Glyphosate Separating and Sensing for Precision Agriculture and Environmental Protection in the Era of Smart Materials

The present article critically and comprehensively reviews the most recent reports on smart sensors for determining glyphosate (GLP), an active agent of GLP-based herbicides (GBHs) traditionally used in agriculture over the past decades. Commercialized in 1974, GBHs have now reached 350 million hectares of crops in over 140 countries with an annual turnover of 11 billion USD worldwide. However, rolling exploitation of GLP and GBHs in the last decades has led to environmental pollution, animal intoxication, bacterial resistance, and sustained occupational exposure of the herbicide of farm and companies’ workers. Intoxication with these herbicides dysregulates the microbiome-gut-brain axis, cholinergic neurotransmission, and endocrine system, causing paralytic ileus, hyperkalemia, oliguria, pulmonary edema, and cardiogenic shock. Precision agriculture, i.e., an (information technology)-enhanced approach to crop management, including a site-specific determination of agrochemicals, derives from the benefits of smart materials (SMs), data science, and nanosensors. Those typically feature fluorescent molecularly imprinted polymers or immunochemical aptamer artificial receptors integrated with electrochemical transducers. Fabricated as portable or wearable lab-on-chips, smartphones, and soft robotics and connected with SM-based devices that provide machine learning algorithms and online databases, they integrate, process, analyze, and interpret massive amounts of spatiotemporal data in a user-friendly and decision-making manner. Exploited for the ultrasensitive determination of toxins, including GLP, they will become practical tools in farmlands and point-of-care testing. Expectedly, smart sensors can be used for personalized diagnostics, real-time water, food, soil, and air quality monitoring, site-specific herbicide management, and crop control.


Ecological Contamination with Glyphosate.
According to the United Nations Food and Agriculture Organization (FAO), the world's population will attain ∼9.7 billion by 2050, corresponding to a 32% projected growth. 1 A recent meta-analysis of projected global food demand revealed that the total global food demand should increase by 35% to 56% between 2010 and 2050. 2 The food shortage threat belongs to global agriculture's most harmful socio-economic and environmental challenges. 3 Aside from the COVID-19 pandemic, 4 increases in temperature and atmospheric CO 2 concentration, the environmental pollution of agrochemicals arising from ill-considered farming and insufficient fertilizer delivery systems has become a civilization problem. 5 2020-Forecasted global agrochemical annual use was 120 million tons for nitrogen-based fertilizers, 50 million tons for phosphate-based fertilizers, and over 2.6 million tons for pesticides. 6,7 In recent decades, the nutrient-use efficiency (NUE) has dropped significantly, i.e., over 50% of the N, 85% of the P are not assimilated by crops, 8,9 and less than 10% of the applied pesticides reach their targets. 5 If not handled, these usages were estimated to increase by 50−90% by 2050. 5,10 The first global initiative to solve agriculture's challenges began with the Third Agricultural (Green) Revolution in the 1950s and 1960s. 11 This technology transfer involved (i) highthroughput cultivation of high-yielding varieties of cereal seeds, (ii) improvement of the NUE by spatiotemporal biofertilization and microbial biodiversity, and (iii) reduction of the reactive nitrogen species use and nitrogen oxide emission. In the 1970s, it was followed by the Gene Revolution, 12 based on the extensive use of genetically engineered (GE) herbicideresistant crops, especially glyphosate (GLP)-resistant plants, which resulted in yield increases, tillage reduction, and enhanced the high technology-based weed management. 13 Yet, despite several advantages of lowering greenhouse gas emission, 14 the large-scale misuse of GLP-based herbicides (GBHs) and GLP-resistant crops and GLP-resistant weedoriginated single mode-of-action herbicides has caused environmental pollution with GLP, herbicide resistance, superweeds, and pests generation, as well as consuming GE organisms and GBH-contaminated products, which have consequently re-empowered the expensive tillage. 13,15 These outcomes have boosted the international debate on the policy controlling or forbidding GBH exploitation and, on the other hand, developing sensors for GLP contaminants that emerged in the environment over the last 50 years. 16 Emerging malnourishment, inappropriate weed management, and critically imbalanced and anthropogenically altered P-cycle have been recognized by the U.S. National Science Foundation as some of the most crucial challenges of modern and future agriculture and ecology, which require advances both in proper fertilizing and devising portable and sensitive tools of chemical analysis in agriculture and ecology. 3,17,18 1.2. Smart Materials in Precision Agriculture. The latest advancements in agriculture and ecology originated with precision agriculture (PA). According to the International Society of Precision Agriculture (ISPAg) (https://www.ispag. org), 19 PA is "a management strategy that gathers, processes, and analyzes temporal, spatial, and individual data and combines them with other information to support management decisions according to estimated variability for improved resource use efficiency, productivity, quality, profitability and sustainability of agricultural production." PA's most crucial challenge is reducing herbicide resistance in weed management. In 2021, the herbicide market value accounted for ∼30 billion USD and was forecasted to reach ∼40 billion USD by 2027 (https://www.imarcgroup.com/herbicides-market). Expectedly, herbicide-resistant weed management (HRWM) shall comprise almost half of the modern agricultural activities, including tillage, crop and herbicide diversity, and growing GE herbicide-resistant plants. 20 The increase in the NUE and crop productions will be achieved by applying artificial intelligence (AI)-excelled devices and smart materials (SMs), 21 including engineered nanomaterials (ENMs) and biomaterials, 22 and plant wearable (PW) sensors, actuators, and soft robotics. 23 These PA HRWM nanobiotechnological tools enable direct, spatiotemporally targeted, and dose-dependent delivery as well as rapid, ultrasensitive, and selective determination of herbicides.
SMs are well-defined, self-sensing, self-healing, and stimuliresponsive materials that change their properties and act according to their surroundings or external stimuli, including microorganisms, chemical compounds, heat, pH, temperature, electromagnetic field, light, humidity, ultrasounds, pressure, and mechanical factors. 21 The SMs are mostly based on electroactive, piezoelectric, shape memory, and biocompatible polymers. They are integrated with electronic devices implanted or installed in the site of interest. 21,23 Moreover, these devices, e.g., lab-on-chips (LOCs), microrobots, or smartphone-assisted nanosensors, exploit incorporated computational algorithms that allow for acquiring, processing, and analyzing vast amounts of data in real-time, thus providing a model simulation of spatial and seasonal distribution as well as risk assessment of agrochemicals. 3,17,18,24 Regarding PA-dedicated SM sensors, there is a need for advanced inexpensive, highly efficient, multifunctional, and flexible consumer-and operator-friendly tools that can determine analytes, including toxins, pests, herbicides, and microbes in nonlaboratory settings, hardly accessible locations, and diverse agroecosystems. 25,26 Conventional multimodal chemosensors utilize immunochemical receptors, targeted to toxic fertilizers and pesticide residuals, and electrochemical and/or optical transducers. However, because of the environmental systems' complexity, vulnerabilities, and uncertainties, these conventional sensors will hopefully be upgraded or replaced with smart sensors equipped with computational devices. 27 In AI PA smart sensors, these conventional sensors are assembled or virtually connected with computational devices that convey the machine learning (ML), deep learning (DL), artificial neural networks, nanoinformatics, and translational bioinformatics algorithms to integrate, compute, process, analyze, and interpret the massive amounts of spatiotemporal data. 17,3,18 Hence, PA SM sensors, based on smartphones, soft robotics, robotic-automated vehicles, and drones, provide rapid, mobile, and high-throughput analysis. Moreover, they afford high-quality, decision-making outcomes for herbicide misuse control and sustained and profitable agricultural production and weed management. 3,17,18,28,29 For instance, DL-enhanced computing methods enable the analysis of highresolution spectral images of herbicide-sprayed plants mapped by scanning transmission X-ray microscopy or X-ray fluorescence spectroscopy. 30,31 Data science-excelled and PA-targeted SM sensors can generally analyze short-or long-term weather conditions, soil properties, plant diseases, pests, microbial communities, and industrial pollution. 3,17,18 Smart sensors and sensor networks must sense and sustain optimal conditions for plant cultivation, including moisture, temperature, pH, nutrients, and agrochemicals. 3 Environment-friendly AI PA targets the dynamic and complex nature of the local agroecosystem by synergistic application of theoretical prediction models and experimental stimuli-responsive delivery-detection tools. If attached or administrated to the soil, plant, or crops, the AI PA sensor can capture, digitize, and process images, as well as detect, monitor, respond, and regulate physicochemical stimuli and atmospheric conditions in an information-supported decisionmaking manner. 18 Since pests cause 34% of crop loss globally, controlling and increasing crop yields is essential. 32 By predicting the ecosystem components' behavior, PA provides a real-time response to weather conditions, nutrient cycling, crop growth, plant phenotyping, disease diagnosis, weed infestation, insect damage, and food production, as well as emerging contamination with these agents. These data are then correlated to the site-targeted delivery, uptake, detection, retention, performance, interaction, and transformation of nanomaterial-based agrochemicals in plants.
