The observations used for calibration (ObC) critically impact the utility and accuracy of crop growth models. However, targeted insights into ObC selection and quantity for effective model parameterization remain limited. This study evaluates four rice phenology models, including CERES-Rice, ORYZA, RiceGrow, and WOFOST, as well as a multi-model ensemble. These models are used to predict the heading and maturity of the rice cultivar ‘Shanyou 63’ across 48 stations from 1985 to 2010. We tested different ObC quantities and selection strategies, including random selection, latitude-based, phenological duration-based, and K-means clustering, and analyzed the impact of ObC variability on model uncertainty. The results revealed substantial differences in model prediction performance without specific ObC selection, evident in the significant variability in predicted rice phenology. The MME generally outperformed individual models, as expected. Increasing the quantity of ObC improved both accuracy and stability of model predictions, with K-means-selected ObC proving most effective. Additionally, model prediction uncertainty related to ObC variation was quantified using the Mean Squared Error of Prediction (MSEP), showing that model squared bias contributes more to the total uncertainty than parameter uncertainty. These findings underscore that the quantity and selection strategy of ObC significantly influence model performance, highlighting the importance of tailored calibration approaches for reducing uncertainty and enhancing the reliability and applicability of model predictions.

Conventional breeding for ideotypes in target environments remains challenging due to genotype-by-environment interactions and the genetic complexity of key agronomic traits. Traditional multi-environment field trials are costly and time-consuming, limiting rapid genetic gain. These challenges highlight the need for digital tools to support rice breeding. However, two major approaches, genomic prediction (GP) and gene-based crop models (GBCMs), have distinct advantages. In this study, a dataset derived from a natural rice population comprising 210 genotypes, genotyped with about 650,000 markers, rice dry matter, and yield across three environments, was used to develop two genomic prediction models, genomic best linear unbiased prediction (GBLUP) and a convolutional neural network (CNN), together with a gene-based crop modeling framework. The effectiveness of these models in predicting rice traits and assisting in breeding selection was subsequently evaluated. Prediction results indicated that biomass and yield could be effectively predicted by all models, with Normalized Root Mean Square Error (NRMSE) values ranging from 10.60

Early-season prediction of winter wheat yield and grain protein content is essential for guiding fertilizer and irrigation decisions and reducing uncertainty in variable agroecosystems, as yield affects profitability and quality affects market value and nutrition. Although multi-source data and both single-task learning (STL) and multi-task learning (MTL) are widely used for predicting grain yield and quality, the conditions under which each approach performs best remain poorly understood, especially when data availability, noise, and measurement or computational constraints vary. To address this gap, we conducted a three-year field experiment in Henan Province, China, compiling environmental, agronomic, and proximal-sensing variables across five growth stages. Seven subsets were constructed, including environmental, agronomic, sensor, and combined subsets, and STL/MTL variants of Multilayer Perceptron (MLP), Transformer, and Random Forest (RF) were benchmarked. SHapley Additive exPlanations (SHAP) analysis quantified feature- and stage-level contributions and guided construction of compact Top-K subsets for accuracy–efficiency trade-offs. Multi-source fusion substantially improved accuracy over single-source inputs, with the combined agronomic-sensor subset providing the best performance (yield R2 = 0.823; GPC R2 = 0.743). Under the current stage-aggregated multi-source representation, MLPs outperformed Transformers and RFs across configurations, indicating that compact nonlinear models were better suited to the present medium-dimensional tabular setting. MTL provided the greatest benefit with sparse feature sets or imbalanced predictive difficulty, whereas STL performed better when information was abundant and signals were strong. SHAP analysis showed that agronomic and sensor features associated with biomass accumulation, nitrogen status, water availability, and canopy light interception were key drivers of model predictions, particularly during erecting and early grain filling. These findings further show that the value of STL versus MTL depends on data-source composition and information richness, and that SHAP can be used not only for interpretation but also for reduced-feature subset design. Within the present plot-scale setting, this study therefore provides a decision-oriented framework for identifying both accuracy-oriented and efficiency-oriented configurations, with efficiency referring to feature parsimony, reduced input and preprocessing burden, and computational time for winter wheat yield and GPC.

