In the present study, the ground-based vicarious calibration based on in-situ spectroradiometric measurements and a semi-empirical line approach (SELA) using pseudo-invariant targets (PITs) has been applied to convert the raw data (DN(λ)) acquired with a UAV-multispectral imaging system (Phantom DJI Pro 4+) to the ground-surface reflectance factors (GSRF(λ)) in the visible, red-edge and NIR wavebands. To achieve so, spectral measurements were acquired using an ASD considering four types of homogeneous bright and dark PITs (i.e. white ceramic tiles, concrete surface, asphalt, and very dense vegetation cover). Statistical regressions were established (p2≤ 0.97) for all considered bands. While the GSRF were retrieved with an RMSE ≤ 5.4% for visible, and around 3% for red-edge and NIR. The magnitudes of these RMSE are in agreement with the expected absolute radiometric calibration accuracies reported by the literature, providing good quality-assurance to retrieve biophysical and/or geophysical parameters comparable in time and consistent with other calibrated datasets. Undoubtedly, the SELA can be used with confidence ensuring acceptable accuracies for radiometric calibration and atmospheric corrections of imagery data acquired by UAV-mounted imaging system.

The present study analyzes for the first time the usefulness of the synergy between UAV-multispectral data and biomass-chlorophyll indices to discriminate and map carob trees dieback and damages. To achieve so, the UAV flight was performed over a carob forest located in a valley of a watershed in the Moroccan Middle-Atlas Mountain. The UAV data were rigorously pre-processed and twelve biomass-chlorophyll indices were implemented, analysed (spectrally and radiometrically), and validated using the ground truth. Then, a histogram thresholding classification was applied to the index offering the best performance. The results obtained pointed out that the TDVI and CIGindices have similar and good dynamic range values, and better performance than the other indices tested. They are well correlated, completely independent of soil background artefacts, and relatively avoid linearity and saturation problems. They showed a curvilinear relationship between their computed values and the considered classes (i.e., bare soil, healthy, dieback, and dead trees). The validation shows that TDVI and CIGare sensitive to the carob spatial variations, thus allowing an excellent land-use separating power, predicting early warning signals of dieback, and providing useful bio-physiological information about carob tree conditions. Furthermore, the results highlighted the radiometric and spectral performance of the DJI Phantom-4 camera for powerful sensitivities in discriminating carob forest classes. This simple and quick method can be a useful tool for decision support for monitoring and protecting carob forests on a large scale to promote sustainable development.

The purposes of this paper is an evaluation of the accuracy of DEMs and associated products derived from UAV-data acquired at two different flight altitudes over a narrow-deep mountainous valley area. The assessment of quality and accuracy were undertaken, respectively, qualitatively by reference to the field observations and quantitatively using DGPS in situ measurements. The assigned time for processing and products derivations steps was also analyzed. The images were acquired using UAV Mavic Pro-2 RGB digital camera at two flight altitudes of 80 and 150 m providing, respectively, 2 dissimilar spatial resolutions, 2.0 and 3.5 cm. The "PIX4D-Mapper" software was used for photogrammetric processing. Uniformly distributed over the study area, 35 GCPs were measured with accurate 500-RTK GPS (σ ≤ ± 1 cm). Among them, 26 points were applied for the calibration procedure and 9 as checkpoints. The results demonstrated that the low altitude flight at 80 m enabled the acquisition of 482 overlapping images that were processed during 7 h 41 min, yielding an RMSE ˂ ± 2.0 on planimetry and ˂ ± 2.5 cm on altimetry of checkpoints. Whereas, the flight at 150 m allowed the recording of 250 overlapping images which were processed during 3 h 22 min, reaching an RMSE ˂ ± 3.0 and ˂ ± 3.7 cm on planimetry and altimetry, respectively. Visual analysis pointed out that regardless the considered altitude, the derived products (DEMs, DSMs and ortho-image) are very similar and faithfully reflect the reality on the ground. Accordingly, a flight at 150 m is able to provide satisfactory accuracies of DEM and auxiliary products generated from UAV Mavic Pro-2 camera data for applications in narrow-deep mountain valleys. However, an experienced operator of drone flight control is required to avoid the risks of accidents when maneuvering between narrow rock walls and to land successfully the drone in a safe place. Moreover, these operations are also conditioned by the reliability of telemetry control and the availability of GPS (or GNSS) signals in this harsh environment.

The sulfates, carbonates, and chlorides based minerals have been remotely detected on the Martian surfaces. These evaporite deposits are not only indicative of environmental conditions of their formation, but can provide clues indicating their interaction with water. Due to narrow absorption features, the hyperspectral remote sensing is a powerful technology to identify and discriminat evaporite minerals. This study intends to detect evaporite minerals including gypsum and salt deposits on the Ritchey crater region using hyperspectral data acquired by CRISM carried onboard MRO platform. To achieve this, the data were pre-processed to correct several radiometric, atmospheric, spectral and geometric anomalies. Afterwards, the most popular indices for gypsum and salts detection in arid terrestrial environments such as the NDGI and SSSI were implemented and maps derived. The results obtained indicate that both the indices have mapped the evaporite minerals patterns almost identical. However, the NDGI further highlights the gypsum content, while SSSI product show greater sensitivity to evaporit minerals (i.e., gypsum, sulfate chloride, halite, etc.). Likewise, for validation and comparison procedure, SSSI derived product revealed high similarity with that of chloride published by CRISM-NASA team. Moreover, the spectral signature extracted from the pixels with high evaporite minerals content exhibits the largest hydration signature around 1.95 μm, which is associated with several features typical of gypsum and halite between 1.0 and 2.5 μm. The position, shape, and structure of the absorption features at 1.45, 1.75 and 1.95 μm, as well the gradual decrease in reflectance at 2.4 μm provide some resemblance with the gypsum compared to less hydrated forms of calcium sulfates. However, these absorption characteristics look like a mixture among gypsum, bassanite, anhydrite, and halite.

