Series or Parallel? An Exploration in Coupling Physical Model and Machine Learning Method for Disaggregating Satellite Microwave Soil Moisture

Abstract

Remotely sensed soil moisture (SM) dataset with well accuracy and fine spatiotemporal resolution is very valuable in various fields. Downscaling is a promising way to obtain such an SM dataset. There are currently two basic methodologies in the downscaling with satellite datasets, i.e., the machine learning (ML) methods and the physical or semi-physical (PH) models. This study focuses on exploring feasible ways to integrate them for boosting the performance of downscaling. Here, three parallel modes i.e., arithmetic average (Ari), geometric average (Geo), and weighted average (Wei), and two series modes i.e., first PH then ML (PH-ML) and first ML then PH (ML-PH) were developed and evaluated. A representative of PH, DSCALE_mod16, and three ML algorithms Random Forest, XGBoost, and LightGBM were used in this study. Microwave SM datasets from Soil Moisture Active and Passive Mission (SMAP), satellite datasets from MODIS, and in-situ SM observations spanning from April 2015 to December 2018 were employed in the evaluation. Spatial dynamic range, energy conservation, and precision preservation analyses were conducted as evaluation methods. Results demonstrated the PH-ML series mode outperformed the other coupling modes, which was even better than the better one of the PH and ML. The PH can first generate a valuable initial estimate, and the initial estimate's advantages can be preserved while its errors can be suppressed by the ML. Resultantly, the ideology of first using the PH to obtain an initial estimate and then putting the initial estimate into the ML is suggested for further microwave SM downscaling.

Enhanced Generalized Regression Neural Network With Backward Sequential Feature Selection for Machine-Learning-Driven Soil Moisture Estimation: A Case Study Over the Qinghai-Tibet Plateau

“…To evaluate the performance of different L-band SM products over the tropical and the rainforests areas, a comprehensive evaluation was conducted [11]. Recently, a new research stream is to integrate the machine learning methods and the physical or semiphysical models for boosting the performance of SM estimation or downscaling [12], [13], [14], [15].…”

A Subsurface Microwave Imaging System Based on the Combination of Sub-Band-Subspace Images

“…It is mentioned in this paper that random forests overestimate and underestimate the very low and very high moisture values respectively. In [6], a combinative technique is proposed to enhance deep learning by incorporating features from physical models. This study examines scenarios where the physical model is applied both before and after the ML process, utilizing random forest , extreme gradient boosting , and light gradient boosting machine as the ML approaches.…”