Maps serve as the primary representation of national territory, and their normative expression is closely tied to na-tional sovereignty and territorial integrity. However, most existing R mapping packages rely on international geo-graphic datasets and therefore face limitations in providing a complete and compliant representation of China’s territory, controlling map symbology, and configuring projection parameters, making them insufficient for com-pliance-oriented scientific cartography. To address these challenges, the “ggmapcn” package was developed within the “ggplot2” grammar of graphics framework. It integrates administrative division data from the National Plat-form for Common GeoSpatial Information Services and employs a layered visualization architecture to enable compliant mapping of both China and the world, while supporting flexible projection transformations and inte-grated visualization of vector and raster data. This paper systematically introduces the functional architecture and usage of “ggmapcn” and illustrates its application in producing thematic maps that comply with national standards through representative examples. The results demonstrate that “ggmapcn” can generate map outputs in R that con-form to official national boundary standards for China and other countries, thereby enriching the R ecosystem for spatial cartography. Finally, the paper outlines future directions for functional improvement and broader applica-tion of the package.

Plant phenology is a key indicator of how ecosystems respond to global climate change. Although automated imaging technologies can generate vast amounts of time-series phenological data, the accurate automated recognition of discrete phenological periods remains a significant methodological challenge, limiting the scalability of phenological monitoring. In this study, we focused on Rhododendron hypoglaucum, one of dominant species in the evergreen broad-leaved mixed forests of Shennongjia, China. Using 4624 automatically captured time-lapse images collected continuously from 2022 to 2025, we implemented and evaluated 3 representative object detection algorithms—Faster R-CNN, YOLOv11, and RT-DETR to develop automated recognition models for key phenophases. The performance of the models was compared to identify the most effective algorithm for establishing a high-frequency, long-term phenology recognition method. All 3 algorithms successfully detected six phenological traits: floral buds, leaf buds, flowers, new leaves, fruits, and senescent leaves. The YOLOv11 model performed best, achieving a precision of 0.785, recall of 0.745, mAP50 of 0.788, and mAP50-95 of 0.501. Based on the results of the optimal model, the phenological periods duration of R. hypoglaucum—including the duration of floral bud growth, leaf bud growth , flowering, new leaf growth, fruit growth, and yellow leaf—were automatically determined, showing a high level of consistency with manual visual interpretation. This study demonstrates that deep learning-based object detection methods can effectively realize the automated extraction of phenological traits and quantitative information from long-term, in-situ observations. It provides a novel approach for high-frequency, precise, and automated phenological recognition at the individual plant level. In the future, through multi-target collaborative monitoring and model optimization, the applicability and robustness of this method are expected to be further enhanced.

Abstract Aim This study aims to develop a methodological framework for plant trait extraction using large language models (LLMs) and systematically evaluate the effects of model type, prompt design, and parameter settings on task performance. Methods Using the Flora of China as the data source, we constructed a standardized benchmark dataset containing 27,922 manually annotated records and designed a unified data preprocessing pipeline and systematic prompt templates. Two models (DeepSeek-V3 and DeepSeek-R1-0528-Qwen3-8B) were selected, and a three-factor experimental design was established by combining model type, temperature setting, and prompt strategy. Multiple repeated experiments were conducted for plant growth form and life span classification tasks. Model performance was comprehensively evaluated using missing rate, mismatch rate, accuracy, precision, recall, specificity, F1-score, and AUC, and analysis of variance was used to test the significance of main effects and interaction effects. Important findings The results showed that different models varied in parameter sensitivity and error-type control. DeepSeek-V3 showed more stable overall performance, whereas DeepSeek-R1-0528-Qwen3-8B was more sensitive to systematic prompts. A lower temperature setting generally helped reduce hallucination rates and the risk of output instability, whereas a higher temperature setting (0.6) may improve recall in some scenarios. Systematic prompts significantly reduced missing and hallucination rates, improved output completeness and category validity, and enhanced classification performance in some tasks and metrics. Compared with simple model performance comparison, parameter configuration and prompt strategy had a more critical influence on the practical applicability of LLMs in botanical text parsing tasks. The proposed framework provides a reusable technical pathway for automated biodiversity information extraction.