Field-Level cereal yield estimation using Machine Learning (ML) models poses a significant challenge especially when applied across large areas. A large sample size is required to represent the high yield variability caused by varying topographic and climatic conditions. To enhance ML-based prediction accuracy, we propose to decompose the complexity of agricultural landscape using landforms and agro-ecological zones and use these classes as spatially explicit constraints to partition field samples. We trained three ML models using remote sensing data to estimate wheat yield. When training ML models without the mentioned spatial constraints, we achieved an R2=0.58 and RMSE=840kg/ha. Training ML separately across various landform classes increase the accuracy. For instance, wheat yield cultivated in plain areas was predicted with R2=0.72, and RMSE=809kg/ha. These results emphasized the potential of training ML separately across main landform classes for improving the accuracy of yield predictions across diverse geographical contexts.

Evapotranspiration (ET) plays a crucial role in the water cycle, influencing hydrology, agriculture, and climate. Capturing its spatial and temporal distribution at broad scales is vital for effective water resource management. This study validates five ET estimation products, using direct measurements from Eddy covariance towers across different crops in Morocco's Haouz plain. The aim is to ascertain each product's performance at the field scale, including their usage over irrigated land. Findings indicate variable performances of these products depending on the crop type and irrigation scheme. Given the variance across spatial and temporal resolutions, and differing irrigation schemes, no single product uniformly excelled, showing the necessity of selecting a suitable product based on specific needs.

Food and Agriculture Organization (FAO) dual crop coefficient (FAO-2Kc) is one of the widely used formulation to estimate soil evaporation (E) due to its operationality and simplicity. The FAO-2Kc method could explicitly distinguish the contribution of E and plant transpiration, separately. However, systematic and random uncertainty in E observations still exist. In this vein, this paper attempts to improve FAO-2Kc evapotranspiration estimates through assimilating SMAP-based disaggregated soil moisture (SM) into FAO-2Kc E component via the soil evaporation coefficient. Sequential data assimilation methods (Kalman filter) was used for this purpose, where E is strongly linked to SM especially under arid atmospheric conditions where energy is not the limited factor. The proposed approach is applied over a semi-arid site in central Morocco. Results revealed that the assimilation approach provides better results in term of evapotranspiration by decreasing the root mean square error from 0.98 mm/day to 0.75 mm/day compared to the standard FAO-2Kc.

A precise estimate of surface evapotranspiration (ET) is fundamental in water science for determining the crop water needs and for optimizing water management practices and irrigation regimes. FAO-56 dual crop coefficient (FAO-2Kc) based on a water balance model allows the partitioning between bare soil evaporation (E) and plant transpiration (Tr). However, its performance for estimating the water use efficiency is limited by uncertainties in the modeled evaporation/transpiration partitioning [1], [2] due to its simplicity. This paper aims to improve the accuracy of the ET components, through assimilating remotely sensed data. Remotely sensed soil moisture (SM) and land surface temperature (LST) are simultaneously assimilated into FAO-2Kc. SM was used to improve the E component while LST to update the plant Tr element. SM and LST data were derived from SMAP and Landsat 7/8 remotely sensed observations during the 2015–2016 wheat season, respectively. The standard FAO-2Kc yields an error of 0.98 mm/day with a bias of 0.47 mm/day. Assimilating combined SM and LST into FAO-2Kc leads to an improvement of the ET prediction with an error of 0.73 mm/day compared to eddy correlation measurements.

Surface soil moisture (SM) is an essential component for crop water stress detection and irrigation management. It controls soil evaporation and plant transpiration. SM dynamics is temporally related to root zone soil moisture which is the primary measure of the plant's water status. The aim of this work is to assess the robustness of high-resolution SM product derived from remote sensing on the energy balance based latent and sensible heat fluxes. Radar SM products retrieved from Sentinel-1 data only combined with Landsat Normalized Difference Vegetation index and land surface temperature are used together to feed the energy balance model TSEB-SM to estimate turbulent fluxes. The model estimates have been evaluated against the Eddy-covariance measurements over an irrigated wheat field situated in the Haouz plain in the center of Morocco. The results are very encouraging, with few observed anomalies meanly linked to the retrieved Priestley Taylor coefficient that is affected by SM.

