Early detection of Ips typographus infestations by using Sentinel-2 satellite images in windthrow affected Norway spruce forests in Smolyan region, Bulgaria

Strong winds uprooted more than 100 thousand m 3 of coniferous trees in natural forest stands nearby the town of Smolyan (the Western Rhodopes) in January 2018. Although damaged trees were quickly removed from the stands, the European spruce bark beetle ( Ips typographus ) attacked the healthy Norway spruce trees near the windthrow areas in August 2020. Our hypothesis was that the trees were infested by the pest in previous years when no symptoms of attacks were observed. This study was conducted in three spruce stands, located near the windthrow areas and attacked by I. typogra-phus , and in three control (healthy) stands located 5-10 km from the affected areas. We used satellite images captured by Sentinel-2 in September 2017-2020. It was established that in September 2017 (a year before the windthrow), the mean values of the Normalized difference vegetation index (NDVI) in the attacked stands were concentrated between 0.60 and 0.75 (with a maximum at 0.70), indicating that the trees were in good health. During the period 2018-2020 the distribution of mean values of NDVI was stretched between 0.35 and 0.75, which is an indication of evidence of pest attacks on the individual trees. The detail comparison of pixel values of the NDVI in the attacked and control sample plots was made on the base of images captured on 27.06.2020. The mean NDVI values in the three control plots (0.74-0.79) were much higher than the mean values in the sample plots attacked by the pest (0.57-0.65). These results showed that the values of NDVI based on satellite remote sensing data


Introduction
The State Forestry Smolyan is located in the Western Rhodopes at a territory of 28368.3 ha with mountainous landscape, different climatic characteristics and plant species biodiversity. Coniferous stands occupy 75.4% of the forested area with main tree species Norway spruce (Picea abies (L.) Karst.) -35%, Scots pine (Pinus sylvestris L.) -27%, Austrian pine (Pinus nigra Arnold) -8%, and European silver fir (Abies alba Mill.) -5%.
Natural disturbances caused by abiotic factors periodically occur in mountain forest ecosystems. Severe damages from wet snow were recorded in Scots pine plantations in the region of Smolyan in 2015. In 2018 strong winds brought down more than 100 thousand m 3 of coniferous trees in natural forest stands nearby the town . The damaged trees were removed in less than one year from the affected stands. Nevertheless, the presence of fresh food substrate (stumps and stem residues) contributed to the multiplication of the European spruce bark beetle (Ips typographus Linnaeus, 1758) (Coleoptera: Curculionidae, Scolytinae) attacking healthy Norway spruce trees near the windthrow area. The first visible symptoms of the pest infestations were observed in 2020.
The application of remote sensing data based on images captured by satellites or unmanned aerial vehicles (UAV) has consistently increased for assessment the health status of forest ecosystems. Despite the appearance of new equipment and methods, satellite-borne multispectral sensors are the most commonly used technology for monitoring and assessment of the forest health condition (Torres et al., 2021). An integrated approach including the use of RGB orthophotos, multispectral data and terrestrial verification has been successfully applied in Bulgaria for mapping the bark beetle spots of Ips typographus (Dimitrov et al., 2019; and of Ips acuminatus (Gyllenhal, 1827) .
The aim of this study was to investigate the possibility of using the satellite images for early detection of Ips typographus infestations in Norway spruce stands around the windthrow areas in Smolyan region.

Material and Methods
The study was conducted in three spruce stands infested by Ips typographus near the windthrow areas (SP), and in three control (healthy) stands (CP) with similar structural characteristics, located 5-10 km from the damaged areas. The main characteristics of the sample plots are shown in Table 1. Satellite images captured by Sentinel-2 of the European Space Agency in September 2017-2020 were used to assess the earliest infestation of the pest in the studied spruce stands. Sentinel-2 has a multispectral instrument of a passive type, recording sunlight reflected from the Earth's surface with a spatial resolution of 10 m for visible and near-infrared bands, and 20 m and 60 m for short wave infrared bands. Imaging is performed in 13 spectral bands in the visible and near-infrared part of the electromagnetic spectrum (VNIR from 400 to 1100 nm) and short-wave infrared part (SWIR from 1100 to 3000 nm). The recording frequency is 5 days for the equator and 3 days for the other latitudes. The data is freely accessible. The images were acquired using the application Earth Explorer of US Geological Survey. Raster image processing was performed with QGIS-3 -a professional open source GIS application.
Pre-processing of the used images was carried out in order to remove the influence of water vapor and other particles located in the atmospheric layers above the studied objects. Shape files of the surveyed spruce stands were created.
Based on the spectral characteristics of the registered multispectral images, the Normalized difference vegetation index (NDVI) was calculated, which is the main indicator of the health status and vitality of the vegetation. NDVI is a mathematical combination between the red band (B4) and the nearinfrared (NIR) band (B8) that produces values between -1 to +1, according to the following expression: where B 8 is a spectral band of Sentinel-2 with a wavelength of 665 nm; B 4 -with a wavelength of 842 nm.
The comparison of pixel values of NDVI in the attacked and control plots was made by satellite images captured on 27.06.2020. Extraction of the investigated raster images was performed with subsequent raster analysis of all pixel values in each raster layer.
Statistical evaluation of NDVI values was performed using descriptive statistics of Statistica 12.0 for Windows (StatSoft). The results were cited together with their standard error. Ttest for independent samples was applied to compare the means.
The terrain verification of Ips typographus attacks was performed in the sample plots on 06.08.2020.

