Search Results (6,139)

Highly perishable strawberries demand strict post-harvest time management, making accurate short-term yield prediction central to optimizing modern greenhouse production and supply chain scheduling. However, existing models that rely excessively on isolated environmental factors exhibit delayed responsiveness to actual crop physiological dynamics and struggle with integrating multimodal data. To overcome these limitations, we propose a short-term method for predicting greenhouse strawberry yield that integrates visual phenology with meteorological sequences. The proposed method was validated using a multimodal dataset acquired from 150 tracked greenhouse strawberry plants over a 72-day monitoring period (11 December 2025, to 20 February 2026), incorporating continuous microclimate records and an image repository of 784 original images annotated into five distinct phenological classes (flower, green, white, pink, and red). First, using our improved YOLO11-SC model, we effectively resolve challenges of complex illumination and dense foliage occlusion, achieving high-precision automated extraction of five consecutive strawberry phenological stages. Second, by fusing these visual markers with meteorological time series (e.g., temperature, humidity, and light intensity), we construct a multimodal spatiotemporal feature matrix. To accommodate diverse smart agriculture application scenarios, we designed two distinct prediction architectures: on servers with ample computing power, a Bidirectional Temporal Convolutional Network with self-attention (BiTCN-SA) to achieve highly accurate predictions; and for resource-constrained IoT edge nodes, a lightweight machine learning ensemble (Stack-LGR). Experimental results demonstrate that, in predicting the cumulative mature fruit yield within the next harvesting cycle, BiTCN-SA achieves strong performance with a coefficient of determination ($R^2$) of 0.958 and a root mean square error (RMSE) of 3.154. Simultaneously, the edge-deployed Stack-LGR ensemble maintains stable prediction accuracy ($R^2$ = 0.892) while ensuring acceptable inference latency. This study mitigates the latency limitations of single-environment-driven models. It provides a solution for precise crop yield prediction and tiered computational deployment, with good predictive performance, deployment adaptability, and methodological reference value. Full article

A process-based developmental model was developed to predict leaf appearance and expansion in romaine lettuce (Lactuca sativa L. var. longifolia) using hourly air temperature and daily light integral as environmental inputs. The model consisted of two linked modules representing leaf appearance and individual leaf expansion as distinct biological processes. The leaf appearance module calculated the hourly leaf tip appearance rate as the product of a beta-type nonlinear temperature response function and a developmental stage weighting function based on growing degree days accumulated above a base temperature of 4 °C. The leaf expansion module estimated individual leaf expansion using photothermal age, a Gompertz growth function, and a leaf length–area allometric relationship, with potential leaf length determined by the mean growing temperature and leaf rank. The optimum temperatures for leaf appearance rate and potential leaf length differed by approximately 6 °C (26.7 °C vs. 20.4 °C), indicating distinct temperature response patterns between the two developmental processes. Model calibration was performed using datasets (n = 437) collected from a temperature-gradient greenhouse with a nutrient film technique hydroponic system across the spring, summer, and autumn growing seasons, yielding an overall model efficiency (EF) of 0.92 and a root mean square error (RMSE) of 4.26 leaves for leaf appearance, and an EF of 0.80 and an RMSE of 448.1 cm² for leaf expansion. Independent model evaluation was also performed using datasets (n = 132) obtained from a commercial greenhouse with either a deep flow technique hydroponic system or a perlite-based substrate system across the same three growing seasons, yielding an EF of 0.96 and an RMSE of 2.24 leaves for leaf appearance, and an EF of 0.92 and an RMSE of 216.8 cm² for leaf expansion. These results demonstrate that the model effectively described leaf appearance and expansion in romaine lettuce across the tested soilless culture systems under greenhouse conditions, highlighting its potential as a leaf development module for integration into canopy photosynthesis and biomass production models. Full article