This study aims to establish a quantitative framework for soil nitrogen (N), phosphorus (P), and potassium (K) availability in crop fields, using hyperspectral remote sensing imagery and 30m resolution landmark map charting soil fertility across the whole of Africa that has been recently developed by a group of international scientists (iSDA soil). The iSDA map data will be analysed against PRISMA derived Level-2 surface reflectance imagery. We expect to highlight spectral absorption features of N, P and K that relate with their variability in soil. We will investigate the performance of random forest, principal component analysis, support vector regression, partial least squares regression and artificial neural networks in estimating soil total N and extractable P and K quantities from hyperspectral imagery. These models will be assessed using descriptive statistical indices and analysis of variance. We expect to demonstrate the suitability of remote sensing in precision for African agriculture.

In this letter, it is proposed to use TerraSAR-X data for analysis and estimation of soil surface texture. Our study is based on experimental campaigns carried out over a semiarid area in North Africa. Simultaneously with TerraSAR-X radar acquisitions, ground measurements (texture, soil moisture, and roughness) were made on different test fields. A strong correlation is observed between soil texture and a processed signal from two radar images, with the first acquired just after a rain event and the second corresponding to dry soil conditions, acquired three weeks later. An empirical relationship is proposed for the retrieval from radar signals of clay content percent. Soil texture mapping is proposed over the study site, which includes bare soils and olive groves.

Land surface temperature data are rarely available at high temporal and spatial resolutions at the same locations. To fill this gap, the low spatial resolution data can be disaggregated at high temporal frequency using empirical relationships between remotely sensed temperature and fractional green (photosynthetically active) and senescent vegetation covers. In this paper, a new disaggregation methodology is developed by physically linking remotely sensed surface temperature to fractional green and senescent vegetation covers using a radiative transfer equation. Moreover, the methodology is implemented with two additional factors related to the energy budget of irrigated areas, being the fraction of open water and soil evaporative efficiency (ratio of actual to potential soil evaporation). The approach is tested over a 5 km by 32 km irrigated agricultural area in Australia using airborne Polarimetric L-band Multibeam Radiometer brightness temperature and spaceborne Advanced Scanning Thermal Emission and Reflection radiometer (ASTER) multispectral data. Fractional green vegetation cover, fractional senescent vegetation cover, fractional open water, and soil evaporative efficiency are derived from red, near-infrared, shortwave-infrared, and microwave-L band data. Low-resolution land surface temperature is simulated by aggregating ASTER land surface temperature to 1-km resolution, and the disaggregated temperature is verified against the high-resolution ASTER temperature data initially used in the aggregation process. The error in disaggregated temperature is successively reduced from 1.65$^{\circ}\hbox{C}$to 1.16$^{\circ}\hbox{C}$by including each of the four parameters. The correlation coefficient and slope between the disaggregated and ASTER temperatures are improved from 0.79 to 0.89 and from 0.63 to 0.88, respectively. Moreover, the radiative transfer equation allows quantification of the impact on disaggregation of the temperature at high resolution for each parameter: fractional green vegetation cover is responsible for 42% of the variability in disaggregated temperature, fractional senescent vegetation cover for 11%, fractional open water for 20%, and soil evaporative efficiency for 27%.