Accurate estimation of gross primary production (GPP) is essential for understanding the carbon cycle in terrestrial ecosystems. Although numerous models have been developed to estimate GPP, many require a large number of input variables and parameters. For example, light-use efficiency (LUE) models typically rely on multiple inputs, such as photosynthetically active radiation (PAR), the fraction of absorbed PAR, air temperature, vapor pressure deficit, and biome-specific parameters. This dependence on multiple data sources can introduce additional uncertainties into GPP estimates. Here, we developed and evaluated an efficient chlorophyll-based canopy photosynthesis model (CPM) for estimating GPP using the Sentinel-3 OLCI-derived canopy chlorophyll content (CCC) and the canopy chlorophyll absorption coefficient in the red-edge band (alpha(RE)). In the CPM framework, GPP is estimated from the product of potential incident PAR (PAR(pot)) and remotely sensed indicators of photosynthetic capacity, such as CCC or alpha(RE). By using PAR(pot) instead of actual PAR, CPM reduces its dependence on ancillary environmental inputs. We evaluated the CCC-based and alpha(RE)-based CPM models using data from 301 eddy covariance flux sites distributed globally. Based on leave-one-site-out cross-validation, the global CPM models accurately estimated GPP using CCC & times; PAR(pot) (R-2 = 0.76, RMSE = 1.79 gC.m(-2).d(-1), nRMSE = 6.73%) and alpha(RE)& times;PAR(pot) (R-2 = 0.73, RMSE = 1.91 gC.m(-2).d(-1), nRMSE = 7.15%). The CPM models outperformed the widely used near-infrared reflectance of vegetation (NIRv)-based method and the MODIS GPP product. We further evaluated the ability of CCC & times; PAR(pot) to capture global photosynthetic patterns using TROPOMI sun-induced chlorophyll fluorescence (SIF) as a benchmark and found a strong correlation between CCC & times; PAR(pot) and SIF (R-2 > 0.80). These findings demonstrate that the CPM framework relying solely on CCC and alpha(RE) provides accurate and consistent estimates of GPP, offering a practical approach for improving assessments of the global carbon cycle and ecosystem responses to climate change.

Spatial crop growth models are critical for agronomic decision-making, especially in assessing climate change impacts on productivity. However, regional applications of point-scale models are limited by insufficient input data and computational resources. This study aims to address these challenges by developing WG4GEE—a regional winter wheat growth model integrated with Google Earth Engine (GEE). The GEE platform provides high-resolution weather and soil data for the model, which runs on GEE's cloud computing environment. Validated using 2014–2019 winter wheat yield data from China, WG4GEE completed regional yield simulation in 7 h, achieving an R2 of 0.62 and an NRMSE of 15.57%. Simulated LAI time series accurately reflected wheat growth dynamics. Compared with traditional models, WG4GEE simplifies data acquisition, enhances efficiency via cloud computing, and reduces local resource reliance. This study confirms GEE's value for regional agronomic models, offering a tool for large-scale crop monitoring, yield prediction, and climate-adaptive management.

