Parcel-level cropland maps are an essential data source for crop yield estimation, precision agriculture, and many other agronomy applications. Here, we proposed a rice field mapping approach that combines agricultural field boundary extraction with fine-resolution satellite images and pixel-wise cropland classification with Sentinel-1 time series SAR (Synthetic Aperture Radar) imagery. The agricultural field boundaries were delineated by image segmentation using U-net-based fully convolutional network (FCN) models. Meanwhile, a simple decision-tree classifier was developed based on rice phenology traits to extract rice pixels with time series SAR imagery. Agricultural fields were then classified as rice or non-rice by majority voting from pixel-wise classification results. The evaluation indicated that SeresNet34, as the backbone of the U-net model, had the best performance in agricultural field extraction with an IoU (Intersection over Union) of 0.801 compared to the simple U-net and ResNet-based U-net. The combination of agricultural field maps with the rice pixel detection model showed promising improvement in the accuracy and resolution of rice mapping. The produced rice field map had an IoU score of 0.953, while the User‘s Accuracy and Producer‘s Accuracy of pixel-wise rice field mapping were 0.824 and 0.816, respectively. The proposed model combination scheme merely requires a simple pixel-wise cropland classification model that incorporates the agricultural field mapping results to produce high-accuracy and high-resolution cropland maps.

Rice is one of the most important staple food sources worldwide. Effective and cheap monitoring of rice planting areas is demanded by many developing countries. This study proposed a weakly supervised paddy rice mapping approach based on long short-term memory (LSTM) network and dynamic time warping (DTW) distance. First, standard temporal synthetic aperture radar (SAR) backscatter profiles for each land cover type were constructed on the basis of a small number of field samples. Weak samples were then labeled on the basis of their DTW distances to the standard temporal profiles. A time series feature set was then created that combined multi-spectral Sentinel-2 bands and Sentinel-1 SAR vertical received (VV) band. With different combinations of training and testing datasets, we trained a specifically designed LSTM classifier and validated the performance of weakly supervised learning. Experiments showed that weakly supervised learning outperformed supervised learning in paddy rice identification when field samples were insufficient. With only 10% of field samples, weakly supervised learning achieved better results in producer’s accuracy (0.981 to 0.904) and user’s accuracy (0.961 to 0.917) for paddy rice. Training with 50% of field samples also presented improvement with weakly supervised learning, although not as prominent. Finally, a paddy rice map was generated with the weakly supervised approach trained on field samples and DTW-labeled samples. The proposed data labeling approach based on DTW distance can reduce field sampling cost since it requires fewer field samples. Meanwhile, validation results indicated that the proposed LSTM classifier is suitable for paddy rice mapping where variance exists in planting and harvesting schedules.

Large-scale crop phenology monitoring is critical for agronomic planning and yield prediction applications. Synthetic Aperture Radar (SAR) remote sensing is well-suited for crop growth monitoring due to its nearly all-weather observation capability. Yet, the capability of the dual-polarimetric SAR data for wheat phenology estimation has not been thoroughly investigated. Here, we conducted a comprehensive evaluation of Sentinel-1 SAR polarimetric parameters’ sensibilities on winter wheat’s key phenophases while considering the incidence angle. We extracted 12 polarimetric parameters based on the covariance matrix and a dual-pol-version H-α decomposition. All parameters were evaluated by their temporal profile and feature importance score of Gini impurity with a decremental random forest classification process. A final wheat phenology classification model was built using the best indicator combination. The result shows that the Normalized Shannon Entropy (NSE), Degree of Linear Polarization (DoLP), and Stokes Parameter g2 were the three most important indicators, while the Span, Average Alpha ($\overline{{\alpha }_2})$, and Backscatter Coefficient ${\sigma }_{VH}^0$ were the three least important features in discriminating wheat phenology for all three incidence angle groups. The smaller-incidence angle (30–35°) SAR images are better suited for estimating wheat phenology. The combination of NSE, DoLP, and two Stokes Parameters (g2 and g0) constitutes the most effective indicator ensemble. For all eight key phenophases, the average Precision and Recall scores were above 0.8. This study highlighted the potential of dual-polarimetric SAR data for wheat phenology estimation. The feature importance evaluation results provide a reference for future phenology estimation studies using dual-polarimetric SAR data in choosing better-informed indicators.

Retrieval-augmented generation (RAG) addresses the problem of knowledge cutoff and overcomes the inherent limitations of pre-trained language models by retrieving relevant information in real time. However, challenges related to efficiency and accuracy persist in current RAG strategies. A key issue is how to select appropriate methods for user queries of varying complexity dynamically. This study introduces a novel adaptive retrieval-augmented generation framework termed Layered Query Retrieval (LQR). The LQR framework focuses on query complexity classification, retrieval strategies, and relevance analysis, utilizing a custom-built training dataset to train smaller models that aid the large language model (LLM) in efficiently retrieving relevant information. A central technique in LQR is a semantic rule-based approach to distinguish between different levels of multi-hop queries. The process begins by parsing the user’s query for keywords, followed by a keyword-based document retrieval. Subsequently, we employ a natural language inference (NLI) model to assess whether the retrieved document is relevant to the query. We validated our approach on multiple single-hop and multi-hop datasets, demonstrating significant improvements in both accuracy and efficiency compared to existing single-step, multi-step, and adaptive methods. Our method exhibits high accuracy and efficiency, particularly on the HotpotQA dataset, where it outperforms the Adaptive-RAG method by improving accuracy by 9.4% and the F1 score by 16.14%. The proposed approach carefully balances retrieval efficiency with the accuracy of the LLM’s responses.

The accurate forecasting of vegetable prices is crucial for policy formulation, market decisions, and agricultural market stability. Traditional time-series models often require manual parameter tuning and struggle to effectively handle the complex non-linear characteristics of vegetable price data, limiting their predictive accuracy. This study conducts a comprehensive analysis of the performance of traditional methods, deep learning approaches, and cutting-edge large language models in vegetable-price forecasting using multiple predictive performance metrics. Experimental results demonstrate that large language models generally outperform other methods, but do not have consistent performance for all kinds of vegetables across different time scales. As a result, we propose a novel vegetable-price forecasting method based on mixture of expert models (VPF-MoE), which combines the strengths of large language models and deep learning methods. Different from the traditional single model prediction method, VPF-MoE can dynamically adapt to the characteristics of different vegetable types, dynamically select the best prediction method, and significantly improve the accuracy and robustness of the prediction. In addition, we optimized the application of large language models in vegetable-price forecasting, offering a new technological pathway for vegetable-price prediction.

Understanding ecosystem service characteristics along urban-rural gradients is vital for enhancing the well-being of urban and rural residents. Despite this importance, prior research has neglected the dynamic evolution of urban-rural gradients during urbanization. This study investigates the spatiotemporal variations of four ecosystem services—habitat quality, carbon sequestration, water yield, and soil retention—along the urban-rural gradient in Jinghong City, China. We propose a method for identifying the gradient using the inverse S function of urban land density distribution and concentric analysis. From 2000 to 2020, ecosystem service supply capacity in Jinghong City continuously declined, indicating degradation over the two decades. The urban-rural gradient zone is classified as core area, inner urban area, suburban area, and urban periphery, each experiencing outward expansion, reflecting significant urbanization. Changes in ecosystem services along the gradient revealed consistently high losses in habitat quality, carbon sequestration, and overall services in the inner urban area, while water yield and soil retention suffered the greatest losses in the urban periphery. As urbanization expanded outward, the loss of these services shifted from the inner urban area to the suburban and urban periphery. These results support decision-making in urban planning and sustainable development for urban-rural regions.