Comparison of Different Machine Learning Approaches for Monthly Satellite-Based Soil Moisture Downscaling over Northeast China

Yangxiaoyue Liu,Xiafang Yue,Yaping Yang,Wenlong Jing

doi:10.3390/rs10010031

Abstract

Although numerous satellite-based soil moisture (SM) products can provide spatiotemporally continuous worldwide datasets, they can hardly be employed in characterizing fine-grained regional land surface processes, owing to their coarse spatial resolution. In this study, we proposed a machine-learning-based method to enhance SM spatial accuracy and improve the availability of SM data. Four machine learning algorithms, including classification and regression trees (CART), K-nearest neighbors (KNN), Bayesian (BAYE), and random forests (RF), were implemented to downscale the monthly European Space Agency Climate Change Initiative (ESA CCI) SM product from 25-km to 1-km spatial resolution. During the regression, the land surface temperature (including daytime temperature, nighttime temperature, and diurnal fluctuation temperature), normalized difference vegetation index, surface reflections (red band, blue band, NIR band and MIR band), and digital elevation model were taken as explanatory variables to produce fine spatial resolution SM. We chose Northeast China as the study area and acquired corresponding SM data from 2003 to 2012 in unfrozen seasons. The reconstructed SM datasets were validated against in-situ measurements. The results showed that the RF-downscaled results had superior matching performance to both ESA CCI SM and in-situ measurements, and can positively respond to precipitation variation. Additionally, the RF was less affected by parameters, which revealed its robustness. Both CART and KNN ranked second. Compared to KNN, CART had a relatively close correlation with the validation data, but KNN showed preferable precision. Moreover, BAYE ranked last with significantly abnormal regression values.

Highlights

Soil moisture (SM) is a key indicator for characterizing agricultural drought, hydrological processes, land surface evapotranspiration, and regional climate change [1,2,3,4]
There are various types of satellites that provide near real-time SM products, such as the passive microwave-based Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) [21], Advanced Microwave Scanning Radiometer 2 (AMSR2) [22], Soil Moisture Ocean Salinity (SMOS) [23], WindSat [24], active microwave-based Advanced Land Observation Satellite-Phased Array type
Its superiority is embodied in relative fast training speed, and the performance optimization process improves the accuracy of the random forests (RF) model [65]

Summary

Introduction

Soil moisture (SM) is a key indicator for characterizing agricultural drought, hydrological processes, land surface evapotranspiration, and regional climate change [1,2,3,4]. The SM observations provided by ground-based networks Ecosystem Research Field Observational Stations Network [11]) have been effective sources of long time series of regional soil water data [12,13,14]. Considering that there are a limited number of ground stations with uneven distribution in an observation network, it is hard to reflect the SM of an entire region on the same scale [15,16,17,18]. Because every single station only represents the SM of a restricted homogeneous region and all the in-situ measurements could hardly cover the entire required time span, it is less suitable to use ground station data for broad and long-term analysis [12,13,14]. IntroductionsIntroductions of the four downscaling employed intothis study the to compare the performance of downscaling

Methods

Results

Discussion

Conclusion