Peach firmness is a critical indicator of fruit texture, maturity, and shelf life. Existing unimodal nondestructive methods lack sufficient accuracy and robustness for multi-variety prediction. To address this limitation, multimodal data were collected from peaches using a laser Doppler vibrometer and a visible/near-infrared spectrometer in this study. Consequently, a cross-attention gating fusion framework (CAGFusion) was proposed for peach firmness prediction based on multimodal acoustic vibration and spectral data. CAGFusion employs dualstream frozen backbone networks for independent feature extraction, a multi-head cross-modal attention module for heterogeneous feature alignment and inter-modal interaction, and an adaptive gating fusion head for regression prediction. The results demonstrated that CAGFusion outperformed the unimodal acoustic vibration and spectral models, with an R2P and an RMSEP of 0.8457 and 0.8703 N/mm, respectively. Compared with a decision-level fusion baseline, CAGFusion further increased the R2P by 5.0%. Ablation experiments showed that removing the adaptive gating module or the cross-modal attention mechanism reduced the R2P by 2.32% and 3.68%, respectively, confirming the essential contribution of each component. Gradient-weighted class activation mapping confirmed that the acoustic vibration and spectral modalities capture complementary physical and chemical information linked to fruit firmness. Overall, this study provides a feasible and generalizable multimodal framework for the nondestructive firmness evaluation of multi-variety peaches.

Oil content is a critical indicator for evaluating the economic value of Camellia oleifera seeds. Given the insufficient mechanistic interpretation in macroscopic oil content detection and visualization studies, this study aimed to explore the feasibility of using hyperspectral microscope imaging (HMI) to evaluate oil content and visualize its spatial distribution at the cellular level in Camellia oleifera seed kernels. Practical constraints associated with equipment and experimental conditions result in limited sample data, thereby adversely affecting model performance. To address this limitation, Wasserstein generative adversarial network with gradient penalty (WGAN-GP) was employed to augment spectral and oil content data, and qualitative and quantitative methods were used to systematically evaluate the generated data and the real data. Partial least squares regression (PLSR) and convolutional neural network regression (CNNR) models were constructed by adding different proportions of generated data to the original calibration set. The results showed that WGAN-GP outperformed traditional DCGAN in data augmentation and improved the prediction performance of the PLSR and CNNR models, with Rp2 reaching 0.6981 and 0.8220, respectively, representing increases of 31.89% and 13.19%. This study not only provides new microscopic mechanistic analysis for the macroscopic nondestructive detection of Camellia oleifera seeds, but also offers an important reference for the accurate prediction of physicochemical traits using HMI under small sample conditions.

Field water status is a critical variable for agricultural water management. In recent years, the development of space–air–ground multi-platform collaborative observation and data fusion technologies has provided new options for precision monitoring. However, challenges in applicability, robustness, and transferability persist. This study employs bibliometric analysis to systematically synthesize the literature, revealing that research has evolved from single-point observations to multi-platform synergy. Satellite, unmanned aerial vehicle (UAV), and ground-based monitoring are analyzed, as well as challenges in multi-source data fusion, including scale mismatch, error propagation, and uncertainty quantification. Finally, applicability and other barriers are evaluated across three typical agricultural scenarios: large-scale surface soil moisture monitoring, crop root zone soil moisture retrieval, and paddy field water depth estimation. The results indicate that space–air–ground collaborative observation constitutes a mature framework, with satellite and ground-based monitoring as core components and UAV technology as a supplement. However, scale transformation and error propagation mechanisms in multi-source data fusion remain unresolved. Currently available vertical water information is limited, and quantitative retrieval has yet to achieve the reliability required for operational applications. This limitation is particularly evident in paddy field water depth retrieval and root zone soil moisture retrieval. This review provides a theoretical reference for precision field water status monitoring and identifies future research priorities, including the integration of physical mechanisms with machine learning (ML) in multi-source data fusion, as well as error quantification and paddy field water depth retrieval.