Plant secondary metabolites remain indispensable for pharmaceuticals, nutraceuticals, and cosmeceuticals, yet conventional plant culture systems are increasingly limited by inconsistent yields, poor scalability, and inadequate capacity for real-time process monitoring. Microfluidic technologies and organ-on-a-chip (OoC) platforms, originally developed for mammalian biology, are now emerging as powerful tools to overcome these constraints. These systems enable laminar flow, precise gradient generation, single-cell resolution, and biosensor integration, providing unprecedented control over the cellular microenvironment and supporting non-destructive, real-time metabolomic monitoring. While recent reviews have surveyed plant microfluidics broadly covering developmental biology, single-cell phenotyping, and root–microbe interactions, this review provides, to our knowledge, the first synthesis focused specifically on organ-on-a-chip approaches for plant secondary metabolite biosynthesis and real-time metabolomic monitoring. Advances in device fabrication, including PDMS, paper-based, hydrogel, and thermoplastic materials, surface engineering, gradient-based elicitation strategies, and integration of optical, electrochemical, and mass spectrometric detection systems have also been critically examined. Special emphasis is placed on root-on-a-chip, shoot meristem, protoplast, callus, and 3D organoid platforms for studying cell wall mechanics, vacuolar dynamics, cytoskeletal responses, and signalling cascades. However, challenges remain in long-term culture stability and scalability; nonetheless, these technologies offer a roadmap toward programmable ‘plant biosynthetic factories’ to produce high-value natural products. Full article

Improving water productivity while maintaining technological quality is a major challenge for sugar beet production under increasing water scarcity. This study was conducted for only one year at one location during the 2023/2024 winter season near the Village Kafraljaraydh to evaluate the effects of irrigation water type (normal vs. magnetized), irrigation levels (100 and 75% ETc) combined into four irrigation treatments, and NPK fertilization levels (100, 80, 60, and 40% of the recommended dose) on growth, yield, quality, and water relations of sugar beet (Beta vulgaris L.) under field conditions in Egypt (2023/2024 season; strip-plot design; three replicates). Magnetized water treatment (MWT) enhanced vegetative growth, root development, and biomass accumulation, improving yield. The highest root yield (80.34 t ha⁻¹) and sugar yield (16.62 t ha⁻¹) were obtained under magnetized irrigation at 100% ETc; 80–100% NPK, representing a 3–6% increase over normal irrigation. Technological quality improved, with higher sucrose content, total soluble solids, apparent purity, and reduced impurity-related parameters (α-amino N, K, Na, and molasses sugar). Deficit irrigation reduced yield but had limited effects on quality. Water use efficiency was mainly driven by irrigation level, with higher values under 100% ETc, while MWT provided a moderate improvement. Reducing fertilization to 80% NPK maintained yield and improved quality under MWT. MWT improved crop performance and quality without increasing water use, supporting sugar beet production. Full article

Silicon (Si) is a beneficial, non-essential element used in cannabis (Cannabis sativa L.), yet cannabis-specific evidence is scattered across hemp and drug-type cannabis. Because the marketable organ in drug-type and medicinal cannabis is the inflorescence, whose cannabinoid and terpene quality and strict residue limits make every fertigation input unusually consequential, a cannabis-focused synthesis of silicon use is needed. This comprehensive review aims to synthesize that evidence and clarify when, how, and in what form Si supplementation is justified across the crop cycle, based on a four-database search (covering the literature up to May 2026, with the cannabis-specific evidence base spanning 2019–2026). Cannabis is an intermediate Si accumulator depositing silica in bast fibers and trichomes, with benefits clearest in propagation and vegetative growth. Root Si lowers cadmium and zinc uptake and supports antioxidant defense; foliar nano-Si aids drought tolerance, and Si suppresses powdery mildew while raising tissue Si ~2.1-fold and inflorescence biomass ~1.2-fold without reducing cannabinoid or terpene quality. On current evidence, Si appears best deployed as a resilience-enhancing, stage-specific input within an integrated program rather than a universal additive. Future research should prioritize genotype- or chemotype-resolved dose–response studies on cannabinoid and terpene yields and late-flower-application safety. Full article

Apple replant disease (ARD) is a significant problem in regions with intensive fruit production. ARD deteriorates the production properties of soil, reduces yield, and results in lower fruit quality. The effects of ARD can be mitigated with organic amendments applied to soil to improve its productivity. Our experiment was conducted on a farm in western Poland. Biochar was added to replanted soil at 5%, 10%, and 20% by volume and its effect on the physicochemical and biological properties of the soil, the growth vigour, and the quality of strawberries was analysed. Strawberry plants were planted in soil which either had been used for nursery production (replanted soil) or had never been used for this purpose (agricultural soil). The results of our study confirmed earlier reports on the negative consequences of ARD. The replanted soil was more saline and contained less organic matter than the agricultural soil. The plants growing on the replanted soil developed worse. When biochar was added to the replanted soil, depending on its percentage content, the organic matter content and soil respiration rate almost doubled, whereas the count of bacterial taxonomic units increased. As a result of these changes, the root mass of the plants growing in the replanted soil increased by nearly 50%, whereas the root collar mass of shrubs and their leaf area increased by 25 to 30%. The average fruit weight also increased by about 40%. The optimal biochar content in the replanted soil was 10%. The biochar content of 20% did not cause significant changes in most of the parameters under analysis. Full article

