ИСТИНА |
Войти в систему Регистрация |
|
ИСТИНА ЦЭМИ РАН |
||
The Russian breeding population of the Barnacle goose increased from 70 thousand in 1980 to 1.2 million individuals in 2015. The population growth coincided with the expansion of the breeding range to the southwest, as well as an increase in the diversity of nesting habitats used. New nesting colonies appeared on the Yugorsky Peninsula, Kolguev Island, Kanin Peninsula, Dolgiy and Golets Islands, as well as along the Barents Sea coast. Barnacle geese began to nest on coastal marshes and sandbars, later colonies began to appear on steep slopes under the protection of nesting Peregrine falcons (Falco peregrinus) and Rough-legged buzzards (Buteo lagopus), as well as in sedge-moss bogs further from the coast. This was accompanied by a rapid increase in the size of nesting colonies, as can be observed on Kolguev Island, where the colony in the Peschanka River delta had grown from a couple of hundred breeding pairs in 1994 to 70 thousand pairs in 2019. This is now the largest known Barnacle goose colony in the world. Such a rapid growth of the colony size was facilitated by the unique conditions on the island, where, due to the absence of rodents and the relatively stable pressure of predators, the nesting success of Barnacle geese is exceptionally high (more than 90% in some years). Estimating the abundance of such a large colony is associated with both methodological and resource difficulties. The first attempts to estimate the abundance of the colony were made in 2006 by means of counting nests on transects and counting plots in different parts of the colony. Counts were repeated in 2012 and 2019, also by the method of transects and plots. In 2022, in addition to calculating the nesting density of Barnacle geese on the counting plots, we applied new methods for colony mapping using unmanned aerial vehicles (UAVs). A section of the Peschanka River delta with an area of about 10 km2 was surveyed. The survey was conducted using DJI quadcopters from a height of 40 m above the ground. Machine learning methods were used to process the survey materials. The YOLO (You only look once) algorithm version 5, which combines object classification and identification, was used for searching for Barnacle goose nests in the images. More than 800 manually labeled examples of target objects were used to train the algorithm. When implementing the method, the PyTorch software platform in the Python programming language was used. To verify the recognition of Barnacle goose nests in the images, the counting plots were surveyed, where all nests had been mapped. The results of this work made it possible to obtain data on the distribution of Barnacle goose nests in different habitats, which will allow to extrapolate the obtained nesting densities to similar habitats in other parts of the island and estimate the total number of the breeding population of Barnacle goose on Kolguev Island. The collection and processing of field material was supported by the Russian Science Foundation grant No. 22-17-00168, https://rscf.ru/project/22-17-00168/.