Określenie defoliacji świerka pospolitego na podstawie zobrazowań lotniczych pozyskanych przy pomocy bezzałogowego statku powietrznego

Autorzy

• Szymon Stańczyk Instytut Badawczy Leśnictwa, Laboratorium Ochrony Przeciwpożarowej Lasu
  Sękocin Stary, ul. Braci Leśnej 3
  05-090 Raszyn

Abstrakt

The assessment of tree crown defoliation is a key component of forest health monitoring and forms the basis for reporting under the ICP Forests programme. Currently, direct field assessment remains the standard, but the development and increased availability of unmanned aerial vehicles (UAV’s) offer the possibility of partially replacing field assessment with remote methods. The aim of this study was to assess the suitability of RGB images obtained by drone for determining the degree of defoliation of Norway spruce (Picea abies) crowns in accordance with ICP Forests guidelines and to compare these results with standard field assessment.
The study included spruce trees (≥ 60 years) from two populations: mountain (Szklarska Poręba Forest District) and lowland (Płońsk Forest District). For each tree, two types of photographs were taken: vertical (from above the top, at a height of approx. 5 m) and oblique (from the side, showing the entire crown). The sample included individuals assigned to five categories of crown damage, including dead trees. The photographs were used for remote assessment, and sample images as well as and a diagram of the photography method are presented in Figures 1A/1B–2A/2B. The results were compared with a reference assessment conducted in the field for the same trees, carried out in parallel with the aerial surveys under comparable lighting and phenological conditions.
Imaging was performed using a DJI Mavic 2 Zoom drone. A team of 12 experts (assessors with monitoring experience), who were briefly trained and independently estimated crown defoliation based on 120 images for each type of shot. The assessments were recorded on forms and then compared with the reference assessment. Statistical analyses included both percentage compliance and compliance with damage class assignments (Tables 1 – 2). Operationally, it took about 1–2 minutes to take a complete set of photographs of one tree, saving time compared to standard field inspection.
The results indicate that oblique images more accurately reflect the condition of tree crowns than vertical images. The average agreement between damage classes and field assessments for oblique images was 56% and 46% for vertical images. For oblique shots, the highest agreement was observed in damage category 3 (66%), while for vertical shots it was in category 5. The lowest accuracy was recorded in category 2 (oblique images) and category 1 (vertical images), (Tables 1–2 respectively). Error analysis showed a moderate tendency to overestimate the degree of defoliation (14 cases, average +6.79%) compared to underestimations (10 cases, average −3.84%) (Figures 3–4). Regarding population, the differences between sites indicated a smaller overestimation in the mountain population (2.87%) than in the lowland population (7.47%) (Figures 5–8). Furthermore, 62% of all assessments (for both site types) differed from the reference assessment by ≤ 5 percentage points, confirming the practical usefulness of the method.
The results show that the using BSP in defoliation assessment can significantly support forest health monitoring, especially in stands with dense canopy cover and in areas that are difficult to access. The method is repeatable, operationally fast and feasible in practice, although its use may be limited by compositional factors such as crown overlap and obstructed side visibility in dense stands. Further research should take into account the influence of lighting conditions, flight altitude and configuration.

Słowa kluczowe

• tree stand health
• drones
• dron
• świerk pospolity
• defoliacja
• defoliation
• Norway spruce
• stan zdrowotny