1.3. Artificial Intelligence-Excelled SM Sensors. AI PA technologies have excelled in using self-powered wireless sensor networks (WSNs), weed patches, and PW sensors for remote and in situ monitoring. 3,23 They involve nanosensors and nanomaterial-based delivery systems, PWs, and nanorobots that leverage online databases and algorithms of cheminformatics and translational bioinformatics. 3,18 As a result, they enable spatiotemporal management of crops and livestock, low-cost in situ nutrient-sensing, precise agrochemical/fertilizer placement, and soil and plant tissue testing. By constantly measuring vegetative indices and evapotranspiration rates, AI PA helps control various and diverse plant growth conditions, nitrogen uptake, and secretion of nitrogen into the rhizosphere.
Before applying these sensors, one must overcome potential environmental pollution and mechanical breakdown risks. The plant-implantable or PW sensors should be constructed from biocompatible, mechanically resistant, nanotechnology-excelled materials, thus allowing well-defined interaction with microbiomes and metabolites exudated by plant leaves and roots. 33−35 Although these sensors can be biotransformed or biodegraded in the phyllosphere, they are vulnerable to mechanical destruction, because of sudden weather changes. It is crucial to understand the (sensor material)-plant interactions, off-target performance, and long-term toxicity and persistence of the nanomaterial in the plant vascular structure and organelles, especially since the nanomaterials may penetrate the phloem and xylem, thus changing the fluid composition and flow rate. 18 Finally, because of large-scale demands, energy storage costs must be well-considered. The efficient use of smart sensors requires efficient handling of enormous amounts of digitized output images and biophysical and physiological comprehension to solve real problems of scheduling sowing and controlling pesticide usage, as well as tracking and predicting crop growth and quality. 36,37 Most advanced AI PA technologies propose intelligent selfpowered WSNs and PW. 3,23 Characterized by low-power demand, long-duration processing, and near-field communication (NFC), WSNs are coarsely distributed and interconnected sensors, typically exploited for environmental in situ and realtime monitoring. The WSNs, equipped with fully rechargeable batteries or battery-free solar energy harvesting systems for power supply, are easily controlled and power-supplied by smartphones, provided the distance from the source is less than several meters. However, they have limitations, including reduced light in the shaded, bushy, and foliaged places or prolonged signal transmission resulting from far distances from the aboveground microstrip antenna to the underground sensor. 38,39 A similar technology underlies the modus operandi of PWs, i.e., the electronic devices implanted into the plant tissue for tracking and transmitting physiological parameters. Being an example of nanobionics, the PWs enable precise and continuous in-site or remote sensing of microclimate and tissue microenvironment changes, thus informing about the plant health and agrochemical performance. 40 An outstanding recent review presents the state of the art of the nano biotechnologically excelled PW sensors. 23 The present review focuses on developing analytical methods using conventional and SM-based sensors for GLP determination in water, food, crop, and soil samples. Expectedly, conventional sensors will soon be replaced by SM-based sensors. Examples of these sensors, including smartphoneintegrated sensors, are briefly introduced and discussed herein.

ENVIRONMENTAL AND HUMAN TOXICITY OF GLYPHOSATE
2.1. Glyphosate Structure and Properties. Glyphosate, N-(phosphonomethyl) glycine (GLP), is a widely used broadspectrum, nonselective, and postemergent herbicide for crop desiccation. 41 Since the approval for agricultural use by U.S. Environmental Protection Agency (EPA) in 1974 and the authorization in the EU in 2002, GLP-based surfactants (GLP-SH) and GBHs have been commercialized as Roundup and RangerPro in over 140 countries, covering 350 million hectares of crops, with an annual turnover of 11 billion USD worldwide. 42,43 According to the FAO, GBHs' global market constitutes 18% of the total pesticide active ingredients (ActIs) and 92% of herbicide ActIs, with a global annual revenue accounting for 11 billion USD. 43 As a powerful tool of modern crop management, prior-harvesting large-scale GBH interventions, called "green burndowns," are expected to meet the growing demands for food and crops production, which are projected to increase to 100−110% by 2050. 41,44 2.2. The Weed-Killing Activity of Glyphosate. GLP reversibly inhibits the 5-enolpyruvynyl-shikimate-3-phosphate synthase (EPSPS, EC 2.5.1.19) involved in the biosynthesis of aromatic amino acids in plants 45 and some microorganisms. 46−49 Roundup Ready crops are GLP-tolerant because they are genetically modified to carry the CP4 EPSPS gene derived from Agrobacterium sp. strain CP4, a naturally GLPresistant rhizosphere bacteria. 46 Moreover, the harmful efficacy of GLP against plant development strongly depends on the daily and circadian rhythms of the plant cells, 42 and plants die within 4−20 days after crop spraying. 50 This chronotherapeutic responsiveness of plants to GLP brings hope for optimized crop protection and safe food production security.
