State-of-the-art technologies for remote sensing of crops water status and nutrients in agriculture: A review

. The research relevance is predetermined by the need to introduce effective methods and tools for monitoring water resources and tracking soil nutrient levels to improve agricultural production and sustainable use of natural resources. The research aims to provide a comprehensive overview of the latest technologies and techniques used in RS for accurately assessing water status and nutrient levels in crops, aiming to enhance agricultural productivity and sustainability. The latest advancements in remote sensing techniques that enable precise monitoring and assessment of water levels and nutrient conditions in crops, crucial for optimizing agricultural practices, were studied. The literature research was conducted by adapting the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Methods. The current study provides an overview of RS technology, with a special focus on establishing crops’ water and nutrient status in agriculture. A thorough review of research focused on the applications and technologies of RS in agriculture, using a broad-to-narrow approach, was also conducted. The scientific studies analysed provide the following:

reflectance monitoring as a tool for water stress detection in greenhouses.They reviewed past research and recent advances regarding the sensors and approaches used for crop reflectance measurements and the indices used for crop water and nutrient status detection.A. Damm et al. (2018) conducted a review of remote sensing of plant-water relations.They reviewed suitable remote sensing approaches to assess plant-water relations ranging from pure observational to combined observational-modelling approaches.They used a combined energy balance and radiative transfer model to assess the explanatory power of pure observational approaches focusing on plant parameters to estimate plant-water relations, followed by an outline for a more effective use of remote sensing by their integration into soil-plant-atmosphere continuum (SPAC) models.It is possible to conclude that there are similar studies, but they are not categorized as in this publication -from the general to the particular, from the global scale to the leaf level.
The research aims to present a thorough examination of contemporary technologies and methodologies employed in remote sensing to precisely evaluate the water conditions and nutrient concentrations in crops.The ultimate goal is to improve agricultural efficiency and promote sustainable practices.

REMOTE SENSING CROPS ON LARGE SCALE LEVEL
Improved satellite data resolution from platforms like Landsat 8 and Sentinel-2 has opened up new possibilities for monitoring crop conditions and mapping yields on a larger scale in agriculture.J. Clevers et al. (2017) utilized Sentinel-2 satellite images to estimate LAI, LCC, and CCC of potato crops using vegetation indices (VIs).Their study found that Sentinel-2 bands with a 10 m spatial resolution were sufficient, eliminating the need for red-edge bands with a 20 m resolution.J. Miao et al. (2022) employed machine learning models to estimate leaf nutrient levels in mangrove forests using Sentinel-2 images.They compared XGBoost, RF, and LightGBM models and mapped nutrient levels for multiple seasons, highlighting the importance of monitoring seasonal changes.A. Hama et al. (2022) introduced a low-cost LiDAR approach to estimate leaf water content (LWC) by measuring reflectance in the 905 nm band, which was found to be closely related to leaf structure.T. Dong et al. (2020) utilized Landsat 8 and Sentinel-2 data to estimate crop biomass in Manitoba, Canada, by parametrizing a crop growth model using remotely sensed LAI.Lastly, I. Sanches et al. (2014) used spectral feature analysis to detect plant stress in visible/near-infrared wavelengths, identifying a new index, PSDI, effective for detecting stress in hydrocarbon-impacted plants based on field and airborne data.Table 1 presents the above data in a summarized form:

