Accurate spatiotemporal monitoring and modeling of soil moisture (SM) is of paramount importance for various applications ranging from food production to climate change adaptation. This study deals with modeling SM with the random forest (RF) algorithm using datasets comprising multispectral data from Sentinel-2, Landsat-8/9, and hyperspectral data from the CoSpectroCam sensor (CSC, licensed to AgriWatch BV, Enschede, The Netherlands) mounted on an unmanned aerial vehicle (UAV) in Iran. The model included nine bands from Landsat-8/9, 11 bands from Sentinel-2, and 1252 bands from the CSC (covering the wavelength range between 420 and 850 nm). The relative feature importance and band sensitivity to SM variations were analyzed. In addition, four indices, including the perpendicular index (PI), ratio index (RI), difference index (DI), and normalized difference index (NDI) were calculated from the different bands of the datasets, and their sensitivity to SM was evaluated. The results showed that the PI exhibited the highest sensitivity to SM changes in all datasets among the four indices considered. Comparisons of the performance of the datasets in SM estimation emphasized the superior performance of the UAV hyperspectral data (R² = 0.87), while the Sentinel-2 and Landsat-8/9 data showed lower accuracy (R² = 0.49 and 0.66, respectively). The robust performance of the CSC data is likely due to its superior spatial and spectral resolution as well as the application of preprocessing techniques such as noise reduction and smoothing filters. The lower accuracy of the multispectral data from Sentinel-2 and Landsat-8/9 can also be attributed to their relatively coarse spatial resolution compared to the CSC, which leads to pixel non-uniformities and impurities. Therefore, employing the CSC on a UAV proves to be a valuable technology, providing an effective link between satellite observations and ground measurements.

Enhancing digital and precision agriculture is currently inevitable to overcome the economic and environmental challenges of the agriculture in the 21st century. The purpose of this study was to generate and compare management zones (MZ) based on the Sentinel-2 satellite data for variable rate application of mineral nitrogen in wheat production, calculated using different remote sensing (RS)-based models under varied soil, yield and crop data availability. Three models were applied, including (1) a modified “RS- and threshold-based clustering”, (2) a “hybrid-based, unsupervised clustering”, in which data from different sources were combined for MZ delineation, and (3) a “RS-based, unsupervised clustering”. Various data processing methods including machine learning were used in the model development. Statistical tests such as the Paired Sample T-test, Kruskal–Wallis H-test and Wilcoxon signed-rank test were applied to evaluate the final delineated MZ maps. Additionally, a procedure for improving models based on information about phenological phases and the occurrence of agricultural drought was implemented. The results showed that information on agronomy and climate enables improving and optimizing MZ delineation. The integration of prior knowledge on new climate conditions (drought) in image selection was tested for effective use of the models. Lack of this information led to the infeasibility of obtaining optimal results. Models that solely rely on remote sensing information are comparatively less expensive than hybrid models. Additionally, remote sensing-based models enable delineating MZ for fertilizer recommendations that are temporally closer to fertilization times.

Early-season area estimation of the winter wheat crop as a strategic product is important for decision-makers. Multi-temporal images are the best tool to measure early-season winter wheat crops, but there are issues with classification. Classification of multi-temporal images is affected by factors such as training sample size, temporal resolution, vegetation index (VI) type, temporal gradient of spectral bands and VIs, classifiers, and values missed under cloudy conditions. This study addresses the effect of the temporal resolution and VIs, along with the spectral and VIs gradient on the random forest (RF) classifier when missing data occurs in multi-temporal images. To investigate the appropriate temporal resolution for image acquisition, a study area is selected on an overlapping area between two Landsat Data Continuity Mission (LDCM) paths. In the proposed method, the missing data from cloudy pixels are retrieved using the average of the k-nearest cloudless pixels in the feature space. Next, multi-temporal image analysis is performed by considering different scenarios provided by decision-makers for the desired crop types, which should be extracted early in the season in the study areas. The classification results obtained by RF improved by 2.2% when the temporally-missing data were retrieved using the proposed method. Moreover, the experimental results demonstrated that when the temporal resolution of Landsat-8 is increased to one week, the classification task can be conducted earlier with slightly better overall accuracy (OA) and kappa values. The effect of incorporating VIs along with the temporal gradients of spectral bands and VIs into the RF classifier improved the OA by 3.1% and the kappa value by 6.6%, on average. The results show that if only three optimum images from seasonal changes in crops are available, the temporal gradient of the VIs and spectral bands becomes the primary tool available for discriminating wheat from barley. The results also showed that if wheat and barley are considered as single class versus other classes, with the use of images associated with 162 and 163 paths, both crops can be classified in March (at the beginning of the growth stage) with an overall accuracy of 97.1% and kappa coefficient of 93.5%.