Irrigated agriculture is the largest consumer of water worldwide, and especially in semi-arid areas such as southern Mediterranean countries. Agricultural water use is difficult to estimate over large areas, since several irrigation networks may co-exist and since farmers operate private pumping stations. In this context, the objective of this work was to estimate irrigation through the combination of a crop/water balance model and optical remote sensing data. The methodology was tested on a 2800 ha irrigated area located in the Tensift-Marrakech plain, central Morocco. Remote sensing data consists in the time series of FORMOSAT-2 images. This sensor operates in four bands (blue to near infrared) and allows to acquire daily images at 8 m spatial resolution. Images are used in conjunction with a simple crop model to monitor the nature of crops, their phenological states and the average level of water stress. This information and other ground data (soil map, climatic data) are introduced into a water balance model, to estimate irrigation at the pixel resolution. The results are evaluated over irrigation units (25 to 200 ha size) with data collected by the regional agency in charge of the distribution of dam water. At a seasonal scale, the accumulated irrigation water was found to vary between 0 and 300 mm, with accuracy of 25%. The approach may offer perspectives for a better estimate of the quantity of ground water used for agriculture. Indeed, it works without requiring any a priori data on agricultural practices. This makes it very attractive for operational application at a regional scale.

This paper develops two alternative approaches for downscaling passive microwave-derived soil moisture. Ground and airborne data collected over the Walnut Gulch experimental watershed during the Monsoon'90 experiment were used to test these approaches. These data consisted of eight micrometeorological stations (METFLUX) and six flights of the L-band Push Broom Microwave Radiometer (PBMR). For each PBMR flight, the 180-m resolution L-band pixels covering the eight METFLUX sites were first aggregated to generate a 500-m ldquocoarse-scalerdquo passive microwave pixel. The coarse-scale-derived soil moisture was then downscaled to the 180-m resolution using two different surface soil moisture indexes (SMIs): (1) the evaporative fraction (EF), which is the ratio of the evapotranspiration to the total energy available at the surface; and (2) the actual EF (AEF), which is defined as the ratio of the actual-to-potential evapotranspiration. It is well known that both SMIs depend on the surface soil moisture. However, they are also influenced by other factors such as vegetation cover, soil type, root-zone soil moisture, and atmospheric conditions. In order to decouple the influence of soil moisture from the other factors, a land surface model was used to account for the heterogeneity of vegetation cover, soil type, and atmospheric conditions. The overall accuracy in the downscaled values was evaluated to 3% (vol.) for EF and 2% (vol.) for AEF under cloud-free conditions. These results illustrate the potential use of satellite-based estimates of instantaneous evapotranspiration on clear-sky days for downscaling the coarse-resolution passive microwave soil moisture.

A new physically based disaggregation method is developed to improve the spatial resolution of the surface soil moisture extracted from the Soil Moisture and Ocean Salinity (SMOS) data. The approach combines the 40-km resolution SMOS multiangular brightness temperatures and 1-km resolution auxiliary data composed of visible, near-infrared, and thermal infrared remote sensing data and all the surface variables involved in the modeling of land surface-atmosphere interaction available at this scale (soil texture, atmospheric forcing, etc.). The method successively estimates a relative spatial distribution of soil moisture with fine-scale auxiliary data, and normalizes this distribution at SMOS resolution with SMOS data. The main assumption relies on the relationship between the radiometric soil temperature inverted from the thermal infrared and the microwave soil moisture. Based on synthetic data generated with a land surface model, it is shown that the radiometric soil temperature can be used as a tracer of the spatial variability of the 0-5 cm soil moisture. A sensitivity analysis shows that the algorithm remains stable for big uncertainties in auxiliary data and that the uncertainty in SMOS observation seems to be the limiting factor. Finally, a simple application to the SGP97/AVHRR data illustrates the usefulness of the approach.

The issue of deriving cross-scale aggregation rules has been extensively investigated over the last two decades. A widely used approach consists of formulating grid-scale surface radiances using the same equations that govern the patch-scale behavior but whose arguments are the aggregate expressions of those at the patch-scale. This approach derives the area-averaged or effective radiative surface temperature as might be observed using low spatial resolution satellite data. The problem however is that such satellite data exhibit large directional effects and no study has addressed this issue. The present work tackles this problem in the thermal infrared domain. The directional effects are studied by modeling. Thus, an infrared sensor observing a two-dimensional (2-D) heterogeneous plane surface is modeled. The 2-D heterogeneous plane surface is simulated by a grid with two homogeneous elements (vegetation-bare soil). The angular properties of the local surfaces, assumed homogeneous, are calculated by a multiple scattering model. The equivalent angular radiance of the complete heterogeneous scene is then determined by applying the aggregation method. This radiance is very sensitive to the surface heterogeneity, especially when the spatial variation of the surface temperature is significant and when the directional behavior of the surface is non-Lambertian. As a result, an angular variation of 6% on radiance was obtained on a heterogenous surface between a zenith angle of 70/spl deg/ and on-nadir measurements.