GBHs are commonly used in agriculture, industry, forestry, and weed management. From 1974 to 2014, the GBHs' use increased 100-fold. However, the current regulations for the safety standards of GBH handling still rely on the studies performed in the late 1980s. Although the ∼90% growth of all GE seeds produced by Monsanto in 1996 is considered safe, 51 and despite extensive research conducted in human biomonitoring, hazard assessments, epidemiological studies including occupationally exposed workers, pregnant women, and their offspring, and evaluation and standardization, the GBH safety to humans and the environment is still questioned. 52 The major issues concern the environmental pollution that affects crops, soil, surface and groundwaters, sediment, and the spreading through wind and erosion, thus threatening wildlife and human occupational activities on farmlands and GBH factories. 53 43,57 Average urinary GLP levels for occupational exposure and nonoccupational exposure range from 0.26 to 73.5 μg/L and 0.13−7.6 μg/L, respectively. 58 As an organophosphate (OP), GLP efficiently transmits orally, dermally, conjunctively, gastrointestinally, and via respiratory routes. 56 According to the EFSA, GLP has low acute toxicity owing to the absence of the EPSP-metabolic pathway in vertebrates and the rapid degradation of GLP in mammals (half-life time of ∼5−10 h). 57,58 However, occupational poisonings have become a global medical issue because of environmental pollution. 59 Regarding genetic modification of GLP-tolerant crops, it is crucial to delineate the potential genotoxicity of the transgenic plants 60−63 from the chemical toxicity of commercial GBH coformulants, including GLP isopropylamine (GLP-IPA) salt, polyoxyethyleneamine (POEA), and ppb traces of heavy For example, a study on aquatic microorganisms (bacteria, microalgae, protozoa, and crustaceans) revealed the highest toxicity for POEA and GBH, whereas the only effect of GLP resulted from its acidity. 70 Analogically, commercial GLP-SHs occurred more toxic than GLP-IPA in short-term acute toxicity trials (24 and 48 h). 71 Moreover, in rats, 12-week exposure to GBH caused significant increases in kidney biomarkers, oxidative stress markers, and membrane-bound enzymes, indicating the accumulation of GLP residues in the kidneys, while GLP alone caused no nephrotoxicity. 72 Concerning human toxicity, ingesting substantial GLP-SH volumes (100−500 mL) is associated with a human death rate up to 29.3%, depending on patients' characteristics, such as age and intent of exposure. 73,74 The uncoupling oxidative phosphorylation and POEA-or (heavy metals)-induced cardiovascular and cardiopulmonary harms are major lethal causes. 74−76 According to a survey of medical reports of 107 patients, the ingestion of GLP-SH caused hypotension (47%), deterioration (38.6%), respiratory failure (30%), acute kidney injury (17.1%), and arrhythmia (10%). Interestingly, these complications depend on the volume of GLP-SH ingested and not the type of surfactant ingredient of the GBH. 77 Emergency treatment of those intoxications comprises gastric lavage followed by hemodiafiltration and direct hemoperfusion, enabling the removal of GLP-IPA (228 Da) and surfactants (over 500 Da), respectively. 78 Intensive care is required in severe GLP-SH intoxication, including dehydration, oliguria, paralytic ileus, hypovolemic shock, cardiogenic shock, pulmonary edema, hyperkalemia, and metabolic acidosis. 79,78

SENSORS AND SEPARATION SYSTEMS FOR GLP DETERMINATION
Nowadays, SM sensors are robustly applied to solve the most challenging environmental, socio-economic, and biomedical issues. Those involve rapid response to climate changes, realtime water, food, soil, and air quality monitoring, biotic and abiotic stress factors, nutrient recycling, sustained crop growth, biofuel production, point-of-care (POC) devices fabrication, and defense against bioterrorism. 22,80 Receptor items of the modern SM sensors are emerging-field-deployable chemo-and biosensors and synthetic biology tools capable of detecting pathogenic pollution in various ecosystems and industrial environments. They involve molecularly imprinted polymers (MIPs), aptamers, viruses, prokaryotic or eukaryotic cells, and individual plants or insects that are able to sense single molecules or cells of emerging contaminants at pico-and femtomolar concentrations. Single-molecule/cell-based nanosensors allow determining agrochemical toxins as well as soil, plant, and insect-associated microbial communities. As such, they have become a powerful PA tool to track the impact of herbicides, pesticides, and insecticides on crop-associated microbiomes and ecosystems. 81,82 Restricted concentrations of GLP in drinking water and food are 0.7 mg/L and 0.1−310 ppm, respectively. 83 Various nanomaterial-based sensors for GLP have recently been prepared for detecting, adsorbing, and degrading GLP in real samples (Table 1). Conducting, semiconducting, and nonconducting polymers were prepared. 84 For example, polyaniline-zeolite (PANI/ZSM-5 and PANI-FeZSM-5) compositebased adsorbents of efficient GLP adsorption capacity were prepared by oxidative polymerization. 85,86 Polydopamine (PDA) was used to synthesize BiVO 4 /PDA-g-C 3 N 4 photocatalyst sheets for exploiting dopamine self-polymerization, ultrasonic dispersion, and self-assembling. Under visible light irradiation, the BiVO 4 /PDAg-C 3 N 4 photocatalysts degraded GLP more actively than the control composites prepared without PD. 87 Fluorescent porous N-benzyl (carbazole derivative)-based polymer GLP detectors were synthesized in a one-step polymerization. The polymers of tunable pore sizes emitted bright cyan, blue, and green light upon ultraviolet (UV)-light excitation. GLP and other pesticides quenched the fluorescence of polymers according to the Stern−Volmer kinetics, thus demonstrating pesticide-specific recognition and determination. 88 An intriguing report on a proteinoid polymer composite-based sensor for GLP was presented. 89 The proteinoid polymer nanoparticles (NPs) were prepared by thermal step-growth polymerization of natural and unnatural amino acids in the presence of various agrochemicals. The agrochemicals interacted with the hollow NPs by encapsulation, integrating with the crude shell, or bound covalently/ physically to the NP surface. Once hydrophobized and exploit modern GLP-sensitive materials to detect and analyze, as well as to extract and degrade, GLP traces. Hybrid nanomaterials used in these approaches include enzymes (AChE, ESPS), deoxyribonucleic acid (DNA)-or antibody-based aptamers, immune-magnetic conjugates, MIPs, and inorganic materials. Moreover, we herein discuss recent reports concerning the GLP determination by AI-excelled smartphone-integrated sensors, 154−158 and an automated vegetable analyzer for targeted GLP delivery. 159 3.1. Chromatography and Adsorptive-Extracting Systems for GLP. Chromatographic techniques are traditional tools for determining pesticides in liquids, foods, and clinical and environmental samples. They enable the pretreatment, separation, detection, or degradation of pesticides among other contaminants (micropollutants) of emerging concern, including endocrine-disrupting chemicals, plasticizers, artificial sweeteners, pharmaceuticals, personal care products, pyrethroid insecticides, and halogenated or organophosphorus retardants. 160−162 For example, an intelligent postacquisition sample validation following mixed-mode solid-phase extraction (SPE) and ultraperformance liquid chromatography quadrupole-time-offlight mass spectrometry (UPLC-Q-ToF-HRMS/MS) was employed for wide-scope target screening of 2316 emerging pollutants in wastewater samples collected from the Wastewater Treatment Plant of Athens. Upon validation of the method, it was employed to detect and quantify the influent and effluent wastewater connect of 398 selected contaminants of pesticides, opiates, and opioids, stimulants and sympathotomimetics, cannabinoids barbiturates, benzodiazepins, tranquilizers, analgesics, antibiotics, steroids, and industrial chemicals. This method allowed for determining the contents as low as 0.3 ng/L (perfluoroundecanoic acid), 0.4 ng/L (acetochlor, N-2,4-dimethylphenylformamide), and 0.5 ng/L (haloperidol, perfluoroheptanesulfonic acid). 163 Moreover, HRMS-based suspect screening integrated with national monitoring data was recently applied in aquatic toxicology by investigating the presence of 16 not-well-explored pesticides and 242 pesticide transformation products in Swedish agricultural areas and streams. The study confirmed the occurrence of 11 transformation products and 12 tentatively identified ones. 164 Furthermore, gas chromatography coupled to electron ionization mass spectrometry (GC-EIMS) was employed to determine levels of OP ethers in air and soil samples. 165 Chromatographic analysis of hazardous agrochemicals usually involves two steps. First, the analyte is pretreated with optically active or polystyrene-coated magnetic NPs and then separated using GC or high-performance liquid chromatography (HPLC) coupled with a UV spectroscopic, fluorescent (FLD), or mass spectrometric (MS) detector. Most advanced adsorptive/separating chromatographic systems include nanotechnologically excellent adsorptive systems, engaging MIPs or silica NPs (SiNPs) of various porosity, developed surface area, and sorbing properties. Moreover, they are often modified with ionic liquids, silanes, amines, enzymes (AChE, carboxylesterases, laccases, or OP hydrolases), fluorescent, electrochemiluminescent, or surface-enhanced Raman spectroscopy (SERS) labels and magnetic beads. Finally, these systems are prepared as beads, wires, and sheets to improve their sensitivity, stability, amenability to modifications, and "onsite" applicability. 166,167  Novel chromatographic techniques applied to GLP determination involve GC or HPLC coupled with UV spectroscopic, 106 FLD, or electrospray ionization mass spectrometric (ESI-MS) detectors, 104   3,6-dimethoxy-9-phenyl-9H-carbazole-1-sulfonyl chloride (DPPC-Cl) to determine GLP in soybean with the LOD of 0.02 ng/mL and the extraction recovery exceeding 95%. 