The early-season area estimation of the winter wheat crop as a strategic product is important for decision makers. Classification of multi-temporal images is an approach which is affected by many factors like appropriate training sample size, proper frequency and acquisition times, vegetation indices (VIs) type, temporal gradient of spectral bands and VIs, appropriate classifier and missed values because of cloudy conditions. This paper addresses the impact of appropriate frequency and acquisition times and VIs type along with the spectral and VI gradient on random forest (RF) classifier when missed values exist in multi-temporal images. To investigate the appropriate temporal resolution for image acquisition, the study area was selected on an overlapping area between two LDCM paths. In our developed method, the miss values of cloudy bands for each pixel are retrieved by the mean of k-nearest ordinary pixels. Then the multi-temporal image analysis is performed by considering different scenarios provided by decision makers in terms of desired crop types that should be extracted at early-season in the study areas. The classification results obtained by the RF decrease by 1.6% when temporally missed values retrieved by the proposed method, which is an acceptable result. Moreover, the experimental results demonstrated that if temporal resolution of Landsat 8 increased to one week the classification task can be conducted earlier with almost better results in terms of OA and kappa. The impact of incorporating VIs along with the temporal gradients of spectral bands and VIs as new features in RF demonstrated that the OA and Kappa are improved 3.1% and 6.6%, respectively. Furthermore, the obtained result showed that if only one image from seasonal changes of crops is available, the temporal gradient of VIs and spectral bands play the main role to discriminate remarkably wheat from barley. The experiments also demonstrated that if both wheat and barley merge to a single class the crop area can be estimated two months earlier with 97.1 and 93.5 in terms of OA and kappa, respectively.

Pear cultivation in Greece, with an annual production of approximately 81,000 tonnes, constitutes a significant segment of the national fruit industry, particularly in Northern regions such as Macedonia and Thessaly. Despite ranking 8th in the EU in terms of pear production, Greece’s cultivated area is slightly declining, and adoption of smart agriculture technologies (SAT) remains limited. In this context, the present study investigates the preferences, patterns, and barriers of SAT adoption within the Greek pear sector, aiming to lay the groundwork for more effective digital transformation in the agri-food domain. Using a structured interview-based survey, data were collected from 30 pear growers, revealing critical insights into the technological landscape of the sector. A central challenge that emerged was the insufficient internet connectivity in rural farming areas, highlighting the urgent need for improved digital infrastructure to support SAT deployment. Furthermore, the study emphasizes the importance of targeted education and awareness programs to bridge the digital knowledge gap among pear farmers. An especially notable finding concerns the role of the chosen tree training system in influencing SAT uptake: more than 50% of adopters utilize the palmette training system, suggesting a strong correlation between orchard design and technological readiness. Among the SAT categories, Data Analytics and Farm Management Software were the most widely adopted, a trend partly driven by attractive governmental subsidies of €30 per hectare. Importantly, all respondents who had implemented SAT (100%) reported a measurable increase in farm income, reinforcing the technologies’ impact on productivity and profitability. Foremost among the challenges encountered is the deficit in technical knowledge and training. In conclusion, this study offers a comprehensive overview of Greek pear producers’ perceptions, challenges, and emerging opportunities related to smart agriculture.