Volume 13, Issue 2 (10-2025)
Ecol Iran For 2025, 13(2): 143-157 |
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:
Zandi S, Bonyad A E, Hasanzad Navrodi I, Alavi S J, Ghorbanzadeh N. (2025). Estimation of Canopy Cover and Ground Biomass in Afforestation Areas of Western Guilan using Remote Sensing and Geostatistical Methods (Study Area: Pilembera). Ecol Iran For. 13(2), 143-157. doi:10.61882/ifej.2025.578
URL: http://ifej.sanru.ac.ir/article-1-578-en.html
URL: http://ifej.sanru.ac.ir/article-1-578-en.html
Solmaz Zandi *1 
, Amir Eslam Bonyad2, Iraj Hasanzad Navrodi2, Seyed Jalil Alavi3, Neda Ghorbanzadeh1
, Amir Eslam Bonyad2, Iraj Hasanzad Navrodi2, Seyed Jalil Alavi3, Neda Ghorbanzadeh1
1- Faculty of Natural Resources of Soumesara, Guilan University, Guilan, Iran
2- Department of Forestry, Faculty of Natural Resources, Soumesara, Guilan University, Guilan, Iran
3- Department of Forestry, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Mazandaran, Iran
2- Department of Forestry, Faculty of Natural Resources, Soumesara, Guilan University, Guilan, Iran
3- Department of Forestry, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Mazandaran, Iran
Abstract: (793 Views)
Extended Abstract
Background: The reduction of natural forests due to various factors has made afforestation particularly important, aiming to increase forest area and wood production. Estimating forest characteristics using remote sensing data has been proposed as a tool for sustainable natural resource management. Remote sensing and geostatistical methods can be useful and cost-effective in estimating canopy cover and aboveground biomass of forests for management and conservation purposes. In the Hyrcanian forests, canopy structure and biomass are considered important criteria for studying and monitoring changes in the forest. This study aimed to estimate and examine the canopy cover and aboveground biomass of the afforested areas of Pilemebra, located in Guilan Province, using Landsat 9 satellite imagery, remote sensing methods, and geostatistics.
Methods: To implement ground sample plots (in summer 2023), the plot grid was randomly and systematically selected with a distance of 300 by 300 meters. A total of 120 sample plots were collected. The data obtained from field surveys and statistical measurements were entered into Excel to perform calculations related to canopy cover and aboveground biomass. Additionally, vegetation indices (including NDVI, EVI, DVI, GNDVI, RVI, ARVI, SAVI, OSAVI, and IPVI) were calculated using Landsat 9 images, and then the corresponding values for each sample plot were extracted from each of the indices. Correlation and regression analyses were conducted on these data. The Information Gain Ratio (IGR) technique and Average Merit index were used to evaluate the predictive power of factors affecting canopy cover and biomass in the afforestation areas of the Pilmabra region. The correlation between canopy cover and aboveground biomass in the ground sample plots and their corresponding spectral values in satellite data was examined using SPSS software and Pearson’s correlation analysis. Inverse Distance Weighting (IDW), ordinary kriging, and radial basis functions (RBF) were employed for estimating the percentage of canopy cover and aboveground biomass. Furthermore, Random Forest (RF) regression models and support vector machines were used for variable selection and estimating the percentage of canopy cover and aboveground biomass in the study area, which were analyzed and evaluated using Weka software. In this study, 80% (96 sample plots) of a set of 120 existing sample plots were used as training datasets. The ROC curve was utilized to assess the performance of geostatistical methods and remote sensing in modeling.
Results: The results of IGR and the Mean Competence Index (MA) indicated that the highest significant correlation was observed between the percentage of canopy cover and the variables OSAVI (0.918) and NDVI (0.915), as well as the highest significant correlation between aboveground biomass (tons per hectare) and the variables SAVI (0.913) and NDVI (0.902). The presence or absence of spatial dependence of the canopy cover percentage and biomass, after fitting several semivariogram models based on the residual sum of squares (RSS) and the coefficient of determination (R²), showed that the exponential model for aboveground biomass (R² = 0.64) and the spherical model for canopy cover (R² = 0.76) were selected as the most suitable models. The validation results of the canopy cover percentage and biomass using RMSE and bias criteria indicated that the RF algorithm and Support Vector Machine (SVM) had a better status and validity than the geostatistical approach due to having the lowest mean squared error and bias values. Additionally, both the RF and SVM models performed well; however, the RF model (92%) had a higher accuracy and sensitivity index in predicting the percentage of canopy cover than the SVM model (83%). The results of this study showed a significant relationship and correlation between the percentage of canopy cover and biomass (tons per hectare) with over 80% of the studied variables. The ROC curve was used to evaluate the performance of geostatistical and remote sensing methods in modeling. In this study, the AUC values based on machine learning models for canopy cover and biomass (RF = 0.94, SVM = 0.93) outperformed geostatistical methods for canopy cover (0.71) and biomass (0.76) in terms of predicting and modeling these indices. Furthermore, there was no significant difference at the 99% and 95% probability levels between the estimated values of canopy cover and aboveground biomass and the actual values.
