Investigating dieback intensity trend of Hyrcanian boxwood based on time series studies of vegetation index (Case study: Boxwood stands of Rezvanshahr city, Guilan province)

Background and objectives: Boxwood is a valuable and unique species of the Hyrcanian forests that has been exposed to various pests for several years. About 72,000 hectares of forest areas in the north of the country are occupied by boxwood habitats. Currently, the boxwood blight disease and the box tree moth pest (Cydalima perspectalis) have infected these habitats. In recent years, extensive research has been conducted on the applications of remote sensing in monitoring forest pests, especially by using two physiological characteristics of leaf loss and color change. The purpose of this study is to comprehensively examine the time series satellite images available in the boxwood stands of Rezvanshahr city to derive vegetation indicators and investigate the trend of long-term changes in the vegetation.
Methodology: The study area is the Shemshad forests, with an area of 400 hectares in Rezvanshahr city, in Guilan province. Since boxwood is an evergreen species and is located in the deciduous Hyrcanian forests, only the months from September to December with no canopy and snowfall were used for image analysis in this study.
In this research, Landsat satellite images from 2010 (the time of the outbreak) to 2016, and Sentinel-2 satellite images from 2017 to 2019 were used. After obtaining satellite images, radiometric and atmospheric corrections were performed on the Landsat images using ENVI software. For the Sentinel-2 images, since they already include atmospheric corrections at the L2A level, there was no need for further correction. To convert the spatial resolution of satellite images from 30 meters to 15 meters, image integration and pansharpening methods were applied. After the necessary corrections, the Normalized Difference Vegetation Index (NDVI) was extracted for Landsat and Sentinel-2 images using ENVI and SNAP software, respectively.
To verify the accuracy of the data obtained from satellite images, random sampling was carried out at several points, based on the pixel dimensions of the Sentinel-2 image, which is ten meters by ten meters, in three zones: contaminated, healthy, and completely dry areas for vegetation index control. Using Ordinary Least Squares (OLS) and Mann-Kendall (MK) regression methods, the long-term trend of vegetation index changes was calculated.
Results: Based on the indices extracted from the satellite images, the average index was calculated during the period from 2010 to 2019. Additionally, the minimum, maximum, average, and standard deviation values for 10 satellite images were examined and calculated. Based on the average vegetation index, the year 2017 had the highest density of boxwood vegetation (with an average of 0.63) and the least drying, while the year 2011 had the highest level of boxwood dieback (with an average of 0.08) in the forests of Rezvanshahr. According to the results, out of the total study area of 400 hectares, 269 hectares showed a 67% increase in vegetation index from 2010 to 2019. Meanwhile, a decrease of 18% was observed in 72 hectares of vegetated areas. In the present study, only 59 hectares (15%) maintained a constant growth trend during these years. The result of long-term trend analysis using OLS and the Mann-Kendall method in most of the forest areas of boxwood trees shows a positive slope (green spectrum), though very small, indicating a greening trend. The low slope values indicate the absence of a clear trend in vegetation changes, and their positive values confirm a very slight increase in NDVI, which suggests the occurrence of very weak greenness and predominant drying. The results of the long-term trend analysis using OLS and the Mann-Kendall method indicate a low slope in the trend of greenness.
Conclusion: Based on the obtained results, only 15% of the vegetation covered with boxwood exhibited a constant trend of change. Through analyses using the Mann-Kendall method and Ordinary Least Squares, and by extracting the Normalized Difference Vegetation Index, we identified a very limited greening trend in boxwood forests. We also found that NDVI is the most effective index for detecting and predicting vegetation changes under pest pressure. It is also suggested that other satellite image sources be used to monitor the trend of dieback changes.