111 The LC coupled to isotope-ratio MS (LC−IRMS) enabled the GLP determination in 21 commercial herbicide samples, revealing δ 13 C values between −24 ‰ and −34 ‰ in the submicrogram concentration range. 107 With porous (graphitized carbon absorbent)-based chromatography coupled to a three-quadrupole MS detector, the 6 ng/mL LOD of GLP in aqueous solutions was attained. 105 In a similar study, with an HPLC coupled with an inductively coupled plasma (ICP-MS) detector or a diode array detector (DAD), GLP was determined with the LOD of 8.2 and 300 μg/L, respectively. 108 Chromatographic analysis was employed for the GLP determination in real-life water samples from 10 agricultural provinces of China during various meteorological conditions. 112  Inevitable progress will be made to address well-known limitations of GLP-targeted separating systems. Modern separation techniques, including capillary electrophoresis, GC, or LC, provide relatively low sensitivity (nanomolar) compared with optical, electrochemical, or immunochemical techniques that offer the detection of subnanomolar concentrations. Additionally, the blocky construction of modern systems disallows them for in-field use. Expectedly, the selectivity of future systems will be improved using tools of nanoinformatics, providing models and structures of analytereceptor complexes of higher affinities. Finally, coupling these systems, based on aptamers, ionic liquids, NPs, and MIPs, with MS or FLD detectors, followed by their integration into the mobile microfluidic devices, shall enhance both sensitivity and applicability. 168 3.1.1. MIP-Based Chromatography of GLP. Various chromatographic methods developed to remove or degrade GLP traces from environmental, biological, or grocery samples exploited the adsorptive-extracting properties of porous MIPs. 169 The maximum adsorption capacity of GLP-selective MIPs, formulated via free radical polymerization (FRP) of acrylamide (AA) and ethylene glycol dimethacrylate (EGDMA), was evaluated as high as 3.37 mg/g 90 ( Figure  1G and 1H). A series of dual-templated methacrylic acid MAA-MIPs, imprinted with herbicides, including GLP, were fabricated by precipitation polymerization for water treatment. Efficiency, expressed by binding factors (BFs), K MIP /K NIP , where K is the partition coefficient, of chosen GLP-selective MAA-MIPs in tap water for GLP solution, ranged between 2.12 and 2.55. 92 Moreover, based on 1-allyl-2-thiourea (ATU), GLP-detecting MIP cartridges were prepared to assess the (UPLC-MS-MS)-mediated GLP recovery from mineral and underground waters. The herbicides were totally retained from these real matrices, spiked with 0.5 μg/L GLP. 97 The selective sorptive extraction of GLP from river water and soil samples, with mean recoveries ranging from 90.6 to 97.3%, was demonstrated for ATU-and 2-dimethyl aminoethyl methacrylate (DMAEM)-based MIPs, prepared by UV lightactivated FRP 93 (Figure 1A-1F). Eventually, in a recent study, positively charged (quaternary ammonium cation)-MIP quartz crystal microbalance (QCM) sensors were constructed for the (electrostatic interaction)-mediated binding of GLP from river waters. 101 The next-generation chromatographic sensors shall unite traditional multimodal separation and detection techniques, including thin-layer chromatography (TLC) or immunochromatography, with microfluidics and AI tools. Recent studies indicate the ultrasensitive on-site determination of toxins and pesticides in real samples using smartphone-coopted. Personal low-cost AI tools offer high-resolution photoimaging, portability, and instant online data availability that can be immediately shared. For example, an open-source smartphone-imaging app was developed to excel TLC screening and quantify pharmaceuticals. 170 In another study, smartphonebased dual-channel immunochromatographic test strips labeled with polymer carbon QDs were fabricated for on-site simultaneous biomonitoring of cypermethrin, a pyrethroid pesticide, and its metabolite, 3-phenoxybenzoic acid, determined with LODs of 0.35 and 0.04 ng/mL, respectively. 171 Expectedly, similar tools will soon be commercially available for GLP, as has already been demonstrated for other pesticides.

Optical Sensors for GLP.
Recently, a series of optical strategies have been developed for pesticide sensing. Those include photonic, photoluminescent, photoelectric, electrochemiluminescent, and colorimetric methods. Over the past five years, carbon dot (CD)-based optical sensors equipped with aptamers, antibodies, enzymes, gold and silver nanoclusters (AuNCs and AgNCs), and nanoparticles (AuNPs and AgNPs), and MIPs as recognition units have been used as herbicide-derived signal indicators, catalysts, coreactants, and electrode surface modifiers. 169,172 Moreover, novel carbonbased SERS biosensors with similarly excellent sensing properties were fabricated. They primarily include 0D carbon quantum dots (QDs), 1D carbon nanotubes (CNTs), 2D graphene, graphene oxide (GO), 3D carbon nanomaterials, and core−shell nanostructures. The SERS sensors were devised for selective and quantitative in situ analysis of agrochemicals by exploiting so-called localized "hotspots" produced during the application. 173 QD-based chemosensors were used for the highly sensitive and selective detection of pesticide poisons in the clinical and forensic toxicological analysis of gaseous, anionic, phenolic, metallic, drug, and pesticide specimens. A breakthrough has been made by continuously applying whispering gallery modes (WGMs) to biosensing. 174−176 Miniaturized size and excellent lasing properties of WGM-based microlasers, called resonators, were exploited in constructing label-free aptasensors 177 and applied to detect single molecules, particles, cells, and molecular electrostatic changes at biointerfaces and barcodetype tagging and tracking. 178−180 Furthermore, remarkable advances have been made in developing electrochemiluminescent (ECL) and photoelectrochemical sensors to analyze food quality. In most recent works, nanomaterial-based ECL luminophores have been synthesized and incorporated into immunoassay-, aptasensor-, and microfluidic systems for lowcost ultrasensitive determination of heavy metals, illegal additives, microbes, and pesticide contaminants in complex matrices. 181 Food safety issues can be effectively solved by using brand-new photoelectrochemical biosensors. These Environmental Science & Technology pubs.acs.org/est Critical Review devices, equipped with photoactive nanomaterial-based recognition units, quantitatively determine mycotoxins, antibiotics, and pesticides with high sensitivity at a significantly low signalto-noise ratio. 182 Modern optical methods of GLP sensing mainly exploit SERS, interferometry, chemiluminescence, colorimetry, fluorescence, and Forster resonance energy transfer (FRET). A SERS-based method, which exploits organometallic osmium carbonyl cluster-conjugated AuNPs, was used for AChEmediated GLP determination with the LOD below 0.1 ppb 118 ( Figure 2D−2H).
Regarding SERS sensors for GLP, an air-stable sensor was devised by using reduced GO (rGO)-wrapped dual-layer AgNPs on TiO 2 NT arrays as a SERS substrate. This sensor's adsorption capacity and SERS enhancement were excellent thanks to the tremendous electromagnetic field and chemical enhancement generated by localized SPR excitation of the dense dual-layer AgNPs uniformly deposited onto the TiO 2 NTs and to facilitation of the charge transfer between the extensive π−π conjugations in the rGO. Because the GLP molecule does not have a specific chemical group, it is hardly detectable using a conventional SERS method. In the discussed study, GLP was determined by the enhanced adsorption area of the nanocomposite. That enabled GLP determination in environmental samples of waters and soils in the LDCR of 0.005−50 mg/L with the LOD of 3 μg/L, i.e., lower than the limit specified by the U.S. EPA and the European Union. 120 Two UV−vis colorimetric sensors for GLP were devised by linking 3-chloro-4-methylpyridine with 4-(dimethylamino) benzaldehyde or 4-(dimethylamino) cinnamaldehyde in a one-step synthesis, resulting in 4-(2-(3-chloropyridin-4-yl) vinyl)-N,N-dimethylaniline (BP-Cl) or 4-(3-chloropyridin-4yl) buta-1,3-dien-1-yl)-N,N-dimethylaniline (CP-Cl), respectively. In the GLP reaction with these sensing compounds, the N atom of GLP interacted with the Cl atom on the pyridine ring, resulting in highly sensitive and selective naked-eye detection of GLP, observed as color changes ranging from colorless to yellow (BP-Cl) and from yellow to orange (CP-Cl). In a naked-eye analysis, GLP was determined in tap water and potato samples with the LOD of 15 and 10 μM for BP-Cl and CP-Cl, respectively, whereas using UV−vis spectrophotometry, these LODs were of 0.847 and 1.23 μM, respectively, in the LDCR of 1−40 μM. 121 GLP was determined by using interferometry and chemiluminescence. Picomolar traces of GLP were detected using an EPSP-decorated interferometric sensor with GLPattached poly(ethylene glycol) (PEG)-based soft colloidal probes 117 (Figure 2A,B). In a chemiluminescent-assisted method of GLP determination, the LOD of 46 ng/mL was reached using poly(vinyl chloride) (PVC)-MIP microbeads prepared by FRP of acrylamide and subsequent conjugation with PVC. 103 Fluorescent sensors rely on the analyte-induced triggering or quenching of the receptor's fluorescence, called the "ON/OFF strategy." In this strategy, pesticides, including GLP, act as either direct or indirect quenchers or triggers of fluorescence. In the direct approach, GLP binding by the receptor modifies its electronic structure and fluorescent properties, thus enabling direct optical sensing and determination of the GLP content. In contrast, in the indirect approach, GLP interacts with the molecular trigger/quencher to subsequently turn on/ off the receptor. Recent studies on GLP-based fluorescence modulation demonstrate examples of both approaches.
