PUBLISHEDDD PAPERR

International Journal of Remote Sensing ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tres20 Soil moisture mapping from SMOS: evaluating the accuracy of the operational product in a Mediterranean setting Triantafyllia Petsini & George P. Petropoulos To cite this article: Triantafyllia Petsini & George P. Petropoulos (2024) Soil moisture mapping from SMOS: evaluating the accuracy of the operational product in a Mediterranean setting, International Journal of Remote Sensing, 45:2, 508-531, DOI: 10.1080/01431161.2023.2295838 To link to this article: https://doi.org/10.1080/01431161.2023.2295838 Published online: 25 Jan 2024. Submit your article to this journal View related articles View Crossmark data Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=tres20 INTERNATIONAL JOURNAL OF REMOTE SENSING 2024, VOL. 45, NO. 2, 508–531 https://doi.org/10.1080/01431161.2023.2295838 Soil moisture mapping from SMOS: evaluating the accuracy of the operational product in a Mediterranean setting Triantafyllia Petsini and George P. Petropoulos Department of Geography, Harokopio University of Athens, Athens, Greece ABSTRACT ARTICLE HISTORY The present study provides a thorough evaluation of the SMOS surface soil moisture (SSM) product in a typical Mediterranean setting in Greece. For this purpose, a total of 4 agricultural sites were used for which co-orbital in-situ measurements from ground SSM sensors were available for year 2020. In this context, the effect of topographical and geomorphological features, land use/cover and the satellite orbit type and the Radio Frequency Interference (RFI) were also examined. A series of statistical metrics were computed, which allowed evaluating the agreement between the 2 datasets. In overall, results showed a reasonable agreement in specific land use/cover types between the SMOS product and the corresponding in-situ measurements obtained from the 0–5 cm soil moisture layer. In most cases, Root Mean Square Difference (RMSD) was close to 0.15 m3 m−3 (minimum 0.126 m3 m−3). Tomato and vineyard showed a satisfactory agreement compared to walnut and cotton crops. The autumn period had the highest agreement for tomato crop. The effect of RFI was also quite noticeable, as after the exclusion of pixels with high RFI, statistical agreement was noticeably improved. This study is, to our knowledge, one of the few that investigates in a Greek setting the accuracy of the SMOS product. The study results can contribute to the understanding of the practical value of the SMOS product in agricultural and arid/ semi-arid Mediterranean environments while support efforts ongoing globally aiming at improving the SMOS SSM product accuracy. Received 20 October 2023 Accepted 8 December 2023 KEYWORDS Soil moisture; remote sensing; SMOS; validation; agriculture; Mediterranean setting 1. Introduction Soil moisture is a key element of the hydrological cycle as it links atmospheric precipitation and underground water (Babaeian et al. 2019; Li et al. 2021). It has been long recognized as a key state variable of the global energy and water cycle due to its control on exchanges of energy and matter and physical processes such as the partitioning of available energy at the Earth’s surface into latent (LE) and sensible (H) heat exchange (Bao et al. 2018; Deng et al. 2019). Furthermore, it can help with sustainable water resource management, understanding ecological processes and ecosystems (Maltese et al. 2015), enhancing productivity and plant growth and determining plant water demands (Fuzzo et al. 2019; Shi and Liang 2014). The knowledge of spatiotemporal variability of SSM is of crucial importance in disaster events, such as drought, floods, as it can aid in enhancing CONTACT Triantafyllia Petsini [email protected] Venizelou St. 70, Athens 17671, Greece © 2024 Informa UK Limited, trading as Taylor & Francis Group Department of Geography, Harokopio University of Athens, El. INTERNATIONAL JOURNAL OF REMOTE SENSING 509 agricultural production, managing water resources and developing a climate circulation model globally and regionally (Gupta et al. 2021; Liu and Yang 2022). SSM is also required in decision-making tools, at national level, helping in the use and water supply for agricultural production and food availability (Srivastava et al. 2013). Because of climate change and global concerns associated to water security and global food (Deng et al. 2019), receiving information on the spatiotemporal availability of SSM is of vital importance now, particularly so in water-limited environments. Worldwide, water scarcity poses a serious problem (World Economic Forum 2020). In recent years, water consumption has been increasing by 1% annually (UNESCO 2020). A quarter of the world’s population is concentrated in areas with water scarcity (FAO 2015). At the same time, climate change is a factor affecting water quantity and precipitation and generates considerable uncertainty. For those reasons, it is required to find solutions for water management, particularly in semi-arid areas on which there is limited availability of water. In those areas, water scarcity is worsening as a result of the climate change, which clearly affects the production of ecosystems and crops as well as the health of living beings (O’Neill and Boyer 2020). Increased water scarcity will be a major challenge for adaptation to climate change (Molénat et al. 2023). Key societal challenges are to be faced by regions in relation to food and water security since it is climate change that affects their availability due to its impact on the food production process. It is necessary to achieve climate change mitigation to ensure and preserve food security and nutrition for all people (IPCC 2019). The Mediterranean is an example of one of the most vulnerable regions to climate change impacts due to its semi-arid environment characterized by increased temperatures and decreased precipitation annually (Sun 2016). Soil moisture measurements can be obtained using a variety of methods utilizing ground instrumentation such as probes and gravimetric measurements. At smaller scales, there are numerous approaches utilizing ground instrumentation (Petropoulos, Ireland, and Barrett 2015). Generally, the use of ground instrumentation has some advantages and disadvantages. Key advantages include the easy installation, the relatively direct instrument, operation and maintenance, the ability to provide measurement at different depths, the relative maturity of the methods and also the instrument portability (Petropoulos 2013). On the other hand, due to its varied spatial and temporal distribution, in-situ measurements of soil moisture may not be helpful for large-scale applications (Gupta et al. 2014). This happens because those techniques are labour-intensive, expensive, complex and in some cases, can be destructive, such as in gravimetric sampling (Zhang et al. 2014). In addition, extensive equipment deployment is often required to be set up if information on this parameter is required at larger spatial and temporal scales. Earth Observation (EO) is regarded nowadays as the optimum way to obtain spatiotemporal estimates of SSM across different sizes of geographical areas (Petropoulos et al. 2018). The increased availability of EO data particularly over the last decade or so, has led to the development of more and new methods for measuring soil moisture exploiting spectral information acquired at different parts of the electromagnetic spectrum. The scientific maturity of the developments in EO technology is evidenced and from the fact that several operational products of SSM are currently available from different sensors. Many studies have demonstrated the use optical and thermal infrared EO datasets, as well as active and passive microwave (MW) to estimate SSM. In particular, the most widely used spectral region for soil moisture estimation is considered to be the MW part of 510 T. PETSINI AND G. P. PETROPOULOS electromagnetic radiation. The direct relationship between soil moisture and soil dielectric constant explains this development. The protected MW range 1–2 GHz (L-band) is most promising for SSM estimation while the majority of MW sensors operate at higher frequencies (>5 GHz). The factors that explain this characteristic are the resistance of the sensor to surface roughness, the presence of vegetation canopies, atmospheric effects and the