This study presented a method integrating portable hyperspectral imaging and machine learning to enable non-destructive, in-field visualization of honey peach maturity. The effects of spectral preprocessing, feature selection, dimensionality reduction, and classification models on maturity discrimination performance were systematically investigated. Random Forest (RF) and XGBoost maturity classification models were constructed based on full spectral data, with their performance compared across different preprocessing methods. The RF model combined with wavelet transform preprocessing achieved 96.43

Plants are critical for global food security and ecological stability. Although real-time plant health data is essential for precision irrigation, early disease detection, and yield optimization, conventional monitoring techniques lack continuous, nondestructive tracking capabilities, often missing early physiological stress signals. Flexible and miniaturized plant wearable sensors address this limitation by enabling in situ, real-time monitoring throughout the entire plant growth cycle. This review systematically explores the latest advances of wearable sensing technology for monitoring physical signals, chemical signals, and the local microenvironment. Innovative approaches that enhance the adaptability, autonomy, and intelligence of plant wearable sensors are highlighted. The development of origami-inspired stretchable 3D sensors, biomimetic winding structures, functional hydrogels, gas-sensitive nanocomposites, and self-powered triboelectric nanogenerators enables continuous measurement of plant health status under field conditions. Furthermore, the integration of miniaturized sensing chips has enabled the noninvasive profiling of internal biochemical components. Despite persisting challenges in environmental adaptability, cost-effectiveness, and complex data management, the convergence of advanced nanomaterials, biomimetic designs, self-powered systems, and AI-driven analytics is poised to elevate the performance and intelligence of plant wearable sensors, thereby driving the development of precision agriculture.

Chicken meat was widely favored by consumers for its excellent texture and reasonable price.However,the fre-quent occurrence of white striping(WS)myopathy significantly impacted its quality and marketability.In this study,the comprehensive myopathy index(CMI)was developed as a quantitative predictor of WS myopathy using multiple physico-chemical parameters,including pH,shear force,cooking loss,drip loss,and color.In combination with hyperspectral imag-ing(HSI),a CMI prediction model was developed to achieve rapid and non-destructive identification of WS myopathy.To address the limitation of insufficient samples,a regression generative adversarial network(RGAN)was proposed.Spectral data and the corresponding CMI values were generated simultaneously.And the generated data were highly consistent with the real data in spectral features,t-distributed stochastic neighbor embedding(t-SNE),and CMI values.The comparison of models showed that the convolutional neural network regression(CNNR)model achieved the best performance(of 0.835,RMSEP of 0.057)when 300 generated samples were added.This result confirmed the effectiveness of RGAN in im-proving modeling accuracy and generalization.In conclusion,the proposed CMI prediction method integrating HSI with RGAN enabled accurate quantification of WS myopathy in chicken breast.It would also provide an effective solution for intelligent detection under limited sample conditions.

Soluble solids content (SSC) is a key intrinsic indicator of honey peach flavor, ripeness, and commercial value. This study proposes a nondestructive method for predicting SSC in honey peaches using bioimpedance spectroscopy (BIS) and machine learning. Bioimpedance data from ‘Hujingmilu’ honey peaches were collected over the frequency range of 1 Hz to 100 kHz. A Single-Cole-Warburg equivalent circuit model was constructed to extract seven physiologically relevant electrical parameters. Subsequently, five regression models were trained and evaluated under four feature combinations: partial least squares regression (PLSR), support vector regression (SVR), random forest regression (RFR), extreme gradient boosting (XGBoost), and convolutional neural network regression (CNNR). Among the models evaluated, CNNR, XGBoost, and RFR showed comparable and superior performance. CNNR achieved an Rp of 0.864, an R2 p of 0.728, and an RMSEP of 0.982 on the prediction set, closely followed by XGBoost (Rp = 0.860) and RFR (Rp = 0.858). Shapley additive explanations (SHAP) analysis further indicated that equivalent circuit parameters and phase-angle features were the primary contributors to the model’s predictions. Moreover, four feature selection methods, namely genetic algorithm (GA), ant colony optimization (ACO), competitive adaptive reweighted sampling (CARS), and successive projections algorithm (SPA), were used to construct simplified models. Simplified models based on only five frequency points retained comparable predictive performance. Among them, the GA-CNNR model performed best (Rp = 0.856, R2 p = 0.706, RMSEP = 1.029). These results indicate that BIS data contain substantial information redundancy. Overall, BIS combined with machine learning provides an effective approach for nondestructive SSC assessment in honey peaches.

