Soybean–corn strip intercropping is a national strategy for easing China's protein-crop deficit while protecting wheat. Turning a county pledge (Huojia County's 3,000-to-10,000 mu expansion; 1 mu ≈ 0.0667 ha) into a parcel-level plan couples a band-shaped area target, wheat protection, threshold-based site-selection compliance, and a non-linear subsidy-augmented economic objective. We propose a parcel-level Proximal Policy Optimization (PPO) framework that keeps every cultivated non-strip parcel, including every rule violator, in the decision pool, with compliance entering only through the reward via an auditable rate ρ and an adjustable acceptance floor ρ* (default 0.85). Because per-mu conversion gains are largest for rule-violating parcels, compliance must be learned against the economic gradient, and a counterfactual (difference-reward) advantage estimator repairs the credit-assignment failure of episode-level PPO over 8,833 simultaneous decisions. On Huojia County (17,833 parcels), the learned policy designates 31 parcels (8,828 mu) within the 7,000–10,000 mu band, preserves the wheat floor, and raises gross margin by 3.79 million yuan (+ 3.4%) over the 2022 baseline; it audits at ρ = 0.871 versus 0.000 for a profit-greedy heuristic on the same pool, separating compliant from violating parcels almost perfectly (AUC ≈ 1.0) across three seeds. Compliance tracks the dial: ρ* = 0.95 yields 0.957 and a zero-tolerance fine yields exactly 1.000. A neighbourhood-conditioned contiguity variant reaches up to 76% spatial coherence at matched area, versus 6% for a size-sorted heuristic that moreover overshoots the band. We claim no economic superiority: a correctly targeted size-sorted heuristic ties the learned plan, and under zero tolerance the policy reproduces that heuristic's identical 8,601 mu / 27-parcel plan; the contribution is that compliance, the tunable floor, and the contiguity coupling emerge jointly from one reward formulation rather than a hand-chosen sort key. The framework outputs GIS-ready designation shapefiles; a subsidy-response sweep and a network-architecture comparison remain future work.

As water scarcity intensifies in arable land, accurate estimation of regional agricultural water requirements is becoming increasingly important. Where direct measurement of actual water use is impractical, agronomists and land managers commonly rely on crop water requirement estimates to quantify water demand and thereby guide irrigation scheduling and manage water allocations. A widely used approach for calculating crop water requirements is to determine the crop coefficient (Kc) from satellite-derived NDVI within a functional form such as Kc = Kc max × f(NDVI), which, when multiplied by reference evapotranspiration (ETo), yields crop evapotranspiration under standard conditions (ETc) for field and agronomic crops. In the absence of local measurements, FAO-56 climate-adjusted mid-season coefficients (Kc mid) may be used, but these often deviate from field conditions due to site-specific variability in canopy, soil, and management. To address this limitation, this study proposes a Kc correction method that integrates satellite NDVI with a two-source energy balance model (TSEB-SM) to determine corrected K*c−TSEB−SM. The corrected coefficients were tested within the Kc-ETo framework using field data from ten cropland sites (10 crop types) across North America, Europe, and China, where crops were grown under local management practices. Results showed that the K*c−TSEB−SM approach improved ETc estimation accuracy relative to the climate-adjusted FAO-56 Kc mid, with RMSE reduced by 40

Spring frost is a major meteorological disaster for winter wheat, causing substantial yield losses. Rapid and largescale monitoring is essential for effective post-disaster loss assessment and management. However, most existing frost indices focus primarily on temperature, overlooking the critical modulating roles of soil moisture and phenological stages in influencing frost damage severity at regional scales. To address this gap, this study proposed the Multi-factor Integrated Frost Index (MIFI), which extends prior temperature-based approaches by incorporating low-temperature intensity and duration, soil moisture mitigation, and phenological sensitivity. In contrast to studies reliant on meteorological station data, MIFI leverages widely accessible multimodal remote sensing and reanalysis datasets, thereby enabling more spatially continuous and mechanistically comprehensive large-scale monitoring. Historical frost damage records indicated that 4 degrees C is the optimal low-temperature threshold for frost occurrence in the study area. With the optimized soil moisture adjustment coefficient, the fixed effects model relating county-level MIFI to winter wheat yield showed strong performance, and regression analysis confirmed a significant negative relationship between Delta MIFI and Delta yield. Monitoring results from 2014 to 2024 showed that northern Henan Province, southern Shandong Province, and eastern Jiangsu Province are high-severity and high-frequency areas for spring frost. The spatial pattern of the severe frost event between April 3 and 7, 2018, closely corresponds with historical records. This study provides scientific support for timely, large-scale monitoring of winter wheat spring frost damage and supports improved agricultural disaster risk management in the Huang-Huai-Hai Region.