high sensitivity of the sensor to soil moisture. In recent years, there are many satellites in orbit providing data suitable for soil moisture estimation. Those include the Advanced Scatterometer (ASCAT) mission, NASA’s Soil Moisture Active Passive (SMAP) mission and European Space Agency’s Soil Moisture Ocean Salinity (SMOS) mission. ASCAT has been launched, specifically for soil moisture retrieval, with a spatial resolution of ~ 25-50 km2 and daily revisit period. Continuing, SMAP is characterized by a a spatial resolution of ~36 km2 and a 2-day revisit period. The European Space Agency’s (ESA) Soil Moisture Ocean Salinity (SMOS) mission includes the first operational L-band radiometer capable of measuring SSM at a spatial resolution of 40 km (Djamai et al. 2015; Kerr et al. 2012). Liu and Yang (2022), one can find a brief overview of available operational products for soil moisture estimation followed by their main properties. The L-band is one of the most suitable microwave for soil moisture assessment as it has low sensitivity to vegetation cover and surface roughness. While soil moisture measurement has been found to be effective at a global level, the systematic study of the SMOS MIRAS instrument has so far been addressed by very few studies. Numerous studies have been conducted focusing on evaluating the accuracy of SSM retrievals provided by a range of operational products (Deng et al. 2019; Piles et al. 2016). In the majority of them, ground measurements from validated observational networks have been employed as reference data against which satellite-derived SSM predictions were compared. For instance, Srivastava et al. (2013a), examined the soil moisture content of SMOS satellite data at the watershed scale for hydrological applications. Another such study, by Deng et al. (2019a), investigated the accuracy of the SMOS SM product for two continents, Europe and U.S.A., and for different types of land cover. SMOS satellite data were compared to FLUXNET network in situ SSM observations. In another study, recently Wang et al. (2021) utilized new soil moisture estimates (CLDAS-BPNN) to assess the SMOS operational product with an innovative manner, fusing a soil moisture product of the land surface model (CLDAS) with field observations through a back propagation neural network (BPNN) method. It showed that the use of neural network method improves the results compared to the two previous ones and thus minimizes the soil moisture measurement uncertainty. Such studies have shown that the SMOS SSM product shows reasonable agreement with the field observations. Petropoulos et al. (2014) compared the SMOS operational product’s accuracy to sites in comparison with stations from the REMEDHUS International Soil Moisture Network (ISMN), in Spain. Petropoulos et al. (2015b) assessed the accuracy of the global operational soil moisture SMOS product using a selection of European stations encompassing a range of meteorological, environmental, and seasonal factors. SMOS estimates were compared with the CarboEurope ground-based observation network’s matching field measurements. These two studies were evaluated on the basis of seasonality and Radio Frequency Interference (RFI). In both cases, the accuracy was high, especially when the comparison was performed by filtering out high interference days. In a similar approach Lamptey et al. (2023), evaluated the performance of SMOS satellite estimates compared to in situ measurements, for three INTERNATIONAL JOURNAL OF REMOTE SENSING 511 different monitoring station depths, in West Africa. From this study, it was found that as the depth of the measurement instrument increases, the error and bias deteriorate. Also, the analysis was carried out for both ascending and descending orbit estimates as well as for the dry and wet season of the year. The overall results gave high agreement mainly for the dry season. In the published literature, elimination of high RFI values has shown an improved accuracy of results. The International Validation Protocol (IVP) for operational products was used in these studies to compute their statistical metrics. However, to our knowledge, although many studies have focused on the assessment of soil moisture products either through the SMOS satellite or other sensors, studies on the assessment of soil moisture from SMOS in a Mediterranean environment are scarce. Thus, this study can help to evaluate the SMOS product algorithm in Greece and at the same time lead to a rapid development of the sectors of the Greek agricultural economy by improving the management of water resources in the Mediterranean basin. In the purview of the above, this study aims at exploring – to our knowledge – for the first time the accuracy of the soil moisture product SMOS-derived in a typical agricultural Mediterranean environment located in Greece. The main study objectives are to: (1) evaluate the accuracy of SM operational products from SMOS satellite data in a semi-arid region between annual and tree crops for the year 2020, and, (2) assess the effect of Radio Frequency Interference (RFI), land use/cover, seasonality, satellite orbit, and topographical and geomorphological features as well as rainfall on SSM retrievals by the SMOS satellite. 2. Experimental set up 2.1. Study area and in-situ measurements In-situ soil moisture measurements were acquired from a network of field monitoring stations for the Greek area which measures a number of parameters, including SSM. The measurements included in this total data set have been taken by the ‘Drill and Drop (DnD)’ and ‘Environscan’ instruments at a depth 5 and 10 cm, respectively. Further information on how the in-situ soil moisture measurements were measured and recorded is provided by the company providing the data. In this study, field measurements were collected from the ‘DnD’ soil moisture sensor from network locations in the Larisa area for the whole year 2020. Larisa is in the central part of Greece and belongs both geographically and administratively to the region of Thessaly. It covers an area of 5.381 km2 and occupies a geographic position of major importance, as it has a missing link between the north and south of Greece. A remarkable feature of Larisa’s terrain is the presence of extensive lowland areas, with the plain of Larisa being one of the largest, covering an area of 589 km2. A key characteristic of this plain is its fertile soil. The altitude of each site presents low fluctuations, and the slopes of the topography are observed to be relatively flat (Figure 1). The climate of Larisa is characterized as continental with average annual temperature 15– 17°C and average annual rainfall is 426.2 mm. The rainfall that occurs in the region causes serious impacts on crops and agricultural land causing excessive soil moisture and making farmers’ work more difficult. The land cover is dominated by crops. It is an area of intense agricultural activity and due to its rich geomorphology and hydrographic network is considered an ideal place for the evaluation of satellite SSM retrievals and in-situ measurements. Thessaly is one of the most economically developing regions in Greece due to its vast 512 T. PETSINI AND G. P. PETROPOULOS Figure 1. An overview of the elevation of study area based on digital elevation Model (DEM). The DEM is taken from USGS EarthExplorer. arable land. The region represents about 10% of the farms and 12% of the cultivated land in the whole of Greece. A wide variety of crops are grown in Larisa such as cotton, wheat, cereals, tomatoes as well as other products such as nuts and olives. Specifically, regarding the production of cotton in Greece, Larisa produces about 40%. At the same time, at European level, the production of cotton in Greece covers 80%. The area consists of 24 SM measuring stations installed at a variety of land uses. Among them, four important sites were chosen to be used in the present study, covering mainly annual and tree crops. The selection of experimental sites was based on the satisfaction of several criteria that have been used in previous studies (Deng et al. 2019a, 2019b). In particular, the criteria used in selecting the test sites included the following: (1) stations with insufficient data, in the study year, were removed. (2) Also, it is important to there is homogeneity of each site within a radius of 1 km. The 1 km radius allows better homogeneity in the cover