The topsoil aggregates diameter size (ADS) distribution reflects the micromorphological formation mechanisms of soil layers and serves as a quantitative indicator of pore structure, water retention, and soil mechanical properties. To further improve the efficiency, field safety, and parameter standardization of ADS analysis, this study proposes an auto-detection framework by integrating an improved LMTPI strategy, auto-DL, and model transfer in field. First, the vanishing maximum phenomenon was discovered in MTPI (Maximum Transfer Potential Index) auto-DL framework, so the improved LMTPI (Local MTPI) was used to extend its applicability to microscopic ADS datasets. The results demonstrate that the improved LMTPI reduces the required training dataset to 10.40% while decreasing the training time by 89.6%, which drives the subsequent auto-DL optimization process. Subsequently, an auto-DL framework based on LMTPI was employed to train an optimal detection model for ADS, and its results indicate that the automatically selected best Seg-ResNet50, with a mean Average Precision (mAP) of 98.43% and only 40.99M parameters. Finally, using a migration dataset including lighting changes, shooting angles, and foreign object interference, LMTPI-based transfer learning was further optimized for the pretrained model, where results show that model achieves 93.08% accuracy in indoor simulation environments and maintains 90.73% detection accuracy in actual field deployment of wheeled-legged robot. This research provides a technical reference for the auto-DL on-site soil detection and environmental assessments.

The significant differences in insects trapped by pest detection lamps lead to low classification accuracy of existing models for rice pests. To address this issue, this paper proposes a small pest target detection and classification model (ViT-YOLOv5p) by integrating the YOLO backbone and Transformer module. First, the number of training samples is expanded through data augmentation during model training. Furthermore, appropriate noise data are introduced to enhance the robustness and generalization ability of the model. Before detection and classification, image cutting and stitching strategies are adopted to improve the detection accuracy of small objects. The bounding box of the pest is determined by the YOLO backbone, and the corresponding region is fed into the Transformer model to obtain the classification result. Finally, YOLOv5, Faster R-CNN, YOLOv4, and the proposed ViT-YOLOv5p are trained on the same dataset, with average detection time (ADT) and classification accuracy employed as evaluative metrics. The results show that ViT-YOLOv5p achieves the highest classification accuracy of 91.89% with an ADT of 50.41 ms. Compared with the commonly used Faster R-CNN, YOLOv5, and YOLOv4 models, the accuracy is improved by 1.50%, 8.71%, and 9.74%, respectively. This study provides a reference for agricultural pest detection, automatic insect classification systems, and deep learning-based detection of small agricultural targets.

Chicken manure composting is often limited by insufficient humification and substantial nitrogen loss. This study investigated the stage-dependent effects of four treatments during 50-day composting: CK (control without additive), M (Trichoderma longibrachiatum FSJ-F2 inoculation), Mn (manganese sulfate amendment), and MMn (combined FSJ-F2 inoculation and manganese sulfate amendment). M mainly accelerated early precursor depletion and substrate transformation, whereas Mn contributed more strongly to maintaining a favorable degradation environment, promoting nitrogen retention, and driving deeper molecular reorganization during the thermophilic and maturation stages. Consequently, MMn achieved the best overall performance. By day 50, humic acid content in MMn was 13.8% higher than that in CK, while the germination index increased by more than 81.8% over CK. Apparent total nitrogen loss in MMn was 30.1% lower than in CK. MMn also enriched T. longibrachiatum and functionally related bacterial groups, formed a tighter positive cross-domain network, and more consistently promoted radish aboveground growth under equal compost mass application. Overall, integrating microbial inoculation with mineral regulation improved compost humification, apparent nitrogen preservation, and integrated fertilizer performance.