Soluble solids content (SSC) is an important indicator for determining the commercial value of peaches. Visible/ near-infrared (Vis/NIR) spectroscopy combined with chemometric methods is a primary technique for predicting peach SSC. However, the interference of fruit color with spectral signals makes it challenging to accurately detect SSC across different varieties. This study explored the feasibility of fusing spectral and image data to achieve accurate SSC prediction for multiple peach varieties. Diffuse reflectance spectra and images of three peach varieties ('Hujing', 'Jinqiuhong', and 'Dongxue') were collected. Multiple feature-level fusion strategies for spectral and image data were proposed. Partial least squares regression (PLSR) and support vector regression (SVR) models were developed based on the multimodal fusion data to predict the SSC of individual and multiple varieties, respectively. Their predictive performance was compared with that of models established using spectral data alone. To further improve the generalization ability of the multi-variety models, a spectrum-image fusion network (SIFNet) was proposed by extracting and leveraging high-level image features and integrating them with spectral information. The results showed that the SIFNet achieved superior performance in predicting the SSC of multi-variety peaches, with R2P, RMSEP, and RPDP of 0.8235, 1.0514, and 2.5208, respectively.

The honey peach (Prunus persica L.) is highly valued by consumers due to its distinctive flavor and nutritional richness. However, its quality is strongly influenced by the maturity stage, which directly affects the optimal harvest time and market value. To achieve rapid, non-destructive, and quantitative maturity assessment, hyperspectral imaging (HSI) was utilized to collect spectral data from samples across the visible-near-infrared (VIS-NIR) range. A comprehensive maturity index (CMI) was subsequently established by integrating key physicochemical parameters, including fruit weight, size, color, firmness, pH, moisture content, and soluble solids content (SSC). This index provided a holistic representation of maturity status. To address the limitation of a small sample size, a regression generative adversarial network (RGAN) was introduced. This model jointly generated spectral data and their corresponding CMI values. The generated samples exhibited high consistency with real samples in terms of spectral feature distribution, as visualized by t-distributed stochastic neighbor embedding (t-SNE), and CMI value distribution. Comparative evaluation of prediction models indicated that data augmentation with 800 generated samples enabled the convolutional neural network regression (CNNR) model to achieve optimal performance, with Rp of 0.952 and RMSEP of 0.187. This approach significantly improved predictive accuracy and generalization capability. In summary, the proposed method, which integrates HSI with RGAN for CMI prediction, allows for accurate quantification of honey peach maturity. And it offers an efficient and intelligent solution for maturity grading and quality control.

Maturity critically impacts rational harvesting of Xanthoceras sorbifolia Bunge, influencing yield, oil quality, and postharvest handling. This study pioneered visible/near-infrared hyperspectral imaging (HSI) for non-destructive maturity assessment of Xanthoceras sorbifolia Bunge using a proposed maturity characterization factor (MCF) generated from nine key indicators. All 628 hyperspectral images across four maturity stages of Xanthoceras sorbifolia Bunge were collected, and 19 maturity-related physicochemical parameters were measured. Band variability was analyzed using principal component analysis (PCA), and maturity discrimination models were developed by partial least squares discriminant analysis (PLS-DA), genetic algorithm-support vector machine (GA-SVM), residual network (ResNet), and convolutional neural network-Transformer (CNN-Transformer), incorporating spectral preprocessing. For simplified models, feature wavelengths were extracted using principal component (PC) loading, successive projection algorithm (SPA), and stability competitive adaptive reweighted sampling (sCARS). Developed partial least squares regression (PLSR) showed that model based on second-order derivative spectra, SPA, and PLS-DA (FD2-SPA-PLS-DA) performed well (Accuracy = 98.73%, Kappa = 0.9830). Performance was then significantly improved by combining spectral, color, and texture features (Accuracy = 99.36%, Kappa = 0.9915). For MCF prediction, model based on first-order derivative spectra, SPA, and PLSR (FD1-SPA-PLSR) yielded the optimal results with R2 P = 0.7534, root mean square error of prediction (RMSEP) = 0.1853, and relative percent deviation (RPD) = 2.0138. Finally, an MCF distribution map generated using the preferred simplified model successfully visualized maturity heterogeneity in Xanthoceras sorbifolia Bunge. Our work demonstrates the first application of HSI for rapid, non-destructive Xanthoceras sorbifolia Bunge maturity assessment, establishing it as an effective tool for further field application.