area of each station. Choosing a larger radius could lead to significant variation within the area while choosing a smaller radius may not record enough diversity in the environment and may lead to a higher density of ground stations which may increase the complexity of the study and data management. The homogeneity of stations is quite important for the study as it increases the validity of the results because stations in similar environments are compared. This ensures that the data collected are representative of the surrounding area. (3), Also, there is necessarily a variability in vegetation type so that there are both tree and annual crops. (4) Another factor of selected sites was the avoidance of proximity to one station to another. The distance between stations is estimated at a minimum of 30 km. (5) The last reason for selecting these stations was their dispersion spatially throughout on the whole study area (Figure 2). INTERNATIONAL JOURNAL OF REMOTE SENSING 513 Figure 2. Overview of the study areas using images from google earth: (a) Site_211-walnut, (b) site 181-cotton, (c) site 180-tomato, (d) site 353-vineyard. A detailed description of the validation sites is provided in Table 1. Briefly, ID Sites 211 and 353 are tree crops and ID Sites 180 and 181 are annual crops. Table 1. Main characteristics of study sites that used SMOS L2 SM estimates. Location ID 211 181 180 353 Location Name Elassona 4 Larisa 16 Larisa 15 Larisa 28 Land Cover Type Walnut Cotton Tomato Vineyard Start date 1/1/2020 11/6/2020 20/6/2020 17/6/2020 End date 31/12/2020 16/9/2020 15/9/2020 31/12/2020 In addition, rainfall data from the Power Data Access Viewer are utilized as supplementary data, in this study, to assess the product accuracy as the study is carried out over a one-year period and can thus offer helpful insights for future algorithm improvement. 2.2. SMOS soil moisture product Soil Moisture and Ocean Salinity (SMOS) is one of the main missions of the European Space Agency (ESA). This is the first L-band MW mission to provide observations of soil moisture over land and ocean salinity worldwide because of the constant fluctuation in the Earth’s water cycle between the atmosphere, land and oceans. The SMOS satellite was launched in November 2009 in the framework of their Earth Explorer Opportunity Missions. Its orbit is almost circular ranging from 761.3 to 788.4 km with equatorial local crossing times of 6 a.m. (ascending) and 6 p.m. (descending), and it is still functional surpassing 5 years of its proposed service period. The SMOS platform’s main instrument is 514 T. PETSINI AND G. P. PETROPOULOS MIRAS. This instrument is a dual polarized L-band 2-D passive interferometer radiometer, spaceborne, polar orbiting, and operating in the 1.4 GHz protected band. This sensor is designed to provide soil moisture estimates with global near-surface coverage and 4% accuracy, a spatial resolution of about 40 km, a three-day revisit at the equator and neardaily at the pole (Kerr et al. 2001; Petropoulos, Ireland, and Srivastava 2015; Srivastava, Petropoulos, and Kerr 2016). The SMOS SSM product is distributed in the Discrete Global Grid (DGG) which is an Icosahedral Snyder Equal Area Earth grid (ISEA-4H9) with equally spaced nodes at roughly 15 km. However, the SMOS mission centre offers sets of data for four levels characterized as Level 1 (L1), Level 2 (L2), Level 3 (L3) and Level 4 (L4). In this study, the SMOS Soil Moisture Level 2 User Data Product (SMDUP2) was used to acquire satellite soil moisture estimates. A total of 167 SMOS images (product version 07) was acquired covering the entire year of 2020. Both ascending and descending orbits were selected. All images were acquired from the ESA portal (https://earth.esa.int/eogate way/missions/smos/data). 3. Methodology 3.1. Data pre-processing After the SMOS SSM images were obtained, the data were loaded into the SNAP software, and the SMOS toolbox was used for further processing. SMOS L2 images were loaded into the software, in four batches according to the stations selected from the field data. The next step was to enter the geographical coordinates (longitude and latitude) of the sites as ‘Pins’ in the software. In this way, for the selected sites, the extraction of SM estimates was carried out. Then, ’.csv’ files were created which were extracted from the output data for Excel analysis. For statistical analysis, Excel worksheets with the extracted files from the SNAP software were created. Evaluating the accuracy of SMOS L2 with the field measurements and considering that the satellite recording time is measured in UTC, the measurement times of the ground data corresponding to the ascending (6.00 am) and descending (6.00 pm) times of the SMOS satellite downloads were used (Dente, Su, and Wen 2012). Then, the soil moisture values of the field data were converted to m3 m−3 to correlate with the retrieved corresponding values from the satellite data. The conversion was carried out through Excel calculations on the original value of the measured soil moisture which was expressed as a percentage. In Figure 3, the spatiotemporal variability of soil moisture between two pre-processed images is shown. Quality control based on the available quality flags is usually necessary before the statistical comparisons are carried out (Wang et al. 2021). The output data from SMOS SSM is supplemented with auxiliary data that reflects the accuracy of the retrieved parameters. Data Quality Index (DQX) is the quality flag that characterizes the product. Thus, after multiple filter applications for the appropriate DQX value, the soil moisture values with DQX < 0.04 were selected as they had the highest accuracy. INTERNATIONAL JOURNAL OF REMOTE SENSING 515 Figure 3. Two examples of pre-processed images, SMOS SM distribution in the ground station network for SMOS L2, obtained for a day on July 21, 2020 – Summer (top) and for a day on September 12, 2020 – Autumn (bottom). Blue points indicate the ground stations under consideration. Another index, known as RFI, was utilized in addition to the parameterization of the quality index (DQX). Three bands were used to calculate the RFI parameter, which are N_RFI_X and N_RFI_Y and M_AVAO. The total number of the measurements of brightness temperature is represented by each band for that node (Dente, Su, and Wen 2012). The RFI fraction equation is defined as the ratio of the sum of the N_RFI_X and N_RFI_Y bands with the M_AVAO band. N_RFI_X is the 516 T. PETSINI AND G. P. PETROPOULOS RFI detected in L2 test X polarization (the count of deleted views), N_RFI_Y is the RFI detected in L2 test Y polarization (again, the count of deleted views), and M_AVAO is the total number of views available (ATBD, SMOS 2019). RFI ¼ N RFI X þ N RFI Y M AVAO (1) As a result, values with low RFI presence (RFI <0.2) were calculated independently. The RFI threshold was set based on similar studies (Gruber et al. 2020; Gupta et al. 2021; Petropoulos et al. 2014). Soil moisture values with high RFI were also discarded. In the generated Excel, the RFI ratio was computed for each pixel via Equation (1). The data were then grouped and were analysed initially by all days of comparison, seasonality, orbit and LULC. It is important to note that the unshaded land pixels in Figure 3, represent regions where SMOS failed to obtain recommended SSM retrievals. 