International audienceA meta-analysis data-driven approach is developed to represent the soil evaporative efficiency (SEE) defined as the ratio of actual to potential soil evaporation. The new model is tested across a bare soil database composed of more than 30 sites around the world, a clay fraction range of 0.02-0.56, a sand fraction range of 0.05-0.92, and about 30,000 acquisition times. SEE is modeled using a soil resistance ($r_{ss}$) formulation based on surface soil moisture ($\theta$) and two resistance parameters $r_{ss,ref}$ and $\theta_{efolding}$. The data-driven approach aims to express both parameters as a function of observable data including meteorological forcing, cut-off soil moisture value $\theta_{1/2}$ at which SEE=0.5, and first derivative of SEE at $\theta_{1/2}$, named $\Delta\theta_{1/2}^{-1}$. An analytical relationship between $(r_{ss,ref};\theta_{efolding})$ and $(\theta_{1/2};\Delta\theta_{1/2}^{-1})$ is first built by running a soil energy balance model for two extreme conditions with $r_{ss} = 0$ and $r_{ss}\sim\infty$ using meteorological forcing solely, and by approaching the middle point from the two (wet and dry) reference points. Two different methods are then investigated to estimate the pair $(\theta_{1/2} ; \Delta\theta_{1/2}^{-1})$ either from the time series of SEE and $\theta$ observations for a given site, or using the soil texture information for all sites. The first method is based on an algorithm specifically designed to accomodate for strongly nonlinear $\text{SEE}(\theta)$ relationships and potentially large random deviations of observed SEE from the mean observed $\text{SEE}(\theta)$. The second method parameterizes $\theta_{1/2}$ as a multi-linear regression of clay and sand percentages, and sets $\Delta\theta_{1/2}^{-1}$ to a constant mean value for all sites. The new model significantly outperformed the evaporation modules of ISBA (Interaction Sol-Biosph\`{e}re-Atmosph\`{e}re), H-TESSEL (Hydrology-Tiled ECMWF Scheme for Surface Exchange over Land), and CLM (Community Land Model). It has potential for integration in various land-surface schemes, and real calibration capabilities using combined thermal and microwave remote sensing data

Radar data have been used to retrieve and monitor the surface soil moisture (SM) changes in various conditions. However, the calibration of radar models whether empirically or physically-based, is still subject to large uncertainties especially at high-spatial resolution. To help calibrate radar-based retrieval approaches to supervising SM at high resolution, this paper presents an innovative synergistic method combining Sentinel-1 (S1) microwave and Landsat-7/8 (L7/8) thermal data. First, the S1 backscatter coefficient was normalized by its maximum and minimum values obtained during 2015-2016 agriculture season. Second, the normalized S1 backscatter coefficient was calibrated from reference points provided by a thermal-derived SM proxy named soil evaporative efficiency (SEE, defined as the ratio of actual to potential soil evaporation). SEE was estimated as the radiometric soil temperature normalized by its minimum and maximum values reached in a water-saturated and dry soil, respectively. We estimated both soil temperature endmembers by using a soil energy balance model forced by available meteorological forcing. The proposed approach was evaluated against in situ SM measurements collected over three bare soil fields in a semi-arid region in Morocco and we compared it against a classical approach based on radar data only. The two polarizations VV (vertical transmit and receive) and VH (vertical transmit and horizontal receive) of the S1 data available over the area are tested to analyse the sensitivity of radar signal to SM at high incidence angles (39°-43°). We found that the VV polarization was better correlated to SM than the VH polarization with a determination coefficient of 0.47 and 0.28, respectively. By combining S1 (VV) and L7/8 data, we reduced the root mean square difference between satellite and in situ SM to 0.03 m 3 m-3 , which is far smaller than 0.16 m 3 m-3 when using S1 (VV) only.

A precise estimate of the evapotranspiration (ET) partitioning is fundamental for determining the crop water needs and optimizing irrigation management. The plant transpiration (T) is generally considered to be the most desirable component, while reducing the soil evaporation (E) could be one of the most important water-saving actions in semi-arid agricultural regions. Given the lack of reference method to estimate the E/T partitioning of wheat crop, this study inter-compares four different methods based on eddy covariance, sap flow and lysimetry measurements and FAO modeling. The objectives are: i) to quantify the systematic and random uncertainty in E and T observations, ii) to evaluate the partitioning ratio (T/ET) at the daily/field scale and iii) to assess the performance of the FAO model over two drip irrigated wheat fields. Results indicate that despite the small surface sensed by minilysimeters, the partitioning ratio is evaluated more precisely (19% relative error) with lysimetry than with the other systems (any combination of eddy covariance, lysimetry and sap flow measurements). Moreover, stem-scale T measurements from sap flow sensors are subject to representativeness issues at the field scale, and to systematic errors during water-stress and senescence periods. The lysimeter-derived partitioning ratio increases from about 0.50 to 0.85 during the growth stage and rapidly drops towards 0 during senescence. Its dynamics is found to be significantly correlated (R>0.7) with the 5-cm soil moisture. By comparing FAO simulations with observations, it is found that the FAO method overestimates T and underestimates E, while keeping satisfying ET estimates for drip irrigated wheat. This study suggests that different independent measurement techniques should be implemented to both quantify and reduce uncertainties in the T/ET ratio, and that accurate observations are still needed to improve the modeling of E/T components.