GLP-induced fluorescence was directly quenched using carbazole-based porous polyaminals 88 ( Figure 2C). Moreover, various hybrid fluorescent-magnetic immunosensors were fabricated to detect GLP in liquids. For example, water-in-oil microemulsion-based Co−V/SiO 2 NPs, doped with rhodamine, enabled immunoassaying GLP-dsDNA double target/ probe core−shell NPs with the LOD of 0.35 nM. 116 Likewise, a magnetic-assisted oligonucleotide aptamer probe labeled with 6-carboxy-fluorescein was used for GLP sensing with the LOD of 88.8 ng/L. 115 Finally, recently devised FRET-based sensors for GLP explored GLP-induced turn-on fluorescence following the aggregation of positively charged cysteamine-AuNPs and negatively charged CdTe QDs capped with thioglycolic acid. This FRET assay utilized GLP detection in apples with a 9.8 ng/kg LOD. 114 In another study, the fluorescent carbon QD probe operating in the AND logic gate was successfully quenched by GLP, resulting in the GLP determination with the LOD of 0.6 μM. 119 Finally, the GLP-induced FRET switching in a self-assembled nanosensing system, formulated from p-tert-butylcalix [4] arene-grafted ruthenium(II) bipyridine-doped SiNPs, was used for GLP determination with the LOD of 0.791 μM. 113 The indirect-modulation fluorescence sensor mainly relies on the chelating properties of GLP, provided by its phosphonate and carboxyl groups, as well as the monoprotonated secondary amine nitrogen atom. A Cu 2+ -modulated DNA-templated AgNCs was used to determine GLP by GLP reaction with the Cu 2+ , a primary quencher. Upon chelation, the DNA AgNCs fluorescence was recovered, which enabled the stoichiometric determination of GLP in real samples in the LDCR of 15−100 μg/L and with the LOD of 5 μg/L. 122 A similar approach was exploited in fluorescent and colorimetric sensing of Cu 2+ and GLP by the (o-phenylenediamine)-SiNPs FRET interaction. The Cu 2+ oxidation of o-phenyaldiamine disabled the fluorescence of SiNPs by FRET. Hence, by chelating Cu 2+ , GLP served as a secondary quencher by hindering the FRET donor's oxidation and restoring the FRET acceptor's emission. This approach allowed determining GLP in the LDCR of 0.15 to 1.5 μg/L with the LOD of 0.003 μg/ mL. 123 Consistently, the concept of the Cu 2+ -GLP system was used in a fluorescent 4-butyl-3-thiosemicarbazide-labeled CDsbased sensor. Cu 2+ -quenched fluorescence of the sensor was recovered by the GLP addition in a dose-dependent manner, which enabled GLP determination in real samples with the LOD of 0.27 μM. 124 Likewise, a Cu 2+ -modulated 4dihydroxyanthraquinone-CD nanosensor enabled for ultrasensitive sensing of GLP in vegetable samples with the LDCR of 50 to 1300 ng/mL and the LOD of 0.8 ng/mL. 125 Moreover, GLP was rapidly determined in real samples of agrifood products (tea, soybean, wheat, cucumber) using (rhodamine B)-embedded amino-functionalized iron-based MOFs bonded with Cu 2+ via Lewis interactions that resulted in fluorescence quenching. The addition of GLP resulted in Cu 2+ chelation via hydrogen bonding, thus turning on the fluorescence of the nanosensor. That allowed for the GLP determination with the LOD of 0.18 μM in the 0.6−45 μM LDCR. 126 Similarly, a hierarchical, highly porous fluorescent nanocomposite, UiO-67/Ce-PC, consisting of UiO-67 NPs grown on Ce-MOF-derived porous carbon, was devised to determine GLP in soybean, wheat, and corn. A large specific surface area and abundant metal active sites that alleviated the diffusion barrier and enhanced the GLP preenrichment ensured the determination. Additionally, the competitive coordination effect between GLP's phosphonate groups and metalloorganic ligands (hydroxyl groups) attenuated the ligand-to-metal charge transfer between metallic nodes and organic struts, thus providing a dose-dependent fluorescence recovery upon GLP detection. The plot of the fluorescence enhancement response of UiO-67/Ce-PC toward GLP was linear in the range of 0.02−30 μg/mL with the LOD of 0.0062 μg/mL. 127 An optical/temperature nanozyme platform, fabricated from nitrogen-doped CDs anchored onto Zr-based ferrocene MOF nanosheets, was devised for the sensitive and portable determination of GLP. The nanozyme mimicked peroxidase activity, i.e., oxidation of colorless 3,3′,5,5′-tetramethylbenzidine into a blue product in the presence of H 2 O 2 , which GLP readily suppressed. That enabled a trisignal response of fluorescence enhancement, absorbance, and temperature decrease. These features were exploited to construct a portable mini-photothermal device capable of colorimetric and fluorescent GLP determination in the 0.039−3.19 μg/mL LDCR with the LOD of 0.0131 μg/mL and the 0.0088−3.98 μg/mL LDCR with the LOD of 0.0015 μg/mL, respectively. 128 Because of the analyte-triggered AChE inhibition, the optical sensors for GLP, based on the single molecule "turn on/off" on-site sensing mode, belong to the most sensitive and selective. Future optical technologies, compatible with smartphone-assisted kits and LOCs, include paper-, liquid-,

Environmental Science & Technology pubs.acs.org/est
Critical Review and gel-based sensors. Double-signal fluorescence, phosphorescence, chemiluminescence, lateral flow immunoassay, or enzymatic fiber-optic biosensing are envisioned as the most promising operation strategies, enabling the in-field red-greenblue (RGB)-determination of GLP. Most recent advancements in optical sensors involve metasurface-based devices. Metasurfaces are 2D composite micro-or nanomaterials of the subwavelength thickness and desired geometry, allowing for adjusting the material's refractive index of the material to positive, near-zero, or negative values (so-called negative refraction). Moreover, the nonlinear metasurface can transform the infrared signal into the visible signal, which the human eye or smartphone CCD can visualize. 183 These exquisite features allow for devising a surface-enhanced infrared absorptionoperating sensor for passive trapping and detecting glucose and proline with the LOD of ∼1 pg. 184 In this context, the LODs of the devices discussed in the present review are comparable or one order of magnitude lower. Besides, devising optical sensors for agricultural use shall necessitate integrating ultrasensitive (of the order of ∼fM) devices with smartphones or portable meters, similar to commercial smart Raman spectroscopic alcohol meters.

Electrochemical Sensors for GLP.
The contemporary electrochemical sensors provide a portable, simple, and sensitive determination of herbicides and pesticides in real water, food, and soil samples. 185,186 Recently fabricated multimodal sensing devices contain aptamers, enzymes, graphene, CNTs, polymers, viruses, and cells as recognition units integrated with optical, piezoelectric, and microgravimetric transducers. 187,188 The GLP concentration is mainly measured using amperometry, cyclic voltammetry (CV), differential pulse voltammetry (DPV), nonfaradic electrochemical impedance spectroscopy (EIS), and electrochemical surface plasmon resonance (SPR) spectroscopy.