Rice blast (RB), a devastating fungal disease, causes severe yield losses worldwide and demands accurate severity quantification for effective management. Remote sensing has been demonstrated useful in disease monitoring and offers a scalable solution, but the phenology challenges the robustness of the model built for spectroscopic severity quantification. Since the variations induced by phenology are closely confounded with the infection progression, it is crucial to identify the specific plant traits that explain the inconsistency in disease severity (DS) estimation while mitigating the phenological influence. To address this issue, this study proposed a novel approach by extending the PROSPECT+SAIL model to account for the optical effects induced in RB-infected rice plants. By introducing DS into PROSPECT simulations based on spectral mixture analysis and lesion optical measurements, the use of RB-extended PROSPECT decreased the leaf simulation errors by up to 36.3 % in the crucial spectral regions for RB monitoring. Subsequently, such an extension enabled the generation of synthetic datasets for disentangling phenological versus RB-induced physiological effects. The sensitivity and disentanglement analysis revealed that leaf chlorophyll content was the primary factor that compromises the relationship between DS and the rice blast index (RIBInir), which was designed for RB severity quantification. After correcting for these effects by normalizing RIBInir with an optimized chlorophyll-sensitive vegetation index (nRIBInir), estimation accuracies significantly improved with an increment of R2 from 0.67 to 0.79, and rRMSE decreased by 9 %, particularly for vegetative samples with mild infection (R2 increased by 0.51). Consequently, the proposed nRIBInir overcame the underestimation of severe infection areas in both severity quantification and spatial mapping. The adapted nRIBInir for drone and satellite sensors also exhibited great performance in DS estimation. Our findings suggest that RB-extended PROSAIL simulations facilitate mitigating the phenological influence with reliable validations and mechanistic interpretation. Moreover, the adaptation flexibility and robustness of nRIBInir ensured its potential in practical applications including resistance breeding, disease tracking, and precision fungicide management at various scales.

CONTEXT: Reconciling agricultural production with environmental sustainability requires precise quantification of nitrogen (N) system operating boundaries. However, determining regionally differentiated Environmentally Safe Nitrogen Thresholds (ESNT) remains challenging due to complex interactions between management intensity and environmental constraints. OBJECTIVE: This study aimed to develop a hybrid machine learning framework integrating Graph Convolutional Networks (GCN) to capture non-linear dependencies of environmental drivers and determine county-scale ESNT for intensive rice systems in the Middle and Lower Reaches of the Yangtze River. METHODS: Based on a comprehensive dataset of 2385 field observations (2000-2023), we constructed a feature interaction graph to encode dependencies among soil, climate, and management variables. This topological structure was integrated into a hybrid prediction model to quantify four N loss pathways. Subsequently, we simulated three management scenarios (S1-S3) combined with ESNT regulation to assess system-level environmental impacts and potential agronomic feasibility based on empirical N demands. RESULTS AND CONCLUSIONS The GCN-enhanced framework explicitly captured feature interactions, achieving high predictive accuracy (R2 = 0.80-0.91) for NH3 volatilization, N runoff, N leaching, and N2O emissions. Scenario analysis revealed that while Integrated Management (S2) and Advanced Efficiency (S3) both reduced total N losses by approximately 6%, S2 optimized soil physicochemical properties, effectively expanding environmental carrying capacity and raising the mean ESNT to 158.20 kg N ha-1. Consequently, coupling Integrated Management with ESNT regulation (S2 + ESNT) permits higher N inputs than conventional restrictions (S1 + ESNT) while maintaining environmental safety. SIGNIFICANCE: This "active regulation" paradigm balances empirical crop nitrogen requirements with pollution control, offering a scientifically grounded and actionable pathway for the sustainable intensification of regional agroecosystems.