3.2. Statistical comparisons To compare SMOS L2 estimates with the field SM measurements, statistical parameters were calculated. Table 2 summarizes the statistical parameters calculated which are Sample size (Ν), Bias, Scatter, Mean Absolute Error (MAE), Root Mean Square Difference (RMSD), Correlation coefficient (R), Slope and Intercept of the Major Axis Regression linear fit and the index of Agreement. Those statistical metrics were selected in line to the Surface Albedo Validation (SALVAL) protocol and Validation practices for satellite soil moisture retrievals (EOLAB SALVAL tool 2022; Gruber et al. 2020). In addition, the same metrics have also been utilized in numerous similar studies evaluating the SSM product accuracy of different EO products including SMOS (e.g. Deng et al. 2019, 2022; Gupta et al. 2021). Table 2. Statistical metrics used for evaluating the accuracy between the SMOS SM data and the in-situ observations. ‘N’ represents the in-situ observations, ‘P’ represents the ‘predicted’ values, and ‘O’ represents the ‘observed’ values. Subscripts i = 1. The horizontal line represents the mean value. Name Bias/MBE Description Bias (or Mean Bias Error) Scatter/SD Scatter (Standard Deviation) MAE Mean Absolute Error RMSD R Root Mean Square Difference Correlation coefficient D-index Index of Agreement Mathematical Equation N P bias ¼ MBE ¼ N1 ðPi Oi Þ ri¼1 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P 2 N ðPi Oi Þ �¼1 scatter ¼ SD ¼ N N P MAE ¼ N 1 jPi Oi j i¼1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RMSD ¼ bias2 þ scatter 2 PN � ðP P�ÞðOi O ffiffiffiffiffiffiffiffiffiffii¼1 ffiffiffiffiffiffiiffiffiffiffiffiffiP ffiffiffiffiffiffiffiffiffiffiÞffiffiffiffiffiffiffiffiffiffiffi R ¼ qffiP 2 N N � Þ2 O O ðPi O� Þ ð i i¼1 �i¼1 PN � ðOi Pi Þ2 PN i¼1 D¼1 2 i¼1 ðjOi OjþjPi PjÞ INTERNATIONAL JOURNAL OF REMOTE SENSING 517 The analysis was carried out for different land cover types of the four selected sites like mentioned. Subsequently, the agreement was assessed for the four seasons: autumn (September–November), winter (December–February), spring (March-May) and summer (June–August) as well as for both data categories of satellite orbits. Regarding the topographic characteristics of the area, the data analysis was not evaluated as the sites of interest have similar slope values and their elevation is characterized by small to negligible variations. The analysis was performed on two subsets, without the RFI threshold and then with the parameterization adjustment. 4. Results 4.1. Comparisons for all days Statistical scores were determined to evaluate the agreement of SMOS L2 data with field soil moisture measurements for all selected sites. The main results are presented in Table 3 and Figure 4. Various parameters such as seasonality, satellite orbit and land cover types were Table 3. Agreement between SMOS L2 SM data and in-situ measurements for all the different types of comparisons performed. ALL DATA All sites Site_211 Site_181 Site_180 Site_353 RFI <0.2 All sites Site_211 Site_181 Site_180 Site_353 MBE SD MAE RMSD R Slope −0.197 −0.219 −0.269 −0.122 −0.175 0.104 0.244 0.123 0.066 0.073 0.198 0.219 0.269 −0.122 0.175 0.223 0.328 0.296 0.138 0.189 −0.013 −0.076 −0.116 0.245 0.216 −0.009 −0.035 −0.049 0.397 0.303 −0.191 −0.209 −0.285 −0.107 −0.173 0.106 0.112 0.124 0.068 0.076 0.192 0.209 0.285 0.118 0.173 0.218 0.237 0.311 0.126 0.189 0.001 0.067 −0.196 0.331 0.261 0.0004 0.032 −0.093 2.470 0.387 Intercept D-index N 0.091 0.086 0.092 0.005 0.024 0.373 0.382 0.352 0.220 0.304 668 256 98 102 212 0.098 0.073 0.108 −0.406 0.002 0.358 0.418 0.332 0.116 0.317 276 44 54 40 138 Figure 4. Agreement between SMOS L2 SM and in-situ (m3 m−3) based on all days. 518 T. PETSINI AND G. P. PETROPOULOS calculated on both datasets. Their results are presented in Tables 4, 5 and 6. For the pooled datasets (all sites), there was generally moderate agreement between the two datasets as there was a slight underestimation of the satellite data by the field measurements (Mean Bias Error (MBE) = − 0.197 m3 m−3, Standard Deviation (SD) = 0.104 m3 m−3, Root Mean Square Deviation (RMSD) = 0.223 m3 m−3). The underestimation of SMOS data from ground measurements is a major characteristic of this study. Regarding the results separately for each site, ID Sites 211-Walnut and 181-Cotton performed a relatively low agreement with high errors (RMSD = 0.328/0.296 m3 m−3, respectively). In contrast, the high agreement between SMOS product data and in-situ measurements showed ID Sites 180-Tomato (RMSD = 0.138 m3 m−3) and 353-Vineyard (RMSD = 0.189 m3 m−3). Table 4. Agreement between SMOS L2 SM data and in-situ measurements based on seasonality, ND means “No Data”. Seasonality ALL DATA All sites Site_211 Site_181 Site_180 Site_353 RFI <0.2 All sites Site_211 Site_181 Site_180 Site_353 Subsets MBE SD MAE RMSD R Slope Intercept D-index N Autumn Winter Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Summer −0.150 −0.220 −0.306 −0.210 −0.143 −0.281 −0.306 −0.244 −0.148 ND ND −0.286 −0.150 ND ND −0.117 −0.158 −0.151 ND −0.201 0.077 0.108 0.046 0.108 0.083 0.095 0.046 0.096 0.172 ND ND 0.107 0.024 ND ND 0.070 0.059 0.078 ND 0.078 0.150 0.220 0.306 0.211 0.143 0.281 0.306 0.244 0.148 ND ND 0.286 0.150 ND ND 0.122 0.158 0.151 ND 0.201 0.169 0.245 0.310 0.236 0.166 0.297 0.310 0.262 0.227 ND ND 0.305 0.152 ND ND 0.136 0.168 0.170 ND 0.216 0.238 −0.515 0.279 −0.084 0.024 −0.508 0.279 −0.033 −0.360 ND ND −0.057 0.439 ND ND 0.233 0.389 0.166 ND 0.0002 0.180 −0.922 0.271 −0.050 0.013 −0.668 0.271 −0.013 −0.113 ND ND −0.028 6.653 ND ND 0.363 0.540 1.927 ND 0.0002 0.052 0.442 −0.032 0.093 0.080 0.340 −0.032 0.065 0.107 ND ND 0.084 −1.295 ND ND 0.018 −0.031 −0.443 ND 0.082 0.413 0.202 0.180 0.356 0.379 0.150 0.180 0.360 0.413 ND ND 0.326 0.024 ND ND 0.256 0.342 0.050 ND 0.272 228 64 48 328 106 34 ND 68 12 ND ND 86 16 ND ND 86 94 30 ND 88 Autumn Winter Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Summer −0.151 −0.142 ND −0.220 −0.138 ND ND −0.099 ND ND ND −0.295 −0.138 ND ND −0.099 −0.153 −0.094 ND −0.210 0.070 0.094 ND 0.116 0.009 ND ND 0.074 ND ND ND 0.114 0.009 ND ND 0.074 0.054 0.043 ND 0.080 0.151 0.142 ND 0.223 0.138 ND ND 0.113 ND ND ND 0.295 0.138 ND ND 0.113 0.153 0.094 ND 0.210 0.166 0.170 ND 0.249 0.138 ND ND 0.124 ND ND ND 0.317 0.138 ND ND 0.124 0.162 0.104 ND 0.225 0.381 −0.739 ND −0.146 −0.072 ND ND 0.329 ND ND ND −0.165 −0.072 ND ND 0.329 0.560 0.010 ND 0.004 0.350 −1.405 ND −0.093 −0.360 ND ND 2.393 ND ND ND −0.088 −0.360 ND ND 2.393 0.849 0.050 ND 0.004 0.017 0.655 ND 0.108 0.138 ND ND −0.382 ND ND ND 0.106 0.138 ND ND −0.382 −0.112 0.206 ND 0.078 0.427 0.177 ND 0.340 0.016 ND ND 0.150 ND ND ND 0.322 0.016 ND ND 0.150 0.370 0.116 ND 0.275 92 22 ND 162 8 ND ND 32 2 ND ND 52 8 ND ND 32 62 14 ND 62 INTERNATIONAL JOURNAL OF REMOTE SENSING Figure 5. Agreement between SMOS L2 SM and in-situ (m3 m−3) based on seasonality. 519 520 T. PETSINI AND G. P. PETROPOULOS The results of the seasonality analysis are summarized in Table 4 and Figure 5. Predicted and observed data were compared allowing the algorithm to perform under different land cover types and climatic conditions during different seasons of the year. Between datasets, the autumn period was found the highest agreement (MBE = − 0.150 m3 m−3, SD = 0.077 m3 m−3, RMSD = 0.169 m3 m−3). The spring period was recorded the lowest accuracy (RMSD = 0.310 m 3 m −3), followed by winter (RMSD = 0.245 m3 m−3) season and summer season (RMSD = 0.236 m3 m−3) where there is very high error. At Site_211, high agreement was found during the autumn period (RMSD = 0.166 m3 m−3) despite the low accuracy of the station in total. In these sites, the high MBE values indicated that the SMOS product was underestimated which explains the low performance of the product at these sites. However, Site_180 had high accuracy (RMSD = 0.136 m3 m−3) in the summer period and Site_353 (RMSD = 0.168 m3 m−3), in the autumn period. In the majority of analyses, absence of data was mainly observed in the spring period. In Table 5 and Figure 6 are presented the results in terms of orbit analysis. At all sites, the metrics were showed poor performance with no significant differentiation between ascending and descending orbits. The same results are observed at Sites 211 and 181. A relatively good accuracy in product performance was observed with the descending orbit showing the best values for ID 180 and 353 sites (RMSD = 0.121/0.178 m3 m−3 respectively). Table 6 and Figure 7 illustrate the results for different land cover type. There is variation in the value of RMSD depending on the type of each vegetation. Tomato, an annual crop, was given the best accuracy (RMSD = 0.138 m3 m−3) between soil moisture datasets followed by vineyard, a tree crop (RMSD = 0.189 m3 m−3). The lowest performance of algorithm was found walnut (RMSD = 0.328 m3 m−3) and cotton (RMSD = 0.296 m 3 m −3 ) crop, respectively. Unexpectedly, cotton Table 5. Agreement between SMOS L2 SM data and in-situ measurements based on satellite orbit. Satellite orbit ALL DATA All sites Site_211 Site_181 Site_180 Site_353 RFI <0.2 All sites Site_211 Site_181 Site_180 Site_353 Subsets MBE SD MAE RMSD R Slope Intercept D-index N Asc Desc Asc Desc Asc Desc Asc Desc Asc Desc −0.203 −0.190 −0.214 −0.224 −0.310 −0.230 −0.136 −0.100 −0.184 −0.161 0.101 0.107 0.104 0.109 0.119 0.116 0.061 0.068 0.069 0.077 0.204 0.191 0.214 0.224 0.310 0.230 0.141 0.104 0.184 0.161 0.227 0.218 0.238 