Global soil moisture dynamics since 1980: datasets biases, trends, and science-informed selection

Area of Science:

• Earth and Environmental Sciences
• Climate Science
• Hydrology

Background:

• Soil moisture is crucial for climate prediction, ecological management, and disaster warning.
• Existing multi-source soil moisture datasets exhibit spatiotemporal inconsistencies and limited regional applicability due to algorithmic and observational constraints.

Purpose of the Study:

• To statistically assess the performance and spatiotemporal variations of 23 global surface soil moisture datasets.
• To evaluate dataset applicability across diverse global regions, classified by Köppen climates and IPCC land uses.

Main Methods:

• Analysis of 23 global surface soil moisture datasets spanning 1980-2023, including reanalysis, land surface models, and microwave remote sensing.
• Dual validation against 992 in situ stations and a filtered subset of 483 highly representative stations.
• Statistical performance assessment using Pearson correlation (R) and analysis of spatiotemporal trends.

Main Results:

• A slight global surface soil moisture decline (-4.30 × 10^-4 m³ m^-3 a^-1) was observed from 1980-2023, with some datasets showing short-term wetting post-2010.
• Most products performed moderately well (R ≈ 0.5-0.7), with microwave remote sensing (especially SMAP) showing superior temporal dynamics capture (R ≈ 0.7).
• Spatial representativeness error significantly impacts validation; filtering stations improved tropical region validation (mean R increase of 0.41).

Conclusions:

• Dataset performance varies, necessitating a science-informed, application-specific selection approach for hydrological and climatic research.
• Microwave remote sensing products demonstrate higher reliability for capturing temporal soil moisture dynamics.
• Addressing spatial representativeness is critical for accurate soil moisture dataset validation and application.