Spatial-statistical modeling of deforestation from an ecogeomorphic approach in typical Hyrcanian forests, Northern Iran

Abstract

Deforestation is a significant environmental concern of the present century. This issue has received serious attention from global and regional communities due to its relationship with various environmental issues such as climate change, erosion, water quality, and biodiversity. Concerns about deforestation in arid and semi-arid countries, such as Iran, are quite tangible since it confronted with a shortage of forest resources on the one hand and acute issues caused by flood and erosion hazards on the other hand. The Talesh forests in the north of Iran, which are a manifestation of the ancient Hyrcanian forests with huge ecological reserves, is exposed to forest loss due to the expansion of human activities such as agriculture, wood harvesting, livestock grazing, and mining, despite the implementation of watershed management plans and forest conservation. The conservation and restoration of these forest ecosystems necessitate knowledge of the location and rate of deforestation, as well as its driving factors, through a systemic and interdisciplinary approach that has not yet been taken into account. By using an ecogeomorphic approach to model the deforestation event, we attempted to investigate the link between the ecological process of deforestation and geomorphological processes by combining the spatial terrain analysis with the statistical logistic regression. Given the approach in deforestation modeling, we were able to explain the effects of both physical and anthropogenic factors on deforestation only by incorporating physical variables (geomorphology), which can easily be derived from available digital elevation models (DEMs). First, we succeeded in mapping deforestation points in 12 catchments over 32 years using Landsat images acquired in 1991 and 2022 through change detection technique. The results of the assessment of negative changes in forest cover from 1991 to 2022 showed that about 90 km² (4.5% of the total area of catchments) has been deforested. The percentage of deforestation area varied from 7.7% in Haviq catchment to 1.8% in Dinachal catchment. We used the spatial logistic regression model to explain the relationship between geomorphological variables and deforestation probability, since the model is the most efficient predictive model to gather a group of independent variables of different natures without the need for their normal distribution. Geomorphologically independent variables included altitude, slope, topographic position index (TPI), northness, eastness, plan curvature, profile curvature, length of slope (LS) factor, slope length, topographic wetness index (TWI), contributing area, distance to stream, and terrain ruggedness index. The results of logistic regression analysis by revealing the direction of multivariate relationships showed that the probability of forest loss is higher in such places: low altitude and valleys, low slopes, divergent flow points, convex surface, downstream section, flat areas with homogeneous, and dry zones with low moisture. In addition, determining the intensity of the relationships between the independent variables and the dependent variable through regression test showed that the variables of slope, altitude, and ruggedness index with coefficients of β equal to − 2.82, − 2.1, and 1.92, respectively, are among the most important variables explaining deforestation. In contrast, the variables of eastness, northness, distance from the river, and slope length with coefficients of β equal to − 0.000017, − 0.000031, and − 0.000124, respectively, were identified as the insignificant variables in explaining the deforestation and were excluded from the final prediction model. Furthermore, the results of evaluating the efficiency of the logistic regression model through pseudo-R² (0.19) and relative operating characteristic (0.75) statistics indicated a good fit and an acceptable agreement between the actual map and the predictive map of deforestation. However, the use of more accurate and high-quality DEMs with different spatial resolutions was recommended for future studies. Regional, urban, and rural policymakers and planners should pay attention to the geomorphic environments that have a high probability of deforestation based on the results of this study. The need for more care and protection is evident in these areas, and any human interference in them must be done consciously and in accordance with environmental sustainability principles.

Graphical Abstract