بررسی پتانسیل تصاویر سنتینل-2 در برآورد زی‌توده جنگلی

Introduction
land use/cover Change, especially the destruction of forest lands, is one of the most important factors in increasing atmospheric carbon emissions, resulting in global warming and climate change. On the other hand, forest ecosystems play an important role in absorbing and maintaining the global carbon balance. Because forest biomass constitutes about 90% of the biomass of the earth's vegetation and is considered an important parameter for evaluating the amount of carbon absorbed by the forest. Therefore, in the current situation where climate change and global warming have caused many environmental crises, evaluating the functioning of the forest ecosystem requires an accurate estimation of biomass and its changes, which should be done with minimum cost and time and without destruction. The total biomass includes Aboveground (trees, shrubs, etc.) and underground (living roots, etc.) Biomass research is concentrated in the Aboveground biomass sector. Various methods have been provided to estimate it. The most accurate method is based on ground measurements, but the destruction and economic costs of ground measurements are very high and it is not suitable for large-scale measurements. Remote sensing-based measurement methods, in addition to having the mentioned advantages, also make suitable estimates of forest Aboveground biomass possible. Because, the methods based on remote sensing, only require a small number of samples in the forest for ground measurement and include the use of various sensors, images, data, and different processing algorithms, the results have acceptable. Biomass prediction models based on satellite images can be obtained from radar data, multi-spectral bands, and vegetation indices (for example, Normalized Difference Vegetation Index (NDVI) and variables Vegetation biophysics such as leaf area index). These models can be developed with or without secondary thematic data (e.g. height). Therefore, this study aims to estimate the aboveground biomass of part of Ardabil Fandoghlo forests using vegetation indices and regression models and to evaluate the efficiency of Sentinel-2 images in estimating the aboveground biomass of the forest.
Methodology
First, by conducting a forest tour, the characteristics of the mass in terms of area, topography, homogeneity and heterogeneity, and density were investigated. Then, the dimensions of the statistical network were determined to be 150 x 100 meters and the dimensions of the sample pieces were 20 x 20 meters. Taking into consideration the above-mentioned matters and to calculate the amount of biomass, 14 square samples with dimensions of 20 x 20 meters were established in a random-systematic way using a GPS device. Then, the information about the trees, including the species, diameter using a caliper and diameter measuring tape, and height using Suunto, were collected and entered into statistical forms and then into Excel. In the next step, due to the branching of the trees, the diameter weight of each group was obtained as the square root of the sum of the square of the diameter at the height of half a meter of all the trees of each group. Finally, the amount of aboveground biomass (AGB) was calculated in terms of tons per hectare. In order to carry out this research, the cloud-free image of the Sentinel-2 satellite was prepared on June 2019. QGIS 3.10, Arc GIS 10.3, SPSS, and Google Earth software were used to conduct this research. Then, to ensure the quality of the data and bands, the image used in this research was corrected for atmospheric errors using the Dark Object Subtraction (DOS) method in the QGIS 3.10 software environment, and then AVI, NDVI, DVI, SI, RVI, IPVI, SAVI, and BI indicators were calculated. A linear regression model was used to investigate the relationship between the calculated amount of biomass and the value of each sample plot in the above indices. So that the biomass measured in each plot with ground data as the dependent variable (Y) and the numbers extracted from the calculated indices as the independent variable (X) were entered into the model by means of the linear regression model. The amount of biomass corresponding to each of the indicators was obtained. In the following, in order to evaluate the accuracy between the measured and estimated biomass values, the coefficient of determination (R2) and the root-mean-square-error (RMSE) was used. It should be noted that with 80% of the data, modeling and Accuracy assessment were done with 20%. Finally, after evaluating the accuracy, the biomass map of the region was prepared using the SAVI index.

Conclusion
The results showed that the SAVI index with a coefficient of determination of 0.78 and a root-mean-square-error of 2.45 compared to other indices calculated in this study is more accurate in estimating aboveground biomass. For this reason, the biomass map of the study area was prepared using the SAVI index. Also, the amount of biomass in the area was estimated at 132.443 tons per hectare. The results also showed that the linear regression model is one of the common models for estimating biomass using ground data and the value of vegetation indices extracted from satellite images. Because it uses more limited data and less time than other models to provide more accurate results. It is still an important tool for estimating forest aboveground biomass. In addition, the high potential of remote sensing methods in estimating forest biomass with less time and cost was confirmed in this study. However, only a limited number of these methods are useful due to their high correlation with biomass content. It can be said that the role of bands and spectral textures in modeling aboveground biomass depends on the complexity of the forest structure. Forest aboveground biomass is one the essential data to evaluate the role of carbon storage in studies related to climate change and global warming. Therefore, measuring and estimating forest aboveground biomass using different methods based on remote sensing is possible with the least cost and time and is also without degradation. Finally, based on the results of this study, it can be said that Sentinel-2 images have acceptable accuracy in estimating aboveground biomass of forest ecosystems and also vegetation indices such as the SAVI index, which consider soil coefficients, are more accurate than Indicators that do not take these coefficients into account.