Results
The first visible infestations by Ips typographus in the healthy Norway spruce stands near the windthrow sites were observed in late July -early August 2020. The main symptoms of the pest development were observed: discoloration of tree crowns (Fig.  1A) and the preliminary fall of needles on the ground. The development of I. typographus's second generation under the bark of stems was also established (Fig. 1B).
The presence of I. typographus infestations in the spruce stands adjacent to the windthrow in August 2020 was a reason to hypothesize that the same trees were infested by the pest in previous years, when no symptoms of attacks were observed.
In September 2017 (before the windthrow), the mean values of NDVI in attacked stands were concentrated between 0.60 and 0.75 (with a maximum of 0.70), indicating good vitality of spruce stands (Fig. 2). In next years (September, 2018-2020), the main distribution of mean values of NDVI was between 0.35 and 0.75), which is an indication of pest attacks on individual trees.
The NDVI models of the studied areas on the base of satellite images captured on 27.06.2020 (before the appearance of visible symptoms of I. typographus attack) showed that in the control plots that were away from the windthrow the spruce stands were in good health status (Fig. 3 A-C), in contrast to the sample plots close to the damaged areas (Fig. 3 D-F). The mean NDVI values in the three control plots (CP1=0.74±0.02; CP2 and CP3=0.79±0.03) were much higher than the mean values in the sample plots with pest attacks (SP1=0.57±0.09; SP2=0.65±0.05; SP3=0.59±0.05) (Fig. 4).
The differences in the mean NDVI values from three control plots without I. typographus attacks (CP1 -CP3) and the sample plots attacked by the pest (SP1 -SP3) were statistically proven (p<0.001).

Discussion
Ips typographus is one of the most destructive xylophagous insect pests of Norway spruce species in Europe (Kaminska et al., 2021). Its outbreaks are usually caused by disturbances or extreme weather conditions (Wermelinger, 2004). Several outbreaks of the pest were recorded in the 20 th century in various parts of Europe where over 100 million m³ of spruce trees were killed (CAB International, 2021).
In Bulgaria, I. typographus does not usually cause strong damage to the managed spruce forests because of the intensive sanitary and control measures. Conversely, in protected areas, the pest develops on fallen and broken fresh trees, increases in number and attacks healthy stands. In 2001, a tornado affected mature spruce stands on 62 ha in Bistrishko Branishte Biosphere Reserve in Vitosha Mountain, after which I. typographus population first developed in windthrown timber, and from 2003 to 2005, spruce trees on over 200 ha were killed by the pest near the windthrow in the reserve (Rossnev et al., 2005).
In the Smolyan region, it was found that the use of NDVI from multispectral UAV and satellite data is not suitable for long-term monitoring of windfall forest areas due to the overgrowth of damaged stands with grass and shrub vegetation .
It is well known that the symptoms of the advanced stages of bark beetle infestation (i.e. red attack), can be observed using the visible part of the electromagnetic spectrum (400 to 700 nm), but the detection of early stage symptoms (i.e. green attack) is more effective using sensors in near-infrared (700 to 1300 nm) and shortwave infrared (1300 to 2500 nm) spectrum (Abdullah, 2019). In the Polish part of the Białowieża Forest, I. typographus outbreak was estimated by mapping dead spruce stands on a tree level using airborne hyperspectral and laser scanning data obtained from HySpex VNIR-1800 camera and Riegl LMS-Q680i scanner (Sterenczak et al., 2019(Sterenczak et al., , 2020. Satellite optical and thermal data of Landsat-8 and Sentinel-2 were successfully used to investigate the early stage of I. typographus infestation in the Bavarian Forest National Park in Germany (Abdullah et al., 2018(Abdullah et al., , 2019. The authors established that the majority of spectral vegetation indices calculated from Sentinel-2, particularly red-edge dependent indices (NDRE 2 and 3) and water-related indices (SR-SWIR, NDWI, DSWI and LWCI) are able to discriminate healthy from infested plots. In addition, satellite images of WorldView-2, Pléiades 1B, and SPOT-6 were also used for detection of early stage of pest infestation in the Czech Republic (Abdollahnejad et al., 2021). The authors determined that identifying physiological stress in earlier stages could be more feasible, with higher reliability and accuracy, using hyperspectral sensors in wavelengths that multispectral sensors cannot cover.
Multispectral UAV-based imagery is capable of classifying tree decline during a bark beetle infestations in Scots pine  and Norway spruce (Junttila et al., 2022). Recent studies of spruce bark beetle attacks using Sentinel-2 multispectral data and different vegetation indices (including NDVI) in the Italian Alps show that the two stages of the epidemic (i.e. early and late) can be detected with an overall accuracy of 83.4% (Dalponte et al., 2022).
In conclusion, the present study showed that periodic observations and analyzes based on NDVI from satellite data of Sentinel-2 can serve reliably for early detection of I. typographus infestations in spruce stands around windbreak and windfalls in mountain areas. From a management point of view, the possibility to detect the attack of the European spruce bark beetle at an early stage using freely available satellite data is very important for rapid planning and implementing of sanitary measures in order to reduce the pest attacks.