Volume 18, Issue 3 (1-2025)
مرتع 2025, 18(3): 451-466 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Iranmanesh M, Esmaeilpour Y, Gholami H, Moradi N. Applying Ensemble Modeling for Species Distribution Forecasting of Ferula assa-foetida in Southern Iran. مرتع 2025; 18 (3) :451-466
URL: http://rangelandsrm.ir/article-1-1273-en.html
URL: http://rangelandsrm.ir/article-1-1273-en.html
Department of Natural Resources Engineering, Faculty of Agriculture and Natural Resources, University of Hormozgan, Bandar Abbas
Abstract: (553 Views)
Background and objectives: Changing environmental conditions and climate change significantly affect the geographical distribution of plant species. Accurate weather forecasting is essential for the sustainable production of valuable plant species, crucial for protecting soil and vegetation resources. Given the high costs of field surveys, distribution prediction modeling serves as an effective method for the proper exploitation and conservation of special medicinal plants. This research aims to predict the distribution of Ferula assa-foetida L. using an ensemble modeling approach to manage the habitat of this valuable plant species.
Methodology: Sampling for the presence of F. assa-foetida was conducted during the 2019 growing season using GPS in the study area. Environmental and climate data were obtained from relevant websites and processed using the RUDMM package to eliminate collinearity. From 28 variables, 14 were selected and modeled without collinearity issues. Twelve models (ANN, FDA, CTA, GLM, GAM, GBM, MARS, SARE, RF, XGBOOST, MaxENT, MaxNET) were used in this study. These models were initially evaluated individually based on ROC and TSS curves, then combined for continued modeling with TSS above 0.94. The relative significance of variables was assessed in both single and combined models. Four algorithms (EMca, EMci, EMmean, EMcv) were employed in the ensemble models. The study utilized R programming software and the Biomod2 package.
Results: Detailed examination of single function models indicated that the RF model (Random Forest) had high accuracy, and the combined models with TSS above 0.94 provided the basis for this research. The relative significance of variables showed that bio19, bio18, bio4, and earth-clay were the most important for modeling. Bio19, representing the impact of the coldest season rainfall, contributed 40%, with minimal impact from the slope direction (0.06%). Precipitation's total contribution rate was 13.1%, temperature's total contribution rate was 2.7%, NDVI (vegetation index) total contribution rate was 0.4%, soil change total contribution rate was 4.5%, and topography's (SRTM2) total contribution rate was 0%. The habitat suitability analysis provided clear maps from four combined model algorithms, showing high suitability for species growth in high and mountainous areas, specifically in northern Hormozgan and southern Fars provinces, and Lar city.
Conclusion: The consensus models outperformed single models in representing variable contributions, response curves, and habitat suitability for F. assa-foetida. Ensemble modeling produced high-resolution maps predicting species distribution probabilities. The findings are crucial for the conservation, introduction, and optimal use of F. assa-foetida in the study area, addressing gaps in previous research.
Methodology: Sampling for the presence of F. assa-foetida was conducted during the 2019 growing season using GPS in the study area. Environmental and climate data were obtained from relevant websites and processed using the RUDMM package to eliminate collinearity. From 28 variables, 14 were selected and modeled without collinearity issues. Twelve models (ANN, FDA, CTA, GLM, GAM, GBM, MARS, SARE, RF, XGBOOST, MaxENT, MaxNET) were used in this study. These models were initially evaluated individually based on ROC and TSS curves, then combined for continued modeling with TSS above 0.94. The relative significance of variables was assessed in both single and combined models. Four algorithms (EMca, EMci, EMmean, EMcv) were employed in the ensemble models. The study utilized R programming software and the Biomod2 package.
Results: Detailed examination of single function models indicated that the RF model (Random Forest) had high accuracy, and the combined models with TSS above 0.94 provided the basis for this research. The relative significance of variables showed that bio19, bio18, bio4, and earth-clay were the most important for modeling. Bio19, representing the impact of the coldest season rainfall, contributed 40%, with minimal impact from the slope direction (0.06%). Precipitation's total contribution rate was 13.1%, temperature's total contribution rate was 2.7%, NDVI (vegetation index) total contribution rate was 0.4%, soil change total contribution rate was 4.5%, and topography's (SRTM2) total contribution rate was 0%. The habitat suitability analysis provided clear maps from four combined model algorithms, showing high suitability for species growth in high and mountainous areas, specifically in northern Hormozgan and southern Fars provinces, and Lar city.
Conclusion: The consensus models outperformed single models in representing variable contributions, response curves, and habitat suitability for F. assa-foetida. Ensemble modeling produced high-resolution maps predicting species distribution probabilities. The findings are crucial for the conservation, introduction, and optimal use of F. assa-foetida in the study area, addressing gaps in previous research.
Type of Study: Research |
Subject:
Special
Received: 2024/07/10 | Accepted: 2024/08/31 | Published: 2025/01/29
Received: 2024/07/10 | Accepted: 2024/08/31 | Published: 2025/01/29
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
