Enhancing Soil Moisture Data Reliability in South Korea: Advanced Quality Control and Ensemble Gap-filling of the KIHS Network

High-resolution soil moisture data is a critical component for understanding the hydrological cycle and establishing climate adaptation strategies, particularly in the complex mountainous terrains of the Far East Asian region. Recognizing the significance of this data within the southern part of the Korean Peninsula, the Korea Institute of Hydrological Survey operates in-situ soil moisture monitoring networks to provide standardized, high-quality hydrological data. Located in mountainous regions with long-term operational history, these networks are co-located with evapotranspiration and streamflow stations, facilitating efficient and integrated water balance studies.

To ensure high data reliability for global research applications, KIHS implements a multi-stage quality control (QC) framework for its SM datasets. We have developed an automated outlier detection system based on the International Soil Moisture Network (ISMN) protocols to identify and filter physical anomalies such as spike, break and constant values. Furthermore, to provide continuous data, KIHS utilizes a hybrid framework of statistical methods and machine learning algorithms for gap-filling. This framework integrates CDF Matching, Kalman Filter, and SARIMAX with non-linear models like Random Forest and KNN, ensuring robust and continuous time-series data even under challenging field conditions.

These high-quality datasets are shared internationally through ISMN and are highly recommended for the calibration and validation of satellite products such as SMAP and Sentinel, particularly during the non-frozen period from April to November. The objective of this presentation is to present KIHS's soil moisture monitoring networks and QC methodologies and to demonstrate the academic significance of soil moisture observation stations in the Korean Peninsula.

keywords : soil moisture, the Korea Peninsula, mountainous terrain, monitoring networks, long-term operation, QC frameworks