Global soil moisture (SM) products are currently available from passive microwave sensors at typically 40 km spatial resolution. Although recent efforts have been made to produce 1 km resolution data from the disaggregation of coarse scale observations, the targeted resolution of available SM data is still far from the requirements of fine-scale hydrological and agricultural studies. To fill the gap, a new disaggregation scheme of Soil Moisture Active and Passive (SMAP) data is proposed at 100 m resolution by using the disaggregation based on physical and theoretical scale change (DISPATCH) algorithm. The main objectives of this paper is (i) to implement DISPATCH algorithm at 100 m resolution using SMAP SM and Landsat land surface temperature and vegetation index data and (ii) to investigate the usefulness of an intermediate spatial resolution (ISR) between the SMAP 36 km resolution and the targeted 100 m resolution. The sequential disaggregation approach from 36 km to ISR (ranging from 1 km to 30 km) and from ISR to 100 m resolution is evaluated over 22 irrigated field crops in central Morocco using in-situ SM measurements collected from January to May 2016. The lowest root mean square difference (RMSD) between the 100 m resolution disaggregated and in-situ SM is obtained when the ISR is around 10 km. Therefore, the two-step disaggregation is more efficient than the direct disaggregation from SMAP to 100 m resolution. Moreover, we propose a moving average window algorithm to increase the accuracy in the 100 m resolution SM as well as to reduce the low-resolution boxy artifacts on disaggregated images. The correlation coefficient between 100 m resolution disaggregated and in situ SM ranges between 0.5–0.9 for four out of the six extensive sampling dates. This methodology relies solely on remote sensing data and can be easily implemented to monitor SM at a high spatial resolution over irrigated regions.

Modeling soil evaporation has been a notorious challenge due to the complexity of the phenomenon and the lack of data to constrain it. In this context, a parsimonious model is developed to estimate soil evaporative efficiency (SEE) defined as the ratio of actual to potential soil evaporation. It uses a soil resistance driven by surface (0 − 5 cm) soil moisture, meteorological forcing and time (hour) of day, and has the capability to be calibrated using the radiometric surface temperature derived from remotely sensed thermal data. The new approach is tested over a rainfed semi-arid site, which had been under bare soil conditions during a 9-month period in 2016. Three calibration strategies are adopted based on SEE time series derived from 1) eddy-covariance measurements, 2) thermal measurements, and 3) eddycovariance measurements used only over separate drying periods between significant rainfall events. The correlation coefficients (and slopes of the linear regression) between simulated and observed (eddy-covariance-derived) SEE

Evapotranspiration (ET) plays an essential role for detecting plant water status, estimating crop water needs and optimising irrigation management. Accurate estimates of ET at field scale are therefore critical. The present paper investigates a remote sensing and modelling coupled approach for monitoring actual ET of irrigated wheat crops in the semi-arid region of Tensift Al Haouz (Morocco). The ET modelling is based on a modified Penman-Monteith equation obtained by introducing a simple empirical relationship between surface resistance (rc) and a stress index (SI). SI is estimated from Landsat-derived land surface temperature (LST) combined with the LST endmembers (in wet and dry conditions) simulated by a surface energy balance model driven by meteorological forcing and Landsat-derived fractional vegetation cover. The proposed model is first calibrated using eddy covariance measurements of ET during one growing season (2015-2016) over an experimental flood-irrigated wheat field located within the irrigated perimeter named R3. It is then validated during the same growing season over another dripirrigated wheat field located in the same perimeter. Next, the proposed ET model is implemented over a 10 x 10 km 2 area in R3 using a time series of Landsat-7/8 reflectance and LST data. The comparison between modelled and measured ET fluxes indicates that the model works well. The Root Mean Square Error (RMSE) values over drip and flood sites were 13 and 12 W m-2 , respectively. The proposed approach has a great potential for detecting crop water stress and estimating crop water requirements over large areas along the agricultural season.

