**1. Introduction**

Soil moisture (SM) is an essential parameter for the hydrology and energy cycle. Rapid acquiring and accurate monitoring of terrestrial SM is not only required in the hydrological research but also a significant benefit to water management and agricultural production. Since the L-band microwave has a strong sensitivity to the change of surface SM and can more easily penetrate the atmosphere and vegetation canopy, it has been widely used as the main soil moisture remote sensing frequency band in the satellite-based radiometer and radar missions [1]. Such as the European Space Agency's (ESA) Soil Moisture and Ocean Salinity (SMOS) mission and the National Aeronautics and Space Administration's (NASA) Soil Moisture Active Passive (SMAP) mission, both can provide global SM measurement with the spatial resolution on the order of 40 km and coverage every 2–3 days using carried L-band radiometer. Spaceborne global navigation satellite system-reflectometry (GNSS-R) is an innovative and sustainable low-cost technique with high spatial and temporal resolution [2], which operates as a passive bistatic forward scattering radar. The observe system directly receives the pre-existing signals transmitted by the GNSS satellites reflected off the Earth's surface [3], and the received scattering signals are typically expressed in a delay-Doppler map (DDM) for Earth's surface geophysical parameters retrieval [4], which provide a new paradigm in the land remote sensing to cover the space-time gap of the traditional high-cost dedicated monostatic active or passive satellite missions.

**Citation:** Dong, Z.; Jin, S. Evaluation of the Land GNSS-Reflected DDM Coherence on Soil Moisture Estimation from CYGNSS Data. *Remote Sens.* **2021**, *13*, 570. https:// doi.org/10.3390/rs13040570

Academic Editor: Hugo Carreno-Luengo

Received: 7 December 2020 Accepted: 1 February 2021 Published: 5 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

In the past decade, spaceborne GNSS-R has undergone rapid development with successfully deployed satellite missions, such as the UK Technology Demonstration Satellite-1 (TDS-1, launched in July 2014), NASA's cyclone global navigation satellite system (CYGNSS, launched in December 2016), China's Bufeng-1 A/B mission (launched in June 2019) [5]. Although all missions were originally designed for ocean surface wind speed retrieval, they also provided a large number of land observations for terrestrial remote sensing applications, such as soil moisture retrieval, forest biomass estimation, and wetland extent detection [6]. However, there are many differences between GNSS-R land and ocean applications [7]. Before using GNSS-R for geophysical parameters retrieval, the key issue is to determine the scattering mechanisms of observed DDM. Over the sea surface, the surface height standard deviation is at least a significant fraction of the signal wavelength under windy conditions and increases with the wind speed [8,9], so the L-band GNSS signal echoes from the ocean are purely incoherent, which can be well explained by the Z-V model [10]. Compared to the ocean surface, the L-band signal scattering from the land surface is more complicated, and the GNSS signal returns are affected by many factors, such as soil moisture, vegetation, surface roughness, inland water, topographic relief, and soil texture. The DDM generated after noncoherent integration loses phase information, and the land surface small-scale roughness is variant in space and time, which is extremely difficult to be determined. As a result, it is hard to distinguish the coherence of land reflected DDM, which affects its subsequent land applications.

In previous GNSS-R land applications, it has been generally assumed that the coherent component dominates the land scattering field, and the incoherent component is negligible. The coherence defined here refers to reflected signals from the first Fresnel zone arriving at the GNSS-R receiver with similar phase shifts [9]. Many studies have proved that coherent DDM derived reflectivity is sensitive to the change of soil moisture and forest biomass [11–15]. However, due to the sensitivity of coherent and incoherent observation on the land geophysical parameters is different, it is important to distinguish the coherence of observations for quantified parameter retrieval. Theoretical simulations have revealed that the roughness of the land surface was close to 5 cm, where only incoherent scattering will occur [9]. Meanwhile, the effect of topography is independent of surface roughness, and the topographic relief can mitigate the reflectivity [16]. Different DDM observables have been used for GNSS-R sea ice detection based on the difference of coherent reflected signal from the sea ice surface and diffuse scattering from the sea surface [17,18]. However, it is relatively difficult to verify the coherence of current ground CYGNSS data. The coherence of a single complex DDM look can be robustly distinguished based on the differences of coherent and noncoherent integration from the "raw IF" signal [19] because the correlated power of a perfectly coherent signal will increase over the given period from longer integration lengths, while the incoherent will not. Unfortunately, the CYGNSS mission only recorded very few I/F signals limited by its storage capability. Nevertheless, with the help of these I/F signals from the land surface, different estimators have been characterized in the studies for DDM coherence detection [7,20], and the results show that the purely coherent reflection only occurs over the inland water surface in spaceborne GNSS-R observation. The problem is that the differences in estimator performance can lead to different results, and the I/F signal dataset used is too small, lacking sufficient persuasive power. Based on the different assumptions, several SM inversion methods have also been developed, such as spatial averaging, combine linear regression method, machine-learning method, and the global inversion accuracy of SM can reach about 0.05 cm3/cm3 [21–30].

In this paper, a statistical method is developed to detect the coherence of CYGNSS level-1 DDM from the land. We assume that the delay-Doppler-spreading features of incoherent DDM from the ocean and land scattering are similar, which all present a typical "horseshoe" shape, only the magnitude of the absolute scattering power differs. The defined estimators are used to determine the flag of coherence in terms of known incoherent DDM from the windy ocean surface, and the inversion accuracy of GNSS-R derived soil moisture with high confidence coherent DDM is evaluated and validated. The paper is organized as follows, Section 2 introduces the scattering theory over the sea surface and smooth soil surface and the definition of the coherent classification estimators based on the difference of typical coherent and incoherent dominated DDM. Section 3 shows the classification performance of different estimators, the distribution characteristics of coherent and incoherent observation over the land surface, and presents soil moisture retrieval results. Section 4 discusses the impact of coherent and incoherent DDM on SM remote sensing applications. Finally, conclusions are summarized in Section 5.