Hydroponic and soilless systems are increasingly adopted as low-cost, sustainable solutions for global food production, yet they remain highly susceptible to contamination by potential toxic elements (PTEs), particularly arsenic. While biochar is widely recognized as an effective amendment for mitigating PTE contamination in soil-based systems, its ability to alleviate PTE stress in hydroponic environments has been largely overlooked. The gap reveals a critical and underexplored frontier in controlled-environment agriculture, where extending biochar-based mitigation strategies could yield substantial benefits. Here, we evaluated whether Acacia auriculiformis wood biochar could alleviate arsenic (As) toxicity in lettuce (Lactuca sativa) grown in a continuous-flow hydroponic system. Using a completely randomized factorial design (arsenic species × dose × biochar) with three independent replicates per treatment, we tested biochar under 0.2 and 0.8 mg/L of As(III) and As(V). Arsenic significantly (p < 0.05) reduced lettuce growth, with As(III) being more toxic than As(V). Biochar significantly (p < 0.05) improved morphological traits (2.4–103%), cell membrane stability (5.5–12%), photosynthetic pigments (3–73%), and stress indicators proline (8–11%) and malondialdehyde (8–14%). Arsenic accumulated mainly in roots (1.7–2.63 mg/kg) and shoots (0.76–1.36 mg/kg), but biochar reduced accumulation by 28–47% in roots and 33–48% in shoots. Additionally, biochar enhanced nutrient uptake (K, P, Ca, Mg, B, Zn, Cu, S, Mn) at both arsenic levels. Overall, the results indicate that Acacia biochar can substantially reduce arsenic toxicity and improve plant physiological responses in continuous-flow hydroponics, highlighting its promise as a viable and scalable mitigation tool for safeguarding soilless food production systems against PTE contamination. Full article

Background/Objectives: Drought stress is a primary constraint on maize productivity, yet the role of rhizosphere microbial communities in modulating cultivar-specific drought resilience remains poorly understood. This study aimed to investigate the physiological and microbiome-mediated responses underlying differences in drought tolerance between contrasting cultivars to better understand drought tolerance mechanisms. Methods: Two maize cultivars with contrasting drought tolerance—NK718 (tolerant) and Zhongdan 808 (sensitive)—were subjected to drought stress at the V12 stage. We assessed yield components, oxidative stress indicators (Malondialdehyde (MDA)), and antioxidant enzyme activities (Superoxide Dismutase (SOD), Peroxidase (POD), Catalase (CAT)). Metagenomic sequencing was employed to analyze structural and functional shifts in the rhizosphere microbiota. Results: Drought significantly suppressed yield and physiological performance in both cultivars. However, the sensitive cultivar suffered more pronounced yield losses and severe oxidative stress, indicated by elevated Malondialdehyde (MDA) and decreased antioxidant enzyme activities. Conversely, the tolerant cultivar maintained superior physiological homeostasis. Metagenomic sequencing revealed drought-induced microbial shifts, including decreased Proteobacteria and Ascomycota, alongside increased Actinobacteriota and Mucoromycota. Notably, the drought-tolerant cultivar exhibited enhanced microbial community stability and more complex co-occurrence networks. Furthermore, it enriched specific functional pathways, such as phenylpropanoid biosynthesis, which positively correlated with yield stability and antioxidant capacity. Conclusions: Maize drought tolerance is underpinned by the coordinated regulation of plant physiological adaptation and the structural and functional stabilization of the rhizosphere microbiome. These findings offer a theoretical framework for developing breeding strategies that leverage root-microbe interactions to optimize maize yields under water-limited conditions. Full article