INTRODUCTION
The timely and reliable information on crops' water and nutrient status is of critical importance in the conditions of modern precision agriculture and smart farms.By 2050, the world population is expected to increase to almost 10 billion people, and to feed them, existing food production must be increased by 59-98% according to estimates by the Food and Agriculture Organization (FAO, 2023).It is also estimated that agriculture will use over 70% of the world's freshwater reserves, with about half of it being lost or wasted.Up to 95% of the world's food production depends on soil.As a result of unsustainable agricultural practices, overexploitation of natural resources, and growing populations, one-third of the soils have already degraded, and experts estimate that soil erosion could lead to a 10% loss of crop yields by 2050.Soils are also full of life and contain approximately 25% of the world's biodiversity.Globally, remote sensing (RS) has been supporting decision-making in the field of agriculture for many years.
Several previously published literature reviews partially examine the application of the RS in the crops' water status and nutrients determined in agriculture.M. Weiss et al. (2020) conducted a meta-review of agronomical variables and plant traits that can be estimated from remote sensing.They described different methodological approaches to retrieve them, discussed how these variables are employed by different stakeholders for specific applications and concluded with an overview of caveats and future challenges.A. Konings et al. (2019) presented a review of microwave remote sensing observations sensitive to plant water content.They introduced the principles behind microwave remote sensing observations to illustrate how they are sensitive to plant water content and discussed how various sensor types can be leveraged for specific applications depending on the spatiotemporal resolution needed.M. Bacco et al. (2019) provided a survey of the most recent research activities in the area of digitalization of agriculture, in the form of both research projects and scientific literature, to show the already achieved results, the current investigations, and the still open challenges, both technical and non-technical.G. Lassalle (2021) conducted a meta-review about the advances achieved in monitoring natural and anthropogenic plant stressors by hyperspectral remote sensing over the last 50 years.He presented advances in hyperspectral monitoring of plant stress with a total of 466 peer-reviewed articles.
N. The generated models produced estimations for chlorophyll a, b, and total with calibration errors of 0.98, 0.58, and 1.47µg ml -1 cm -2 , respectively.For prediction, the errors are 1.03, 0.67, and 1.49 µg ml -1 cm -2 .In terms of leaf nitrogen content, the calibration error is 1.49 g kg -1 , while the prediction error is 3.

REMOTE SENSING CROPS ON THE GREENHOUSE LEVEL
Q. Li et al. (2022) investigated the relationship between soil water content (SWC), drought stages of wheat, canopy temperature, and spectral response characteristics.They found that red-valley position (RVP) and red-edge position (REP) parameters showed significant shifts during certain drought stages of wheat, while vegetation water indexes effectively distinguished different levels of water stress across various growth stages.A. Bianchi et al. (2017) developed a water balance model based on the FAO-56 method to schedule irrigation for greenhouse spinach crops.They found that nitrogen treatment had minimal impact on crop development and irrigation requirements.C. Ru et al. (2020) examined the effectiveness of using the CWSI based on leaf temperature to assess the water status of grapevines.They determined the optimal time and conditions for observing CWSI values and concluded that CWSI was more effective than Tc-Ta (canopy temperature minus air temperature) in monitoring plant water stress, with stomatal conductance (gs) showing the strongest correlation with CWSI.Table 3 presents the above data in a summarized form:

Water requirements
The FAO-56-GH model demonstrated favourable performance during both the validation and calibration periods, as indicated by the evaluation metrics.In the validation period, the model exhibited a coefficient of determination (R 2 ) of 0.80, an RMSE of 0.41 mm day -1 , and an NSE of 0.78.Similarly, during the calibration period, the model achieved an R 2 of 0.84, an RMSE of 0.21 mm day -1 , and an NSE of 0.83

Early water stress detection
The accuracy of identifying three different water treatments is found to be 80.95%, while the accuracy of detecting water stress is 90.39% Source: developed by the author

REMOTE SENSING ON LEAF LEVEL
In 2015, Dhillon introduced the "Leaf Monitor," a continuous leaf monitoring system specifically designed for measuring the PWS of almond and walnut trees.The Leaf Monitor system utilized a thermal infra-red sensor to measure leaf temperature and integrated additional sensors to monitor various environmental factors such as air temperature, relative humidity, photosynthetically active radiation (PAR), and wind speed.To optimize its performance, the system included a leaf holder, a solar radiation diffuser dome, and a wind barrier.These systems were interconnected in a wireless mesh network, allowing data collection and transmission every 16 minutes through the internet.

