Open Access Open Access  Restricted Access Subscription Access

Ecological Niche Modelling of an Industrially Important Mushroom - Ganoderma lucidum (Leys.) Karsten: A Machine Learning Global Appraisal


Affiliations
1 ICAR- Central Arid Zone Research Institute, Jodhpur 342 003, Rajasthan, India
2 Jodhpur Institute of Engineering and Technology, Jodhpur 342 802, Rajasthan, India

Species Distribution Modelling (SDM) involves utilizing observations of a given species and its surrounding environment to produce a sound approximation of the species' potential distribution. The intricate relationships between organisms and their surroundings, coupled with the profusion of data, have captured the attention of ecologists and statisticians alike. Consequently, they have directed their efforts towards exploring the potential of machine learning techniques. Our study employs an ensemble machine learning approach to simulate the global ecological niche modelling of Ganoderma lucidum fungus. This involves the utilization of various environmental predictors and the averaging of multiple algorithms to achieve a comprehensive analysis. 563 spatially thinned presence points of G. lucidum were projected with three bio-climatic time frames, namely current, 2050, and 2070, and four Representative Concentration Pathways (RCPs), namely 2.6, 4.5, 6.0, and 8.5, as well as non-climatic variables (surface soil features, land use, rooting depth and water storage capacity at rooting zone). We observed excellent model qualities as the Area Under the receiver operating Curve (AUC) approached 0.90. Random Forest was identified as the best individual algorithm, while the Maxent entropy was identified as the least effective for Ecological Niche Modelling (ENM) of G. lucidum. Globally, under the current bio-climatic and non-bioclimatic projection, optimum habitat for this fungus covers 12510876.3 km2 area while, maximum area (13248546.9 Sq. km.) under this habitat class with future projections was recorded with RCP of 8.5 in 2070. The primary determinants of its current global distribution were ecosystem rooting depth, water storage capacity, and precipitation seasonality. While, with two future bioclimatic time frames and RCPs, Isothermality was identified as the most influential predictor. Based on our assessment, it has been determined that this particular fungus is exhibiting a persistent pattern of proliferation across the regions of Europe, America, and certain areas of India. The present investigation sought to underscore the importance of discerning the native habitats of this species, taking into account both current and anticipated climatic shifts. This knowledge is essential for effectively coordinating the artificial cultivation and natural harvesting of G. lucidum, which is necessary to meet the ever-increasing industrial demands.

Keywords

Bioclimatic variables, Ecosystem rooting depth, Ensemble machine learning, Random forest algorithm, Representative concentration pathways
User
Notifications
Font Size

Abstract Views: 12




  • Ecological Niche Modelling of an Industrially Important Mushroom - Ganoderma lucidum (Leys.) Karsten: A Machine Learning Global Appraisal

Abstract Views: 12  | 

Authors

Manish Mathur
ICAR- Central Arid Zone Research Institute, Jodhpur 342 003, Rajasthan, India
Preet Mathur
Jodhpur Institute of Engineering and Technology, Jodhpur 342 802, Rajasthan, India

Abstract


Species Distribution Modelling (SDM) involves utilizing observations of a given species and its surrounding environment to produce a sound approximation of the species' potential distribution. The intricate relationships between organisms and their surroundings, coupled with the profusion of data, have captured the attention of ecologists and statisticians alike. Consequently, they have directed their efforts towards exploring the potential of machine learning techniques. Our study employs an ensemble machine learning approach to simulate the global ecological niche modelling of Ganoderma lucidum fungus. This involves the utilization of various environmental predictors and the averaging of multiple algorithms to achieve a comprehensive analysis. 563 spatially thinned presence points of G. lucidum were projected with three bio-climatic time frames, namely current, 2050, and 2070, and four Representative Concentration Pathways (RCPs), namely 2.6, 4.5, 6.0, and 8.5, as well as non-climatic variables (surface soil features, land use, rooting depth and water storage capacity at rooting zone). We observed excellent model qualities as the Area Under the receiver operating Curve (AUC) approached 0.90. Random Forest was identified as the best individual algorithm, while the Maxent entropy was identified as the least effective for Ecological Niche Modelling (ENM) of G. lucidum. Globally, under the current bio-climatic and non-bioclimatic projection, optimum habitat for this fungus covers 12510876.3 km2 area while, maximum area (13248546.9 Sq. km.) under this habitat class with future projections was recorded with RCP of 8.5 in 2070. The primary determinants of its current global distribution were ecosystem rooting depth, water storage capacity, and precipitation seasonality. While, with two future bioclimatic time frames and RCPs, Isothermality was identified as the most influential predictor. Based on our assessment, it has been determined that this particular fungus is exhibiting a persistent pattern of proliferation across the regions of Europe, America, and certain areas of India. The present investigation sought to underscore the importance of discerning the native habitats of this species, taking into account both current and anticipated climatic shifts. This knowledge is essential for effectively coordinating the artificial cultivation and natural harvesting of G. lucidum, which is necessary to meet the ever-increasing industrial demands.

Keywords


Bioclimatic variables, Ecosystem rooting depth, Ensemble machine learning, Random forest algorithm, Representative concentration pathways