EUVIMEDOpdag EUVIMED
Open access · CC BY 4.0

A Spatiotemporal Interpolation Method for Regional Precipitation Data Based on a Spatiotemporal Decay Graph Model

Li Liu, Chuhan Lu, Julong Huang, Feng Zhang, Guangyu Qu, Lu Guo, Runze Luo

Climate · 2026-06-24 · 14 (7) · 136

DOI: 10.3390/cli14070136 ↗

At a glance

Journal article in Climate, published online on 2026-06-24. Written by Li Liu and 6 co-authors. 16 references. The full text is freely available (open access, CC BY 4.0). Funded by: Longnan Science and Technology Plan Projec, Qinglan Project of Jiangsu Province, Wuxi University Research Start-up Fund for Introduced Talents.

AbstractBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Traditional meteorological data spatial interpolation methods often rely on linear or static assumptions, which are inadequate for complex terrain and fail to exploit continuous spatiotemporal variation information. This paper proposes a Spatiotemporal Graph Network with Adaptive Temporal Decay (DG) that integrates a learnable graph convolution module and a temporal attenuation mechanism, enabling accurate precipitation estimation for target stations or regions at consecutive time steps. The method is evaluated using daily precipitation data from nine stations in Longnan City, Gansu Province, China, along with ERA5 (0.25°) and GPCP (0.5°) gridded reanalysis products. In the station-to-station interpolation scenario, DG significantly outperforms ordinary Kriging (OK), reducing the average RMSE from 1.4 mm/day to 1.2 mm/day, with a 28.6% improvement at mountainous stations. The DG model also exhibits superior performance in grid-to-station interpolation, achieving an average RMSE of 1.9 mm/day (OK: 2.5 mm/day). On heavy precipitation days (≥20 mm/day), DG reduces the RMSE nearly by half (11.7 mm/day) compared to OK (23.2 mm/day). A temporal-only LSTM baseline and three ablation variants (spatial-only OSI, temporal-only OTI and dgcn-only OD) are also compared, and DG consistently outperforms them, confirming the essential role of spatiotemporal integration. Additional baselines including IDW and Co-Kriging further validate the superiority of DG. The proposed method offers a promising new approach for high-precision spatiotemporal interpolation of meteorological elements in complex terrain.

Funding

  • Longnan Science and Technology Plan Projec
  • Qinglan Project of Jiangsu Province
  • Wuxi University Research Start-up Fund for Introduced Talents

Cite

APA

Liu, L., Lu, C., Huang, J., Zhang, F., Qu, G., Guo, L., & Luo, R. (2026). A Spatiotemporal Interpolation Method for Regional Precipitation Data Based on a Spatiotemporal Decay Graph Model. Climate, 14(7), 136. https://doi.org/10.3390/cli14070136

Vancouver

Liu L, Lu C, Huang J, Zhang F, Qu G, Guo L, et al. A Spatiotemporal Interpolation Method for Regional Precipitation Data Based on a Spatiotemporal Decay Graph Model. Climate. 2026;14(7):136. doi:10.3390/cli14070136

BibTeX
@article{Liu2026,
  author = {Liu, Li and Lu, Chuhan and Huang, Julong and Zhang, Feng and Qu, Guangyu and Guo, Lu and Luo, Runze},
  title = {A Spatiotemporal Interpolation Method for Regional Precipitation Data Based on a Spatiotemporal Decay Graph Model},
  journal = {Climate},
  year = {2026},
  volume = {14},
  number = {7},
  pages = {136},
  doi = {10.3390/cli14070136}
}
Start for free →