Accurate and timely information on rice cultivation areas and cropping intensity is essential for precision crop management, food security and environmental sustainability. However, the generation of high-quality rice products is often hindered by the lack of ground truth samples, particularly in the regions with complex cropping patterns. While most rice mapping methods rely on the use of remotely sensed flooding signal from the transplanting period to distinguish rice from other crops, they face significant challenges from the increasingly prevalent direct-seeded rice for which the flooding signal at the beginning of season is weak due to the unique management measures. To address these issues and achieve the comprehensive extraction of rice under diverse cropping patterns and establishment methods, this study proposed the phenological knowledge-guided automatic rice mapping approach using optical and synthetic aperture radar (SAR) data (PHAROS). This method first determined the cropping intensity of rice and its concurrent crops using harmonized multi-resource Normalized Difference Vegetation Index (NDVI) time series and phenological knowledge. Subsequently, the Double Phase SAR Index (DPSI) was constructed to extract candidate rice samples by synthesizing the early and peak growth phases retrieved from NDVI time series and the corresponding SAR polarization features. Consequently, training data were generated automatically and fed into a machine learning classifier for rice mapping. The effectiveness of the PHAROS was evaluated over the Middle and Lower Reaches of the Yangtze River (MLRYR) of China and four other major rice production regions in Asia. Furthermore, the earliest timing of early, middle and late rice in the MLRYR was also quantified via the PHAROS and cross-year model transfer. The results demonstrated that the PHAROS could identify rice of diverse cropping pattern with an overall accuracy (OA) from 0.965 to 0.976 in the MLRYR from 2019 to 2023. The classification maps exhibited welldelineated rice parcels and clear separation between single- and double-cropping rice. The PHAROS also yielded OA values of 0.902-0.979 and similar distribution pattern with the reference rice products in Suihua of China, Sataka of Japan, An Giang of Vietnam and Punjab of Pakistan, which represent diverse climatic conditions and cropping patterns across Asia. Compared to its counterpart methods, PHAROS demonstrated significant improvements by 0.022-0.235 in OA in rice planting area extraction and cropping intensity detection. The early, middle and late rice could be identified as early as at tillering, tillering-jointing and sowing/transplanting stages, respectively. This study reveals the necessity of handling the overlooked weak flooding signal from direct-seeded rice and offers a viable solution for large-scale rice cropping intensity detection and mapping under diverse establishment methods.

Global climate warming is characterized by diurnal and seasonal asymmetry, with greater increases at nighttime and in winter and spring. Growing evidence has recognized that night-warming in winter and spring significantly impacts winter wheat production. Pre-crop straw returning is the principal method for straw utilization, but the interactions between straw returning and night-warming on wheat yield and N use efficiency (NUE) remain unclear. Here, a consecutive three-year field experiment with two straw treatments (S0, straw removal; S1, straw returning) and two warming treatments (W0, no warming control; W1, night-warming) found that both S1 and W1 improved wheat grain yield and NUE, with W1 exhibiting more pronounced improvements. Notably, the interaction between S1 and W1 (S1W1) further enhanced yield and NUE by 13.0 and 16.5%, respectively, compared to S0W0 through increasing grain number and 1,000-grain weight (threeyear average). Additionally, root growth and topsoil inorganic N content decreased in S1 before jointing, thereby reducing plant dry matter and N accumulation. However, W1 exhibited an opposite trend, thereby mitigating these negative effects. Simultaneously, under S1W1, increased N translocation to grain and post-anthesis dry matter accumulation, driven by greater N distribution to leaves and higher N metabolism enzyme activity, enhanced both yield and NUE. This improvement was supported by better root morphology and biomass, particularly in the 0-40 cm soil layer, boosting plant N absorption. Additionally, elevated soil N-acquiring enzyme activity after jointing increased the net N mineralization rate and microbial biomass N, enhancing soil N-supply capacity. As a result, post-jointing inorganic N content rose in the 0-20 cm layer while decreasing at 20-60 cm, thus reducing the apparent N surplus. Collectively, straw returning, night-warming, and their interactions enhanced root distribution and N-supply capacity after jointing in the topsoil layer, thereby increasing plant N uptake and its translocation to grains, along with post-anthesis dry matter accumulation, ultimately improving grain yield and NUE.