The Upper Minjiang River Basin is a critical ecological barrier in the upper Yangtze River, where forest ecosystems play a vital role in carbon sequestration, water conservation, and soil retention. Given that different forest types exhibit significant variations in community structure, species composition, and ecological processes, their ecosystem service (ES) supplies and trade-off/synergy relationships are also expected to show distinct heterogeneity. However, systematic research on the trade-offs and synergies of ESs across different forest types remains limited, constraining the development of precision forest management and differentiated management strategies. To deal with this, we used the InVEST model and calculated five key services across the basin: carbon stock (CS), water yield (WY), soil conservation (SC), habitat quality (HQ), and forest stock volume (FSV). We then applied Spearman’s correlation, root mean square deviation (RMSD), and the GeoDetector model to analyze trade-offs and uncover driving mechanisms. Finally, we used spatially constrained K-means clustering to map different management zones. The results indicate that the Upper Minjiang River Basin stored 1.78 × 10⁸ t of carbon, retained 2.98 × 10⁸ t of soil, produced 6.48 × 10⁹ m³ of water yield, maintained a mean habitat quality of 0.78, and supported a forest stock volume of 1.20 × 10⁸ m³. Coniferous forests exhibited the highest CS (181.07 t ha⁻¹) and FSV (176.37 m³ ha⁻¹), whereas shrublands contributed the largest share (52.17%) of regional water yield. At the regional scale, CS and FSV showed the strongest synergy (r = 0.71, p < 0.01), while WY displayed significant trade-offs with most other services. GeoDetector analysis revealed that forest type acts as the primary driver shaping the relationships among services, while elevation and precipitation play supporting roles. Based on the ES bundles identified via spatially constrained K-means clustering, the Upper Minjiang River Basin was divided into four distinct management zones: a carbon sequestration core zone, an ecological balance zone, an ecologically fragile zone, and a multifunctional conservation zone. Therefore, findings from the Upper Minjiang River Basin may provide insights applicable to other mountain forest ecosystems facing similar environmental and management challenges. Full article

Background/Objectives: Rice (Oryza sativa) is the staple food for over half the global population, and nitrogen availability is the primary limiting factor determining rice yield. As the rate-limiting enzyme in nitrogen assimilation and allocation, glutamate synthase (GOGAT) plays an irreplaceable role throughout the plant life cycle. The evolutionary history, natural genetic variation, and regulatory networks of the GOGAT family in rice source–sink tissues during grain filling remain largely elusive. Methods: Here, we combined comparative genomics, population genetics, transcriptomic and biochemical approaches to systematically characterize the GOGAT gene family. Genome-wide identification was performed across 12 angiosperm species, followed by haplotype analysis using resequencing data from ~2000 rice accessions. Transcriptomic, enzymatic activity and metabolite content determination were integrated to investigate their responses to three nitrogen gradient treatments in source (roots, flag leaves) and sink (developing embryos) tissues. Results: A total of 48 GOGAT genes were identified, clustered into two ancient subfamilies (GLU/GLT), with a Poaceae-specific duplication event generating GLT1 and GLT2 subgroups. Specifically, three rice GOGAT genes exhibited distinct domestication signatures: Fd-GOGAT showed strong indica-japonica subspecific differentiation, while NADH-GOGAT2 harbored tropical japonica-specific haplotypes. Furthermore, tissue-specific and developmental stage-dependent nitrogen response patterns were revealed, identifying 5 days after pollination as the critical metabolic switch point. OsGOGAT promoters are enriched with light-, ABA- and stress-responsive cis-elements, suggesting coordinated hormonal and environmental regulation. Conclusions: This study provides comprehensive insights into the functional divergence of the plant GOGAT gene family and coordinated strategies that rice employs under exogenous nitrogen stress, and identifies elite haplotypes for nitrogen-efficient rice breeding. Full article