Farm-level crop allocation is a sequential, constraint-rich decision problem in which farmers must reconcile biological rotation requirements, subsidy-driven economic incentives, and externally imposed crop-expansion targets. Traditional mathematical-programming and population-based evolutionary approaches require explicit reformulation for each new constraint and lack the ability to learn from sequential parcel-by-parcel decisions; heuristic rules in turn provide no formal handle on the trade-offs between competing objectives. We address this gap by formulating the annual crop allocation problem as a Markov Decision Process (MDP) over parcels and training a Proximal Policy Optimization (PPO) agent whose reward jointly encodes target-crop expansion, rotation compliance, economic surplus, and fertilizer-budget guards. The framework is evaluated on Hongxing Farm (Bei’an, Heilongjiang Province, Northeast China), a 308,443-mu farm with 553 parcels, across two policy-driven scenarios. In the soybean-expansion scenario the learned policy achieves a 4,599 mu soybean increase (92% of the 5,000 mu target), maintains rotation compliance at 85.71% (above the 85% policy threshold), and improves total economic return by 4.91% with only a 6.31% increase in fertilizer expenditure. The same trained policy supports built-in subsidy counterfactual analysis through a single-parameter retraining sweep, positioning the framework as a forward-looking complement to ex-post supply-response analyses of recent policy-reform studies. Benchmarked against the farm's own realized 2023 and 2024 allocations, the learned plan is the only one of the three that satisfies every policy constraint, restoring rotation compliance to 85.71% (above the 0.85 floor) where the unguided 2024 outcome falls to 78.84%.

Efficient, accurate, and large-scale acquisition of planting patterns is crucial for food security. Although previous research has successfully achieved large-scale crop mapping using time-series features derived from remote sensing imagery, classification schemes integrating crop-specific rhythms and phenological knowledge remain insufficient under target-label-free conditions in the study area. This study proposed a Crop Multiclass Sampling Optimization Transfer Adaboost (CMSTA) method, which utilized a geoscience-informed instance-feature-domain adaptation approach to combine large-scale crop features with fine-scale phenological features. First, we used time-series Sentinel-1/2 imagery to identify high-quality simulated samples within the target domain through a combination of crop characteristics and spatial analysis. Next, the Phenological Feature Transformer Vector (PFTV), a feature-space transformation technique, was applied via transfer learning to augment diverse samples suitable for the fine-scale target domain. Finally, large-scale crop mapping was performed using a Random Forest classifier integrating both crop and phenological features. The experimental results demonstrated that: (1) Utilizing crop features, 476 credible samples were initially generated for the Yellow River Delta (YRD). These were subsequently expanded to 5,091 through phenological features and instance-feature-domain transfer. Integrating phenological and crop features notably enhanced model performance, increasing the learning curve score by 0.11; (2) The CMSTA method achieved superior crop mapping results within the YRD, with an Overall Accuracy (OA) and Weighted F1-score of 0.90. Compared with models using solely feature-domain transfer (OA: 0.44) and instance-domain transfer (OA: 0.69), CMSTA improved OA by 46% and 21%, respectively.; (3) When applied to large-scale mapping in the North China Plain (NCP), the crop mapping results were validated as reliable, achieving R2 values greater than 0.6 against county-level statistical data. Specifically, wheat exhibited high consistency, achieving an R2 value of 0.88. Overall, CMSTA facilitates scalable, highly accurate multi-crop classification with minimal samples, offering a robust solution for unified crop mapping across extensive regions.