For the last twenty years, the needs of water has increased and implied an overexploitation of the aquifer causing an important drop of the water table. The principal objective of this study, included in the multidisciplinary project "SudMed", a French and Moroccan cooperation, is the estimation of water volumes used to irrigate cultivated surfaces and to quantify the total catchment by pumping. First, a better knowledge of the land cover of the plain is necessary. For this purpose, the use of remote sensing proves to be necessary. This land cover map allows us to calculate surfaces of the main crop types, then to estimate the water needs for irrigation.

The use of satellite data to monitor the snow contribution to runoff in a small semi-arid mountainous sub-watershed is explored. The Rehraya watershed, 228km/sup 2/ is located in the Atlas range. Its altitude range from 1068 to 4084m and its hydrological regime is characterized by an important snow contribution. The "soil and water assessment tool" has been selected to model the watershed's functioning and first results showed some limitations. To partly overcome the lack of climatic data over this area, we extract the snow extent from satellite images and we integrate this information in the model which should lead to better results.

The SUD-MED project aims at monitoring water resources over Mediterranean regions. As part of the project, this paper presents a method we developed for estimating cereal water requirement. The method consists in driving the simple model developed by the FAO with remotely-sensed data. It was tested on an little area cultivated with wheat in the semi-arid Marrakech plain (Morocco). We use a time series of high spatial resolution images acquired by SPOT-4/HRVIR during the 2001/2002 agricultural season. The method outlines the spatio-temporal patterns of crop cycles. The associated maps of phenological variables and seasonal evapotranspiration appear consistent with regional rainfall and irrigation features. Perspectives of improvement are finally discussed.

This study is part of the SUDMED project from IRD (Research Institute for Development), CESBIO, University of Marrakech and Moroccan administrations in charge of agriculture, forestry and water management. The objective of this project is the hydro-ecological modeling of hydrological resources on the Marrakech region. In this context, it is important to characterize the snow cover and the melting dynamics as it is the main water source for the plain. Evaluate the snow cover using satellite images at a short temporal scale is a first step to estimate water quantity available during the melting season, in spring. The objective of this study is to calculate the snow cover area and to follow its evolution during the winter season using both high and low satellite imagery. For this purpose, several snow cover indices (SCI) calculated from low resolution images SPOT/VEGETATION (pixel size of 1 km/sup 2/) have been compared with the snow cover percentage (SCP) calculated derived from high spatial resolution sensors (in particular LandSat-TM). Particular attention was given to the classification procedure and the co-registration problem. Then, pixel-based regressions between SCP and SCI have been calibrated for several dates (when both VEGETATION and TM images were available) with the objective of finding the most robust relationship. The previous relation is applied to a time series of 30 SPOT/VEGETATION images acquired from December 1998 to May 1999. The dynamics of snow cover is compared with rainfalls and flow chronicles observed on three mountainous watersheds. The main interest of this study is to show that low resolution precision can be easily improved using high resolution imagery to obtain reliable quantitative information on snow cover. This last information is an important input of snowmelt runoff models.

Remote sensing and crop models have proved to be useful to monitor vegetation and estimate above ground biomass. In this study, NDVI from VEGETATION, MODIS, and Landsat reflectance data were compared with field measurements. The phenology was inferred and identified the main stages. LAI obtained from reflectance was used with the STICS model to give estimates of grain yield within about 5% of field measurements. No clear relationship was established between the sowing date and yield. Temperature seems to be the most important driver of wheat phenology.