The expansion of remote sensing data and products necessitates a thorough examination and evaluation of these products for applications across various land resource topics. This importance is heightened for biomes such as forests, which are recognized as one of the most significant biomes providing ecosystem services worldwide. This study examines the accuracy and performance of six global land cover products, including ESA WorldCover, ESRI land use/land cover, Dynamic World, GLC_FCS30, GlobeLand30, and Forest/Non-Forest classification, as well as their fusion with classifications produced for the Hyrcanian forests using satellite data and machine learning algorithms. The Hyrcanian forests were classified using Landsat-8 and Sentinel-2 satellite imagery and k-nearest neighbor, Support Vector Machine, Random Forest, and Classification and Regression Trees algorithms in 2020. A decision-level fusion approach was employed to compare their accuracy with land cover products and fuse them. This fusion approach, using a voting mechanism, aimed to achieve the best result in assessing the ecosystem services of the Hyrcanian forests. The results of this study indicate that the area of the Hyrcanian forests ranges from approximately 1,765,640 to 1,852,033 ha, with the total annual economic value of their ecosystem services estimated between $27.491 billion and $28.836 billion. Iran, with 1,664,178 ha (94%) of these forests, holds the largest share compared to Azerbaijan, which has 101,461 ha (5%). These findings underscore the importance of conserving and sustainably managing this valuable ecosystem. To preserve and enhance these values, the development and implementation of effective conservation programs and management strategies are essential.