Among other OP pesticides, including glufosinate and AMPA, GLP was determined amperometrically using a threeelectrode sensor fabricated by (UV laser)-inscribing of OPselective Cu NPs on a polyimide film. For GLP determination in natural water samples, the sensor exhibited the LOD of 3.42 μM, whereas for glufosinate and AMPA, the LOD was 7.28 and 17.78 μM, respectively. Moreover, the sensor prevailed in pesticide selectivity in the presence of ion and organic interferences in natural water. 138 Recently, a very sensitive, highly conducting sensor for GLP was constructed based on Cu-benzene-1,3,5-tricarboxylate-loaded 2D Ti 3 C 2 T x nanosheets deposited on a glassy carbon electrode (GCE). This nanocomposite was fabricated upon in situ copper component growth on chemically etching nanosheets that were subsequently dispersed and vacuum-dried on the GCE. Because of the Cu ions' high affinity to GLP, the sensor sensitivity was excellent, resulting in the exquisitely low LOD of 26 fM and a broad LDCR of 100 fM to 1 μM. 139 Carbon-based electrochemical sensors have recently become a powerful tool for tracing metabolism. The significant advantage of these materials is the enhancement of the electrochemical sensing performance by enlarging an active surface area. For example, a probe-free screen-printed carbon electrode (SPCE) was used for direct GLP sensing in tap water with micromolar sensitivity. 131 An ECL horseradish peroxidase (HRP)-based sensor, formulated on a sulfonate polymer matrix, was used for the GLP determination with the LOD of 1.7 μg/L. 132 In a recent study, a simple and accurate (pencil graphite electrode)-supported sensor containing an HRP enzyme, immobilized on a multiwalled CNTs (MWCNTs)doped polysulfone membrane, was devised to detect GLP in the river and drinking water samples. This highly selective sensor was readily applied to the in-field determination of GLP, showing reproducible and repeatable CV and amperometric readouts in the LDCR of 0.1−10 mg/L and the LOD of 0.025 mg/L. 140 Moreover, functionalized single-walled CNT (SWCNT)-based nanomaterials were used in electrochemical GLP sensing by providing transducing layers for water-gated transistor-based sensors. In particular, GLP was selectively determined with a sensor composed of networks of semiconducting, monochiral (6,5) SWCNTs featured with polyfluorene-bipyridine copolymer and a Cu 2+ -selective membrane. The functionality of this semiconducting sensor relied on the n-doping resulting from Cu 2+ complexation by bipyridine. Adding GLP suppressed this complexation by competitive chelation of Cu 2+ , thus enabling the stoichiometric quantification of the herbicide analyte at nanomolar concentrations. 137 Likewise, GLP was selectively determined using an amperometric sensor containing glycine oxidase, a flavoenzyme, immobilized on a platinum-decorated laser-induced graphene scaffold. Exquisite electronic and solid properties, including the LDCR of 10−260 μM and the LOD of 3.03 μM, allowed for selective determination of GLP with only minimal interference of common herbicides and insecticides, including atrazine, 2,4-dichlorophenoxyacetic acid, dicamba, parathionmethyl, paraoxon-methyl, malathion, chlorpyrifos, thiamethoxam, clothianidin, and imidacloprid. In the future, the sensor may enable food mapping and determining GLP in complex river waters and crop residue fluids. 136 GLP was sensed by using electrochemical immunoassays. Recently, a label-free, portable, selective, and highly sensitive (LOD of 0.1 ng/mL in the LDCR of 0.1−72 ng/mL) sensor was devised to determine GLP in human urine. 141 Its electrochemical platform consisted of a portable printed-circuit circular board with gold working and reference electrodes enabling nonfaradic EIS measurements. Its immunoassaybased platform included a monolayer of dithiobis(succinimidyl propionate), a thiol-based cross-linking monomer modified with a GLP antibody, and a coated gold electrode. The selectivity was assessed using typical herbicide interferences, including malathion, 3-phenoxybenzoic acid, and chlorpyrifos. 141 Likewise, GLP and chlorpyrifos were determined in low-and high-fat food matrices using a two-plex, portable electrochemical-immunoassay nonfaradic ESI-based sensor. Both sides of this sensor were functionalized with the respective antibody. In low fat, the sensor determined GLP and chlorpyrifos with a 1-ng/mL LOD in the 0.3−243 ng/mL GLP/chlorpyrifos LDCR, whereas in high fat, the LOD was 1 ng/mL in the LDCR of 1−243 ng/mL. 142 GLP was sensed with the LOD of 0.11 nM using sophisticated ECL sensors based on HRP-assisted in situ generations of ZnS QDs on ordered mesoporous carbon substrates. 133 A bioconjugate of urease-AuNPs and an agaroseguar gum-entrapped biocomposite membrane was devised to detect the enzyme-inhibiting activity of GLP, enabling the determination of this herbicide in ppm traces. 134 Similarly, GLP was determined with a sensitivity of 0.01 ppm (10 ng/ mL) in fruits and vegetables using a field-deployable electrochemical immunosensor based on a polymer-metalized interdigitated two-electrode system, functionalized with a commercial GLP-antibody, and equipped with a portable

Environmental Science & Technology pubs.acs.org/est
Critical Review reader and a machine-learning binary classifier 130 (Figure 3). Likewise, an SPCE immunosensor, enabling the GLP determination with the 5 ng/L LOD, was fabricated using anti-GLP-IgG-modified magnetic beads and an HRP-conjugated-GLP tracer. 129 Moreover, an electrodeposited monocrystalline silicon-PANI-HRP conjugate was constructed to evaluate an immunoassay for GLP sensing with the LOD of 5.44 μg/L. 135    GLP, in water samples in the LDCR of 0.025−500 μM and with the LOD of 3.1 nM and 4 nM, respectively. 144 A typical base-catalyzed sol−gel transition incorporation was performed to incorporate GLP and graphene QD labels into mesoporous organosilica MIPs. The GLP-induced quenchingmediated detection enabled GLP quantification at a subnanomolar level (0.017 ppb, 17 pg/mL). 94 GLP was determined with the 0.35 ng/L LOD using a pencil graphite electrode dip-coated with AuNPs, then modified with the fabrication of glyphosate-glufosinate double-template MIP using atom transfer radical polymerization in the presence of templates and MWCNTs. 96 Finally, GLP was ultrasensitively (the LOD of 92 ng/mL) determined by DPV using a composite of urchin-like AuNPs, Prussian Blue, and PPy-based GLP-MIPs, deposited by electropolymerization on the indium−tin oxide electrode. 100 Electrochemical sensors for GLP are advantageous because of their excellent reproducibility and accuracy as well as high selectivity and sensitivity with a linear output among others. However, these features may deteriorate over time, as extended exposure to the target analyte usually shortens or limits the sensor lifetime, especially at variable temperatures. Once incorporated into the smart sensor, this device would require temperature compensation, which exploits the battery energy

(Atomic Force Microscopy)-Based Sensors for GLP.
Despite extreme usefulness, robustness, and sensitivity, atomic force microscopy (AFM) has hardly been used as a sensing tool. The main reasons included dimensions and the complicated curvature of the AFM tips. Recently, tip functionalization with various species has been optimized so that AFM has become an efficient technique for quantitatively determining chemicals, including herbicides and pesticides. Recently, atomic force spectroscopy, an AFM-derived technique, has been exploited to analyze imazaquin, metsulfuron-methyl, and atrazine samples using tips functionalized with the acetolactate synthase and antiatrazine antibody. This tip functionalization increased markedly (over 130, 140, and 175%, respectively) the adhesion force between the functionalized tips and the herbicides, distinguishing nonspecific and specific interactions between the tip-located biomolecules and the herbicides quantitatively. 190 Regarding AFM-assisted GLP determination, a peroxidasebased AFM nanobiosensor was devised to evaluate GLP content in 0.01 to 10 mg/mL in zucchini extracts. This biosensor for GLP modus operandi relied on detecting GLPinduced changes in surface tension caused by GLP adsorption, followed by a conformational change in the peroxidase structure. The LOD was as low as 0.0238 mg/L. 145 Low sensitivity, long scanning time, limited spatial resolution, and tip damage, i.e., significant drawbacks of AFM sensors, are expected to be addressed in the future. The construction of tuning-fork-balanced tips shall enable chiplike probing of biological and soft material samples. Applying tipsynchronized time-resolved electrostatic force microscopy will also allow for monitoring charge generation, transfer, and recombination, which is crucial for ultrasensitive enzymatic GLP determination in real samples.