Assessing canopy chlorophyll content (CCC) is crucial for evaluating light capture and photosynthetic capacity, as well as for diagnosing and managing maize health. This study aims to develop a robust CCC estimation model using in situ canopy spectral data collected from two regions over a three-year period. The model employs fractional order differential (FOD) and partial least squares regression (PLSR) at multiple spectral resolutions (1 nm, 5 nm, 10 nm, 20 nm, and Sentinel-2 broadband). To mitigate the uncertainties associated with single models and enhance estimation accuracy, a hierarchical weighted combination model integrating k-means clustering and genetic algorithm (GA) is proposed. The results indicate that the CCC estimation models constructed from differential spectra generally outperform those based on original spectra across most orders. The optimal estimation orders are typically within the range of 1.2-1.6 (step: 0.2). For each resolution, the root mean square error (RMSE) of the optimal order in the test set is reduced by 1.65-25.04 % compared to the original spectra. After constructing the hierarchical weighted combination prediction model, the R2 and RMSE of the combined model for each resolution are superior to those of the single models. Specifically, the RMSE of the test set is further reduced by 0.38-9.87 % compared to the optimal FOD order. Moreover, we achieved better monitoring results when we migrated the method to remotely sensed images. These findings suggest that the hierarchical weighted combination prediction model driven by fractional order differential spectra can achieve more accurate CCC estimation in maize. This method provides a basis for applying FOD to multi-resolution sensors, and this achievement contributes to the precise regulation of fertilizer and water during maize growth, offering a new technical approach for improving crop yield and resource use efficiency.

Study region: The Heihe River Basin, China. Study focus: Irrigation data are often from census surveys at coarse administrative or river basin scale, and as such, the amount of water used for agricultural irrigation difficult to quantify. We improve the Soil Moisture to Rain (SM2RAIN) method to estimate irrigation water use in the Heihe River Basin from 2003 to 2020 using thermal infrared and microwave satellite data. The results showed that this approach has satisfactory performance in estimating the annual irrigation water volume (mean volume=0.657 km3/year, R2=0.83, RMSE=0.03 km3/year) when compared with the field measurements at irrigation district administrative scale, due to its reliability in determining the infiltrated water around the root zone used by crops. New hydrological insights for the region: Through an analysis of irrigation water use trends, the results indicate that most farmland areas exhibited a declining trend in water use per hectare (-55 m³/ha/yr). Interestingly, we observed that while water use efficiency improved significantly at the field scale, overall irrigation efficiency showed a decreasing trend. This study reveals a paradox in the Heihe River Basin, where enhanced irrigation efficiency rarely translates into reduced total water consumption at river basin scale. Our study advances agricultural irrigation volume estimation and irrigation mapping across district and river basin scales in arid and semi-arid areas, which should assist in irrigation scheduling and water resource management.

Generative Artificial Intelligence (GAI) is advancing rapidly and is increasingly integrated into visual communication design education. How to effectively and sustainably leverage GAI to support visual communication design teaching has thus become a critical issue faced by educators. While prior studies have focused on GAI’s impact on student learning outcomes and creativity, limited research has explored its effects on emotions and student engagement. This study aims to investigate the impact of customized GAI integration on visual communication design students’ learning engagement and to qualitatively explore the emotions that occur throughout the learning process. Using a quasi-experimental design, 96 students were randomly assigned to either a control group using traditional instruction or an experimental group using a customized GAI. Student engagement was measured using pre- and post-assessment scales, and semi-structured interviews were conducted to analyze students’ emotional changes. The results show that customized GAI integration effectively enhanced students’ cognitive, emotional, and behavioral engagement. Moreover, students experienced diverse and dynamic emotions during the learning process, which influenced their engagement. This study provides empirical support for the application of GAI in visual communication design education, highlighting the importance of balancing technology integration with emotional regulation, thereby informing the responsible and sustainable integration of GAI in design education.