Context The increasing frequency of droughts and freshwater competition exacerbates the need for robust, plant-based crop water status diagnosis that functions across various phenological phases. Allometric relationships between plant saturated water accumulation (SWAP) and dry mass (PDM), and between ear saturated water accumulation (SWAE) and dry mass (EDM) exist for diagnosing winter wheat water status during the vegetative and reproductive growth phases, respectively. However, an integrated whole-season framework and clear post-anthesis diagnostic approach remain unresolved. Objective This study aimed to construct the critical SWAP curves based on accumulated growing degree day (AGDD) and PDM during the whole growth period of winter wheat under different nitrogen treatments, identify the time boundary of SWAP from accumulation to decline, and establish an integrated whole-season water status diagnostic framework for winter wheat. Methods: A four-year rainout-shelter experiment with two nitrogen and four irrigation levels was conducted to determine critical SWAP points and fit AGDD-based double logistic and PDM-based piecewise (power function and Weibull decay function) model. The water diagnostic index (WDI) was calculated as the ratio of the observed to the critical SWAP value, and its relationship with yield components was evaluated. Results SWAP accumulated first and then declined with time or PDM accumulation. Piecewise functions successfully described the SWAP accumulation and decline processes respectively under nitrogen-limited (N1) and non-nitrogen-limited (N2) conditions (N1: SWAP = 5.34PDM0.77, 23.71exp[-(PDM/12.78)10.20]; N2: SWAP = 7.37PDM0.83, 49.01exp[-(PDM/17.02)12.56]). N2 treatment had a higher SWAP vertex and steeper accumulation and decline segments. The AGDD-based critical SWAP curves showed that the time boundary of SWAP accumulation and decline was about AGDD = 1600 °C d (anthesis) regardless of nitrogen status. The decrease in plant WDI under water stress may be associated with multiple processes, including drought-related soil indirect nitrogen deficiency and premature senescence. However, during the reproductive growth period, the correlation between plant WDI and 1000-grain weight was lower than the ear WDI constructed in previous study. Conclusions With AGDD = 1600 °C d or anthesis as boundary, the plant WDI based on the critical SWAP curve and the ear WDI based on the critical SWAE curve can provide the best water status diagnosis for winter wheat during the vegetative and reproductive growth periods, respectively. Significance/Implications This integrated method can provide time continuous, size-independent water status diagnosis under different water and nitrogen conditions, which will help to provide an effective tool for precise irrigation decision-making and improve wheat productivity.

Precise acquisition of crop phenotypes requires integrating three-dimensional (3D) point clouds with two-dimensional (2D) imagery. However, current systems that capture both point clouds and images simultaneously are typically expensive, limiting phenotypic analysis. To address this challenge, we developed and open-sourced a low-cost RGB image and point cloud integrated scanning system (IPCISS). IPCISS provides a hardware platform for spatio-temporal sensor registration, streamlining the fusion of RGB images and point clouds. Furthermore, an indirect cross-modal extrinsic calibration method based on laser spots was proposed to enable precise spatial registration between an RGB camera and a single-point laser range finder (LRF). The calibrated system enabled accurate 3D scanning and phenotypic analysis of wheat. The 3D scanning performance of IPCISS was evaluated in an indoor environment using a color‑textured calibration box with known geometry. Field experiments were conducted to compare the performance of IPCISS and a terrestrial laser scanner (FARO) in estimating wheat canopy height and leaf area index. Results showed that IPCISS achieved high measurement accuracy and stability. At one-tenth the cost of FARO, IPCISS provided comparable geometric performance, with a ranging accuracy of 0.007 m and a precision of 0.009 m. With the proposed integrated scanning, fusion, and analysis workflow for crop phenotyping, IPCISS enabled the simultaneous estimation of wheat structural and physiological traits. IPCISS achieved centimeter-level accuracy in wheat canopy height estimation (R² = 0.972), comparable to that of FARO (R² = 0.992). By combining structural features with vegetation indices, both IPCISS (R² = 0.922) and FARO (R² = 0.918) achieved improved accuracy in LAI estimation. The low-cost system enables stable 3D scanning and reliable crop phenotyping. This open-source solution provides a reliable and practical tool for building 3D crop datasets, sharing data, and performing integrated phenotypic analysis.

