A novel approach to retrieving the surface soil freeze/thaw state in the Qinghai-Tibetan Plateau using the seasonality of CYGNSS time series

Soil freeze–thaw (F/T) processes are a typical physical phenomenon on the Qinghai-Tibetan Plateau (QTP), significantly impacting regional climate change and the hydrological cycle. This study presents a Seasonal-Trend Decomposition using Loess and Long Short-Term Memory (STL-LSTM) method to detect s...

Full description

Saved in:
Bibliographic Details
Main Authors: Qi Liu, Shuangcheng Zhang, Zhongmin Ma, Xin Zhou, Tao Wang
Format: Article
Language:English
Published: Elsevier 2025-03-01
Series:International Journal of Applied Earth Observations and Geoinformation
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1569843225000755
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Soil freeze–thaw (F/T) processes are a typical physical phenomenon on the Qinghai-Tibetan Plateau (QTP), significantly impacting regional climate change and the hydrological cycle. This study presents a Seasonal-Trend Decomposition using Loess and Long Short-Term Memory (STL-LSTM) method to detect spatiotemporal variations in soil F/T on the QTP using time series data from the Cyclone Global Navigation Satellite System (CYGNSS). The model was validated against ERA5 soil temperature data (0–7 cm) and independent in-situ observations, demonstrating good consistency. The SHapley Additive exPlanations (SHAP) model was integrated into the STL-LSTM framework to quantitatively evaluate the contributions of input features to F/T retrieval, revealing that time features contributes the most to retrieval results, followed by surface reflectivity. Moreover, spatiotemporal analysis of QTP F/T dynamics shows prominent seasonal patterns, with topography-induced shielding delaying thawing in central QTP regions and freezing trends extending from low (28°N) to high latitudes (36°N). The proposed method offers a new pathway for monitoring freeze–thaw transitions in high-latitude regions and holds potential for expansion into future high-frequency and multi-polarization GNSS-R missions.
ISSN:1569-8432