At the end of the last century, the expansion of agricultural land in the arid and semi-arid regions of northern China intensified the conflict between agricultural development and ecological protection. Accurately mapping abandoned cropland is crucial for balancing these competing interests. This research evaluates the effectiveness of an innovative remote sensing method for producing 30-meter-resolution long-term maps of abandoned and reclaimed croplands in Inner Mongolia, China, using a temporal segmentation approach developed with Google Earth Engine. The method integrates ground sample collection of major crops and inactive cropland with Normalized Difference Vegetation Index (NDVI) analysis during key growth stages, enabling precise classification of cultivation status. By employing a binary classification strategy and adaptive optimization, the efficiency of sample generation improved, providing more effective samples for the Random Forest algorithm. Cropland status maps were successfully generated for Inner Mongolia from 2000 to 2022 with annual accuracy between 97% and 99%. The Temporal Segmentation of Abandoned and Reclaimed Cropland (TSARC) method created time series maps of abandoned and reclaimed cropland at a 30-meter resolution, achieving an overall accuracy of 87.61%. The proposed remote sensing methodology reveals spatiotemporal trends in abandonment rates across arid and semi-humid regions, offering valuable insights for agricultural and environmental management in Inner Mongolia. Considering regional climatic, hydrological, and phenological conditions improves sample collection efficiency and cropland status monitoring. While designed for northern China, this method is also applicable to other single-season agricultural regions for varied agricultural land use monitoring.

Amid growing global food security concerns and frequent armed conflicts, real-time monitoring of abandoned cropland is essential for strategic planning and crisis management. This study develops a method to map abandoned cropland accurately, crucial for maintaining the food supply chain and ecological balance. Utilizing Sentinel-1/2 satellite data, we employed multi-feature stacking and machine learning to create the ARCC10-IM (Abandoned and Reclaimed Cropland Classification at 10-meter resolution in Inner Mongolia) dataset, which tracks annual cropland activity. A novel temporal segmentation algorithm was developed to extract cropland abandonment and reclamation patterns annually, using sliding time windows over several years. This research differentiates cropland states—active cultivation, unstable fallowing, continuous abandonment, and reclamation—providing continuous, regional-scale maps with 10-meter resolution. ARCC10-IM is crucial for land planning, environmental monitoring, and agricultural management in arid areas like Inner Mongolia, enhancing decision-making and technology in land use tracking.

Accurate and timely crop yield forecasts are critical to realizing global food security, balancing international grain trade, and promoting sustainable agricultural development. By providing consistent and large-scale observations, remote sensing technology has become indispensable in crop yield estimation across local, regional, and global scales. Over the past four decades, numerous crop yield forecasting approaches have been developed, including regression-based statistical models, machine learning, semi-empirical models, crop model-data assimilation (DA), and advanced deep learning (DL) approaches. This review comprehensively explores the latest advancements in these methodologies, critically evaluating their strengths and limitations in practical applications. In particular, this article highlights the challenges associated with spatiotemporal variability, environmental stress factors, and model scalability, offering potential solutions to enhance the accuracy and reliability of regional and global crop yield predictions. Besides, a selection strategy is also outlined, providing guidance on choosing the most appropriate yield estimation methods tailored to specific application objectives, data availability, and geographic scales. We also identify key factors affecting crop yield forecasting and offer insights into future trends and directions of development. Furthermore, we underscore the greatest potential of integrating artificial intelligence (AI) and remote sensing technologies with process-based crop growth models through DA techniques. This fusion holds significant promise for addressing the pressing need for accurate and scalable yield forecasts. As the global demand for food intensifies and the need for sustainable agriculture grows, the development and application of these advanced methodologies will be instrumental in ensuring resilient food systems and supporting sustainable agricultural practices.

Medium spatial resolution remote sensing images are widely used for cropland mapping. In areas where cropland is fragmented, however, the limitation of the spatial resolution may lead to inaccurate or even impossible mapping of small croplands. Super-resolution mapping is an effective method to address this issue by transforming coarse-resolution fraction images, derived from spectral unmixing, into fine-resolution land cover maps. In practical applications, a crucial obstacle of this approach is the difficulty in collecting training samples for spectral unmixing and super-resolution mapping. To address this problem, this article proposed a novel super-resolution cropland mapping approach by simulating spectral and spatial training samples. Specially, a mixture spectral simulation method was used to generate training samples for the regression unmixing model to estimate cropland fraction images. A multilevel feature fusion U-NET model was proposed for super-resolution cropland mapping and was trained with simulated training samples considering fraction errors. The proposed method was tested in the Jianghan Plain, China, by generating 2.5-m cropland maps from the 10-m Sentinel-2 images. The results show that the proposed method can accurately extract more smaller and linear land cover features, preserve the spatial structure of the boundaries, and achieve higher accuracy than other cropland mapping methods. This method overcomes the dependency on actual sample collection in traditional methods, better utilizes spectral and spatial features in remote sensing data, and reduces the impact of spectral unmixing errors on the final fine-resolution cropland maps.