Immunoassays and Immunosensors for GLP.
Immunosensors are (affinity-ligand)-based sensing tools exploiting immunochemical antibody−antigen interactions. These interactions are quantified by transducers with immobilized antibodies. Over decades, the use of immunosensors, both labeled and nonlabeled, has become highly trending, as they comprise (single molecule)-operating sensing tools for food safety control, healthcare, and environmental monitoring. 191 Usually, forming an immunocomplex with an analyte generates electrochemical, photochemical, and piezoelectric changes, contributing to multimodal and sensitive detection. 192 Herbicides and pesticides belong to typical immunosensor analytes, as they are easily complexed by peptides and enzymes immobilized on the transducer surface. The affinity-based reactions resemble physiological and biochemical reactions because these analytes act as natural ligands. 193 Besides, using MIPs, well-known as "plastic antibodies," "semisynthetic enzymes," and "artificial receptors," 194 enables low-cost pesticide residue determination in foods, feeds, medicines, and environmental samples. 195 Various immunosensors have been constructed for GLP sensing. These include antibody-and aptamer-based immuno-assisted electrochemical and optical sensors. 54 Most recent examples report conventional immunochemical methods for GLP determination using enzyme-linked immunosorbent assay (ELISA), which provides submicromolar LODs in liquids and animal feeds. 147,150 The ELISA combination with an online SPE, followed by HPLC−MS analysis, enabled the GLP determination with the subnanomolar LOD. 148 Advanced immunoassays utilize an immobilized GLP-ovalbumin conjugate and avian IgY antibodies for sensing ppb traces of GLP 149 or an AuNP oligonucleotide-based biobarcode immuno-(polymerase chain reaction) (PCR) system with the LOD of 4.5 pg/g 146 (Figure 4).
Although contemporary immunosensors for GLP provide ultrasensitive determination, future trends promise the construction of sensors of higher standards. Because of the high production cost of contemporary biologically derived antibodies and enzymes, future immunosensors will be constructed from single atom/molecule or peptoid/enzyme mimetic catalysts containing single noble metal NCs or MOF nanozymes as recognition units. Since immunodetection of various environmentally emerging pesticide residuals can be executed with extremely low LODs (<pM), 196 it is expected that the application of AI-excelled tools shall upgrade these values. As it was extensively summarized elsewhere, 197 several smartphone-assisted immunosensors were so far devised for determining various biomedically or environmentally relevant analytes, including small molecules, macromolecules, viruses, and bacteria.
3.6. Microfluidic Lab-on-Chips for GLP. In recent decades, enormous progress has been made in devising microfluidic-assisted conventional and smart sensors. Technological progress enables microfabricating and miniaturizing microfluidic paper-based analytical devices (μPADs) 198 and devising smartphone-coopted sensors operating in a microfluidic mode. 199 In a traditional design, μPADs are constructed from conventional dipstick or lateral-flow setups. Once coupled with electrochemical immunosensors and LOCs, these eco-friendly devices have become onsite quantitative and semiqualitative equipment for POC medical diagnostics, food safety control, and environmental purity inspection. Particularly, microfluidic-assisted analytical devices enable direct, low-cost, sensitive, and real-time screening of chemical hazards or pathogens, including metal ions, nitrates and nitrites, phenols, pesticides and herbicides, and bacteria. 200−203 Thanks to an impressive advance in biotechnological sciences, sophisticated living-cell-based microfluidic-handled biosensors have also been devised. They are mainly inspired by conventional in vitro bioassays, including the bacterial luminescence toxicity screen and the algal toxicity test by imaging pulse amplitude modulated fluorometry, which was demonstrated useful for pesticide determination in environmental samples. 204 In the most advanced approach, algaebased biosensors sensitively, sustainably, and multiplexed analyzed agro-environmental samples. In these biosensors, whole algal cells and their photosynthetic complexes were used as miniaturized transducers to construct biomicrofluidic devices. 205

Environmental Science & Technology pubs.acs.org/est
Critical Review Modern microfluidic-handled LOC sensors combine different methods, e.g., electrophoresis-coupled disposable microchips with laser-induced fluorescence detection for the rapid on-site interference-free determination of GLP residues in agricultural products, with the LOD of 0.34 μg/L and 84% recovery. 152,153 (Direct injection)-UPLC with triple quadrupole MS was optimized to determine GLP traces in environmental waters with the LOD range of 0.05−0.09 μg/ L and 76.3% recovery. 151 Moreover, Mn-ZnS QD-embedded MIP, combined with a 3D-μPAD sensor, was fabricated for  Figure 5D-H). The next steps in devising future GLP-selective microfluidicassisted sensors involve increasing sensitivity and integrating these sensors with AI tools including smartphones and mobile meters. These sensors will enable on-site ultrasensitive determination of GLP in foods, industrial waters, and biological fluids. Based on an ELISA assessment, such a tool has already been devised for on-site quantifying ppb (μg/kg) levels of aflatoxin B1 in moldy corn. The immunoassay sensor was 3D-printed on a plastic chip attachable to a smartphone. 206 Similar sensitivity was achieved for various pesticides using a portable, smartphone-adaptable origami μPAD-based potentiostat. Relying on the pesticide-inhibited activity of the enzyme immobilized on the sensor's transducing item allowed for chronoamperometric monitoring of paraoxon, 2,4-dichlorophenoxyacetic acid, and atrazine at a ppb level in river water samples. 207 3.7. Smart Sensors and Plant Wearables for GLP. The necessity of developing the PW pest-, insect-, fungi-, and herbicide-specific sensing and delivery systems is crucial because of the off-target toxicity of these biocides. Among heavy metals, antibiotics and other drugs, food-derived growth factors, and industrial hydrocarbon wastes, these biocides are the most toxic environmental pollutants and hazards to human health. 3 Because of various leakage from the target zone to the rhizosphere, waters, and air, relatively high half-time biocides destabilize the trophic chain's matter cycle and endanger ecosystem safety. 208−210 Thus, enormous progress has been made in AI-enhanced mobile or PW biocide-selective sensors. 3,18 Miniaturized smart sensors, incorporated, e.g., in smartphones, PWs, and POCs or field-deployable devices, are equipped with nanophotonic antennas and dielectric metasurfaces that enable few-molecule sensitivity by confining incident light into intense hotspots of the electromagnetic fields, thus delivering strongly enhanced light-matter interactions. 211 Regarding GLP sensing, an in-smartphone-incorporated mobile lab-in-a-syringe platform was designed for the rapid, visual, quantitative determination of organophosphorus pesticides via dual-mode colorimetry and fluorescence measurements. The platform was based on (cetyltrimethylammonium bromide)-coated NPs conjugated with a silica pad modified with red-and green-emission QDs. In the sensing reaction, thiocholine, the product of AChE-mediated hydrolysis of thioacetylcholine, induced the aggregation of NPs, thus giving rise to the color change. During the on-site GLP determination, the pesticide-caused enzymatic AChE inhib- respectively. Moreover, the sensor was integrated with a portable (test strips)-smartphone sensing platform dedicated to POC testing GLP in food samples. 158 The extremely high technological advancement of the presented pioneer examples of selective and sensitive SMbased sensors sets the direction of action in analytical chemistry. Expectedly, the multimodal smartphone-based sensors will equip the agronomic and POC laboratories, providing automated, computerized, and high-throughput toxin determination, despite the relatively high production and maintenance costs. In the near future, smartphoneaugmented sensors will be commonly used as components of nanobionic PW sensors. For example, PWs based on thin nanofilm electronics, such as SWCNTs field-effect transistors, can now detect dimethyl methylphosphonate, a volatile nerve agent, at the ppm level. 212 Moreover, the flexible, stretchable, and adhesive PW sensor, fabricated by direct writing of chitosan-based inks, was active toward plant mechanical injuryresponsive healing and sensed the intoxication with methyl parathion and nitrites. 213 3.8. Robotic Devices for GLP Delivery. The GLPresistant weeds' evolution in the farmlands forced scientists and farmers to develop novel GLP-free technologies for modern weed management. 13,15 Among the development of transgenic crops displaying resistance to auxinic herbicides and herbicides displaying inhibitory activity against acetolactate synthase, acetyl-CoA carboxylase, hydroxyphenylpyruvate dioxygenase or bioherbicides, and sprayable herbicidal ribonucleic acid interference (RNAi) agents, the nonbiochemical solutions have been considered as well. 13,214 Aside from SM AI-enhanced sensors, these innovations involve soft terrestrial robotics that will give rise to robotic weeding in the nearest future. 13 In 2022, the global agricultural robots market, including unmanned aerial vehicles (UAV), drones, automated harvesting system drones, and driverless tractors, was estimated to be worth 5.9 million USD and is anticipated to reach a forecast value of 30.5 million USD by 2032 (https://www.futuremarketinsights.com/reports/agriculturerobotics-market, access on 21.05.2023).