0.249 0.332 0.257 0.149 0.121 0.196 0.178 −0.026 −0.051 −0.050 −0.176 −0.208 0.056 0.102 0.278 0.191 0.215 −0.017 −0.034 −0.022 −0.085 −0.059 0.024 0.205 0.333 0.274 0.278 0.080 0.114 0.073 0.111 0.072 0.090 0.027 0.047 0.020 0.048 0.362 0.381 0.385 0.367 0.335 0.381 0.154 0.330 0.274 0.345 374 294 136 120 48 50 62 40 128 84 Asc Desc Asc Desc Asc Desc Asc Desc Asc Desc −0.196 −0.181 −0.199 −0.226 −0.327 −0.224 −0.111 −0.089 −0.179 −0.161 0.106 0.108 0.100 0.135 0.104 0.131 0.064 0.092 0.076 0.076 0.198 0.182 0.199 0.226 0.327 0.224 0.120 0.110 0.179 0.161 0.223 0.210 0.223 0.263 0.343 0.260 0.128 0.128 0.195 0.178 −0.044 0.006 0.249 −0.292 −0.073 −0.137 0.385 0.028 0.191 0.294 −0.035 0.005 0.136 −0.129 −0.024 −0.054 2.533 0.405 0.297 0.390 0.096 0.119 0.042 0.127 0.057 0.132 −0.422 0.033 0.017 0.020 0.343 0.379 0.437 0.382 0.316 0.387 0.134 0.053 0.280 0.360 182 94 28 16 32 22 32 8 90 48 INTERNATIONAL JOURNAL OF REMOTE SENSING 521 Figure 6. Agreement between SMOS L2 SM and in-situ (m3 m−3) based on satellite orbit. characterized as an area with sparse vegetation gave a high error. This is likely to be a result of the measuring instrument used in the in-situ measurements as well as the characteristics of the area. Similarly, to the results of the above-mentioned analyses, a systematic underestimation of the SMOS product from the field data was also observed in this case recording the highest value in annual crop (MBE = −0.269 m3 m−3). The sample size from site to site varied as evidenced in the Figure 8, this may be due to the growing season of each vegetation type. Similar patterns as in Figures 4, 5 and 6 appear for each site of interest which are not listed for brevity. 4.2. Comparisons for days with RFI < 0.2 RFI threshold has been regarded as a critical issue in product validation which influences the retrieval accuracy. An extended analysis is conducted by considering only days where the RFI value is < 0.2 in order to analyse the effect of high RFI pixel 522 T. PETSINI AND G. P. PETROPOULOS Table 6. Agreement between SMOS L2 SM data and in-situ measurements based on land cover type. Land Cover ALL DATA Walnut Cotton Tomato Vineyard MBE SD MAE RMSD R Slope Intercept D-index N −0.219 −0.269 −0.122 −0.175 0.244 0.123 0.066 0.073 0.219 0.269 −0.122 0.175 0.328 0.296 0.138 0.189 −0.076 −0.116 0.245 0.216 −0.035 −0.049 0.397 0.303 0.086 0.092 0.005 0.024 0.382 0.352 0.220 0.304 256 98 102 212 RFI <0.2 Walnut Cotton Tomato Vineyard −0.209 −0.285 −0.107 −0.173 0.112 0.124 0.068 0.076 0.209 0.285 0.118 0.173 0.237 0.311 0.126 0.189 0.067 −0.196 0.331 0.261 0.032 −0.093 2.470 0.387 0.073 0.108 −0.406 0.002 0.418 0.332 0.116 0.317 44 54 40 138 Figure 7. Agreement between in-situ and SMOS L2 SM (m3 m−3) based on land cover type. Figure 8. Sample size by land cover type. contamination on product accuracy. The statistical analysis performed for RFI < 0.2 between the retrieved data from the SMOS satellite and the field measurements are recorded in Tables 3–6. As it was proved, for all data sets, after filtering out the RFI contaminated pixels, the results were obviously improved. Regarding the aggregated INTERNATIONAL JOURNAL OF REMOTE SENSING 523 datasets, product accuracy was recorded as moderate with product performance for all values (no RFI threshold) but with lower error values (MBE = − 0.191 m3 m−3, SD = 0.106 m3 m−3, RMSD = 0.218 m3 m−3). Considering the four seasons, the product performance when RFI was filtered was better with the winter season (Table 4 and Figure 5) showing the highest decrease from 0.245 to 0.170 m3 m−3 (a difference of 0.075 m3 m−3). This is probably explained by the lack of a large sample size. However, it was observed that MBE and SD values were mostly improved and, in some cases, showed small increases. The ascending and descending orbit results improved with no large price differences. The walnut crop showed the largest decrease in RMSD (decreased by 0.091 m3 m−3) followed by the tomato crop giving a value of RMSD = 0.126 m3 m−3 (Table 6). At the same time, MBE and SD values of each crop were improved in the majority of them. With respect to the analysis of each crop separately, filtering the RFI pixels, ID sites 211 and 181 showed no major differences. As a result, the performance of the product did not prove to be high at these two stations. On the other hand, the good agreement of ID sites 180 and 353 improved even further in all subset analyses. Site_180 showed a decrease of 0.012 in the RMSD value while the corresponding value of Site_353 remained stable (Table 6). On a seasonal basis, in Site_180, the high agreement remained in the summer period (RMSD = 0.124 m3 m−3) while in Site_353, the highest performance of the product was found in winter with a much-improved RMSD = 0.104 m3 m−3. In the orbital analysis, the best agreement was again found in the descending orbit with a slight improvement in values. MBE and SD values were shown to be considerably improved. Overall, the exclusion of values with high RFI had a positive effect on the accuracy and performance of the SMOS product. 4.3. Temporal consistency between the sites Figure 9 shows the temporal trends between predicted data and in-situ SMOS soil moisture measurements as a function of precipitation during the year 2020 for the four seasons at the sites of interest. Ιn-situ SM measurements and the SMUDP2_40 km product capture the rapid response of soil moisture to weather variation and temporal changes over the investigated period. Relatively similar variability trends and effective seasonal agreement are attributed by comparing the cotton datasets. For the complete study period, an underestimation of SMOS values from in-situ observations was evident. Small intervals of variation where the two data sets showed a mismatch were detected during the winter period (November– December) for walnut, tomato and vineyard. In these cases, the field measurements are characterized by stability (Figures 9(i), (iii), (iv)). This may be due to the presence of frost and low temperatures. In general, it can be observed from the figure that each vegetation type reacts differently to the precipitation data. Little to no days of rainfall are found for tomato and cotton, which are annual crops, and their growing season is the summer period (June-early September). In tree crops, the precipitation datasets showed a greater number of rainfall days, particularly in walnut that the growing season is all year round (Figure 9(i)). 524 T. PETSINI AND G. P. PETROPOULOS Figure 9. Temporal variability of daily satellite and in situ SM measurements for different types of land cover in Larisa (i) walnut, (ii) Cotton, (iii) tomato, (iv) vineyard. Purple is the in-situ SM and green is the SMOS SM. The black line represents the values of precipitation (mm day−1). 5. Discussion In this study, the accuracy of the SMOS SM product was evaluated using in-situ measurements from a network of stations for a typical arid/semi-arid Mediterranean ecosystem of primarily agricultural usage located in Greece. Four crop types included namely walnut, tomato, vineyard and cotton. In general, the results showed that they represent a reasonable agreement with the appropriate range of values for specific crop types (tomato and vineyard). The other two types of crops gave error values which show low agreement. Furthermore, in all analysis, a tendency to underestimate the SMOS product from the in-situ data was observed. The removal of high RFI leads to an improvement in the accuracy of the SMOS product and significantly reduces the underestimation of its data. In the present study, results show a slightly lower agreement compared to other studies examining the validation of the same satellite product (e.g. Deng et al. 2019; Gupta et al. 2021; Petropoulos et al. 2014; Suman et al. 2020). Yet, most of those studies were carried out also under a different experimental setup. For example, a different number of stations and with ground datasets obtained from specific observational INTERNATIONAL JOURNAL OF REMOTE SENSING 525 networks (FLUXNET, The International Soil Moisture Network), for different time periods (e.g. one to five years) and in a variety of environments worldwide (e.g. Europe, Asia, Australia and USA). Petropoulos et al. (2014) is a study where a Mediterranean environment was investigated and the results were in agreement with the present analysis. There are several