Climate warming is reshaping hydrothermal resources for rice production, but the climatic responses of different rice-based cropping systems have not been sufficiently compared. This study evaluated the climatic suitability and dominant climatic controls of four major rice cropping systems in China: single-season rice, double-season rice, rice-wheat rotation, and rice-maize rotation. Occurrence records were extracted from ChinaCP and screened using a 30 m rice distribution mask. MaxEnt models were calibrated using agroclimatic variables for the historical baseline and end-century SSP2-4.5 and SSP5-8.5 scenarios. The models showed good presence-background discrimination, with mean training AUC values of 0.868 +/- 0.001, 0.959 +/- 0.001, 0.969 +/- 0.002, and 0.963 +/- 0.003 for the four systems, respectively. Dominant climatic controls differed among systems. Single-season rice, rice-wheat rotation, and rice-maize rotation were mainly associated with heat accumulation, with temperature >= 0 degrees C (AT0) showing the highest permutation importance of 55.9%, 59.8%, and 77.0%, respectively. Double-season rice showed a distinct response constrained by precipitation and monthly temperature conditions, with annual precipitation contributing 69.8%. Scenario-based projections for 2081-2100 indicated a system-specific redistribution of climatic suitability, with northward increases most evident for single-season rice and rice-wheat rotation, localized changes for double-season rice, and increased suitability for rice-maize rotation mainly in Southwest and parts of South China. Changes were stronger under SSP5-8.5. These findings show that rice-based cropping systems should not be treated as a single uniform category in climate-change suitability assessments.

Sustaining wheat production has become increasingly challenging amid widespread soil degradation and mounting reliance on nitrogen (N) fertilizers. Here we quantify how soil quality improvement can simultaneously enhance yields and reduce N inputs across the main winter wheat region of China. Using multi-source datasets from 129 agro-meteorological stations across 1981-2020, we constructed a soil quality index (SQI) integrating soil organic carbon, total nitrogen, pH, and bulk density, and evaluated soil quality improvement pathways by adjusting these properties to higher SQI classes. Wheat yield responses were simulated using a four-model ensemble (APSIM-Wheat, DSSAT-CERES-Wheat, DSSAT-Nwheat, and WheatGrow), and greenhouse gas emissions, carbon and nitrogen footprints and economic benefits were assessed. Baseline SQI assessments showed that the majority of sites fell into the moderate, low, or very low-quality categories. Improving soil quality increased wheat yields by 2.1-4.6% and reduced the yield-scaled carbon footprint by 2.0-4.3%. Importantly, improved soil conditions enabled N fertilizer reductions of 8.3-17.4% while maintaining baseline yield, increasing partial factor productivity and lowering greenhouse gas emissions. These results highlight soil quality improvement as a promising pathway to simultaneously boost productivity, reduce inputs, and deliver environmental co-benefits, providing a foundation for regionally targeted strategies to sustainably intensify wheat production in China.

Context: Rice is a staple cereal crop in China, and nitrogen (N) is a key nutrient for its growth and development. Precision N management using remote sensing is critical for food security and sustainable agriculture. Although unmanned aerial vehicle (UAV)-based multi-spectral remote sensing has been increasingly applied to rice N monitoring, existing studies still face limitations in diagnostic accuracy, adaptability across cultivars and regions, and validation for large-scale applications. Therefore, UAV-based precision N management strategies tailored for rice require further investigation. Objectives: (1) Develop UAV-enabled N topdressing diagnosis and regulation methods suitable for intra-field and field scales; (2) Evaluate agronomic, economic, and environmental outcomes of these regulation approaches; (3) Identify an optimized fertilization strategy that balances multi-objective benefits. Methods: Multi-year experiments (2017-2023) in Xinghua City covered multiple cultivars and N rates. UAV multi-spectral imagery from key stages, combined with agronomic and temperature data, supported topdressing diagnosis through the N nutrition index (NNI) and accumulated N deficit (AND). We developed and validated RFVariable (intra-field, prescription-map based) and RF/CNN-Optimized (field-scale, multi-objective) regulation approaches. Validation was conducted at Xinghua Station, Zhuhong Farm, and Zhouzhuang Farm (2022-2023). Results: Direct inversion achieved robust accuracy (NNI: R2=0.62, RMSE=0.20; AND: R2=0.61-0.64, RMSE=20.79-22.40 kg ha-1). Both approaches proved superior to conventional fertilization, increasing the proportion of rice plants with N-suitable status by 22.07 %-67.01 %. At the intra-field scale, RF-Variable reduced the coefficient of variation of NNI by 9.29 %-15 % and improved N agronomic efficiency (NAE) by 19.02 %- 20.11 %. At the field scale, CNN-Optimized achieved a balanced performance across agronomic, economic, and environmental objectives. Conclusions: Integrating UAV multi-spectral diagnosis with scale-appropriate regulation enables actionable, datadriven N topdressing in rice. RF-Variable is effective for pixel-level, variable-rate application; CNN-Optimized is suitable for field-level decision optimization where uniform application is required. Implications: The proposed framework links remote sensing diagnosis to implementable prescriptions, advancing productive, cost-effective, and environmentally sustainable rice N management in alignment with global sustainability goals. Further multi-site, multi-season deployment will facilitate broader adoption and policy/ extension integration.

