Multitemporal hyperspectral tree species classification in the Białowieża Forest World Heritage site

Multitemporal hyperspectral imaging can significantly enhance the accuracy of tree species classification in temperate forests, with early summer acquisitions achieving the highest precision. A study in the Białowieża Forest World Heritage site demonstrated that while single-date classifications reached overall accuracies of 83–94%, combining multiple time points (early and late summer, autumn) slightly improved results to 84–94%. The research used Support Vector Machines to classify seven species, revealing that tree size influenced classification accuracy for some species like birch and oak. These findings suggest that while multitemporal data can refine species mapping, identifying optimal acquisition windows may still be more efficient given the scarcity of hyperspectral datasets.

Modzelewska, A., Kamińska, A., Fassnacht, F. E., & Stereńczak, K. (2021). Multitemporal hyperspectral tree species classification in the Białowieża Forest World Heritage site. Forestry: An International Journal of Forest Research, 94(3), 464–476. https://doi.org/10.1093/forestry/cpaa048