The color of Camellia oleifera fruits is an essential indicator of quality and maturity, potentially reflecting biochemical changes driven by intercellular processes. However, the microscopic mechanism remains unclear. This study combined hyperspectral microscopic imaging (HMI) with characteristic wavelengths to investigate cellular-level shell color changes. Samples were collected at different stages of maturity. Whole fruit and microscopic shell spectra were collected, RGB channels were extracted, and characteristic wavelengths were selected via the successive projections algorithm (SPA). The PLS-DA model revealed 1st derivative preprocessing achieved the best classification accuracy. Using full-band micro-spectra, QPSO-SVM and Attention-LSTM models achieved 88.3 % and 91.1 % accuracy respectively, outperforming models with SPA-selected spectra (85.6 % and 82.2 %). The Lasso-RFR model revealed strong correlations between RGB data and characteristic wavelengths (R2R-Micro=0.98, R2G-Micro=0.99, and R2B-Micro=0.96). Specifically, the green channel showed the strongest correlation with characteristic wavelengths, linked to chlorophyll and photosynthesis-related chemicals. The red channel reflected pigment and sugar changes, and the blue channel provided information on moisture and minerals. In summary, HMI effectively captured subtle color differences and chemical distributions, while characteristic wavelengths significantly improved accuracy and efficiency of models. This study offers insights into color change mechanisms and supports the application of HSI in nondestructive fruit quality testing.

The study aimed to build a lightweight neural network model for Lycium barbarum provenance discrimination using hyperspectral imaging technology. Firstly, hyperspectral images (336.2-1038.8 nm) acquisition in 256 bands andregions of interest extraction were carried out. The extracted spectral information was then used to develop two neural network models, i.e. SimpleCNN and FPN, and the discriminative accuracies on the test set reached 90.81 % and 92.74 %, with AUC of 0.9902 and 0.9934, respectively. Subsequently, two lightweight approaches were applied to SimpleCNN. The first approach combined the distillation of attention-shifting knowledge, which increased the accuracy of SimpleCNN to 93.93 % on the test set. As for the second approach using a three-step pruning approach, unstructured pruning removed 96 % of connections (retaining 92.15 % baseline accuracy), while structured pruning reduced neurons or channels by 86 % (preserving 92.30 % accuracy). The lightweight method used in this study avoided the problem of excessive decrease in provenance discrimination caused by feature wavelengths selection. The obtained model has more practical application value in and rapidly discriminating Lycium barbarum provenances.

Currently, the application of outdoor hyperspectral imaging technology still primarily relies on manually selecting regions of interest (ROI). Due to the complex growth environment of Camellia oleifera fruits, severe situations such as branch and leaf occlusion, fruit overlapping, and background interference, realizing in-situ automatic segmentation of Camellia oleifera fruits and the precise extraction of fruit spectral information still faces serious challenges. This study proposes a CA-TransUNet++ model that integrates the Coordinate Attention (CA) mechanism with a Transformer-based architecture. By leveraging multi-scale feature extraction and global information modeling, the proposed model significantly enhances the semantic segmentation performance for Camellia oleifera fruit images. Transfer learning was introduced on a small-scale hyperspectral dataset, which can effectively accelerate the model convergence and improve its generalization capability. The experimental results showed that the proposed model achieves Mean Intersection Over Union (MIoU), Mean Pixel Accuracy (MPA), and Dice Similarity Coefficient (Dice) scores of 92.14 %, 96.51 %, and 95.81 %, respectively. Furthermore, by combining spectral clustering methods, the adherent fruit region can be segmented precisely to ensure the accuracy of the fruit spectral information automatically extracted from the raw hyperspectral images. Compared with manual ROI extraction, the spectral information obtained using the proposed method demonstrated high consistency with a coefficient of determination (R2) of 99.03 %, and the extraction time was significantly reduced. This research provides a feasible method for real-time monitoring and refined management of largescale Camellia oleifera orchards, which has the potential to be extended to other economic crops.

Fruit is a vital component of the human diet, and its quality attributes are of major concern to consumers. Fruit quality is primarily enhanced during the on-tree growth phase. In-field fruit quality detection provides critical data for precision orchard management, thereby enhancing fruit quality at the source of the supply chain. Driven by growing demand, research on in-field fruit quality detection is rapidly expanding. Therefore, a comprehensive review is essential to track the state-of-the-art inspection technologies and devices tailored to orchard environments. This review systematically summarizes recent advances in promising in-field fruit quality detection technologies, which primarily rely on instrumentation based on mechanical, acoustic vibrational, optical, and electrochemical principles. It evaluates the application scenarios and performance of current portable devices, wearable sensors, noncontact devices, and flexible robotic manipulators integrated with quality sensing capabilities. Furthermore, potential pathways are outlined to facilitate the transition of fruit quality detection technologies from postharvest applications to field-based implementation. Different detection technologies and devices are suited to specific application scenarios. Portable devices leverage miniaturization and cost-effectiveness for routine spot checks. Wearable sensors enable continuous monitoring of long-term fruit quality changes. Noncontact devices achieve orchard-scale assessments through broad spatial coverage. Flexible robotic manipulators with quality sensing capabilities allow integrated harvesting and grading operations. However, challenges remain in terms of robustness, efficiency, and cost-effectiveness under real orchard conditions. Future developments will focus on achieving intelligent and automatic in-field fruit quality detection through innovations in micro-nano fabrication, structural design, advanced algorithms, and the integration of multiple sensors.