Intensive protected agriculture is highly productive, but its sustainability is increasingly constrained by soil degradation. To explore effective green soil strategies, this study evaluated the effects of Bacillus subtilis (BS), biochar (BC), and their composite biochar-based microbial fertilizer (MF) applied at four rates on soil physicochemical and biological properties, as well as crop growth and yield. A data-driven Soil Quality Index (SQI) framework integrating K-means clustering, principal component analysis (PCA), and canonical correlation analysis (CCA) was developed to quantify soil quality. Furthermore, partial least squares structural equation modeling (PLS-SEM) was employed to elucidate the multi-pathway response mechanisms of the soil-crop system. The results showed that MF treatment outperformed individual BS and BC treatments in improving soil structure, nutrient availability, and microbial activity. Compared with the control (CK), MF reduced soil bulk density by 10.8% and increased field capacity by 19.0%. It also increased available phosphorus and urease activity by 28.9% and 29.1%, respectively, and significantly improved microbial diversity, as evidenced by higher Shannon and Chao1 indices. The MF achieved the highest SQI (0.79), representing a 46.3% improvement over the CK, and increased cucumber yield by 24.23% at an application rate of 1.0% (w/w). Moreover, PLS-SEM further revealed that MF enhanced soil quality and crop yield primarily through three synergistic pathways: strengthening waternitrogen coupling, activating phosphorus and potassium pools, and stimulating organic matter turnover and enzyme activity. Overall, this study demonstrated that MF substantially improved soil quality and cucumber yield through multi-dimensional synergistic mechanisms. The proposed data-driven SQI framework provided a robust and comprehensive approach for quantifying soil quality dynamics, offering both a theoretical basis and practical guidance for sustainable soil management in intensive agricultural systems.

Heavy metal contamination, particularly cadmium, poses a significant threat to plant health and agricultural productivity. While phytohormones modulate stress responses, the combined effects of chemical amendments and beneficial microbes on this regulatory network remain poorly understood. This study evaluated the integrated application of ethylenediaminetetraacetic acid, fertilizers, and Trichoderma harzianum T22 inoculation to mitigate cadmium toxicity in Cosmos bipinnatus and Amorpha fruticosa. The combined treatment significantly improved soil properties, increasing pH from 4.32 to 6.55 and organic matter by 150%, while reducing soil cadmium bioavailability by 74.9%. Plant growth was enhanced, with height increasing by 31% in C. bipinnatus and 29% in A. fruticosa. The amendments restricted cadmium translocation to shoots by up to 68% and boosted photosynthetic efficiency, elevating chlorophyll content by 102% and photosynthetic rate by 102.1%. Phytohormonal analysis revealed up to a 100% increase in jasmonic acid and indole-3-acetic acid, supported by the upregulation of associated biosynthetic and transporter genes. Antioxidant defense systems were strengthened, with activities of superoxide dismutase and catalase increasing by 98% and 105%, respectively, and levels of glutathione and proline rising by 89% and 127%. Furthermore, the treatment promoted a beneficial shift in the rhizosphere microbial community. These results demonstrate that the synergistic strategy effectively enhances plant resilience and cadmium remediation by activating interconnected physiological and molecular pathways. This approach presents a sustainable model for rehabilitating contaminated soils, with promising potential for field-scale application and integration with other phytotechnologies.

The global dependence on organic fertilizers raises concerns about the proliferation of antibiotic resistance in agricultural ecosystems. However, the long-term (decade-scale) dynamics of antibiotic resistance genes (ARGs) and their potential transmission into the food chain via ready-to-eat vegetables remain poorly understood. This study investigates the accumulation of ARGs in soils subjected to 10 years of pig manure application at two intensities, and assesses the subsequent transfer of soil ARGs and human pathogenic bacteria (HPB) to lettuce. Our results demonstrate that long-term manure fertilization drives the linear or exponential accumulation of specific "accumulative ARGs" in soil, with the total abundance increasing by 196% (under 9.0 t/ha) and 145% (4.5 t/ha) over the decade. This accumulation was dose-dependent and linked to the co-enrichment of specific bacterial hosts and mobile genetic elements, establishing persistent soil resistomes. While ARGs in lettuce roots were mainly derived from freshly applied manure rather than from the historically enriched soil ARGs pool, their presence in stems and leaves was negligible. Notably, lettuce cultivated in long-term manure-amended soil did not show an increased load of pathogenic bacteria. Overall, this study establishes a mechanistic framework for understanding the evolution of antibiotic resistance in manured paddy soil, highlighting that fresh manure inputs, rather than historically accumulated soil ARGs, are the primary driver of immediate resistance risk in lettuce.