Table 3, Continued
In a subsequent study published in 2019 by the same research team led by Dhillon, they developed a methodology to predict PWS in almond and walnut trees using the continuous leaf monitoring system.The researchers collected data on leaf temperature and environmental conditions using the Leaf Monitor system.These data were then utilized to calculate a Modified CWSI, which served as an indicator of the plant's water status.Before an irrigation event, MCWSI values were estimated and used to determine the appropriate irrigation amount for low-frequency variable rate irrigation (VRI).The irrigation amounts were based on the correlation between MCWSI and Drought Stress Water Potential (DSWP).By implementing variable rate irrigation, the researchers achieved a significant average reduction of 39% in water consumption compared to the fixed 100% evapotranspiration (ET) replacement irrigation method for all the trees.
Analysing strawberry leaf colour can effectively help evaluate soil status and protect against excess environmental nutrients and financial losses in strawberry crops (Madhavi et al., 2022).The objective of this study was to create ML models, specifically multiple linear regression (MLR) and gradient boost regression (GBR), that can simulate variations in strawberry leaf colour.These colour changes are based on soil physicochemical properties and plant age, using mean values of the RGB channels.Accurate measurements of soil physicochemical components were obtained from the largest and most diverse coloured leaves of the strawberry plants using a multifunctional soil sensor in the rooting zones.Additionally, 400 strawberry leaflets were sampled at different stages of vegetative and reproductive growth, and individual leaves were captured using a digital imaging system.The RGB mean values of the coloured leaf images were extracted using image segmentation algorithms incorporated into image processing techniques.Subsequently, MLR and GBR models were developed to predict the RGB mean values of the leaves based on the soil physicochemical measurements and plant age.The GBR model demonstrated superior performance compared to the MLR model, achieving high-performance metrics.
Several other studies have explored the use of advanced techniques for assessing PWS and health in various crops.