reasons that can be accounted towards explaining the lack of complete agreement between the compared datasets. In particular, one may be related to the number of ground measurements. This results to an increase in spatial mismatch which may create discrepancies and errors in the product estimation (Suman et al. 2020). This is because the SMOS sensor is characterized as coarse resolution and represents a more heterogeneous surface. Generally, in studies where coarse resolution sensors are used, limitations are created in terms of the spatial representativeness of the flow towers. In contrast, rough resolution sensors represent a larger spatial footprint, which is not characterized by a uniform distribution of flow towers (Howells et al. 2021). These problems could be addressed by using dense networks that provide a traceable representation of soil moisture at the trace scale, reducing errors and allowing accurate assessment of the quality of satellite soil moisture data (Gruber et al. 2020). Another factor to be taken under consideration in the comparisons performed is the penetration depth of the L-band which differs from in-situ instruments. The actual L-band depth of the SMOS product is observed at about 2 cm and varies depending on the SM (Wang et al. 2021) while the in-situ sensors are placed at a depth of about 10 cm. The accuracy of soil moisture retrieval can also be affected by the distance from the sea (Bhuyan et al. 2023). In this study, the distance of the stations from the sea is about 30–40 km which is generally acceptable. Errors associated with instrumentation and representation issues such as measurement method, homogeneity of the data set, depth and manner of instrumentation installation, lack of adequate instrument maintenance, calibration methodology and measurement intervals can affect the values of in-situ observations (Dorigo et al. 2013). One of the most significant factors affecting soil moisture dynamics is seasonality, and its unpredictability can have a substantial influence on the overall performance of soil moisture retrieval sensors. Similar studies addressing product validation on an annual scale have shown that soil moisture estimates are to some extent driven by the seasonal cycle (Petropoulos et al. 2018; Qin et al. 2013). The present study results showed lower RMSD values and higher correlation values in the autumn months compared to other seasons. Phenological variations in vegetation throughout the year explain this finding (Srivastava et al. 2013). In terms of time series comparison, the SMOS SM product is underestimated by in-situ measurements throughout the year. In the summer period, there was high error (RMSD) and low correlation (R). This conclusion is in agreement with Deng et al. (2019a) who discovered that the correlation coefficient was lower in the summer and the error was higher. At Site 211, the exception is the high agreement recorded during the summer period (after RFI threshold). The presence of dew, which occurs mainly during summer, spring and autumn when horizontal brightness is more prevalent and increases may account for this variation in values (De Du et al. 2012; Jeu et al. 2005). Atmospheric and climatic parameters significantly affect the accuracy of soil moisture products. These conditions may be responsible for the accuracy of seasonal variations. More specifically, precipitation input is the most influential element on SM because it 526 T. PETSINI AND G. P. PETROPOULOS causes higher evapotranspiration losses which impact soil water balance (Wagner et al. 2007). In some cases, such as Figure 9(i) & 9(ii), it is evident that the SMOS products are not significantly affected by precipitation compared to ground station data where a visible increase is observed after precipitation in SSM. This is because SMOS data can be affected by clouds or other atmospheric components and thus have an uncertainty whereas ground-based measurements are taken directly, thus minimizing the presence of external factors. The increased performance of SMOS SM in the autumn period is in agreement with other studies which achieved similar results (Deng et al. 2019; Petropoulos, Ireland, and Srivastava 2015). The winter period results gave low agreement between the two datasets probably due to the low temperatures, the presence of snow and frozen ground. In particular, the presence of snow can lead to erroneous results as the ground is frozen (Kerr et al. 2012; Running et al. 1999). However, loss of vegetation is typical of this season and can lead to poor performance. The accuracy of the ascending and descending orbits does not show significant differences. Only for specific land uses, tomato and vineyard, the satellite’s accuracy detects high agreement. Previous research showed no differences between ascending and descending orbits (Sanchez et al. 2012). Topographic complexity emerges as the main cause of the different performance of SMOS products in land cover areas with different land uses. Comparing the datasets by land cover type, the variation of retrieval accuracy becomes apparent by crop type. Tomato crop, which is characterized by sparse vegetation, shows a low RMSD and better correlation compared to the other crops. Generally, crops with increased water requirements throughout the year, such as tomato, show the highest overall accuracy of estimating the operational product. The next station showing high agreement, based on land cover, is vineyard, which is a tree crop. The vineyard crop have a high proportion of bare land or open ground, which eliminates the influence of vegetation cover attenuation on the microwave signal and improves product accuracy. On the contrary, walnut and cotton had the highest RMSD among the land cover categories investigated. This is probably due to the number and frequency between predicted and observed soil moisture values. In particular, the cotton crop, as an area with sparse vegetation, should give a low RMSD, as the L-band microwave signal easily penetrates through the canopy of this crop (Leroux et al. 2013; Liu, Yang, and Yue 2018). Similar studies investigated at this topic by assessing the accuracy of SMOS and other SM satellite products in various areas worldwide (Deng et al. 2019; Gupta et al. 2021; Petropoulos et al. 2014; Suman et al. 2020). Most of them are similarly performed because the resulting mistakes are caused by the presence of tree crops in the radiometer field of vision. In the entire analysis, the underestimation observed in the SMOS satellite data from the in-situ measurements may be due to the dry bias identified in this study and occurs more during the summer period (Kang et al. 2016; Oozeer, Fletcher, and Champagne 2020). With respect to the SMOS satellite, there is no strong evidence to suggest a systematic overestimation or underestimation of SSM. Similar validation studies of the SMOS product are consistent with this result (Petropoulos et al. 2014). A significant cause of disturbances has already been shown to be the RFI in natural microwave emissions measured by satellites. Its characteristic is the effect on electromagnetic radiation emitted by an external source (Murray 2013). Data quality can be damaged or diminished by these disturbances and, in extreme situations, might result in INTERNATIONAL JOURNAL OF REMOTE SENSING 527 total data loss, as was seen mostly during the spring season. Choosing an RFI threshold and excluding RFI-filtered pixels was therefore a critical step in the present study to objectively evaluate the product’s accuracy. The results were highly improved when pixels with high RFI were excluded focusing on the low RMSD values. Globally, soil moisture retrieval results are strongly influenced by the RFI value (Deng et al. 2022; Gruber et al. 2020). 