نوع مقاله : مروری
نویسندگان
1 بخش تحقیقات خاک و آب، مرکز تحقیقات و آموزش کشاورزی و منابع طبیعی آذربایجان شرقی، سازمان تحقیقات، آموزش و ترویج کشاورزی، تبریز
2 گروه علوم و مهندسی خاک، دانشکده کشاورزی، دانشگاه تبریز، تبریز
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Over the past few decades, extensive research has been devoted to the development and application of artificial intelligence (AI) in soil science. Although many current applications of AI in soil science are related to machine learning, this technology also encompasses areas such as digital image analysis, natural language processing, expert systems, and knowledge representation. The aim of this review article is to provide a comprehensive overview of the roles and applications of AI in soil science. First, the history and definitions of AI are reviewed, followed by a common classification of AI into three domains: sensing and interaction, reasoning and decision-making, and learning and prediction. Subsequently, the applications and algorithms associated with each domain in soil science are examined. The main findings indicate that: (a) AI applications in soil science are diverse, including decision support systems, image classification, machine learning-based prediction, and expert systems; (b) currently, the use of AI in soil science is almost entirely linked to machine learning; (c) machine learning applications are primarily seen in digital soil mapping and the development of soil transfer functions; and (d) a significant portion of AI applications focuses on predictive goals. However, several notable exceptions extend beyond these applications, particularly in natural language processing, the development of cognitive soil models, and interpretable machine learning. Based on these findings, an overemphasis on AI-based prediction, coupled with reduced explainability and the lack of effective integration of soil knowledge into algorithms, constitutes major challenges in the field. Future directions include leveraging AI to extract data from legacy soil profile texts for new soil information retrieval, as well as applying natural language processing to construct meta-analyses from soil science literature. These emerging applications have the potential to make a substantial contribution to soil science research.
کلیدواژهها [English]
)