Additives are widely used to mitigate greenhouse gas (GHG) emissions during manure composting, but whether additive combinations improve mitigation remains unclear. Here, we compared zeolite, ferrous sulfate, superphosphate, and their combinations in cattle manure composting. Superphosphate alone showed the strongest overall mitigation, reducing cumulative CH4 and N2O emissions by 87% and 38%, respectively, relative to the control. The best-performing combined treatment (FL) reduced cumulative CH4 and N2O emissions by 52% and 48%, respectively, indicating that FL performed better for N2O, whereas superphosphate alone achieved much stronger CH4 suppression and the best overall GHG mitigation. This effect was associated with a relatively lower-pH regime, higher NH4+-N retention, lower NO3- accumulation, and clear shifts in bacterial, archaeal, and functional gene profiles. Stage-specific multivariate analyses further showed that pH was strongly associated with late-stage functional reorganization, while treatment identity retained an independent effect after accounting for measured environmental variables. Overall, effective compost mitigation depended more on establishing a favorable physicochemical regime than on increasing additive complexity, and superphosphate treatment represented the most effective low-emission strategy in this system.

Understanding how long-term phosphorus (P) fertilization reshapes soil microbiomes and their functional interplay with P cycling is crucial for sustainable agriculture. This study investigated the effects of different P levels on soil physicochemical properties, phosphatase activities, and the bacterial community across three geographically distinct tobacco-growing sites (YD, SH, and ZS) in Fujian Province, China. Our results revealed a significant positive correlation between soil available phosphorus (AP) and phytase (R = 0.92), challenging the classical P deficiency induction paradigm. Site-specific environment was the paramount factor overriding contemporary P levels in shaping the bacterial community structure and its P-cycling genetic potential. At the broadest scale, stochastic processes dominated the assembly of regional species pools, leading to distinct community templates across sites. Within each site, however, the local P levels acted as a deterministic filter, fine-tuning community composition primarily through species replacement rather than wholesale restructuring. Functional prediction revealed site-specific P-cycling strategies: the YD community showed a higher predicted genetic potential toward aggressive P acquisition (mineralization and solubilization), whereas SH and ZS communities exhibited a higher predicted potential for P scavenging, storage, and recycling (polyphosphate degradation and transport). Redundancy analysis linked these functional profiles to distinct local soil properties: organic matter (OM) at YD, a combination of OM and multiple nutrients at SH, and potassium at ZS. In conclusion, microbial-mediated P cycling in these soils may be governed by a hierarchical mechanism: site-specific context sets the community template, localized soil properties (OM, K) shape the functional repertoire, and current P management modulates the system through species sorting and likely transcriptional regulation. This underscores the need for site-specific, ecological precision management strategies that target dominant local environmental drivers to foster microbial communities capable of optimizing soil P efficiency.