6. Concluding remarks In this study, an extended evaluation of the SM product of SMOS L2 was conducted related to insitu observations in Larisa area, for four selected validation sites taking into account seasonal trends, satellite orbit and different land uses. In overall, the study results showed a reasonably good agreement of SMOS operational product with in-situ measurements for the two of four sites, tomato and vineyard, in terms of RMSE and the correlation coefficient. Some of the key study findings are summarized as follows: (1) The removal of high RFI values seems to continue to be a challenge for Europe and especially for a Greek setting. In the overall comparison of the two datasets, it was observed that by excluding pixels with high RFI, the accuracy of the product was significantly improved. RMSD decreased even more as well as the bias and the degree of correlation increased. In specific subsets, depending on the vegetation type, the difference in values was significant (tomato, RMSD = 0.138/0.126 m3 m−3). Previous relevant validation studies have largely agreed with the results. (2) Land cover significantly influences the variation in soil moisture because of increasing rainfall interception and transpiration losses. The analysis showed that the hydrophilic crop, tomato, gave the best results with a low RMSD value (0.138 m3 m−3) in all the comparisons performed. As observed, the type of vegetation differentiates the results as these are crops with different characteristics. (3) On a seasonal basis, the best performance of the SMOS SM algorithm was predicted in the autumn and summer period, which can be explained by the presence of dew (all sites, tomato, vineyard, RMSD = 0.166, 0.124, 0.162 m3 m−3). Further research is required before the conclusions reported herein are generalized. For example, a further investigation could also be conducted by applying inter-comparisons with other similar operational products of different approaches, such as ASCAT and SMAP, to extend the validation of the operational product. In addition, more validation sites can be included in the comparisons and also the time period of the datasets analysis can be extended to more years of analysis. All in all, results obtained herein are very interesting and of important scientific value with regard to the practical use of the SMOS SSM product. Our findings can aid in the improvement of the retrieval algorithm directing users towards more effective of operational product data in a variety of research goals and applications such as water resource management and climate change studies. In practical terms, our study findings can also contribute to agriculture by helping farmers to make better or more informed decisions on irrigation practices. That way can be of practical value in the adoption or adaptation of 528 T. PETSINI AND G. P. PETROPOULOS agricultural practices in Greece and potentially help towards the development of applications and policies for the more sustainable use of water resources and their management. This remains to be seen. Acknowledgements The authors wish to thank Neuropublic S.A. for providing the in-situ measurements used in the present study. The participation of George P. Petropoulos was financially supported by the LISTENEO project, implemented in the framework of H.F.R.I call “Basic research Financing (Horizontal support of all Sciences)” under the National Recovery and Resilience Plan “Greece 2.0” funded by the European Union –NextGenerationEU (H.F.R.I. Project Number: 15898). Also authors are grateful to the anonymous reviewers for their comments that helped improving the initially submitted manuscript. Disclosure statement No potential conflict of interest was reported by the author(s). References ATBD (SMOS Algorithm Theoretical Basis Document). 2019. “SMOS Algorithm Theoretical Basis Document (ATBD) for the SMOS Level 2 Soil Moisture Processor Development Continuation Project.” Issue 4.0. Accessed March 20. https://earth.esa.int/eogateway/documents/20142/ 37627/SMOS-L2-SM-ATBD.pdf. Babaeian, E., M. Sadeghi, S. B. Jones, C. Montzka, H. Vereecken, and M. Tuller. 2019. “Ground, Proximal, and Satellite Remote Sensing of Soil Moisture.” Reviews of Geophysics 57 (2): 530–616. https://doi.org/10.1029/2018rg000618. Bao, Y., L. Lin, S. Wu, K. A. K. Deng, and G. P. Petropoulos. 2018. “Surface Soil Moisture Retrievals Over Partially Vegetated Areas from the Synergy of Sentinel-1 and Landsat 8 Data Using a Modified Water-Cloud Model.” International Journal of Applied Earth Observation and Geoinformation 72:76–85. https://doi.org/10.1016/j.jag.2018.05.026. Bhuyan, M. I., S. Mia, I. Supit, and F. Ludwig. 2023. “Spatio-Temporal Variability in Soil and Water Salinity in the South-Central Coast of Bangladesh.” Catena 222:106786. https://doi.org/10.1016/j. catena.2022.106786. Deng, K. A. K., S. Lamine, A. Pavlides, G. P. Petropoulos, Y. Bao, P. K. Srivastava, and Y. Guan. 2019. “Large scale operational soil moisture mapping from passive MW radiometry: SMOS product evaluation in Europe & USA.” International Journal of Applied Earth Observation and Geoinformation 80:206–217. https://doi.org/10.1016/j.jag.2019.04.015. Deng, K. A. K., S. Lamine, A. Pavlides, G. P. Petropoulos, P. K. Srivastava, Y. Bao, D. T. Hristopulos, and V. A. Anagnostopoulos. 2019. “Operational Soil Moisture from ASCAT in Support of Water Resources Management.” Remote Sensing 11 (5): 579. https://doi.org/10.3390/rs11050579. Deng, K. A. K., G. P. Petropoulos, Y. Bao, A. Pavlides, A. A. Chaibou, and B. A. Habtemicheal. 2022. “An Examination of the SMAP Operational Soil Moisture Products Accuracy at the Tibetan Plateau.” Remote Sensing 14 (24): 6255. https://doi.org/10.3390/rs14246255. Dente, L., Z. Su, and J. Wen. 2012. “Validation of SMOS Soil Moisture Products Over the Maqu and Twente Regions.” Sensors 12 (8): 9965–9986. https://doi.org/10.3390/s120809965. Djamai, N., R. Magagi, K. Goïta, M. Hosseini, M. H. Cosh, A. A. Berg, and B. Toth. 2015. “Evaluation of SMOS Soil Moisture Products Over the CanEx-SM10 Area.” Canadian Journal of Fisheries and Aquatic Sciences 520:254–267. https://doi.org/10.1016/j.jhydrol.2014.11.026. INTERNATIONAL JOURNAL OF REMOTE SENSING 529 Dorigo, W. A., A. Xaver, M. Vreugdenhil, A. Gruber, A. Hegyiová, A. D. Sanchis-Dufau, D. Zamojski, et al. 2013. “Global Automated Quality Control of in situ Soil Moisture Data from the International Soil Moisture Network.” Vadose Zone Journal 12 (3): 1–21. https://doi.org/10.2136/vzj2012.0097. Du, J., T. J. Jackson, R. Bindlish, M. H. Cosh, L. Li, B. K. Hornbuckle, and E. D. Kabela. 2012. “Effect of Dew on Aircraft-Based Passive Microwave Observations Over an Agricultural Domain.” Journal of Applied Remote Sensing 6 (1): 63571. https://doi.org/10.1117/1.JRS.6.063571. EOLAB SALVAL tool. (2022). Surface Albedo Validation Protocol. User Guide. Retrieved from: https:// calvalportal.ceos.org/salval . FAO 2015. Climate Change and Food Security: Risks and Responses. https://www.fao.org/3/i5188e/ I5188E.pdf. Fuzzo, D. F. S., T. N. Carlson, N. N. Kourgialas, and G. P. Petropoulos. 2019. “Coupling Remote Sensing with a Water Balance Model for Soybean Yield Predictions Over Large Areas.” Earth Science Informatics 13 (2): 345–359. https://doi.org/10.1007/s12145-019-00424-w. Gruber, A. D., G. De Lannoy, C. Albergel, A. Al-Yaari, L. Brocca, J. Calvet, A. Colliander, et al. 2020. “Validation practices for satellite soil moisture retrievals: What are (the) errors?” Remote Sensing of Environment 244:111806. https://doi.org/10.1016/j.rse.2020.111806. Gupta, D. K., P. S. Kumar, V. N. Mishra, and R. Prasad. 2014. “Soil Moisture Estimation by ANN Using Bistatic Scatterometer Data.” ISPRS Annals of the Photogrammetry, Remote Sensing & Spatial Information Sciences II-8:97–100. https://doi.org/10.5194/isprsannals-ii-8-97-2014. Gupta, D. K., P. K. Srivastava, A. Singh, G. P. Petropoulos, N. Stathopoulos, and R. Prasad. 2021. “SMAP Soil Moisture Product Assessment Over Wales, U.K., Using Observations from the WSMN Ground Monitoring Network.” Sustainability 13 (11): 6019. https://doi.org/10.3390/su13116019. Howells, O. D., G. P. Petropoulos, P. K. Srivastava, D. Triantakonstantis, I. Sandric, and Dimitrios Triantakonstantis & Ionut Sandric. 2021. “Exploring the Potential of SCAT-SAR SWI for Soil Moisture Retrievals at Selected COSMOS-UK Sites.” International Journal of Remote Sensing 42 (23): 9146–9160. https://doi.org/10.1080/01431161.2021.1988185. IPCC. 2019. “Food Security.” Climate Change and Land an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems. https://www.ipcc.ch/srccl/chapter/chapter-5/. Jeu, R., A. M. De, T. R. H. Holmes, and M. Owe 2005. “Determination of the Effect of Dew on Passive Microwave Observations from Space.” In Geophysical Research Abstracts. Vol. 7, edited by M. Owe and G. D’Urso, 04001, pp. 597608. https://doi.org/10.1117/12.627919 Kang, C. S., K. D. Kanniah, Y. Kerr, and C. A. Varotsos. 