Scholarlink

<p>Climate change and its associated variability are expected to alter the geographic viability of land for cultivating various crops. Anticipating both current and future spatial shifts in land suitability for soybean (<i>Glycine max</i>) and cotton (<i>Gossypium hirsutum</i>) is critical for guiding adaptive responses, especially in India’s agriculturally significant black soil zones. This research investigates the cultivation potential of these crops using Species Distribution Models (SDMs) in combination with machine learning algorithms. Through GPS-assisted field surveys, 352 occurrence points for soybean and 251 for cotton were documented. A comprehensive suite of 57 predictor variables—covering climatic, topographic, and soil characteristics—were used to model suitability across three timeframes: present, 2050, and 2070. The most influential predictors included soil parameters like organic carbon, CaCO<sub>3</sub>, channel network base level, precipitation of wettest month (BIO13) and min temperature of coldest month (BIO6). Under random cross-validation, Random Forest consistently outperformed other models for soybean (AUC = 0.89 ± 0.02, TSS = 0.61 ± 0.05, Kappa = 0.64 ± 0.10) and cotton (AUC = 0.85 ± 0.00, TSS = 0.56 ± 0.01, Kappa = 0.54 ± 0.02). In contrast, spatial cross-validation resulted in markedly lower accuracy across all models and crops, with AUC values clustering between 0.55 and 0.62. For soybean, the current suitability map identified approximately 147,203&#xa0;ha as highly suitable, 162,979&#xa0;ha as moderately suitable and 132,014&#xa0;ha as marginally suitable. Similarly, for cotton, 104,999&#xa0;ha area was identified as highly suitable. Future projections under four climate scenarios (RCP2.6, RCP4.5, RCP6.0, and RCP8.5) suggest a general expansion in overall suitability by mid and late century. Nevertheless, highly suitable zones for soybean are expected to shrink by 9.0% under RCP2.6, while cotton suitability remains relatively consistent across all projected scenarios. These insights are crucial for developing climate-smart agricultural strategies and promoting sustainable land resource management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Modeling climate-resilient crop suitability in central India using machine learning and species distribution approaches

  • 期刊论文

  • 发表日期：

    2025年10月22日

查看全文
  • Pravash Chandra Moharana,
  • Ravindra Naitam,
  • Abhay Omprakash Shirale,
  • Sirisha Adamala,
  • Ch. Jyotiprava Dash,
  • B. Amrutha,
  • Mahendra Singh Raghuvanshi,
  • Karunakaran Karthikeyan,
  • Hrittick Biswas,
  • Nitin Gorakh Patil

摘要

Climate change and its associated variability are expected to alter the geographic viability of land for cultivating various crops. Anticipating both current and future spatial shifts in land suitability for soybean (Glycine max) and cotton (Gossypium hirsutum) is critical for guiding adaptive responses, especially in India’s agriculturally significant black soil zones. This research investigates the cultivation potential of these crops using Species Distribution Models (SDMs) in combination with machine learning algorithms. Through GPS-assisted field surveys, 352 occurrence points for soybean and 251 for cotton were documented. A comprehensive suite of 57 predictor variables—covering climatic, topographic, and soil characteristics—were used to model suitability across three timeframes: present, 2050, and 2070. The most influential predictors included soil parameters like organic carbon, CaCO3, channel network base level, precipitation of wettest month (BIO13) and min temperature of coldest month (BIO6). Under random cross-validation, Random Forest consistently outperformed other models for soybean (AUC = 0.89 ± 0.02, TSS = 0.61 ± 0.05, Kappa = 0.64 ± 0.10) and cotton (AUC = 0.85 ± 0.00, TSS = 0.56 ± 0.01, Kappa = 0.54 ± 0.02). In contrast, spatial cross-validation resulted in markedly lower accuracy across all models and crops, with AUC values clustering between 0.55 and 0.62. For soybean, the current suitability map identified approximately 147,203 ha as highly suitable, 162,979 ha as moderately suitable and 132,014 ha as marginally suitable. Similarly, for cotton, 104,999 ha area was identified as highly suitable. Future projections under four climate scenarios (RCP2.6, RCP4.5, RCP6.0, and RCP8.5) suggest a general expansion in overall suitability by mid and late century. Nevertheless, highly suitable zones for soybean are expected to shrink by 9.0% under RCP2.6, while cotton suitability remains relatively consistent across all projected scenarios. These insights are crucial for developing climate-smart agricultural strategies and promoting sustainable land resource management.