Context: Crop growth models (CGMs) have been widely employed to simulate crop growth processes and predict grain yield. Although wheat spike photosynthesis has been validated to contribute substantially to yield formation, few existing models explicitly account for this key process. This limitation may restrict a comprehensive understanding of crop growth dynamics and compromises the accuracy of grain yield prediction. Objective: This study aimed to develop an integrated model incorporating spike photosynthesis, validate the accuracy of spike light interception estimation, and evaluate model performance in plot-scale wheat yield prediction. Method: A two-year (2021-2023) field experiment was conducted using 17 wheat varieties under four management practices. Proximal RGB images were combined with deep learning algorithm to estimate the spike light interception ratio (SIR). A novel multi-layered sunlit-shaded canopy photosynthesis module (2M-TPM) was constructed and integrated with SIR into the WheatGrow model, forming the improved WheatGrow-M framework to predict wheat grain yield at the plot level. Results: SIR was estimated with an accuracy exceeding 90 %, and showed strong consistency with field-measured spike radiation interception, with correlation coefficients ranging from 0.78 to 0.83, and varied significantly among wheat varieties and management practices. Spike layers alleviated leaf photosynthetic saturation under high radiation. The WheatGrow-M model reduced the relative root mean square error (RRMSE) of plot-level yield estimation to below 20 %, and revealed that ignoring spike photosynthesis could introduce up to 40 % uncertainty in CGM-based yield prediction. Conclusion: The WheatGrow-M model provides a promising tool for analyzing non-foliar photosynthetic phenotype-yield relationships under genotype-environment-management interactions, and facilitating smallscale yield variation analysis and data assimilation between remote sensing and crop growth models.

Context: Nitrogen (N) and potassium (K) are essential nutrients in intensive wheat production, but imbalanced fertilization compromises efficiency, yield, and environmental sustainability. Objective: This study aimed to elucidate the effects of N-K interactions on grain yield, plant nutrient requirements, nutrient use efficiency, and soil nutrient residues, with the goal of developing optimized fertilization strategies for sustainable agricultural production. Methods: A four-year field experiment was conducted with multiple wheat cultivars under contrasting N and K fertilization regimes. Grain yield and its components, nutrient uptake, nutrient use efficiency indices, and soil nutrient residues at maturity were measured. Results: N-K interactions indirectly enhancing grain yield (GY) by increasing the spikes per square meter (Spsm) and the number of grains per spike (Gnp). Combined N and K application stimulated nutrient demand, accumulation, and use efficiency; a synergistic effect was observed between N uptake efficiency (NUpE) and K uptake efficiency (KUpE). Regardless of N rate, higher K supply markedly reduced soil nitrate-N content (SNc) while increasing soil potassium content (SKc) and soil residual potassium (SRK) at maturity. Using the relationship between surface SNc and deep soil residual nitrate (SRN), optimal N rates were 124 kg ha-1 under low N conditions and 171 kg ha-1 under high N backgrounds, with corresponding K rates of 106 kg ha-1 and 138 kg ha-1; these combinations produced grain yields of 6.91 t ha-1 and 8.38 t ha-1, respectively. Conclusion: Combined N and K fertilization at 124 kg N/106 kg K2O ha-1 and 171 kg N/138 kg K2O ha-1 sustains high wheat yield while minimizing environmental risks. Implications or significance: The findings provide a practical fertilization and nitrate residue management strategy for wheat systems and benchmark data for next-generation N-K interaction models.

