نهمین کنفرانس ملی و اولین کنفرانس بین المللی فیزیولوژی گیاهی

رفتن به صفحه اصلی

ارائه مجازی مقاله

کد مقاله 1190

عنوان Integrating physiology and machine learning: a novel framework for precision wheat management in water-limited environments

نویسندگان Mohsen Jahan - Mohammad Behzad Amiri - َAinollah Hesami - Atefeh Mirzaeian

چکیده This research introduces a novel data-driven framework that combines hierarchical climate zoning with explainable machine learning (XML) to forecast county-level wheat yield and identify significant climatic factors in Khorasan Razavi, Iran. Based on 20 years of high-resolution climate data (2004–2023), we identified three agroclimatic zones: semi-arid plains, intermontane semi-arid plains, and arid plains/desert. The study utilized Random Forest and XGBoost classifiers to classify yields into Low, Medium, and High categories, demonstrating strong performance (macro F1: 0.72–0.75). A critical SHAP (SHapley Additive exPlanations) analysis indicated that growing season length (GSL) and minimum temperature (Tmin) were the most reliable yield drivers across clusters, accounting for physiological processes such as grain-filling duration and respiratory losses. In arid regions, model predictions were primarily influenced by precipitation-related a variable, which emphasizes moisture scarcity as a primary limitation. In contrast to conventional correlation-based methods, our XML framework offers a mechanistic understanding of yield-climate interactions at a high spatial resolution, facilitating practical strategies like cluster-specific sowing windows and precise irrigation scheduling. This approach presents a scalable decision-support instrument for climate adaptation in wheat-producing arid regions worldwide, and it supports process-based models by providing both transparency and practicality in data-scarce situations.

فایل ها

فایل پوستر pdf مقاله

ارائه مجازی مقاله

محتوای ارائه مجازی مقاله

نظرات