Accurate seasonal forecasting of cereal yields is an important decision support tool for countries, such as Morocco, that are not self-sufficient in order to predict, as early as possible, importation needs. This study aims to develop an early forecasting model of cereal yields (soft wheat, barley and durum wheat) at the scale of the agricultural province considering the 15 most productive over 2000–2017 (i.e., 15 × 18 = 270 yields values). To this objective, we built on previous works that showed a tight linkage between cereal yields and various datasets including weather data (rainfall and air temperature), regional climate indices (North Atlantic Oscillation in particular), and drought indices derived from satellite observations in different wavelengths. The combination of the latter three data sets is assessed to predict cereal yields using linear (Multiple Linear Regression, MLR) and non-linear (Support Vector Machine, SVM; Random Forest, RF, and eXtreme Gradient Boost, XGBoost) machine learning algorithms. The calibration of the algorithmic parameters of the different approaches are carried out using a 5-fold cross validation technique and a leave-one-out method is implemented for model validation. The statistical metrics of the models are first analyzed as a function of the input datasets that are used, and as a function of the lead times, from 4 months to 2 months before harvest. The results show that combining data from multiple sources outperformed models based on one dataset only. In addition, the satellite drought indices are a major source of information for cereal prediction when the forecasting is carried out close to harvest (2 months before), while weather data and, to a lesser extent, climate indices, are key variables for earlier predictions. The best models can accurately predict yield in January (4 months before harvest) with an R2 = 0.88 and RMSE around 0.22 t. ha−1. The XGBoost method exhibited the best metrics. Finally, training a specific model separately for each group of provinces, instead of one global model, improved the prediction performance by reducing the RMSE by 10% to 35% depending on the provinces. In conclusion, the results of this study pointed out that combining remote sensing drought indices with climate and weather variables using a machine learning technique is a promising approach for cereal yield forecasting.

32The use of land surface temperature (LST) for monitoring the consumption and water status 33 of crops requires data at fine spatial and temporal resolutions. Unfortunately, the current 34 spaceborne thermal sensors provide data at either high temporal (e.g. MODIS: Moderate 35 Resolution Imaging Spectro-radiometer) or high spatial (e.g. Landsat) resolution separately. 36 Disaggregating low spatial resolution (LR) LST data using ancillary data available at high 37 spatio-temporal resolution could compensate for the lack of high spatial resolution (HR) LST 38 observations. Existing LST downscaling approaches generally rely on the fractional green 39 vegetation cover (fgv) derived from HR reflectances but they do not take into account the soil 40 water availability to explain the spatial variability in LST at HR. In this context, a new 41 method is developed to disaggregate kilometric MODIS LST at 100 m resolution by including 42 the Sentinel-1 (S-1) backscatter, which is indirectly linked to surface soil moisture, in addition 43 to the Landsat-7 and Landsat-8 (L-7 & L-8) reflectances. The approach is tested over two 44 different sites -an 8 km by 8 km irrigated crop area named "R3" and a 12 km by 12 km 45 rainfed area named "Sidi Rahal" in central Morocco (Marrakech)on the seven dates when S-46 1, and L-7 or L-8 acquisitions coincide with a one-day precision during the 2015-2016 47 growing season. The downscaling methods are applied to the 1 km resolution MODIS-Terra 48 LST data, and their performance is assessed by comparing the 100 m disaggregated LST to 49 Landsat LST in three cases: no disaggregation, disaggregation using Landsat fgv only, 50 disaggregation using both Landsat fgv and S-1 backscatter. When including fgv only in the 51 disaggregation procedure, the mean root mean square error in LST decreases from 4.20 to 52 3.60 °C and the mean correlation coefficient (R) increases from 0.45 to 0.69 compared to the 53 non-disaggregated case within R3. The new methodology including the S-1 backscatter as 54 input to the disaggregation is found to be systematically more accurate on the available dates 55 with a disaggregation mean error decreasing to 3.35 °C and a mean R increasing to 0.75. 56 68 studies (Crow and Wood, 2003). However, there is a limitation in the existing satellite thermal 69 sensors, since those with high revisit cycles (e.g, MODerate resolution Imaging Spectro-70 radiometer 'MODIS') do not offer high spatial resolution (HR), and those offering HR (e.g, 71 Landsat-8) generally have low temporal resolution (Agam et al., 2007b). In contrast, the 72 visible and near infrared (VNIR) reflectance data are available at a resolution finer than that 73 of most thermal sensors (Ha et al., 2013). To bridge the gap, the finer-resolution VNIR data 74 have hence been extensively used as ancillary data to disaggregate low-resolution (LR) LST 75 at HR. 76 Recently, various efforts have been devoted to disaggregate LST to a finer -typically 100 m-77 resolution. Techniques are generally based on a relation...