Botanical Garden-Institute of the Far Eastern Branch of the Russian Academy of Sciences

Geobotany Laboratory

Biogeography Botany Ecology

Head of Laboratory

Krestov, Pavel Vitalyevich

DSc in Biological/biomedical sciences, Associate member of the Russian Academy of Sciences

Publications

Citations

1 364

h-index

Authorization required.

We study the vegetation cover in all its manifestations on a scale from local to global and the factors that determine its development and evolution. We use a whole range of approaches for this: from classical (many geobotanical descriptions - ordination - classification - bioclimatic modeling) to the avant-garde (big data, new computing systems, deep machine learning, neural networks, a lot of Earth remote sensing materials). We try to extract information about the state and development of terrestrial ecosystems from soil samples, age cores, huge sample areas, spectral responses and patterns. Significant results were obtained in the fields of vegetation diversity research, bioclimatic modeling, SDM, reconstruction of tropical cyclone activity based on historical violations of stands, automatic identification of different types of violations and tree species from ultra-high resolution images. Join us.

SDM
Bioclimatic modeling
Dendrochronology
Machine learning
Neural networks
Ordination and classification of vegetation
Geobotanical description

Pavel Krestov 🥼 🤝

Head of Laboratory

Research directions

Modeling the distribution of economically important tree species

Retrospective and predictive models of the distribution of economically significant tree species of temperate and boreal forests of East Asia have been constructed. The species distribution models were developed by the method of deep machine learning (Random Forest) for pessimistic and optimistic scenarios of climate change by 2070, as well as for past climatic epochs – the last glacial maximum (~21 thousand years BC) and the climatic optimum of the Middle Holocene (~7000 years BC).

Machine identification of tree species based on ultra-high resolution remote sensing materials

Neural networks have been created and trained to accurately identify the crowns of Korean pine and whole-leaved fir using ultra-high-resolution remote sensing materials obtained using an affordable unmanned aerial vehicle with an RGB camera. It is shown that this technology is especially relevant in studies of a particularly threatened formation of black fir-cedar-broadleaf forests: in a short period of time, it became possible to establish the number and exact coordinates of whole-leaved fir and Korean pine trees over large areas (Fig. 2). It was found that solving practical problems related to tree recognition requires a multi-stage process including the collaboration of specialists with different skills and experience, the use of a biological and landscape approach in the application of remote sensing methods to improve recognition results.

Reconstruction of the history of the impact of typhoons (tropical cyclones) on the forest ecosystems of the temperate and boreal zones

The transformative effects of tropical cyclones (TC) on boreal forest ecosystems have been revealed for the first time. The analysis of satellite images using a neural network-based multi-stage algorithm for recognizing the species composition of trees and windfall spots caused by shopping malls has opened up the opportunity to understand the mechanisms and assess the scale and speed of changes in forest cover in the boreal zone. It is shown that the increased activity and intensity of TC, which increasingly and further north penetrate into temperate latitudes, will become the main factor in the rapid change of forests with global warming