Drought stress and inefficient resource utilization present considerable obstacles to cotton production. The utilization of Bacillus subtilis signifies a prospective remedy to these challenges. Nevertheless, the precise regulatory systems governing its effects have yet to be identified. This research examined the impact of Bacillus subtilis on cotton output under varying drought stress situations. The experiment utilized cotton as the subject, incorporating two application levels of Bacillus subtilis (0 kg/hm and 45 kg/hm) and two drought stress levels (H, indicating conventional irrigation at 350 mm; L, indicating 80% of conventional irrigation at 280 mm). Each treatment was duplicated thrice. The research assessed the impact of various treatments on dry matter accumulation, photosynthesis, root shape, microbial biomarkers, and root exudates. The findings indicated that the utilization of Bacillus subtilis mitigated the adverse effects of drought stress. In comparison to the control, cotton dry matter mass exhibited a growth of 4.38%-15.24%, the photosynthetic rate rose by 3.47%-11.94%, the transpiration rate augmented by 1.91%-7.53%, stomatal conductance enhanced by 4.54%-8.85%, and intercellular CO2 concentration elevated by 2.43%-4.32%. Moreover, enhancements in soil root morphology indicators resulted in an 8.94%-9.28% increase in cotton output. Structural equation modeling demonstrated that Bacillus subtilis modulates soil microbial populations, subsequently influencing biomarkers and root exudates. These factors collectively affect photosynthetic characteristics and root shape, improving stomatal conductance and elevating photosynthetic rates. This enhances dry matter buildup and optimizes root architecture, hence enabling the movement of water and nutrients. Consequently, cotton plants can amass greater photosynthetic products, resulting in enhanced dry matter accumulation and elevated yield. The findings suggest that Bacillus subtilis increases productivity during drought stress, offering insights for optimizing cotton production and enhancing yield in arid areas.

Introduction:Soil salinization constrains crop production in arid regions, yet the microbial and functional mechanisms underlying organic-mineral co-application in saline-alkali soils remain unclear. Methods:A pot experiment with sorghum-sudangrass was conducted in a saline-alkali soil under five fertilization regimes with equal total N but different proportions of organic N. Soil physicochemical properties were measured at the seedling and maturity stages, and rhizosphere bacterial communities and C, N and P cycling genes at maturity were characterized by 16S rRNA gene sequencing and SmartChip high-throughput qPCR. Results:Organic-mineral fertilization decreased soil pH and total salt content and increased soil organic matter, total N and available P relative to mineral fertilizer alone, with the strongest improvements under the 50% organic-50% mineral N regime. Organic inputs increased bacterial Shannon diversity and evenness and shifted community composition, enriching Actinobacteriota, Firmicutes, Bacillus and Pseudarthrobacter. The balanced regime increased genes involved in C degradation/fixation, N fixation and P mineralization/polyphosphate metabolism (e.g., xylA, acsA, mct, nifH, phoD, ppx), whereas mineral-only fertilization favored nitrification/denitrification and methane oxidation genes (e.g., amoA2, nirK, nirS, pmoA), indicating a higher potential for N losses. Discussion:Multivariate analyses identified soil pH, total salt, organic matter and total N as primary regulators of bacterial communities and functional gene profiles. Moderate organic-mineral co-application, particularly the 50%-50% regime, improves soil conditions and strengthens nutrient-cycling potential in saline-alkali sorghum-sudangrass systems.

The rhizosphere microbiome is critical for plant health, yet how soil type and intensive management jointly govern its assembly remain unclear. Here, we hypothesized that soil type acts as a primary environmental filter, while intensive cultivation (plant growth plus fertilization) imposes additional selective pressures that differentially shape bacterial versus fungal communities. Using flue-cured tobacco (K326) grown in clay loam and sandy loam soils under field conditions, we examined the rhizosphere microbiome at the topping stage. Intensive cultivation significantly altered rhizosphere physicochemical properties. Key nutrients, including organic matter (OM), dissolved total nitrogen (DTN), available phosphorus (AP), and available potassium (AK), were markedly enriched. Rhizosphere soil pH exhibited a bidirectional shift relative to the corresponding bulk soil, converging to a narrow range (7.4–7.8) in both soil types. Root activity and fertilization imposed contrasting selective pressures on the two microbial kingdoms: bacterial diversity declined slightly, indicating strong deterministic selection, whereas fungal diversity increased, reflecting adaptation to root-generated niches. Differential abundance analysis identified 38 bacterial OTUs as a core rhizosphere-adapted microbiome shared across both soil types, demonstrating robust fitness in the nutrient-enriched rhizosphere environment under intensive management. No shared core fungal OTUs were detected, underscoring strong soil legacy effects and higher habitat specificity in fungi. Notably, the core bacterial microbiome was dominated by K-strategists (slow-growing, resource-efficient taxa) that exhibited opportunistic traits capable of rapidly exploiting nutrient pulses in the rhizosphere. Together, these findings reveal that soil type acts as a critical filter modulating plant–microbe interactions under intensive agriculture, while bacteria and fungi employ divergent ecological strategies in response to selection pressures. This work provides both theoretical and practical insights for optimizing tobacco cultivation and sustaining soil microecological health.

