بررسی قابلیت داده های ماهوارههای لندست 8 و سنتینل 2 در برآورد مشخصه های کمّی توده در جنگل های پهنبرگ استان گلستان

Introduction: Obtaining timely information on the qualitative and quantitative characteristics of forests is useful in ecosystem management, designing management and protection plans. Considering the role of Hyrcanian forests in maintaining biological diversity, adjusting climate, environmental and economic values, and protecting water and soil, and the high cost and time-consuming nature of measuring the quantitative characteristics of the forests through field methods, remote sensing application and modeling provide a lot of information in this field so that there is less need to do field work. The purpose of this research is to investigate the capability of Landsat-8 and Sentinel-2 satellites in estimating quantitative characteristics of stands in broadleaved forests of Golestan province. Materials and methods: In this study, 230 circular sample plots with an area of 1000 square meters were measured using a systematic sampling method and a 100x100 meter grid in five sites of Kordkoy, Shasat Kalateh, Zarrin Gol, Sukhdari and Loveh. The geographic center of each plot was recorded using a Differential Global Positioning System (DGPS) device. In each plot the species, the DBH was more than 12.5 cm and the height of the trees were measured. Then the number of trees density (n. ha^-1), volume (m³. ha^-1), basal area (m². ha^-1) located in each sample plot were calculated. After processing the images and creating plant indices (NDVI ،RVI ،EVI ، GNDVI ، PCA ، SAVI ، IPVI and DVI), the numerical values corresponding to the ground sample plots were extracted from the spectral bands. Uncertainty analysis of the results was also carried out using the Monte Carlo method. Data modeling was carried out using two methods: multiple linear regression and the Random Forest algorithm .In this study, a total of 230 sample plots were used, with 175 plots (75%) allocated for model development, and the remaining 55 plots (25%) employed for validation purposes within the data mining algorithms. Results: The results of estimating the quantitative characteristics of basal area (m². ha^-1) and volume (m³. ha^-1) using linear multivariate regression showed that the percentage of root mean square error was 50.72 and 47.06 using Landsat-8 data, respectively and 48.66 and 45/89 using Sentinel- 2 data. The results also showed that the root mean square error percentage for the characteristics of basal area (m². ha^-1) and volume (m³. ha^-1) are 44.53 and 41.28 percent with Using Landsat-8 data, respectively, and 44.21 and 39.66 using Sentinel-2 data and random forest algorithm. Conclusion: In general, the results of the study indicated that the Random Forest algorithm, combined with Sentinel-2 data, provided approximately 6 to 11 percent better estimates of basal area (m². ha^-1) and volume (m³. ha^-1) quantitative characteristics compared to the linear regression method using Landsat 8 data. Furthermore, the uncertainty analysis showed that the mean estimated values of basal area and volume were within the confidence interval (2.5–97.5%), suggesting that the chosen model was appropriate. In summary, the approach used in this study demonstrated a moderate capability in estimating quantitative basal area and volume of the forest.