CUTTING-EDGE TECHNOLOGIES
Industries are moving towards smart factories, and the agricultural sector should also embrace the concept of smart farms.Unmanned ground vehicles (UGVs) play a vital role in the development of smart farms (Gonzalez-De-Santos et al., 2020).While there are similarities between smart factories and smart farms, specific research is needed to address the unique characteristics of UGVs designed for outdoor agricultural tasks.Multi-drone systems consisting of smaller or medium-sized drones can perform tasks with greater accuracy, tolerance, and safety compared to larger machines.Autonomous drones, whether operating alone or in fleets, are essential for the precise application of herbicides and fertilizers.This approach reduces chemical usage, resulting in benefits such as lower costs, improved operator safety, enhanced community health, and higher food quality with reduced toxicity.
The Research Centre for High Technology Greenhouse Plant Production at Ehime University is working on a project called "Intelligent Greenhouse Systems," in which the main concept is the "Speaking Plant" approach (Hiroshige, 2015).The SPA approach suggests that the best way to grow crops is by considering the plants' physiological status.To implement SPA, a chlorophyll fluorescence imaging drone is developed to measure the photosynthetic activity of tomato plants in greenhouses.This drone is designed to be low-cost, easy to use, and suitable for commercial tomato production.The researchers also studied how storage temperature affects tomato fruit colour using multiple regression analysis to improve postharvest management.
This study (Kalaitzoglou et al., 2021) aims to investigate how blue light affects plant growth by measuring light absorption and photosynthesis in tomato plants.The experiment involved growing tomato plants in six different combinations of artificial solar and blue LED light, with varying levels of blue light.The results showed that increasing the blue light fraction led to a decrease in plant growth, which is related to a more compact plant morphology and lower light absorption.However, the blue light did not affect the plants' maximum photosynthetic capacity.The study suggests that increasing blue light in a solar light environment can hinder plant growth, but further research is needed to confirm this in high-light growth environments typically used in tomato production.
A digital twin refers to a virtual representation of a farm that offers benefits such as increased productivity, improved efficiency, and reduced energy consumption and losses.A. Nasirahmadi and O. Hensel (2022) presented a comprehensive overview of digital twin concepts and technologies in the field of agriculture.It covers various aspects including soil management, irrigation, robotics, farm machinery, and post-harvest processing.The review explores data recording, modelling, artificial intelligence, big data, simulation, analysis, prediction, and communication aspects of digital twin systems in agriculture.By continuously monitoring the farm in real time and updating the virtual model to reflect changes in the physical environment, digital twins provide valuable support to farmers.This technology represents the cutting-edge advancement in the digital transformation of the agricultural sector.
C. Pylianidis et al. (2021) used a combination of qualitative and quantitative methods to explore the advantages of digital twins in the agricultural domain.
The study begins with a comprehensive literature review focusing on digital twins in agriculture between 2017 and 2020.By examining 28 use cases and comparing them with examples from other fields, the researchers assess the extent of digital twin implementation in Table 5, Continued agriculture.They analysed reported benefits, service categories, and technology readiness levels to evaluate the adoption of digital twins in the agricultural sector.Furthermore, the study identifies specific characteristics of digital twins that can be beneficial in agriculture and proposes a roadmap for their implementation, considering increasing levels of complexity.Ultimately, the research underscores the distinctive attributes of digital twins that make them particularly valuable in the context of agriculture.
"Agri-Food 4.0" is a term used to describe the latest developments in the agriculture sector, which includes the integration of digital technologies and the Internet of Things.Similar to the concept of "Industry 4.0", Agri-Food 4.0 aims to optimize efficiency and productivity through the use of these technologies.While the agriculture sector has already incorporated electronic controls and data collection methods through sensors and drones, there is still a need to improve supply chain performance and decision-making processes.This survey (Lezoche et al., 2020) reviews over a hundred papers on new technologies and supply chain methods to understand the future of Agri-Food 4.0.
According to the United Nations Development Programme (2021), global agriculture and food systems are under constant enormous pressure.It is estimated that by 2050 the world population will grow to almost 10 billion people and to feed it, current food production must be increased by 59-98% (Domingues et al., 2022).It is also estimated that agriculture uses over 70% of global freshwater supplies, with around half of this being lost and wasted.Globally, about one-third of the food produced is wasted or thrown away.Agri-food is at the heart of the United Nations 2030 Agenda and impacts all 17 Sustainable Development Goals.
In 2019, the twenty-six EU member states signed a declaration of cooperation to support the successful digitization of agriculture and rural areas in Europe.In it, they recognize the potential of digital technologies to help tackle important and pressing economic, social, climate and environmental challenges facing the EU's agri-food sector and rural areas.In 2022, FAO calls for an end to soil degradation.Up to 95% of the world's food production depends on soil.As a result of unsustainable farming practices, overexploitation of natural resources and a growing population, one-third of soils have already been degraded, and experts estimate that soil erosion could lead to 10% crop loss by 2050.Soils are also full of life and contain approximately 25% of the world's biodiversity (FAO, 2023).
The introduction of so-called precision agriculture (PA) and digital technologies in the field of agriculture are ways to optimize and improve the processes in it.New digital technologies make farming more accessible, using a combination of different technologies working together to improve the precision of farming, including high-speed internet (5G) mobile phones, high-resolution RS technologies through images from satellites and unmanned aerial vehicles (UAVs), sensors connected to the Internet of Things (IoT), wireless sensor networks (WSN), robotics and 3D printing ( Kalaitzoglou et al., 2021).
The use of PA is primarily driven by WSN, which consists of interconnected wireless nodes that monitor environmental conditions (Shafi et al., 2019).These nodes are equipped with components, such as transmitters, microcontrollers, sensors, antennas, and electronic circuits to collect and transmit data to a gateway.Nodes are categorized as either data sources or receivers/gateways, with receivers having more computing power.In agriculture, different plants have varying needs even within the same farmland, and sensors are employed to monitor the changing behaviour of plant parameters, including soil properties, plant characteristics, weather conditions, fertilization options, and water requirements.The most commonly used wireless communication protocols in agricultural IoT applications are Cellular, 6LoWPAN, ZigBee, RFID, Wi-Fi, and LoRaWAN (Avşar & Mowla, 2022).
The combination of ML, HPC, and big data technologies opens up new opportunities for data-intensive science in the interdisciplinary field of agrotechnologies.ML algorithms applied to sensor data transform farm management systems into real-time AI-driven programs that offer valuable recommendations to support farmers' decision-making (Liakos et al., 2018).ML approaches were successfully applied in various areas, such as classifying plant diseases from plant images using CNNs for different plant species and diseases, detecting insects on leaves through object segmentation and deep learning techniques (Domingues et al., 2022), and predicting soil moisture levels (Atanasov et al., 2023).In a global context, RS has been supporting decision-making in agriculture for over 45 years ( Ammoniaci et al., 2021;Domingues et al., 2022).