The moisture content of Chinese walnut is a critical quality indicator which directly influences storage properties and the oil content. This study evaluated the feasibility of hyperspectral imaging (HSI) integrated with chemometrics for moisture content assessment of Chinese walnuts in a rapid and non-destructive manner. After acquiring 200 hyperspectral images (188 were retained after outlier removal), average reflectance spectra from regions of interest (ROI) were analyzed. In comparison, partial least squares regression (PLSR) models using raw reflectance spectra (termed R-PLSR) outperformed those with preprocessed or alternative spectral units (absorbance and Kubelka-Munk), yielding Rp = 0.7161, RMSEP = 0.6005, RPD = 1.35, and RER = 8.73. After that, two-dimensional correlation spectroscopy (2D-COS), regression coefficients (RC), and competitive adaptive reweighted sampling (CARS) were investigated as wavelength selection algorithms to simplify the R-PLSR model. Results showed that 9 wavelengths selected by CARS were preferred, and gave the comparable accuracy to the R-PLSR model (Rp = 0.6921, RMSEP = 0.6084, RPD = 1.33, and RER = 8.62). Finally, this CARS-R-PLSR model enabled moisture content prediction and visualization of spatial distribution, and results aligned with actual conditions. Despite challenges posed by the thick shells of Chinese walnuts, results demonstrate that HSI has potential in the rough determination and visualization of moisture content.

Chicken meat is favored by consumers due to its delicious taste and affordability. In recent years, reports related to chicken meat's safety and quality issues have frequently occurred in the international community. The objective of the present research is to investigate whether visible and near-infrared hyperspectral imaging (400-1000 nm) can effectively predict the pH of fresh chicken breast meat (pectoralis major). Collecting hyperspectral data is labor-intensive and time-consuming, and datasets are typically limited, which may compromise the model's performance. To address this, a variant of the generative adversarial network (GAN), named conditional GAN with regression (CGAN-R), was proposed. This model served to augment the spectral and pH data obtained from the fillets. The reliability of the generated spectral information was evaluated using the tSNE method. Furthermore, four models were built on the generated data to evaluate the correlation between spectral information and pH values. Subsequently, the generated data were incorporated into the original dataset for further analysis. The study found that the CNN model performed best when 400 generated samples were added (Rp2 of 0.887, RMSEP of 0.041), with Rp2 increasing by 0.075 and RMSEP decreasing by 0.018 compared to training with only original data. Additionally, 19 key wavelengths highly correlated with pH values were determined using the successive projections algorithm (SPA) to build a simplified CNN model, ultimately achieving the visualization of the spatial distribution of pH values of chicken breast fillets.

Peach firmness is a critical quality attribute, yet conventional destructive measurement methods are unsuitable for batch detection in industrial settings. This study investigated a noncontact method for firmness assessment across multiple peach cultivars based on acoustic vibration technology. Three peach cultivars were mechanically excited via a controlled air jet, and the resulting acoustic vibration responses were captured noninvasively using a laser Doppler vibrometer. The frequency-domain acoustic vibration spectra were used as input for firmness prediction models developed using partial least squares regression (PLSR), support vector regression (SVR), and a one-dimensional convolutional neural network (ISNet-1D) that incorporated Inception and squeeze-and-excitation modules. Comparative analysis demonstrated that the ISNet-1D substantially outperformed the conventional linear and nonlinear methods on an independent test set, achieving superior predictive accuracy, with a coefficient of determination ( RP2) of 0.8069, a root mean square error (RMSEP) of 0.9206 N/mm, and a residual prediction deviation ( RPDP) of 2.2879. The good performance of the ISNet-1D can be attributed to the integration of multi-scale convolutional filters with a channel-wise attention mechanism. This integration allows the network to adaptively prioritize discriminative spectral features, thereby enhancing its prediction accuracy. A hierarchical transfer learning strategy was proposed to improve model generalizability, offering a practical and cost-effective means to adapt to diverse cultivars. In summary, the combination of noncontact acoustic vibration and deep learning presents a robust, accurate, and nondestructive methodology for assessing peach firmness, demonstrating considerable potential for cross-cultivar application in industrial sorting and quality control.