Cross-instrument comparability of O-J-I-P (OJIP) chlorophyll fluorescence transients remains limited because blue and red excitation strategies generate fluorescence curves with different spectral responses and temporal morphology. This study aimed to develop a transfer learning framework for standardizing heterogeneous OJIP data and improving cotton salt stress diagnosis. OJIP measurements from 14,145 cotton leaf samples collected in greenhouse and field experiments were used. FluorPen-FP110 and Pocket PEA data were compared under blue and red excitation. A phase-specific OJIP-Standard Normal Variate (OJIP-SNV) preprocessing method was developed and evaluated with support vector machine (SVM), bidirectional long short-term memory (Bi-LSTM), one-dimensional convolutional neural network (1D-CNN), and Cotton Salt Stress-OJIP-Net (CSS-OJIP-Net). A conditional generative adversarial network framework, FluoToFluo, combined OJIP-SNV preprocessing with a Bi-LSTM transform for cross-instrument fluorescence migration. Blue- and red-excitation measurements showed clear differences in fluorescence intensity distributions, with weak correlations for several intensity parameters but stronger agreement for relative fluorescence parameters. OJIP-SNV improved SVM and Bi-LSTM classification performance. CSS-OJIP-Net achieved 87.80

Trichoderma harzianum (T. harzianum) and Polyaspartic acid (PASP) have been shown to enhance phytoremediation of Cd-contaminated soils. However, their effects on soil bacterial and fungal community structure and network stabilityremain unexplored. This study systematically investigated microbial community dynamics in response to phytoremediation, employing high-throughput 16S rRNA and ITS sequencing coupled with co-occurrence network analysis. The results showed that the combined treatment (PASP + T. harzianum, referred to as AT) increased the Cd removal efficiency in soil from 21.71 % to 38.27 %. Soil CEC, OM and Cd bioavailability were significantly increased. PASP contributed to increasing the node constancy and community composition stability, thereby increasing network stability (natural connectivity, robustness and positive cohesion) in AT treatment. During AT treatment, Trichoderma emerged as the core fungi, while Pseudomonas and Arenimonas became the core bacterial taxa. The ecological strategy transformation and the synergistic effect between T. harzianum (r-strategy) and PASP (K-strategy) increased the stability of microbial community. This then promoted functional enrichment in the function of cis-vaccenate biosynthesis and TCA cycle II (plant and fungi). The expression of the ZIP gene family and TC.HME transporters was significantly upregulated in AT treatment. Fungal network structure and stability directly influenced bacterial network stability and function, which via soil enzyme activity, promoted plant Cd accumulation. Overall, this study emphasises the crucial role of microbial network restructuring and functional transitions in ensuring the long-term effectiveness of phytoremediation. Identifying core microbial taxa and their responsive ecological strategies provides the theoretical basis and practical guidance needed to develop targeted remediation strategies in complex contaminated soils.

UAV image acquisition and deep learning techniques have been widely used in field hydrological monitoring to meet the increasing data volume demand and refined quality. However, manual parameter training requires trial-and-error costs (T&E), and existing auto-trainings adapt to simple datasets and network structures, which is low practicality in unstructured environments, e.g., dry thermal valley environment (DTV). Therefore, this research combined a transfer learning (MTPI, maximum transfer potential index method) and an RL (the MTSA reinforcement learning, Multi-Thompson Sampling Algorithm) in dataset auto-augmentation and networks auto-training to reduce human experience and T&E. Firstly, to maximize the iteration speed and minimize the dataset consumption, the best iteration conditions (MTPI conditions) were derived with the improved MTPI method, which shows that subsequent iterations required only 2.30% dataset and 6.31% time cost. Then, the MTSA was improved under MTPI conditions (MTSA-MTPI) to auto-augmented datasets, and the results showed a 16.0% improvement in accuracy (human error) and a 20.9% reduction in standard error (T&E cost). Finally, the MTPI-MTSA was used for four networks auto-training (e.g., FCN, Seg-Net, U-Net, and Seg-Res-Net 50) and showed that the best Seg-Res-Net 50 gained 95.2% WPA (accuracy) and 90.9% WIoU. This study provided an effective auto-training method for complex vegetation information collection, which provides a reference for reducing the manual intervention of deep learning.