CONCLUSIONS
Remote sensing technologies offer valuable insights for improving agricultural productivity and efficient water resource management.Soil moisture sensors provide accurate and consistent data, enabling objective observations of plant water status, although they have limitations in capturing field-wide moisture variations.Handheld remote sensing devices offer better temporal, spectral, and spatial resolution for on-site monitoring, but their effectiveness is limited for evaluating larger areas compared to sensors on planes and satellites.Proximal sensors require location-specific calibration and may pose data analysis challenges.Leaf physiology methods provide objective assessments of crop water stress but are time-consuming, destructive, and not suitable for real-time monitoring.Modern imaging techniques such as fluorescence, thermography, and multispectral imaging enable fast and non-destructive phenotyping for early detection of water stress.Satellite-based remote sensing has significant potential but can be constrained by spatial and temporal resolution limitations and cloud cover.To overcome these challenges, unmanned aerial vehicles (UAVs) equipped with remote sensing sensors offer a cost-effective alternative, providing high-resolution imagery with flexibility in flight and mission timing.UAVs surpass satellite resolution and offer quick and repeatable deployment.Combining multispectral and hyperspectral images enhances the comprehensive understanding of crop conditions.By integrating these technologies, agriculture can benefit from improved practices and sustainable resource management.
Prospects for future research in state-of-the-art technologies for remote sensing of crop water status and nutrients in agriculture include advancing sensor technologies for improved accuracy and resolution, integrating data from multiple sources for comprehensive analysis, and exploring the integration of remote sensing with precision agriculture techniques for site-specific management.

Сучасні технології дистанційного
F. Hahn et al. (2021) focused on mango trees and employed dendrometers and capacitors with Teflon clamps to measure LWC and examine the influence of water on mango production.H. Skoneczny et al. (2020) investigated the potential of non-invasive proximal hyperspectral remote sensing (RS) to distinguish between healthy, infected, and dry leaves of apple trees using spectral bands and indices.F. Rojo et al. (2016) proposed a new method for calculating CWSI and Modified CWSI using continuous leaf monitor data in almond orchards and vineyards to implement site-specific irrigation management.J. Rodriguez-Perez et al. (2018) compared ordinary least squares regression (OLSR) and functional linear regression (FLR) models to predict LWC using reflectance data and wavelengths in grapevines.T. Zhao et al. (2020) used hyperspectral imaging to evaluate water status in tomato leaves and computed vegetation indices (VIs) to assess LWC in different parts of the plants.These studies highlight the potential of advanced techniques in monitoring PWS and health for improved crop management.

Table 1 .
Katsoulas et al. (2016) conducted a review of crop reflectance monitoring as a tool for water stress detection in greenhouses.They presented challenges of detecting water status in greenhouses by remote sensing, discussed sensors available for reflectance sensing and applications, effects of environmental, canopy structure and other parameters and proposed possible solutions to mitigate the effects of those parameters.U. Ahmad et al. (2021) conducted a review on crop Examples of RS crops on large scale level

Table 2 .
Examples of RS crops on field level 39 g kg -1 LNA) with R 2 values ranging from 0.79 to 0.81 and RMSE values of 1.43 to 1.45 g m −2 .Similarly, for Plant Nitrogen Accumulation (PNA), the red edge VIs demonstrated strong performance with R 2 values ranging from 0.81 to 0.84 and RMSE values of 2.27 to 2.38 g m −2 There is a strong correlation between RGB-based VIs and LAI.The (RF model outperformed other models across the entire observation period as well as during specific growth stages.It exhibited the highest coefficients of determination (R 2 ) ranging from 0.71 to 0.88 and the lowest RMSE values ranging from 0.12 to 0.25 on the test datasets.Following the RF model, the BPNN (Backpropagation Neural Network) model and LR models also showed good performance (de Oliveira et al., 2022) A high ratio between the foliage level and the NDVI can be observed (r=0.97,p-value < 0.01) Source: developed by the author

Table 3 .
Examples of RS crops on the greenhouse level

Table 2 ,
ContinuedThe correlation between the leaf-air temperature difference (T c -T a ) and indicators of PWS (φ s , g s , E) is statistically significant (P<0.05).Among these indicators, the relationship between g s , E, and T c -T a is particularly strong, with R 2 values ranging from 0.530 to 0.604 and from 0.545 to 0.623, respectively.Additionally, there is a reliable linear correlation between the CWSI value and soil moisture at depths of 0-40 cm, with statistical significance (P<0.05)

Table 4 .
Examples of RS crops on the plant level

Table 5
presents the above data in a summarized form.

Table 5 .
Examples of RS crops on leaf level