**1. Introduction**

L-band radiometry is the most established technique for remotely measuring soil moisture. Currently, there are two L-band missions in orbit specifically designed to globally monitor soil moisture: the Soil Moisture and Ocean Salinity (SMOS) and the Soil Moisture Active Passive (SMAP). SMOS, launched in 2009 by the European Space Agency (ESA), uses a synthetic aperture radiometer with a spatial resolution of ~35–50 km to provide global soil moisture maps every three days [1]. The Centre Aval de Traitement des Données SMOS (CATDS) provides several soil moisture products (L2, L3, and L4), which are processed with the algorithms developed by the Centre d'Etudes Spatiales de la Biosphere (CESBIO). In 2015, the National Aeronautics and Space Administration (NASA) launched SMAP. It currently employs a real aperture radiometer with ~40 km resolution to retrieve soil moisture maps with a three-day revisit time [2].

While these remote sensing sensors provide soil moisture data at coarse spatial resolutions, a growing number of applications require knowledge of soil moisture at regional or local scales (from a few kilometers down to several meters). To overcome this challenge, some soil moisture disaggregation approaches have been developed, such as those based on the synergy of passive microwaves with ancillary optical visible/infrared (VIS/IR) [3–6] or active data [7]. Optical data have also been used to indirectly estimate soil moisture [8]. For instance, methods based on apparent thermal inertia (ATI) to estimate soil moisture rely upon the fact that wet soils have a higher thermal inertia and a lower

temperature fluctuation than dry soils. When soil moisture increases, ATI proportionally increases as well, and there is a short-term reduction in the diurnal land surface temperature (LST) range [9–13].

A main drawback of the L-band observations is that although they have larger penetration than those at higher microwaves (S-, C-, X-, or K-bands), they explore approximately 0–5 cm of the topsoil layer, sensing only surface soil moisture (SSM). However, an increasing number of hydrological and agricultural applications require root zone soil moisture (RZSM) information from the soil profile (0–1 m depth), where plant roots develop [14,15]. Additionally, the SSM may not have significant influence in the soil water availability for plants and crops. For this reason, the RZSM is expected to better reflect the actual soil water content storage of the unsaturated zone than the SSM.

There are several methods for obtaining RZSM. One method comprises in situ soil moisture measurements made at the root zone using probes installed either at a specific depth required or along the whole soil profile. In this regard, although neutron attenuation probes were extensively used in the past, sensors based on the soil dielectric constant—capacitance, Frequency Domain Reflectivity (FDR) and Time Domain Reflectivity (TDR) probes—are now being employed in many networks worldwide [16]. Another method involves using cosmic-ray soil moisture probes. These innovative and noninvasive sensors have been implemented in the Cosmic-ray Soil Moisture Observing System (COSMOS) network [17,18]. The cosmic-ray sensors are placed above the soil surface, but its effective soil measurement depth varies from 12 cm for wet soils to 76 cm for dry soils [19]. Nevertheless, only a limited number of current networks of the International Soil Moisture Network (ISMN) provide RZSM compared to the great amount that provide SSM. Another method relies on the use of Ground Penetration Radar (GPR) measurements. In this case, the GPR sensor can be mounted on a vehicle close the soil surface or an airplane to measure soil moisture during experimental field campaigns. The GPR does not require direct contact with the soil, but its signal can penetrate from one meter to several tens of meters [20]. The last method consists of estimating RZSM through more or less complex models that have this variable as the output.

The models used to estimate RZSM can be classified into two main groups: the so-called land-surface models and the hydrological models. They differ in the detail of description of processes that are taken into account, the parameter estimation approaches, and the spatiotemporal resolutions. Land-surface models describe the vertical exchanges of heat, water, and carbon considering the land-atmosphere couplings and are globally applied. Hydrological models are instead more focused on water resources, are traditionally applied at the basin level, and usually have many parameters that need to be calibrated or estimated regionally [21]. Different data assimilation techniques are used in the two cases to incorporate SSM to them [22–24]. Originally, the assimilated SSM data to estimate RZSM had been measured in situ [25,26], but a variety of satellite SSM data have been assimilated in the last decade [27–29]. In the case of SMAP and SMOS missions, two operational satellite-based RZSM products have recently been developed from the assimilation of SSM measurements into their respective land-surface models [30,31].

The Soil Water Index (SWI) is one of the most common models used to estimate RZSM through SSM remote sensing [32]. This simple model has been able to successfully obtain RZSM over regions with different climatic and soil conditions. Apart from SSM measurements, the SWI requires only an input exponential parameter (T), which is related to the transfer time of water along the soil profile. The T parameter had been calculated in different ways depending on the application, study area, and sensor used [33]. The SWI has been applied to in situ SSM databases in several studies to obtain field scale RZSM [34–36] as well as to active and passive SSM observations to generate several satellite-based RZSM estimates, such as those derived from the European Remote Sensing (ERS) scatterometer [37], the Advanced Scatterometer (ASCAT) [38–40], the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) [39,41], the SMOS [42,43], and the Climate Change Initiative (CCI) soil moisture database [41,44].

The aim of this work was to evaluate six RZSM estimates obtained from SMAP, SMOS, and Moderate Resolution Imaging Spectroradiometer (MODIS) from 31 March 2015 to 31 December

2016. The study period (one year and nine months) is limited at the beginning by the SMAP launch and at the end by the availability of two SMOS soil moisture products—the SMOS-CESBIO L4 RZSM and the SMOS-Barcelona Expert Centre (BEC) L4 SSM. The first two RZSM estimates came from the SMAP and the SMOS-CESBIO L4 RZSM products. The other four RZSM estimates were customized products generated after applying the SWI model to two SSM datasets—the SMOS-BEC L4 SSM and the MODIS ATI-derived SSM—together with two alternatives for calculating the exponential T parameter of the SWI. Presently, some studies devoted to validate SMAP and SMOS RZSM have been published, but none have been made validating both RZSM products at the same network or showing an intercomparison between them over the same study area, including a RZSM estimation based on ATI. Thus, the present work will constitute a novelty within this research line. In this study, all satellite-based RZSM estimates were analyzed both temporarily and spatially. For the temporal analysis, the RZSM estimates were compared against in situ RZSM measurements from 14 stations of the Soil Moisture Measurements Station Network of the University of Salamanca (REMEDHUS). The spatial analysis was based on comparisons of RZSM maps from all the analyzed estimates over the entire Iberian Peninsula (~582,000 km2), an area of contrasting environments in both wet and dry periods.
