ارزیابی سلامت بوم‌سازگان جنگل در مقیاس حوضه آبخیز با استفاده از سنجش از دور و روش پایش سلامت جنگل

Abstract
A healthy forest ecosystem is essential to provide a wide range of environmental, social, and economic services. To evaluate forest health level, it is necessary to quantify ecosystem conditions using a variety of indicators. In this study, a combination of six indices: NDVI, EVI, SAVI, NDWI, ARI1, and CRI1 were used to derive the forest health condition map around Shenroud watershed, Siahkal using Landsat 8 OLI image of 2021 and ENVI Forest Health Tool. We evaluated the forest health with five grades (A; very good to E; very poor). Forest health data validation in various forest types in the study area is conducted via field crosscheck and assessment regarding the forest health level using 40 sample plots and FHM (Forest Health Monitoring) method. In this work, based on the multiple linear regression model of the effect of vegetation indices in forest health assessment, the NDVI vegetation index with an R2 value of 0.77 has the highest effect on the forest health level in the watershed, which shows the greenness level and the density level of forest stands. The results of the analysis show that most of the forests in the study area were moderately healthy. 19.4% of the forest areas were classified as “Healthy”, 56.8% as “Moderately healthy”, and 23.8% as “Unhealthy”. Furthermore, different forest types in the study area have different percentages of healthy forests. Plantations, especially broadleaf plantation, has the largest forest area in unhealthy conditions (weak and very weak) (approximately 33%). In contrast, natural forests, including mixed broadleaf forest and beech forest, have the largest forest area in healthy conditions (very good and good) (approximately 21%). In general, the current state of the ecosystem in the study area is mainly in middle health, which is a result of long-term deforestation, soil erosion, and inappropriate human exploitation. Our RS-based health diagnosis of the area was consistent with field survey results. The process could be very useful in mapping the health conditions of the forests of the country accurately. It is suggested to study the spatial and temporal variation of forest health in the watershed under future climate change.


Introduction
Forest management requires information about the state of the forest with a focus on forest health. Because healthy forests are able to perform well compared to unhealthy forests. Forest health monitoring is especially important in creating sustainable forest management and also the integrated watershed management. However, today, the protection and monitoring of forests concerning forest health still do not exist and are limited to small scale studies. Therefore, this study assesses the health conditions of forest ecosystems at the watershed level in the north of Iran. In this research, by using remote sensing data and preparation of vegetation indices along with field data obtained from forest health monitoring in the Siahkol Shenrud watershed, the analysis of these indices and their grading at the forest health level is discussed. The protection and monitoring of forest conditions requires an extensive investigation of different forest types that are located inside a watershed. Therefore, in this study, a method suitable to the conditions of the forests of northern Iran is presented, which is necessary for detecting forest health levels based on the different types in a watershed. This study is conducted to provide useful knowledge in the field of creating a sustainable management strategy in the forests of northern Iran, which is based on watershed management.
Methodology
This research is a combination of remote sensing survey and analysis methods and a geographic information system (GIS) to describe the health status of forest ecosystems in the Siahkol Shenrod watershed. Land use classification was done using the Landsat 8 OLI satellite image of 2021. Supervised classification method and maximum likelihood algorithm were used to prepare land use maps. In order to evaluate the accuracy of the classification, Google-Earth images were used by selecting 30 educational samples. For accuracy assessment, the high value of the Kappa coefficient (0.92) was obtained.
• Vegetation indices calculation
After pre-processing, the Landsat 8 image was used to calculate vegetation indices according to Table 1.
• Forest health index extraction
The six calculated vegetation indices (Table 1) were combined and classified for the input of forest health index (FHI) in ENVI software. In this classification, ten classes were defined for all six indices, and then in calculating the FHI index, it was reduced to five classes by combining both related classes. The FHI index (Gupta and Pandey, 2021) is available in the ENVI forest health toolbox, as follows:

• Field data collection
To validate the forest health levels in different forest types of the watershed obtained from remote sensing data, it is necessary to conduct a field survey. Field survey data were obtained from the study area in September and October 2021. A stratified sampling method was used before the field survey in the GIS environment. Based on this method, 40 square-shaped samples (30 m * 30 m) were built. Observational activities were conducted using the Forest Health Monitoring (FHM) method (USDA Forest Service, 2020). The area with less than 10% damage was classified as a very healthy forest, and the forest damage between 10 and 25% was classified as a healthy forest. 25-50% damage was classified as medium health forest, 50-75% as poor forest health, and 75-100% damage was classified as very poor forest health.

Results
• Vegetation indicators
Six vegetation indices extracted for forests in the study area are presented in Figure 3. Based on the results, 95% of the forests in the study area showed an NDVI index of more than 0.5, which indicates healthy vegetation. Also, healthy vegetation with high chlorophyll content was shown in 91% of the studied area with an EVI index between 0.2-0.8. In this research, according to the multiple linear regression analysis of the effect of vegetation indices used in calculating forest health, the NDVI index with an R2 value of 0.77 has the greatest effect on the level of forest health in the watershed, which is the greenness level. It also shows the density level of forest stands. Similar results of the influence of vegetation indices have been presented by Tuominen et al. (2009). Furthermore, the higher effectiveness of NDVI in this study is similar to the findings of Barkey and Nursaputra (2017) and Oliech (2019) in assessing forest health. After NDVI, R2 values for EVI, SAVI, and AR1 indices were 0.75, 0.73, and 0.72, respectively, and for NDWI and CRI1 indices were 0.63.
• Forest health index
The relationship of six vegetation indices calculated from the Landsat 8 image was used to determine the level of forest health in the Shenrod watershed. Forest health was divided into five categories: 1) very good (very healthy); 2) good (healthy); 3) moderate (under stress); 4) poor (unhealthy); 5) very poor (dead). Based on the results, the overall accuracy was 86%. Also, the kappa value of 0.81 was obtained. Figure 4 presents the health status of forest stands in the Shenrod watershed.
Based on Figure 4, the distribution of forest health classes in the study area is not uniform. The central areas of the watershed have higher levels of forest health, while the southern areas of the watershed have lower levels of forest health. This can be due to the history of heavy wood harvesting, selective cutting, and as a result, the widespread opening of forest stands (Jahdi and Arabi, 2023), in the form of forestry plans implemented in the southern areas of the watershed during the last two decades. Wildfires in this area also affect the structure and composition of the vegetation and are a great threat to the long-term productivity and overall health of the forest. In general, 56.8% of the forest area is in the medium health category. Healthy and unhealthy forests with levels almost close to each other make up 41.7% of the total forest area. 1.5% of the forest area is also in the very poor health category (Figure 4).

Conclusion
In this study, the quantitative index of forest health was evaluated using broadband spectral indices and stress-related pigments before tree fall using remote sensing data. Using forest health analysis, the quantitative health conditions of the forest were described in 2021 with a forest area of 122.2 km2. Of this amount, there are healthy forests with an area of 23.8 km2 and unhealthy forest conditions with an area of 29 km2. The maximum area of the watershed, i.e. 69.4 km2, is also in medium health conditions. The health status of forest ecosystems is mostly average health, which is the result of forest destruction, soil erosion, and human overexploitation. Forest conservation programs can effectively control forest degradation and improve the health of local ecosystems. Forest health detection based on remote sensing in the study area was consistent with the field survey results. This study guided the systematic approach and provided remote sensing techniques for forest health monitoring programs and sustainable forest management.