В статье рассматривается технология оценки качества материалов таксации леса на основе анализа связей таксационных показателей насаждений с их спектрально-отражательными характеристиками. При проведении исследований используются данные спутниковой съемки Sentinel-2 в видимом и инфракрасном диапазоне спектра с пространственным разрешением 10 м. The article considers the technology of assessing the quality of forest taxation materials based on the analysis of the relationship between the taxation indicators of plantations and their spectral-reflective characteristics. The research uses Sentinel-2 satellite imagery data in the visible and infrared spectral ranges with a spatial resolution of 10 m.

This paper presents studies on using Sentinel-1 imagery data to determine the attributes of mixed forest stands. The fieldwork is carried out in Kostroma, Vologda, and Arkhangelsk oblasts and the Udmurt Republic. The study reveals that quantitative and qualitative forest characteristics correlate with radar survey parameters; the value of this correlation is identified. The results make it possible to zone a study area according to the standing volume and forest density.

Polygraphus proximus Blandford is a bark beetle that has recently invaded the dark coniferous forests of Southern Siberia. Over the last decade, this four-eyed fir bark beetle has destroyed Abies sibirica Ledeb stands on more than 500 thousand hectares. This study considers the initial stages of natural reforestation, which can define the restorative succession scenario for completely dead forest stands after the outbreak fades. Areas disturbed by outbreaks have a strong potential for natural coniferous regeneration when the young generation of Abies sibirica dominates in species composition. As early as 5 years after the outbreak, a successful, sufficient for the future forest formation, young generation density (13,000 trees/ha) is observed. Intensive undergrowth development is noted. The high density of young stands in some areas ensures the canopy closure and forms a shady ground cover vegetation type. At the same time, in some disturbed areas there are threats to undergrowth. In open places with thin undergrowth, a considerable amount of forest fuels represented by large wood residues and dead grass accumulates and can cause high-intensity fire emergence and development. Another risk is human activity. In former outbreak foci, where clear-cutting was implemented, healthy trees were cut down and the existing young coniferous generation was destroyed. To date, clearings have been overgrown with secondary species (Betula pendula Roth., Populus tremula L.). Ground cover communities have transformed into open-type phytocenosis where cereal grasses prevail. This has led to turfing and significant litter and dry grass accumulation. To sum up, in faded Polygraphus proximus outbreak areas, strong young forest generation ensures reforestation without dominant species change. To reduce risks of species change and conserve valuable Abies sibirica species, it is necessary to strengthen wildfire protection.

Implementing inventory in remote and hard-to-reach forests is rather challenging. This study develops methods for identifying the qualitative and quantitative characteristics of mixed forests using Sentinel-1 imagery. The relationships between standing volume, forest density, age, and number of trees and radar survey indicators are revealed. The results are compared with forest inventory materials for a part of the study area. For most stands, there is a similarity in the standing volume and forest density definition. The study demonstrates that it is possible to identify quantitative and qualitative forest characteristics using radar survey data.

В статье рассматривается применение нейросетей глубокого обучения для проверки таксационных показателей лесов, полученных при проведении лесоустройства. Авторами разработана технология верификации данных лесоустройства, основанная на использовании оптических спутниковых снимков высокого разрешения, обработка которых происходит посредством нейросетей. В качестве исходных данных взяты снимки территории Меленковского участкового лесничества Владимирской области с аппаратов Sentinel-2, проведена атмосферная коррекция с помощью алгоритма Sen2Cor. Автоматизация процесса лесотаксационного дешифрирования снимков выполнена в программном обеспечении «Геотрон. Лес-валидация» – специализированном геоинформационном ресурсе, разработанном сотрудниками НЦ ОМЗ, ФБУ ВНИИЛМ и ФГБУ«Рослесинфорг». Для обучения нейросетевых моделей использовались библиотеки TensorFlow и Keras, обучающая выборка основана на средних значениях пикселей в границах каждого выдела для каждого канала. После обучения нейросеть определяет таксационные показатели для каждого выдела, а также выдает разницу между данными из лесоустроительной документации и рассчитанными с помощью нейросети, что позволяет выявить отклонения по запасу, полноте, возрасту насаждений. Работа модели продемонстрировала, что значительная часть неточных данных по лесоустройству характерна для сложных насаждений с несколькими породами в составе, а также для редкостойных приспевающих и спелых древостоев с различным породным составом первого яруса. Результат проведенного эксперимента показал, что точность обучения и работы нейросети напрямую зависит от объема данных, в частности, количества выделов, по которым происходит обучение. Приемлемые результаты получаются при количестве выделов, превышающем 5000 штук. Исходя из этого, разработанная технология позволяет получить хорошие результаты по валидации результатов лесоустройства на территории, превышающей площадь 10 тысяч гектаров, что сравнимо со средней площадью одного участкового лесничества в центральной части России. The article discusses the use of deep learning neural networks to verify forest taxation indicators obtained during forest management. The authors have designed a technology for verifying forest management data based on the high-resolution optical satellite images processed applying neural networks. The initial data were images of the Melenkovskoye forestry district in the Vladimir region taken from Sentinel-2 satellites; atmospheric correction was performed with the Sen2Cor algorithm. The process of forest taxation interpretation of images was automated with the «Geotron. Forest Validation» software, a specialized geoinformation resource developed by the NC OMZ, FBU VNIILM and FGBU “Roslesinforg”. The TensorFlow and Keras libraries were used to train the neural network models; the training sample was based on the average pixel values within the boundaries of each plot for every channel. After training, the neural network determined forest characteristics for each section, and also gave the difference between the forest management documentation data and those calculated by the neural network; this allowed identifying deviations in the stock volume, forest density, and stands age. The model's operation demonstrated that a significant portion of inaccurate forest management data is typical for complex stands with several species in the composition, as well as for sparse maturing and mature stands with different species composition of the first tier. The study result showed that the accuracy of neural network training and operating directly depends on the of data amount, in particular, the number of sections for which training occurs. Acceptable results are obtained with a sections number exceeding 5,000 pieces. Considering this, the developed technology allows obtaining good results in validating forest management data on an area exceeding 10 thousand hectares, which is comparable to the average area of one forestry district in the central part of Russia.