Iron nanoparticles (Fe-NPs) have emerged as a revolutionary tool for enhancing the efficiency of plant growth regulators (PGRs) delivery in modern agriculture. This review explores how Fe-NPs address critical challenges in conventional PGR applications, including instability, rapid degradation, and non-target effects. Their unique properties, such as high surface area, magnetic responsiveness, and biocompatibility, enable the precise encapsulation and controlled release of key PGRs, including auxins, gibberellins, cytokinins, and abscisic acid, thereby improving bioavailability and reducing environmental contamination. Fe-NPs demonstrate remarkable potential in enhancing plant growth, stress tolerance (including drought and salinity), and crop productivity through targeted delivery mechanisms. Additionally, their dual role as both PGR carriers and iron micronutrient supplements offers synergistic benefits for plant health. While promising, challenges in scalability, cost-effectiveness, and environmental safety must be addressed for widespread adoption. By integrating nanotechnology with precision agriculture, Fe-NPs-mediated PGR delivery offers a sustainable approach to enhancing crop performance and resilience in the face of climate change and increasing global food demands. The objectives of this review are to highlight current advancements, key mechanisms involved in the target delivery of Fe-NPs, abiotic stress tolerance (including oxidative stress modulation and enhanced metabolic processes), applications, and future directions for harnessing Fe-NPs in next-generation agricultural practices.

Drought is a global issue that affects agricultural productivity and sustainable development. The application of Bacillus subtilis has significant potential in alleviating drought stress and increasing yield. However, it is not yet clear how Bacillus subtilis affects microbial populations, crop yield, and the biochemical characteristics of rhizosphere soil, as well as the interactions among these factors. In this study, cotton was used as the experimental crop, and different application rates of Bacillus subtilis (0 kg·ha−1 and 45 kg·ha−1 (B)) and drought stress levels (H represents conventional irrigation, 350 mm; L represents 80% of conventional irrigation, 280 mm) were set as three replicates per group. The changes in rhizosphere-soil-related variables, microbial community diversity, enzyme activity, and cotton yield were studied. Compared to the control, the available nitrogen content increased by 19.76–62.40%, and soil moisture increased by 2.48–7.72%. The activities of urease, sucrase, and alkaline phosphatase increased, malondialdehyde content decreased, the Soil Plant Analysis Development (SPAD) value increased, and cotton yield increased by 8.94–9.28%. According to the structural equation model, Bacillus subtilis can increase microbial community diversity and network complexity, improve soil nutrients and enzyme activity, and increase cotton yield. This study’s findings may offer a theoretical foundation for enhancing soil quality and raising agricultural yields in arid regions.

