Search

This work aims to assess the effectiveness of machine learning (ML) algorithms and semiempirical models for surface soil moisture (SSM) retrieval by exploring the Sentinel-1 backscatter and interferometric coherence data. First, three commonly used categories of ML algorithms are evaluated using data gathered from diverse rainfed and irrigated wheat fields located in Morocco and Tunisia. Specifically, these algorithms include artificial neural network (ANN), deep neural network, three support vector regression (SVR) models [radial basis function (SVR_rbf), linear (SVR_linear), and polynomial (SVR_quad) kernels], and two tree-based methods [random forest and eXtreme Gradient Boosting (XGBoost)]. The comparison between predicted and measured SSM showed that the best retrieval results were obtained using Sentinel-1 data at VV polarization with R ranging between $ 0.68$ and $ 0.76$ and root-mean-square error (RMSE) of $ \text{0.05}\,\text{m}^{3}/\text{m}^{3}$ and $ \text{0.06}\,\text{m}^{3}/\text{m}^{3}$. Second, to further assess their transferability, the ANN, SVR_rbf, and XGBoost, which demonstrated the most favorable results from each category, were evaluated and compared against the coupled Water Cloud and Oh models (WCM), using a second dataset collected over a drip-irrigated wheat field in Morocco. Overall, the best retrieval results were achieved by ANN and SVR_rbf with R and RMSE of $ 0.81$ and $ \text{0.034}\,\text{m}^{3}/\text{m}^{3}$, respectively. In addition, their performances were consistent with that of WCM, which yielded R and RMSE values of $ 0.81$ and $ \text{0.04}\,\text{m}^{3}/\text{m}^{3}$, respectively. Finally, due to its good compromise between retrieval accuracy of SSM, processing time, and simplicity, SVR_rbf was chosen to generate high-resolution SSM maps from Sentinel-1 data over irrigated wheat fields.

In this paper, three datasets are described. The first dataset is a complete set of GNSS-R (GNSS-R: Global Navigation Satellite System – Reflectometry) airborne data. This dataset has been generated with the data acquired with the GLObal Navigation Satellite System Reflectometry Instrument (GLORI) developed at Centre d'Etudes Spatiales de la Biosphère (CESBIO), during the Land surface Interactions with the Atmosphere over the Iberian Semi-arid Environment (LIAISE) campaign in north-eastern Spain during the summer of 2021. It is the first time to our knowledge that a complete dataset of GNSS-R observables (reflectivity, incoherent component relative to the total scattering signal to noise ratio (SNR) for copolarized (right-right) and cross-polarized (right-left) measurements has been made available.The two other datasets are ground truth sets of measurements which have been acquired simultaneously with the flights. The in-situ measurements dataset consists in soil measurements (surface soil moisture, surface roughness, Leaf Area Index (LAI)) over 24 reference fields). The land use dataset provides a land use map (along with 385 ground truth plots) over the studied site for GLORI data evaluation.The combined datasets are particularly relevant for soil moisture and vegetation retrievals from GNSS-R observables, as well as studies for calibration and validation of bistatic empirical or physical models simulating coherent or incoherent components on agriculture sites, in the context of the preparation of future GNSS-R space missions, such as HydroGNSS, a European Space Agency mission, launch foreseen in 2024.The entire database is archived in the AERIS LIAISE database. One DOI is available for each of the 3 datasets (airborne GLORI dataset, in situ measurements dataset and land use dataset).

This article aims to estimate the water footprint (WF) of cereals—specifically, wheat and barley—in the Kairouan plain, located in central Tunisia. To achieve this objective, two components must be determined: actual evapotranspiration (ETa) and crop yield. The study covers three growing seasons from 2010 to 2013. The ETa estimation employed the S-SEBI (simplified surface energy balance index) model, utilizing Landsat 7 and 8 optical and thermal infrared spectral bands. For yield estimation, an empirical model based on the normalized difference vegetation index (NDVI) was applied. Results indicate the effectiveness of the S-SEBI model in estimating ETa, demonstrating an R2 of 0.82 and an RMSE of 0.45 mm/day. Concurrently, yields mapped over the area range between 6 and 77 qx/ha. Globally, cereals’ average WF varied from 1.08 m3/kg to 1.22 m3/kg over the three study years, with the majority below 1 m3/kg. Notably in dry years, the importance of the blue WF is emphasized compared to years with average rainfall (WFb-2013 = 1.04 m3/kg, WFb-2012 = 0.61 m3/kg, WFb-2011 = 0.41 m3/kg). Moreover, based on an in-depth agronomic analysis combining yields and WF, four classes were defined, ranging from the most water efficient to the least, revealing that over 30% of cultivated areas during the study years (approximately 40% in 2011 and 2012 and 29% in 2013) exhibited low water efficiency, characterized by low yields and high WF. A unique index, the WFI, is proposed to assess the spatial variability of green and blue water. Spatial analysis using the WFI highlighted that in 2012, 40% of cereal plots with low yields but high water consumption were irrigated (81% blue water compared to 6% in 2011).