Poplar (Populus L.) anthracnose is an infectious disease that seriously affects the growth and yields of poplar trees, and large-scale poplar infections have led to huge economic losses in the Chinese poplar industry. To efficiently and accurately detect poplar anthracnose for improved prevention and control, this study collected hyperspectral data from the leaves of four types of poplar trees, namely healthy trees and those with black spot disease, early-stage anthracnose, and late-stage anthracnose, and constructed a poplar anthracnose detection model based on machine learning and deep learning. We then comprehensively analyzed poplar anthracnose using advanced hyperspectral-based plant disease detection methodologies. Our research focused on establishing a detection model for poplar anthracnose based on small samples, employing the Design of Experiments (DoE)-based entropy weight method to obtain the best preprocessing combination to improve the detection model’s overall performance. We also analyzed the spectral characteristics of poplar anthracnose by comparing typical feature extraction methods (principal component analysis (PCA), variable combination population analysis (VCPA), and the successive projection algorithm (SPA)) with the vegetation index (VI) method (spectral disease indices (SDIs)) for data dimensionality reduction. The results showed notable improvements in the SDI-based model, which achieved 89.86% accuracy. However, this was inferior to the model based on typical feature extraction methods. Nevertheless, it achieved 100% accuracy for early-stage anthracnose and black spot disease in a controlled environment respectively. We conclude that the SDI-based model is suitable for low-cost detection tasks and is the best poplar anthracnose detection model. These findings contribute to the timely detection of poplar growth and will greatly facilitate the forestry sector’s development.

Camellia oil has high commercial and nutritional value. This study combined hyperspectral imaging (HSI) technique with deep learning (DL) to realize rapid and accurate prediction of oil content for Camellia oleifera seeds. First, spectral images of Camellia oleifera seeds from the 400-1000 nm rang were captured, and spectral data from the regions of interest were extracted based on a threshold segmentation method. Then, the partial least squares regression (PLSR) model was used to examine the influence of various preprocessing techniques. It was found that the model performance improved by 7.4 % after standard normal variate (SNV) preprocessing. Meanwhile, an attention mechanism (AM) was introduced into the convolutional neural network regression (CNNR) model, known as ACNNR. The prediction performance of the established models based on full spectra using the traditional (PLSR) and DL methods (CNNR and ACNNR) were compared. Specifically, only raw spectra were taken as inputs for the DL models. The study demonstrated that the model constructed using ACNNR achieved satisfactory results, with R2P, 2 P , RMSEP, and RPD values of 0. 816, 2.552, and 2.348 in the prediction set, respectively. Moreover, to reduce data dimensionality, the full spectra were downscaled using the successive projections algorithm (SPA), genetic algorithms (GA), CNN, and ACNN. Compared to traditional modelling and dimensionality reduction methods, DL showed excellent performance. Finally, the experimental results indicated that the PLSR model developed using spectral features extracted by the ACNN method achieved the optimal performance, with R2P, 2 P , RMSEP, and RPD values of 0.829, 2.462, and 2.425 in the prediction set, respectively. The optimal simplified model was utilized to visualize the spatial distribution of oil content in Camellia oleifera seeds. Generally, the HSI technique combined with DL provides a reliable and effective method for achieving nondestructive detection and visualization of oil content in Camellia oleifera seeds.

Due to containing an abundance of essential nutrients, straw has significant potential to mitigate carbon (C), nitrogen (N), phosphorus (P), and potassium (K) deficits in soil. However, a lack of comprehensive and systematic reviews on C, N, P, and K release and conversion from straw and on the impact of available nutrients in soils supplemented using straw-returning (SR) practices is noticeable in the literature. Therefore, we investigated straw decomposition, its nutrient release characteristics, and the subsequent fate of nutrients in soils. At early stages, straw decomposes rapidly and then gradually slows down at later stages. Nutrient release rates are generally in the K > P > C > N order. Nutrient fate encompasses fractions mineralized to inorganic nutrients, portions which supplement soil organic matter (SOM) pools, and other portions which are lost via leaching and gas volatilization. In future research, efforts should be made to quantitatively track straw nutrient release and fate and also examine the potential impact of coordinated supply-and-demand interactions between straw nutrients and plants. This review will provide a more systematic understanding of SR’s effectiveness in agriculture.