Unmanned aerial vehicle (UAV) multispectral and thermal images, combined with machine learning models, have been widely used for high-throughput phenotyping of crop traits and have great potential for evaluating the drought tolerance of winter wheat cultivars. In order to extract the wheat canopy information from UAV images, noise removal is an essential step. Currently, soil and shadow are two of the most common noises in UAV images influencing the extraction of the canopy information, which have been widely studied in previous studies. However, the noise caused by the abnormal canopy temperature in the thermal images has yet to be addressed. Besides, the machine learning-based methods are data-intensive and cannot meet the requirements for rapid evaluation of the drought tolerance of winter wheat cultivars. In order to rapidly evaluate the drought tolerance of winter wheat cultivars, this study proposed a drought tolerance evaluation method for winter wheat cultivars based on multi-criteria comprehensive evaluation and automatic noise removal. The thermal affected zone (TAZ), in which the canopy temperature was abnormally elevated due to thermal radiation from adjacent bare soil, was proposed in this study, and an effective noise removal method was proposed by comparing the accuracy of six automatic image segmentation methods. Canopy vegetation, texture, and temperature indices were extracted from the UAV multispectral and thermal images and selected based on their correlation with the measured yield stability index (YSI). Based on the multiple canopy indices, two multi-criteria comprehensive evaluation methods, i.e., weighted sum based on principal components analysis (PCA-WS) and technique for order preference by similarity to ideal solution based on entropy weight (Entropy-TOPSIS), were used to evaluate the drought tolerance of winter wheat cultivars. The results showed that the automatic image segmentation methods could effectively remove the noises of soil, shadow, and TAZ. Removing the TAZ resulted in a significant decrease in canopy temperature for each irrigation treatment. The total score (TS) and comprehensive evaluation index (CEI) showed a significant linear relationship with the measured YSI, with a maximum R2 of 0.637 and 0.636, respectively. The top five cultivars ranked by the TS and CEI had a consistency ratio of 60–80% with those selected by the measured YSI. This study indicates that the automatic noise removal and multi-criteria comprehensive evaluation have great potential in rapid evaluation of drought tolerance of winter wheat cultivars for large breeding trials.

Apple cultivation is a mainstay industry that promotes agricultural development and boosts farmers' income in Shaanxi Province. Monitoring the flowering date of apple trees is essential for frost damage prevention and yield assessment. However, conventional ground survey methods suffer from high costs and low accuracy, and traditional approaches relying on meteorological data have limitations in spatial resolution. In this study, a set of Normalized Difference Vegetation Index (NDVI) time series, referred to as Crop Reference Curves (CRC), was extracted from pure apple tree MODIS pixels. Subsequently, this CRC was utilized to reconstruct daily 10 m NDVI data from Sentinel-2 imagery. By comparing the spatial phenological variances between the CRC and the reconstructed NDVI sequence, the historical apple flowering date was monitored and mapped with a 10 m spatial resolution in Shaanxi Province. Furthermore, we compared and analyzed the effects of Sentinel-2 images input number (5, 6, 8, and 9 scenes) on the accuracy of flowering monitoring. The results revealed that the scheme using 8 images with an average annual distribution yielded an absolute error of 2 days in monitoring the flowering date in six counties of Yan'an in 2019, indicating an effective fitting effect on monitoring the apple flowering date. The scheme employing 9 images achieved an absolute error of 1.33 days, offering the highest precision in monitoring apple flowering date. Furthermore, when using 9 images, the average error in flowering monitoring remained within 2 days from 2019 to 2021 in four validation study areas, demonstrating strong fitting and practical applicability for monitoring apple flowering dates. This method can be utilized for rapid, efficient and high-precision monitoring of apple flowering date in a wide range with a 10 m spatial resolution. Additionally, the analysis of reconstructed NDVI characteristic can serve as a technical reference for fruit forest classification and growth trend prediction.

