*2.1. SMOS Soil Moisture*

This study uses six years of global SMOS-BEC L3 v.2 daily SM retrievals, starting on 1 June 2010. This product is provided in the 25 km EASE2 equal-area grid and is obtained from the ESA L2 v.620 data, after discarding SM retrievals being potentially affected by radio frequency interferences (RFI flag) or with a Data Quality Index (DQX) greater than 0.07 m3·m<sup>−</sup>3. The DQX is an estimate of the error in the SM retrieval and the brightness temperature measurement accuracy. Products are provided separately from ascending and descending orbits. Data as well as further processing details are available at http://bec.icm.csic.es/. In this work, daily products from both ascending and descending orbits are combined to maximize coverage (further pre-processing is detailed in Section 2.5).

It is relevant to note the impact that RFI has on L-band satellite measurements. Although L-band is an internationally protected band for radio astronomy, it was soon clear after the SMOS launch that anthropogenic RFI exceeded expected levels in many regions worldwide [26]. The situation has now improved, but the presence of RFI still masks SMOS observations, severely limiting its coverage in some regions [27]. The impact of RFI is considerably reduced in SMAP, since a number of hardware and software measures were implemented to detect and where possible mitigate its effects [28]. In this regard, recent studies combining the brightness temperatures of the two missions are particularly promising [29].

## *2.2. Selection of Target Sites*

A set of representative target sites were selected to characterize and illustrate the SMOS SM climatology under contrasting vegetation types and climatic conditions. The analysis of the SMOS STL decomposition at the target sites allowed us to analyze major features of the STL decomposition procedure and adapt its parameterization to SMOS measurements. This is described in detail in Section 2.6. In addition, the target sites were used as a basis for the inter-comparison of SMOS with GLDAS-Noah and ERA5 SM time series.

The 11 terrestrial Transcom regions were used as an initial segmentation of global continental land from where to select representative sites. Transcom regions are based on a 1-degree land cover map and are delimited by climate zones [30]. The advantage of using this classification is that it encloses climates with similar vegetation seasonality into a limited number of classes. After careful analysis of the SMOS SM time series in each region, eight target sites were chosen covering Transcom regions North America Boreal, North America Temperate, South America Tropical, South America Temperate, Europe, Northern Africa, Southern Africa, and Australia. No target sites were chosen from Transcom regions Eurasia Boreal, Eurasia Temperate and Asia Tropical, since availability of SMOS SM data in these regions is very limited due to combined effects of RFI, the presence of snow and high topography. A global map showing the selected target sites and the terrestrial Transcom regions is shown in Figure 1. The specific location of each target site is provided in Table 1. One of the SMOS and SMAP Core Cal/Val sites (REMEDHUS, site E) was chosen as a target site for further analysis *vs* ground-based measurements.

**Figure 1.** Geographical position of the eight locations (triangles, letters A to H) selected to illustrate the main features of SMOS climatology for a variety of vegetation and climatic conditions. Colors represent the 11 terrestrial Transcom regions [30].


**Table 1.** Target sites: name and location.

To establish the actual spatial representativeness of the target sites, a temporal correlation analysis was performed at each selected location. Figure 2 shows, for each target site, the correlation map of its SMOS SM time series with the SMOS SM time series of all the pixels on the globe. It can be seen that the highest correlation is located in the neighborhoods of the chosen pixel (within the Transcom Region), which indicates the target site is representative of their surrounding area, but, as expected, not of the whole region.

**Figure 2.** Correlation maps showing the representativeness of the SMOS time series for the study period at the selected target sites. The exact location of each target site (see Table 1) is marked with a black cross.

#### *2.3. Modeled Soil Moisture*

#### 2.3.1. GLDAS-Noah

Global Land Data Assimilation System (GLDAS) Noah model v.1 top 10 cm soil moisture estimates were used in this study [31]. GLDAS-Noah land surface state and flux products are provided by the Hydrology Data and Information Services Center (HDISC), a component of the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC). GLDAS-Noah is available with a 3-h temporal interval and 0.25◦ spatial resolution, covering dates from 2000 to the present. Data was extracted for the eight selected target sites (see Figure 1) and the 6-year study period. The original unit of GLDAS-Noah SM is kg·m−2, which was converted to volumetric SM (m3·m−3) for this study by considering a top soil layer depth (10 cm) and assuming the density of water in the soil is 1000 kg·m<sup>−</sup>3. After averaging for daily values, 1-day time series were used to construct 18-day temporal average fields every 5 days (the rationale for this filtering is explained in Section 2.5).