Conclusion: Awareness of changes in the canopy of vegetation is very important for application programs, resource management, and the assessment of environmental services. Based on the obtained results, it can be stated that the information derived from the main bands and spectral indices played a significant role in estimating canopy cover and aboveground biomass. Furthermore, the effectiveness of the models used in this study may indicate the selection of predictive mapping techniques for modeling forest aboveground biomass. The ability of random forests to estimate the forest variables mentioned in this research could pave the way for future studies.
Background: The reduction of natural forests due to various factors has made afforestation particularly important, aiming to increase forest area and wood production. Estimating forest characteristics using remote sensing data has been proposed as a tool for sustainable natural resource management. Remote sensing and geostatistical methods can be useful and cost-effective in estimating canopy cover and aboveground biomass of forests for management and conservation purposes. In the Hyrcanian forests, canopy structure and biomass are considered important criteria for studying and monitoring changes in the forest. This study aimed to estimate and examine the canopy cover and aboveground biomass of the afforested areas of Pilemebra, located in Guilan Province, using Landsat 9 satellite imagery, remote sensing methods, and geostatistics.
Methods: To implement ground sample plots (in summer 2023), the plot grid was randomly and systematically selected with a distance of 300 by 300 meters. A total of 120 sample plots were collected. The data obtained from field surveys and statistical measurements were entered into Excel to perform calculations related to canopy cover and aboveground biomass. Additionally, vegetation indices (including NDVI, EVI, DVI, GNDVI, RVI, ARVI, SAVI, OSAVI, and IPVI) were calculated using Landsat 9 images, and then the corresponding values for each sample plot were extracted from each of the indices. Correlation and regression analyses were conducted on these data. The Information Gain Ratio (IGR) technique and Average Merit index were used to evaluate the predictive power of factors affecting canopy cover and biomass in the afforestation areas of the Pilmabra region. The correlation between canopy cover and aboveground biomass in the ground sample plots and their corresponding spectral values in satellite data was examined using SPSS software and Pearson’s correlation analysis. Inverse Distance Weighting (IDW), ordinary kriging, and radial basis functions (RBF) were employed for estimating the percentage of canopy cover and aboveground biomass. Furthermore, Random Forest (RF) regression models and support vector machines were used for variable selection and estimating the percentage of canopy cover and aboveground biomass in the study area, which were analyzed and evaluated using Weka software. In this study, 80% (96 sample plots) of a set of 120 existing sample plots were used as training datasets. The ROC curve was utilized to assess the performance of geostatistical methods and remote sensing in modeling.
Results: The results of IGR and the Mean Competence Index (MA) indicated that the highest significant correlation was observed between the percentage of canopy cover and the variables OSAVI (0.918) and NDVI (0.915), as well as the highest significant correlation between aboveground biomass (tons per hectare) and the variables SAVI (0.913) and NDVI (0.902). The presence or absence of spatial dependence of the canopy cover percentage and biomass, after fitting several semivariogram models based on the residual sum of squares (RSS) and the coefficient of determination (R²), showed that the exponential model for aboveground biomass (R² = 0.64) and the spherical model for canopy cover (R² = 0.76) were selected as the most suitable models. The validation results of the canopy cover percentage and biomass using RMSE and bias criteria indicated that the RF algorithm and Support Vector Machine (SVM) had a better status and validity than the geostatistical approach due to having the lowest mean squared error and bias values. Additionally, both the RF and SVM models performed well; however, the RF model (92%) had a higher accuracy and sensitivity index in predicting the percentage of canopy cover than the SVM model (83%). The results of this study showed a significant relationship and correlation between the percentage of canopy cover and biomass (tons per hectare) with over 80% of the studied variables. The ROC curve was used to evaluate the performance of geostatistical and remote sensing methods in modeling. In this study, the AUC values based on machine learning models for canopy cover and biomass (RF = 0.94, SVM = 0.93) outperformed geostatistical methods for canopy cover (0.71) and biomass (0.76) in terms of predicting and modeling these indices. Furthermore, there was no significant difference at the 99% and 95% probability levels between the estimated values of canopy cover and aboveground biomass and the actual values.
Conclusion: Awareness of changes in the canopy of vegetation is very important for application programs, resource management, and the assessment of environmental services. Based on the obtained results, it can be stated that the information derived from the main bands and spectral indices played a significant role in estimating canopy cover and aboveground biomass. Furthermore, the effectiveness of the models used in this study may indicate the selection of predictive mapping techniques for modeling forest aboveground biomass. The ability of random forests to estimate the forest variables mentioned in this research could pave the way for future studies.
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