Biochar application is a well-recognized strategy to enhance agricultural soil fertility, but its structural heterogeneity leads to inconsistent outcomes in soil improvement, particularly in water and nutrient transport dynamics. In order to ensure the beneficial effects of biochar-amended agricultural soils in terms of water retention and fertilizer fixation, in this paper, we aim to elucidate the effect of the structural heterogeneity of biochar on the hydraulic properties and nutrient transport of agricultural soils. This study compares biochars at millimeter (BMP), micrometer (BUP), and nanometer (BNP) scales using CT scanning, and investigates the effects of different application rates (0.0–2.0%) on soil’s hydraulic properties and nutrient transport using soil column experiments and CDE analyses. The results show that biochar generally decreased soil saturated hydraulic conductivity (SSHC), except for the application of 2.0% BMP, which increased it. Biochar enhanced soil saturated water content (SSWC) and water holding capacity (WHC), with the 2.0% BMP treatment achieving the highest values (SSHC: 49.34 cm/d; SSWC: 0.40 g/g; WHC: 0.25 g/g). BUPs and BNPs inhibited water infiltration due to pore-blocking, while 2.0% BMP promoted infiltration. Convective dispersion equation analysis (CDE) indicated that BUPs and BNPs reduced water and nutrient transport, with 2.0% BMP showing optimal performance. Statistical analyses revealed that biochar’s structural heterogeneity significantly affected soil water repellency, its hydraulic properties, and solute transport (p < 0.05). Smaller particles enhanced water retention and nutrient fixation, while larger particles improved WHC at appropriate rates. These findings provide valuable insights for optimizing biochar application to improve soil functions and support sustainable agriculture.

Drought severely impacts crop productivity and fertilizer efficiency in arid regions, hindering sustainable agriculture. Enhancing plant drought tolerance and fertilizer efficiency is crucial for adaptation. Bacillus subtilis can improve soil structure and rhizosphere activity, boosting nitrogen utilization and crop yields. This study investigates Bacillus subtilis' potential to mitigate drought stress and enhance cotton growth, aiming to establish a water-bacteria interaction-based irrigation model. Cotton (Tahe 2), the popular variety in local, was selected as the experimental crop, and planted in test pits (3.3 mx2 mx3 m) under varying Bacillus subtilis rates (0 kg center dot ha-1 and 45 kg center dot ha-1) and drought stress -levels (H for conventional irrigation, 350 mm and L for 80 % of conventional irrigation, 280 mm). Each treatment has three replicates. The results showed Bacillus subtilis increased soil water retention by 1.07 %-33.08 % and nitrogen use efficiency by 8.94 %-9.28 %. Cotton growth was also improved, with plant height increasing by 6.45 %-10.5 %, stem diameter by 1.2 %-10.5 %, and leaf area index by 5.3 %6.97 %. Photosynthesis was enhanced, with leaf internal water use efficiency up by 1.02 %-4.21 % and instantaneous water use efficiency by 0.33 %-9.7 %. Yields increased by 8.94 %-9.28 %, and water use efficiency by 5.49 %-19.22 %. Furthermore, bacterial network analysis and the neutral community model revealed that Bacillus subtilis altered the microbial community and rhizosphere environment, increasing the complexity of the bacterial network. This optimized the availability of water and nutrients for root uptake, enhanced the biological utilization of carbon and nitrogen, and supported microbial metabolism and plant growth. These effects reduced the adverse impact of drought stress, alleviated environmental pressures, and fostered a healthier and more sustainable soil ecosystem. In conclusion, combining deficit irrigation (280 mm) with Bacillus subtilis (45 kg center dot ha-1) can effectively alleviate water scarcity and increase cotton yield in arid regions, providing valuable insights for sustainable agricultural development.

Deep learning networks have been widely used for vegetation detection in field images to monitor environmental information, but network structures optimized by manual methods require extensive experience and trial and error (T&E). So, this paper proposes an intelligent agent to auto-optimize networks using a Directed Evolutionary Genetic Algorithm (DEGA) and an auto-transfer learning method (FMTPI) for vegetation detection in dry thermal valley. Firstly, the basic conditions to build the agent were determined, showing that the optimal basic conditions were a block type (BT) of ResNetB , an input size (IS) of 753, a intermediate block number (M) of 4, and a kernel size (X) of 5. Secondly, the operation conditions to drive the agent were obtained with the improved FMTPI, showing that the fitting combination of exponential and inverse functions in FMTPI can reduce the time cost by 90.1%. Finally, the agent auto-optimized the pre-trained network, showing that the model size (MS) could be reduced by 87.3% while maintaining a detection accuracy of 92.3%. The agent can auto-optimize networks and improve network efficiency, providing a meaningful reference for vegetation and environment detection.