploring multi-source spatial data validation

With the increasing availability of multi-source spatial datasets, ensuring data consistency and accuracy has become a critical challenge in spatial analysis. Traditional global indicators such as Mean Error (ME), Mean Absolute Error (MAE), Mean Relative Error (MRE), Root Mean Square Error (RMSE), Correlation Coefficient (CC), and Univariate Linear Regression (ULR) have been widely used to evaluate overall discrepancies between pair-wise datasets. However, these global metrics fail to reveal spatial details of discrepancies between these datasets. To address this limitation, we established a spatially explicit and scale aware validation framework by embedding conventional global indicators within a geographically weighted (GW) modeling context by introducing GW ME, GW MAE, GW MRE, GW RMSE, GW CC, and univariate GWR to provide spatially detailed assessments of discrepancies. The study also emphasizes the crucial role of bandwidth selection in determining the scale of analysis, with large bandwidths smoothing local variations and small bandwidths preserving fine-scale details. Using GPWv4 data, the Gridded Population of China Dataset (GPCD) and the Seventh National Population Census (NPC7) of China uniformly collected in 2020 as a case study, we validate the effectiveness of this approach at the county level. Results show that GW indicators successfully reveal spatial heterogeneities in data discrepancies, which are overlooked by global measures. It is noteworthy that the local validation framework is always straightforward to apply to other types of spatial data sets. Overall, this study provides a systematic and scalable method for multi-source spatial data validation to enhance quality control, improve integration strategies, and facilitate more informed decision-making in spatial analysis and geographic studies.