Extreme climate events due to global climate change are posing an increasing threat to crop production. Shading and waterlogging as a result of extreme rainfall events, frequently co-occur during jointing stage of wheat in the Yangtze River Region of China, but their combined effects on micronutrients such as iron and zinc remain poorly understood. A two-year experiment was conducted using two cultivars (cv. Ningmai 13 and cv. Huaimai 33) under three shading levels (100

Rapid and reliable assessment of wheat seedling vigor is essential for guiding early-season field management and ensuring stable crop establishment. However, traditional field assessment methods are labor-intensive, subjective, and difficult to scale under heterogeneous production environments. There is therefore a need for practical, scalable approaches that enable consistent and efficient seedling vigor monitoring in field conditions. This study developed a practical, scalable smartphone image-based approach for wheat seedling vigor classification under field conditions. RGB canopy images were collected from seven winter wheat production sites under diverse environmental and management conditions. A lightweight model combining a mobile-friendly backbone with multi-level attention mechanisms was developed to enhance canopy feature representation. Model performance was evaluated using both a fixed train–validation–test split and a location-based cross-validation strategy. Under the fixed split, the model achieved 89.60% accuracy, 89.83% macro-F1, and a Cohen’s kappa of 0.86. Location-based cross-validation yielded 85.90 ± 7.58% accuracy, 84.26 ± 5.98% macro-F1, and a Cohen’s kappa of 0.81 ± 0.17, indicating robust cross-site generalization. Grad-CAM analysis further showed that the model primarily focused on canopy-related regions, supporting its interpretability. The proposed approach enables reliable classification of wheat seedling vigor from smartphone RGB images and demonstrates good generalization across sites. The results indicate that smartphone RGB imagery can support rapid, low-cost assessment of wheat seedling vigor and assist field scouting and management decisions. However, the study was based on limited sites and seasons, and further validation across more years, cultivars, devices, and production environments is required. In addition, the link between classification results and subsequent crop performance or management outcomes was not directly evaluated.

Accurate monitoring of key growth stages in rice is crucial for precision fertilization and irrigation scheduling, as well as timely harvesting. Remote sensing products often suffer from reduced accuracy when estimating crop phenology. Process-based models perform well at field scale, yet their large-scale deployment is hindered by scarce parameter data and limited real-time forcing. This study integrates remote sensing-estimated growth stage data into a crop model, using the first layer of the two-layer parameter optimization to generate estimates for the entire growth period. Subsequently, the growth stage results estimated based on multiple remote sensing vegetation indices are combined with those generated by the first-layer optimization strategy, and the second layer of the two-layer parameter optimization strategy is applied to further improve the prediction accuracy. The results show that the average RMSE for estimating rice growth stages using remote sensing vegetation indices is 9.2 days. The single-layer parameter optimization strategy fills in the key growth stages that remote sensing methods cannot accurately estimate, resulting in an average RMSE of 9.3 days. The two-layer parameter optimization strategy proposed in this study reduces the average RMSE to 5.3 days, significantly improving the accuracy of full-growth period estimation for rice. Cross-year and multi-variety tests confirmed the method's high robustness, showing an average RMSE reduction of 4.8 days over the single-layer method and demonstrating outstanding performance in early growth stage estimation. Regional-scale results show that the proposed method accurately captures spatial variations caused by climate and management differences, with minimal simulation errors. It demonstrates good performance in improving the accuracy and stability of rice growth stage estimation, providing a novel technical approach for crop growth stage monitoring.