Regarding GLP-oriented systems, a robot-based herbicide delivery system is a more advanced system for in-row weed control purposes. The robot was tailored to facilitate systematic and site-specific delivery of GLP and iodosulfuron droplets to four different weed species in a drop-on-demand manner without affecting carrot crops. The droplets were selectively sprayed on the leaves of the detected weeds within the plant row. Iodosulfuron, a sulfonylurea-based herbicidal inhibitor of acetolactate synthase, was used as an additive herbicide to control the species that GLP insufficiently controls. In indoor pot trials, amounts of 7.6 μg of GLP and 0.15 μg of iodosulfuron per droplet per plant were delivered, whereas in a field trial with the robot system involved, this amount was 5.3 μg of GLP per droplet. Moreover, the GLP amount was 10 times lower than GLP amount used in conventional crop spraying. 159 In these in-field trials, the presented three-wheeled robot GLP delivery system displayed promising properties, including relatively low cost, maintainability, maneuverability, stability, and robustness, when compared to other commercial agricultural robots. 215,216 (Figure 7). However, it would be required to perform quantitative environmental impact assessments using a comparative life cycle assessment of intra-and inter-row weeding management to claim its superior usefulness. 217 Furthermore, worth mentioning the costs of devising and exploiting robotic GLP sensors/delivery systems, including costs of navigation hardware, machine vision technologies, and power consumption, 216 are usually higher than those of conventional broadcast GLP analytical techniques, which must be considered when designing experiments or applying to industry. Not disregarding these cost limitations, in the long term, such robotic-based delivery systems may become a promising alternative for dosing a well-defined and strictly controlled amount of pesticides to the crops, thus reducing the risk of environmental contamination.
Robotics has become fundamental for constructing remote sensors (RSs) for GLP. They have been devised and implemented to monitor the agricultural and environmental quality of soil, crops, and water status. In particular, these tools have been applied to solve major issues of long-term herbicideresistant weed management (HRWM). 36,37 In the first step of modern HRWM, remote sensing is commonly implemented to detect and identify infested crops, followed by site-specific herbicide treatment of the weeds, mostly in their germination stage. 218 Besides, RSs can be used for the man-free determination of herbicide-sprayed weeds and crops to eradicate the former or improve the latter's growth using variable-rate technology (VRT). 219−224

FUTURE PROSPECTIVE
Environment-protective farming (also called organic farming) is a goal of sustainable agriculture. 225 Agriculture sustainability prevents soil degradation based on economic productivity, future crop yield maintenance, limited usage of agrochemicals, and eco-friendly integrated pest management and weed control. It ensures the health of the soil, air, water, livestock, animal, and humans. 226 Concerning the latter, AI-excelled PA is expected to provide tools that enable rapid, mobile, and man-free delivery or detect harmful agrochemicals or biomaterials such as fertilizers and biocides. The AI technology aims to handle weed infestations because of their devastating role in crop cultivation by competing for water, nutrients, light, agricultural space, and air gases and by secreting potentially toxic exudates. 227,228 Because weeds share the same zone and spread among crops, it is incredibly challenging to exterminate them without harming the crops. Besides, spectral images of herbicide-sprayed weeds and crops in early developmental stages are similar, which impedes proper visualization, recognition, and site-specific eradication. 229−231 Nevertheless, AI-enhanced site-specific weed management, including weed patch mapping, will be possible. 20,232 It shall allow for the application of nanoformulated herbicides in geospatially determined, stimuli-responsive, and dose-dependent manners, thus inhibiting weeds' seed germination and decreasing weed biomass. 218,233 Furthermore, tackling weed infestations can be addressed by UAV-handling of farmlands stricken with drought, air pollution, or heat caused by local climate changes that have been more frequent in the last ten years compared to previous decades. 234 The limitation of the GBHs and agrochemical use was the founding myth of environment-protective farming, ecology, and weed science. 225 Recent years' technological advancement has displaced a previously in-force "many little hammers" approach based on the extensive labor and equipmentconsuming field works. 235 Nowadays, AI PA has been dominated by "digital farming," providing the tool for intelligent and site-specific monitoring and improving crop protection. For instance, the controlled use of GBHs in recent years was associated with no-tillage agricultural systems, thus reducing their inevitable deficiencies, including soil degradation and high expenses. Thus, the current ban on GLP and GBH shall force farmers to return to conventional tillage systems as an effective weed management strategy for keeping crop-weed balance. So far, a series of no-tillage and no-GBH strategies have been developed to reduce the over-reliance on a few single modes of action herbicides, competitiveness in weeds, and dependency on herbicides, while enhancing the strategies of crop diversification, harvest weed seed control, and precision agriculture approaches, including RNAi technology, gene editing, robotics, and in-site or remote sensing. 235 Expectedly, future devising of safe-by-design sensors for herbicides will mainly exploit AI tools of nanoinformatics, 236−239 ML, 240−243 and DL. 18,244−247 Nanoinformaticsenhanced drug/sensor devising uses predictive risk assessment frameworks, including integrated approaches to testing and assessment for the in silico prediction, design, and optimization of the parameters of nanomaterials used as sensing components or delivery systems according to their agricultural destination. 18,248 These parameters refer to structure and functionality as well as biocompatibility, stability, biodistribution, and transgenerational phytotoxicity of the material, thus providing prediction and translation of the long-term responses of the agroecosystem to current conditions, including efficient high or low doses of the agrochemical efficiently. This strategy is particularly promising because current experimental studies on nanomaterial toxicity require short-term exposures to relatively high doses of the agrochemical. 18 Particularly, nanoinformatics models combine chemo-and bioinformatics tools with omics databases and in vitro, ex vivo, in vivo, and clinical data of the active agents and target or nontarget species, allowing for the de novo design of ActI or nanomaterials. 237 The integrated web repositories and database were originally created using the data of eight model The ML and DL methods have already been applied to crop sensing and phenotyping at all scales, including lands, fields, canopies, and leaves. 240−243 Thus, presumably, these powerful computing tools, incorporated in the future smart sensors, will be easily harnessed to predict tripartite (e.g., "agrochemicalsoil-plant," "pathogen-treatment-plant") interactions under variable climate conditions. 18,244−247 The DL-enhanced algorithms will process the object and spectral data regarding the images' numbers, types, sizes, and resolutions by grouping proximal pixels (of a few millimeters or centimeters) with homogeneous spectral value, combining them spectrally, topologically, and contextually. 252 Upon spatiotemporal evaluation of the vegetation indices, including temperature, biomass, moisture, weed infestation, and nitrogen balance index (chlorophyll-to-polyphenol ratio) of these objects, the smart sensors shall identify the nature of the stress and apply proper fertilizers, agrochemicals, and irrigation waters in a sitespecific manner using VRT-based equipment. 253 In effect, the computational tool-excelled smart sensors will acquire and process massive amounts of data on "what," "how," "where," and "when" the human intervention should be applied to address the most pressing needs, thus reducing the consumption of herbicides, such as GLP, to an absolute minimum. 244