Potential distribution modeling based on machine learning of Sechium edule (Jacq.) Sw. in Japan

• Published in: Genetic resources and crop evolution Vol. 71; no. 5; pp. 2105 - 2115
• Main Authors: Iñiguez, Jorge Cadena, Barrera Guzmán, Luis Ángel, Arévalo Galarza, Ma. de Lourdes C., Cadena Zamudio, Daniel Alejandro, Watanabe, Kazuo N., Arévalo Galarza, Gustavo Antonio, Aguirre Medina, Juan Francisco
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01-06-2024; Springer Nature B.V
• Subjects: Agriculture; Bioclimatology; Biomedical and Life Sciences; Biometeorology; Generalized linear models; Geographical distribution; Learning algorithms; Life Sciences; Machine learning; Maximum entropy; Plant Genetics and Genomics; Plant Physiology; Plant Sciences; Plant Systematics/Taxonomy/Biogeography; Research Article; Sechium edule; Statistical models; Japan; Temperature; Chayote; TSS; Bioclimatic variables; AUC
• ISSN: 0925-9864; 1573-5109
• DOI: 10.1007/s10722-023-01739-w

Species distribution models identify regions with ideal environmental characteristics for the establishment and proliferation of species. The chayote (
Sechium edule
) is a crop that originated and domesticated in Mexico; however, it is cultivated in different parts of the world due to its nutritional and pharmaceutical importance. The objective of this research was to locate the potential distribution of
S. edule
in Japan supported on seven machine learning models, to also determine which bioclimatic variables influence its distribution, and which are the most suitable regions for its establishment. Thirty-one occurrence points, elevation, and the bioclimatic variables bio1, bio3, bio4, bio7, bio8, bio12, bio14, bio15, and bio17 were used to infer the models. Hundred percent of the occurrence points coincided with the Cfa climate distributed in Acrisol (60.9%), Andosol (17.4%), Cambisol (13%), Fluvisol (4.35%), and Gleysol (4.35%) soil. The maximum entropy model (Maxent) model reported the highest area under the curve (AUC) value (0.93), while the generalized linear model (GLM) obtained the best true skills statistics (TSS) value (0.84); the super vector machine (SVM) model reported the largest suitability area ≥ 0.5 with 100,394.4 km
2
. Temperature-related variables were the major contributors to the models and the ones explaining the distribution limits of
S. edule
in Japan. The coastal eastern prefectures of Kantō, Chūbu, Kinki, Chūgoku, Kyūshū, and Shikoku regions showed a suitability ≥ 0.5.