Water scarcity and environmental degradation have emerged as critical challenges threatening sustainable agricultural development worldwide. This study hypothesized that organic substitution buffers the physiological stress of deficit irrigation on lettuce (Lactuca sativa L.), and that this buffering effect of organic substitution can be quantified using AquaCrop to optimize irrigation schedules. Field experiments were conducted to evaluate the interactive effects of irrigation levels and organic/urea fertilizer rates on soil moisture, lettuce growth and yield during two growing seasons (April to May and September to October 2024) in Songjiang, Shanghai. Six treatments combining two irrigation levels (full irrigation at 100% of crop evapotranspiration (ETc) and deficit irrigation at 60% ETc) with three fertilization levels (300 kg/ha urea, 150 kg/ha urea + 15 t/ha organic fertilizer, and 150 kg/ha urea) were conducted. AquaCrop was calibrated and validated using field observations. The results demonstrated that organic substitution under deficit irrigation preserved soil moisture, canopy development, and fresh yield at levels comparable to full irrigation and nitrogen fertilization, whereas 50% nitrogen alone caused substantial yield reduction. AquaCrop showed satisfactory performance in simulating lettuce growth under organic substitution, with normalized root mean square error (NRMSE) values of 3.07%~21.42%, 5.02%~20.11%, and 8.22%~20.95% for soil water content, canopy cover, and fresh yield, respectively. Using the validated model, scenario analysis across the different precipitation years revealed that lettuce yield was most responsive to irrigation during the seedling stage, followed by the rosette stage and cupping stage. Specifically, 100% ETc during the seedling and rosette stages was suggested for dry years; 100% ETc at the seedling stage combined with 80% to 100% ETc at the rosette stage for normal years; and more flexible strategies (60% to 100% ETc at the seedling stage) for wet years based on model predictions. These irrigation strategies, combined with organic substitution, can achieve above 90% of the potential yield of the full irrigation and nitrogen treatment. These findings offer preliminary guidance for irrigation and fertilization management of lettuce under the experimental climate and soil conditions, while further validation across diverse cultivars and environments is needed before broader application. Full article

The shift from traditional irrigation methods to pressurized irrigation has become essential, particularly considering the water scarcity in the US Southwest. In this context, this study evaluated the effects of different irrigation systems and rates on crop yield (Y) and water productivity (WP) within a multi-cropping system consisting of cantaloupe (Cucumis melo L.), broccoli (Brassica oleracea var. italica), and silage corn (Zea mays L.) under arid conditions in Arizona. Field experiments compared flood (F) and subsurface drip irrigation (SDI) systems at two crop evapotranspiration (ET_c) replacement levels (100% and 80%), resulting in four treatments: F100, F80, SDI100, and SDI80. Seasonal total water applied (TWA), crop yield, water productivity, and silage corn forage-quality parameters were measured. The effects of irrigation systems varied among crops, whereas cantaloupe achieved the highest yield under flood irrigation, with maximum production observed in F100 (63.8 t ha⁻¹). Meanwhile, cantaloupe yields declined under deficit irrigation and SDI treatments (57.2, 39.8, and 27.4 t ha⁻¹ for F80, SDI100, and SDI80, respectively). In contrast, broccoli and silage corn generally performed better under SDI, where more frequent water applications might have improved root-zone moisture conditions and enhanced water productivity. Deficit irrigation substantially increased WP relative to full irrigation without significantly affecting yield for broccoli and silage corn. Broccoli WP ranged from 3.9 kg m⁻³ (F100) to 6.3 kg m⁻³ (SDI80), while silage corn WP increased from 8.7 to 8.8 kg m⁻³ under flood irrigation to 11.6–12.8 kg m⁻³ under SDI. Silage corn forage-quality parameters were not significantly affected by irrigation system or irrigation rate, indicating that moderate water deficits improved seasonal water use without compromising nutritive value. Overall, deficit irrigation reduced seasonal water use and improved WP, with the greatest benefits observed under subsurface drip irrigation. The results demonstrate distinct crop-specific responses to irrigation management, highlighting the necessity of customized optimization strategies. Moreover, our findings highlight that the SDI with moderate deficit irrigation (80% ET_c) can provide an effective balance between water conservation and productivity, enhancing WP without compromising yield or quality under arid conditions. Full article

Accurate evaluation of the shank-to-vertical angle (SVA) is important for optimizing lower-limb alignment during gait, particularly during ankle–foot orthosis (AFO) tuning. This study investigated the validity of a wearable posture sensor for continuous SVA measurement during walking using an optical motion capture system as the reference standard. Nine healthy adults participated, and 88 gait cycles were analyzed. SVA was measured using a shank-mounted wearable sensor and a three-dimensional motion capture system. Gait events for the wearable sensor were identified independently using synchronized tablet-based video recordings rather than inertial signals. Agreement was evaluated at the gait-cycle and participant levels using waveform correlation, mean bias, root mean square error (RMSE), and Bland–Altman analysis. Across the full gait cycle, the trial-level waveform correlation was 0.935 ± 0.064 and the RMSE was 9.74 ± 3.71°. During the individually identified stance phase (mean toe-off, 63.67 ± 1.78% of the gait cycle), waveform correspondence increased to 0.990 ± 0.010 and the RMSE decreased to 6.78 ± 3.51°. Participant-level estimates were similar. For SVA range, participant-level Bland–Altman analysis yielded a bias of −5.58° with 95% limits of agreement from −14.01° to 2.86°, and repeated-measures analysis produced similar estimates. Temporal error analysis showed smaller and more stable deviations during the majority of the stance phase than during swing, with larger deviations in late swing. These findings support the potential use of the wearable posture sensor as a targeted tool for stance-phase SVA assessment, although further validation of pathological gait and actual orthotic tuning is required. Full article