In precision agriculture, accurate DEMs are very important for the characterisation of topographic attributes (elevation, slope, orientation, curvature, flow direction and accumulation) to understand the spatial variability at the field or within-field scale for managing several agricultural parameters including irrigation, fertilization, runoff production, etc. Unlike conventional methods, the synergy between UAV and digital photogrammetry is a new geomatics approach to generate DEMs with centimetre pixel size. This paper aims a multi-scale analysis of DEMs and related products (DSMs and orthomosaics) derived from UAV data acquired at three different flight altitudes over agricultural field. The quality and accuracies assessment were undertaken, respectively, qualitatively by reference to the field observations and quantitatively using DGPS in situ measurements. The assigned time for processing and products derivations steps was also analyzed. The images were acquired using UAV-Camera (DJI Phantom 4 Pro+ V2.0) at three different flight altitudes (50, 100 and 150 m), providing three dissimilar spatial resolutions (1.25, 2.75 and 4.15 cm). The “Pix4Dmapper” software was used for photogrammetric processing. Uniformly distributed over the study area, 19 GCPs were measured with accurate DGPS (σ ≤ ± 1 cm), LEICA 500-RTK. Among them, 6 points were used for the calibration procedure and 13 as check-points. The results demonstrated that the low altitude flight at 50 m enabled the acquisition of 736 overlapping images that were processed during 5 h 55 min, yielding an RMSE < ± 2 on planimetry and < ± 3 cm on altimetry of check-points. Similar accuracies were achieved for the flight at 100 m, after processing 231 stereoscopic images during 2 h 26 min. Whereas, the flight at 150 m allowed the recording of 103 overlapping images which were processed during 1 h 46 min, reaching an RMSE of ± 2 and ± 4 cm on planimetry and altimetry, respectively. Visual analysis pointed out that regardless the considered altitude, the derived products (DEMs, DSMs and orthomosaics) are very similar and faithfully reflect the reality on the ground. Accordingly, a flight at 150 m is certainly able to provide satisfactory accuracies of DEM and auxiliary products generated from UAV-Camera (DJI Phantom 4 Pro+) data for applications in the context of precision agriculture.

Date palm trees play an important role in Oasis economy, particularly in North-Africa and Middle-East regions. Obviously, information about the palm trees number and health conditions is very fundamental for assessing stress and disease, monitoring stage of growth and development, as well as predicting yield. Deep learning is an innovative method, recently adapted and applied in forest environment to detect and delineate individual tree crowns from remote sensing imagery. This paper aims the experimentation of U-N et deep neural network model to detect individual date palm tree crowns from very high spatial resolution images acquired above “Targa N'Touchka” Oasis plantations in south-west of Morocco. To achieve this goal, a scene of Bing-images has been used. It was sampled in 420 of 128x128 pixels and divided into 70% sub-images to train the data and 30% were used to test and validate the output results. The U-Net model was implemented using Tensor-Flow and Keras. The accuracy of results obtained indicates better classification efficiency (96.94%). Overall, individual crowns of date palms are satisfactory delimited and mapped. However, despite its sophistication and performance, the model requires extensive training information and high investment in calibration mode. As well as, the trees with overlapping crowns are not properly recognized as separate entities. Therefore, further investigations are necessary, particularly over large area with dense plantations.

Soil moisture (SM) is an important physical parameter for several hydrological and agricultural applications, weather and climate predictions, as well as early warning of natural hazards such as flood and drought. It varies significantly in space and time, and it is a challenge to map its variations accurately at the regional and local scales. The aim of the present study is a comparison and validation of four models such as NDWI, SIWSI, TOTRAM and OPTRAM for SM mapping before and after flash-flood storm in arid land. The used methodology exploits two pairs of images acquired with Landsat-OLI/TIRS sensors over the study area before and after flood. The first pair was acquired two weeks before the flash flood, and the second one was collected eight days after the flood. These images were radiometrically and atmospherically corrected, as well topographically rectified using SRTM DEM. For validation purposes, before and after storm, two independent SM products were used. Both, they were derived from data acquired simultaneously with those recorded by Landsat-OLI/TIRS. The first is derived from SMOS data, and the second is compiled from rainfall data (SM-RFE) delivered by NOAA climate prediction center RFE (Rainfall Estimator) for Africa. The results revealed that TOTRAM and NDWI models converge towards the same conclusions describing accurately the drastic change of SM before and after flash-flood highlighting the impact of inundation and the mud accumulation over the study area. Their results validation against those of SMOS and SM-RFE products shows a significant agreement (R2 of 0.95). However, in addition to its simplicity and to its easy implementation, the NDWI show a great potential for SM content estimation independently of LST using only optical data such as Landsat-OLI or Sentinel-MSI.