The upcoming Landsat Next will provide more frequent land surface observations at higher spatial and spectral resolutions that will greatly benefit the agricultural sector. Early modeling of the upcoming Landsat Next products for soybean yield prediction is essential for long-term satellite monitoring strategies. In this context, this article evaluates the contribution of Landsat Next’s improved spectral resolution for soybean yield prediction under varying levels of water availability. Ground-based hyperspectral data collected over five cropping seasons at the Brazilian Agricultural Research Corporation were resampled to Landsat Next spectral resolution. The spectral dataset (n = 384) was divided into calibration and external validation datasets and investigated using three strategies for soybean yield prediction: (1) using the reflectance from each spectral band; (2) using existing and new vegetation indices developed based on three general equations: Normalized Difference Vegetation Index (NDVI-like), Band Ratio Vegetation Index (RVI-like), and Band Difference Vegetation Index (DVI-like), replacing the traditional spectral bands by all possible combinations between two bands for index calculation; and (3) using a partial least squares regression (PLSR) model composed of all Landsat Next spectral bands, in comparison to PLSR models using Landsat OLI and Sentienel-2 MSI bands. The results show the distribution of the new spectral bands over the most prominent changes in leaf reflectance due to water deficit, particularly in the visible and shortwave infrared spectrum. (1) Band 18 (centered at 1610 nm) had the highest correlation with yield (R2 = 0.34). (2) A new vegetation index, called Normalized Difference Shortwave Vegetation Index (NDSWVI), is proposed and calculated from bands 19 and 20 (centered at 2028 and 2108 nm). NDSWVI showed the best performance (R2 = 0.37) compared to traditional existing and new vegetation indices. (3) The PLSR model gave the best results (R2 = 0.65), outperforming the Landsat OLI and Sentinel-2 MSI sensors. The improved spectral resolution of Landsat Next is expected to contribute to improved crop monitoring, especially for soybean crops in Brazil, increasing the sustainability of the production systems and strengthening food security in Brazil and globally.

The fraction of absorbed photosynthetically active radiation (fPAR) is an important parameter reflecting the level of photosynthesis and growth status of vegetation, and is widely used in energy cycling, carbon cycling, and vegetation productivity estimation. In agricultural production, fPAR is often combined with the light use efficiency model to estimate crop yield. Therefore, accurate estimation of PAR is of great importance for improving the accuracy of crop yield estimation and ensuring national food security. Existing studies based on vegetation indices have not considered the effects of genetic variety, light, and water stress on fPAR estimation. This study uses ground-based reflectance data to simulate 21 common Sentinel-2 vegetation indices and compare their estimation ability for winter wheat fPAR. The stability of the vegetation index with the highest correlation in inverting fPAR under different cultivars, light, and water stress was tested, and then the model was validated at the satellite scale. Finally, a sensitivity analysis was performed. The results showed that the index model based on modified NDVI (MNDVI) had the highest correlation not only throughout the critical phenological period of winter wheat (R2 of 0.6649) but also under different varieties, observation dates, and water stress (R2 of 0.918, 0.881, and 0.830, respectively). It even performed the highest R2 of 0.8312 at the satellite scale. Moreover, through comparison, we found that considering water stress and variety differences can improve the estimation accuracy of fPAR. The study showed that using MNDVI for fPAR estimation is not only feasible but also has high accuracy and stability, providing a reference for rapid and accurate estimation of fPAR by Sentinel-2 and further exploring the potential of Sentinel-2 data for high-resolution fPAR mapping.

