Research Article
Spatial-Temporal Three-Corned Hat for Soil Moisture Uncertainty Evaluation Over the Qinghai-Tibet Plateau
@INPROCEEDINGS{10.4108/eai.24-11-2023.2343429, author={Ling Zhang and Yongxu Wang and Zhaohui Xue}, title={Spatial-Temporal Three-Corned Hat for Soil Moisture Uncertainty Evaluation Over the Qinghai-Tibet Plateau }, proceedings={Proceedings of the International Conference on Industrial Design and Environmental Engineering, IDEE 2023, November 24--26, 2023, Zhengzhou, China}, publisher={EAI}, proceedings_a={IDEE}, year={2024}, month={2}, keywords={soil moisture (sm); three-corned hat (tch); spatial-temporal fusion qinghai-tibet plateau (qtp)}, doi={10.4108/eai.24-11-2023.2343429} }- Ling Zhang
Yongxu Wang
Zhaohui Xue
Year: 2024
Spatial-Temporal Three-Corned Hat for Soil Moisture Uncertainty Evaluation Over the Qinghai-Tibet Plateau
IDEE
EAI
DOI: 10.4108/eai.24-11-2023.2343429
Abstract
Accurate uncertainty evaluation of soil moisture (SM) products is crucial for maximizing their utility in research and applications in hydro-meteorology and climatology. At present, the uncertainty analysis of SM is mostly carried out from the perspective of temporal domain based on time series. Whereas, the influence of spatial heterogeneity when representing spatial errors is usually ignored. To solve this problem, a novel spatial-temporal three-corned hat (ST-TCH) method is proposed for SM uncertainty evaluation. Firstly, a moving window is used to construct the spatialtemporal data cube of SM within the neighborhood. Secondly, the heterogeneous pixels are eliminated based on Spearman correlation coefficient to avoid the interference of heterogeneous pixels. Finally, the 3D spatial-temporal data is vectorized into a sequence which is fed into TCH to produce the relative uncertainty (RU). Experiments are conducted on four SM products over the Qinghai-Tibet plateau (QTP). To quantitatively verify the performance, four products are merged based on the estimated RU, and the merged products are further validated with the in-situ data. Results demonstrate that RU obtained by ST-TCH is more complete in spatial distribution, and the merged product produced by ST-TCH is more close to the in-situ data with R = 0.769 among all products.