2016. “Analysis of in-Situ Soil Moisture Data and Validation of SMOS Soil Moisture Products at Selected Agricultural Sites Over a Tropical Region.” International Journal of Remote Sensing 37 (16): 3636–3654. https://doi.org/10.1080/01431161. 2016.1201229. Kerr, Y., P. Waldteufel, P. Richaume, J. Wigneron, P. Ferrazzoli, A. Mahmoodi, A. A. Bitar, et al. 2012. “The SMOS Soil Moisture Retrieval Algorithm.” IEEE Transactions on Geoscience and Remote Sensing 50 (5): 1384–1403. https://doi.org/10.1109/tgrs.2012.2184548. Kerr, Y., P. Waldteufel, J. Wigneron, J. Martinuzzi, J. Font, and M. Berger. 2001. “Soil Moisture Retrieval from Space: The Soil Moisture and Ocean Salinity (SMOS) Mission.” IEEE Transactions on Geoscience and Remote Sensing 39 (8): 1729–1735. https://doi.org/10.1109/36.942551. Lamptey, N. L., A. A. Patrick, K. Gyamfi, C. Yeboah, A. Abigail, C. Doe Edinam, and S. Aboagye-Ntow. 2023. Perspective Chapter: Validation of SMOS Satellite Soil Moisture Estimates Using Capacitance Probes Over the Different Ecological Zones in Northern Ghana. Chapter 2. Soil moisture. IntechOpen. Datta, R: Dr. https://doi.org/10.5772/intechopen.109696. Leroux, D. J., Y. H. Kerr, P. Richaum, and R. Fieuzal. 2013. “Spatial Distribution and Possible Sources of SMOS Errors at the Global Scale.” Remote Sensing of Environment 133 (June): 240–250. https://doi. org/10.1016/j.rse.2013.02.017. Li, Z., P. Leng, C. Zhou, K. Chen, F. Zhou, and G. Shang. 2021. “Soil Moisture Retrieval from Remote Sensing Measurements: Current Knowledge and Directions for the Future.” Earth Science Review 218:103673. https://doi.org/10.1016/j.earscirev.2021.103673. Liu, Y., and Y. Yang. 2022. “Advances in the Quality of Global Soil Moisture Products: A Review.” Remote Sensing 14 (15): 3741. https://doi.org/10.3390/rs14153741. 530 T. PETSINI AND G. P. PETROPOULOS Liu, Y., Y. Yang, and X. Yue. 2018. “Evaluation of Satellite-Based Soil Moisture Products Over Four Different Continental in-Situ Measurements. Remote Sens.” (Basel) 10 (7): 1161. https://doi.org/10. 3390/rs10071161. Maltese, A., F. Capodici, G. Ciraolo, and G. La Loggia. 2015. “Soil Water Content Assessment: Critical Issues Concerning the Operational Application of the Triangle Method.” Sensors 15 (3): 6699–6718. https://doi.org/10.3390/s150306699. Molénat, J., K. Barkaoui, S. Benyoussef, I. Mekki, R. Zitouna, and F. Jacob. 2023. “Diversification from Field to Landscape to Adapt Mediterranean Rainfed Agriculture to Water Scarcity in Climate Change Context.” Current Opinion in Environmental Sustainability 65:101336–103435. https://doi. org/10.1016/j.cosust.2023.101336. Murray, D. W. 2013. “Using RF recording techniques to resolve interference problems.” https://doi. org/10.1109/autest.2013.6645046. O’Neill, B. F., and A.-L. Boyer. 2020. “Water Conservation in Desert Cities: From the Socioecological Fix to Gestures of Endurance.” Ambiente & Sociedade 23. https://doi.org/10.1590/18094422asoc20190069r1vu2020L1AO. Oozeer, M. Y., C. D. Fletcher, and C. M. Champagne. 2020. “Evaluation of Satellite-Derived Surface Soil Moisture Products Over Agricultural Regions of Canada.” Remote Sensing 12 (9): 1455. https:// doi.org/10.3390/rs12091455. Petropoulos, G. 2013. Remote Sensing of Energy Fluxes and Soil Moisture Content. CRC Press. https:// doi.org/10.1201/b15610. Petropoulos, G. P., G. Ireland, and B. Barrett. 2015. “Surface soil moisture retrievals from remote sensing: Current status, products & future trends.” Physics and Chemistry of the Earth, Parts A/B/C 83–84:36–56. https://doi.org/10.1016/j.pce.2015.02.009. Petropoulos, G. P., G. Ireland, and P. K. Srivastava. 2015. “Evaluation of the Soil Moisture Operational Estimates from SMOS in Europe: Results Over Diverse Ecosystems.” IEEE Sensors Journal 15 (9): 5243–5251. https://doi.org/10.1109/jsen.2015.2427657. Petropoulos, G. P., G. Ireland, P. K. Srivastava, and P. Ioannou-Katidis. 2014. “An appraisal of the accuracy of operational soil moisture estimates from SMOS MIRAS using validated in situ observations acquired in a Mediterranean environment.” International Journal of Remote Sensing 35 (13): 5239–5250. https://doi.org/10.1080/2150704x.2014.933277. Petropoulos, G. P., P. K. Srivastava, K. P. Ferentinos, and D. Hristopoulos. 2018. “Evaluating the Capabilities of Optical/TIR Imaging Sensing Systems for Quantifying Soil Water Content.” Geocarto International 35 (5): 494–511. https://doi.org/10.1080/10106049.2018.1520926. Petropoulos, G. P., P. K. Srivastava, M. Piles, and S. Pearson. 2018. “Earth Observation-Based Operational Estimation of Soil Moisture and Evapotranspiration for Agricultural Crops in Support of Sustainable Water Management.” Sustainability (Switzerland) 10 (1): 181. https://doi. org/10.3390/su10010181. Piles, M., G. P. Petropoulos, H. Lu, A. Gonzalez-Zamora, and G. Ireland. 2016. “Towards Improved Spatio-Temporal Resolution Soil Moisture Retrievals from the Synergy of SMOS and MSG SEVIRI Spaceborne Observations.” Remote Sensing of Environment 180:403–417. https://doi.org/10.1016/ j.rse.2016.02.048. Qin, J., K. Yang, N. Lu, Y. Chen, L. Zhao, and M. Han. 2013. “Spatial Upscaling of in-Situ Soil Moisture Measurements Based on MODIS-Derived Apparent Thermal Inertia.” Remote Sensing of Environment 138:1–9. https://doi.org/10.1016/j.rse.2013.07.003. Running, S. W., D. Baldocchi, D. P. Turner, S. T. Gower, P. S. Bakwin, and K. Hibbard. 1999. “A Global Terrestrial Monitoring Network Integrating Tower Fluxes, Flask Sampling, Ecosystem Modeling and EOS Satellite Data.” Remote Sensing of Environment 70 (1): 108–127. https://doi.org/10.1016/ s0034-4257(99)00061-9. Sanchez, N. G., J. Martínez-Fernández, A. Scaini, and C. Perez-Gutierrez. 2012. “Validation of the SMOS L2 Soil Moisture Data in the REMEDHUS Network (Spain).” IEEE Transactions on Geoscience and Remote Sensing 50 (5): 1602–1611. https://doi.org/10.1109/tgrs.2012.2186971. Shi, Q., and S. Liang. 2014. “Surface-Sensible and Latent Heat Fluxes Over the Tibetan Plateau from Ground Measurements, Reanalysis, and Satellite Data.” Atmospheric Chemistry and Physics 14 (11): 5659–5677. https://doi.org/10.5194/acp-14-5659-2014. INTERNATIONAL JOURNAL OF REMOTE SENSING 531 Srivastava, P. K., D. Han, M. D. Ramirez, and T. Islam. 2013. “Appraisal of SMOS soil moisture at a catchment scale in a temperate maritime climate.” Canadian Journal of Fisheries and Aquatic Sciences 498:292–304. https://doi.org/10.1016/j.jhydrol.2013.06.021. Srivastava, P. K., D. Han, M. A. Rico-Ramirez, D. Al-Shrafany, and T. Islam. 2013. “Data Fusion Techniques for Improving Soil Moisture Deficit Using SMOS Satellite and WRF-NOAH Land Surface Model.” Water Resources Management. https://doi.org/10.1007/s11269-013-0452-7. Srivastava, P. K., G. Petropoulos, and Y. Kerr. 2016. Satellite Soil Moisture Retrieval: Techniques and Applications. Elsevier. Suman, S., P. K. Srivastava, G. P. Petropoulos, D. Pandey, and P. O’Neill. 2020. “Appraisal of SMAP Operational Soil Moisture Product from a Global Perspective.” Remote Sensing 12 (12): 1977. https://doi.org/10.3390/rs12121977. Sun, H. 2016. “Two-Stage Trapezoid: A New Interpretation of the Land Surface Temperature and Fractional Vegetation Coverage Space.” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 9 (1): 336–346. https://doi.org/10.1109/jstars.2015.2500605. UNESCO. United Nations World Water Development Report 2020: Water and Climate Change; UNESCO: Paris, France, 2019; ISBN 9789231003714. Accessed: https://unesdoc.unesco.org/ark:/ 48223/pf0000372985. Wagner, W., V. Naeimi, K. Scipal, R. De Jeu, and J. Martínez-Fernández. 2007. “Soil moisture from operational meteorological satellites.” Hydrogeology Journal 15 (1): 121–131. https://doi.org/10. 1007/s10040-006-0104-6. Wang, X., H. Lü, W. T. Crow, Y. Zhu, Q. Wang, J. Su, J. Zheng, and Q. Gou. 2021. “Assessment of SMOS and SMAP Soil Moisture Products Against New Estimates Combining Physical Model, a Statistical Model, and in-Situ Observations: A Case Study Over the Huai River Basin, China.” Canadian Journal of Fisheries and Aquatic Sciences 598:126468. https://doi.org/10.1016/j.jhydrol.2021. 126468. World Economic Forum. The Global Risks Report 2020 Insight Report 15th Edition. 2020. Available online: https://www.weforum.org/reports/the-global-risks-report-2020/. Zhang, J., Z. Zhou, F. Yao, L. Yang, and H. Chen. 2014. “Validating the Modified Perpendicular Drought Index in the North China Region Using In Situ Soil Moisture Measurement.” IEEE Geoscience and Remote Sensing Letters 12 (3): 542–546. https://doi.org/10.1109/lgrs.2014. 2349957.