Long-term mapping of winter wheat is vital for assessing food security and formulating agricultural policies. Landsat data are the only available source for long-term winter wheat mapping in the North China Plain due to the fragmented landscape in this area. Although various methods, such as index-based methods, curve similarity-based methods and machine learning-based methods, have been developed for winter wheat mapping based on remote sensing, the former two often require satellite data with high temporal resolution, which are unsuitable for Landsat data with sparse time-series. Machine learning is an effective method for crop classification using Landsat data. Yet, applying machine learning for winter wheat mapping in the North China Plain encounters two main issues: 1) the lack of adequate and accurate samples for classifier training; and 2) the difficulty of training a single classifier to accomplish the large-scale crop mapping due to the high spatial heterogeneity in this area. To address these two issues, we first designed a sample selection rule to build a large sample set based on several existing crop maps derived from recent Sentinel data, with specific consideration of the confusion error between winter wheat and winter rapeseed in the available crop maps. Then, we developed an optimal zoning method based on the quadtree region splitting algorithm with classification feature consistency criterion, which divided the study area into six subzones with uniform classification features. For each subzone, a specific random forest classifier was trained and used to generate annual winter wheat maps from 2013 to 2022 using Landsat 8 OLI data. Field sample validation confirmed the high accuracy of the produced maps, with an average overall accuracy of 91.1% and an average kappa coefficient of 0.810 across different years. The derived winter wheat area also has a good correlation (R2 = 0.949) with census area at the provincial level. The results underscore the reliability of the produced annual winter wheat maps. Additional experiments demonstrate that our proposed optimal zoning method outperforms other zoning methods, including Köppen climate zoning, wheat planting zoning and non-zoning methods, in enhancing wheat mapping accuracy. It indicates that the proposed zoning is capable of generating more reasonable subzones for large-scale crop mapping.

In recent years, the single multiplicative neuron (SMN) model, rooted in polynomial design, has garnered substantial scholarly attention and practical application. Building on SMN research, this study introduces an innovative multiple multiplicative neurons (MMN) model designed to enhance model robustness and adaptability. The MMN model incorporates customized backpropagation (BP), particle swarm optimization (PSO), and a cooperative random learning particle swarm optimization (CRPSO) algorithm, offering significant advancements in optimization efficiency and performance. These augmentations are strategically devised to enhance the versatility and efficacy of the multiplicative neuron model within diverse applications. To empirically substantiate the proposed MMN model's effectiveness, two benchmark time series prediction tasks were undertaken. The ensuing results unequivocally underscore the superior predictive performance of the MMN model relative to both the SMN model and the conventional feed-forward neural network. While the BP algorithm was computationally less efficient and prone to local optima in non-differentiable scenarios, and the PSO algorithm exhibited sensitivity to initial values, the CRPSO algorithm overcame these challenges by maintaining population diversity and enhancing optimization through a cooperative stochastic learning mechanism. The evaluated algorithm (MMN-CRPSO) demonstrates a remarkable efficacy in error reduction within data measurements, achieving a precision level of one in ten thousand. This empirical demonstration not only substantiates the proficiency of the MMN model in time series prediction but also positions it as a promising advancement within the broader landscape of neural network modelling and time series analysis.

The prompt and precise identification of corn and soybeans are essential for making informed decisions in agricultural production and ensuring food security. Nonetheless, conventional crop identification practices often occur after the completion of crop growth, lacking the timeliness required for effective agricultural management. To achieve in-season crop identification, a case study focused on corn and soybeans in the U.S. Corn Belt was conducted using a crop growth curve matching methodology. Initially, six vegetation indices datasets were derived from the publicly available HLS product, and then these datasets were integrated with known crop-type maps to extract the growth curves for both crops. Furthermore, crop-type information was acquired by assessing the similarity between time-series data and the respective growth curves. A total of 18 scenarios with varying input image numbers were arranged at approximately 10-day intervals to perform identical similarity recognition. The objective was to identify the scene that achieves an 80% recognition accuracy earliest, thereby establishing the optimal time for early crop identification. The results indicated the following: (1) The six vegetation index datasets demonstrate varying capabilities in identifying corn and soybean. Among those, the EVI index and two red-edge indices exhibit the best performance, all surpassing 90% accuracy when the entire time-series data are used as input. (2) EVI, NDPI, and REVI2 indices can achieve early identification, with an accuracy exceeding 80% around July 20, more than two months prior to the end of the crops’ growth periods. (3) Utilizing the same limited sample size, the early crop identification method based on crop growth curve matching outperforms the method based on random forest by approximately 20 days. These findings highlight the considerable potential and value of the crop growth curve matching method for early identification of corn and soybeans, especially when working with limited samples.