The effects of different rootstocks on the grape quality and volatile aromatic compounds of ‘Shine Muscat’ were investigated. The basic physicochemical parameters of berries from five scion–rootstock combinations (SM-Beta, SM-SO4, SM-5C, SM-Fercal, SM-YN2) and own-rooted ‘Shine Muscat’ grape were measured, and their volatile aromatic compounds were determined using headspace solid-phase microextraction (HS-SPME) combined with gas chromatography–mass spectrometry (GC-MS). Our results showed that the SM-YN2 combination exhibited the highest berry weight, and increased by 28.13% compared with SM self-rooted seedling. In addition, the ‘5C’ and ‘YN2’ rootstocks could better maintain total soluble solids (TSS), titratable acidity (TA), TSS/TA and pH in the berries, whereas the ‘Beta’, ‘SO4’ and ‘Fercal’ rootstocks impaired TSS. Volatile profiling identified 841 volatile compounds classified into 15 categories, among which terpenoids constituted the predominant contributors to berry aroma. Furthermore, 22 characteristic aroma compounds were screened, and linalool presented the highest concentration among all terpenoids in ‘Shine Muscat’ fruit. Odor activity values (OAVs) and aroma proportion analyses further revealed that linalool is the key compound for the characteristic ‘muscat’ aroma of ‘Shine Muscat’. In addition, the SM-YN2 combination exhibited 31.96% and 4.61% higher concentration and proportion of linalool compared to own-rooted vines. Notably, with the exception of ‘YN2’, ‘Beta’, ‘SO4’, ‘5C’ and ’Fercal’ downregulated the expressions of one or more terpenoid synthesis-related genes. In conclusion, grafting onto rootstock could improve berry quality, particularly the ‘YN2’ rootstock, which yielded ‘Shine Muscat’ grapes with the highest overall quality. As a resource of the East Asian grape group, ‘YN2’ can be utilized as a rootstock in future production. Full article

Accurate atmospheric temperature and humidity profiles are fundamental to weather monitoring and prediction. Geostationary imagers such as the Advanced Meteorological Imager (AMI) provide continuous observations and enable profile retrievals through radiative transfer–based algorithms; however, these products remain affected by systematic biases associated with the limited number of spectral channels and reliance on background fields from numerical weather prediction models. This study presents a data-driven post-processing framework to generate reanalysis-consistent profiles by refining AMI-retrieved temperature, mixing ratio, and relative humidity profiles using Light Gradient Boosting Machine (LGBM) models trained with ERA5 reanalysis data. Using four years (2020–2023) of hourly observations, the refined profiles were evaluated against both ERA5 and independent radiosonde measurements. Relative to ERA5, the refinement yields modest but consistent reductions in root mean square error (RMSE), including approximately 0.04 g kg⁻¹ (6–7%) for mixing ratio and 1.9 percentage points (≈14%) for relative humidity, while temperature shows a smaller error reduction of about 0.02 K (2–3%). When compared with radiosondes, temperature RMSE shows a marginal increase overall (<1%) with a larger increase in the lower troposphere, whereas improvements are observed for mixing ratio (2–3%) and relative humidity (6–7%). Seasonal and diurnal analyses reveal systematic error structures in the original AMI profiles, particularly wet-bias patterns in summer moisture fields, which are partially mitigated by the refinement. Feature-importance analysis using Shapley Additive Explanations (SHAP) identifies the dominant contribution of AMI water vapor channels, consistent with their known vertical sensitivity. Overall, this long-term evaluation demonstrates the feasibility of machine learning-based refinement for geostationary imager atmospheric profiles, while also highlighting inherent limitations related to the information content of current-generation imagers. Full article