Dr. Ali Mahdavi, Msc. Sahel Ramezani, Dr. Hamidreza Naji,
Volume 4, Issue 2 (9-2025)
Extended Abstract
Background and Objectives: Climate change, manifested through fluctuations in key variables such as temperature and precipitation, poses a significant threat to forest ecosystems, particularly in semi-arid regions like the Zagros. Pistacia atlantica (Wild Pistachio) is a valuable native species in these forests, whose growth is highly sensitive to water availability. Quantifying the impact of climatic variables on its radial growth is essential for adaptive forest management. While traditional statistical methods have been used, advanced artificial intelligence models like the Group Method of Data Handling (GMDH) neural network offer superior capabilities for modeling complex, non-linear relationships. The main objectives of this study were to: 1) evaluate the effect of climatic variables (precipitation, temperature, relative humidity) on the diameter growth of P. atlantica, 2) determine the most influential climatic factors, and 3) assess the accuracy of the GMDH neural network model in predicting growth based on climate data.
Materials and Methods: The study was conducted in three habitats of P. atlantica (Darreh Shahr, Abdanan, and Majin) in Ilam province, Iran. A total of 18 tree discs (6 from each site, divided into two diameter classes: <30 cm and >30 cm) were collected from trees with similar topographic conditions. After surface preparation and polishing, high-resolution images were taken, and annual ring widths were precisely measured using Motic image software. Wood density was also determined for each sample. Climatic data (annual precipitation, average, minimum, and maximum temperature, relative humidity) for the past 15 years were obtained from the nearest meteorological stations. The relationship between ring width indices and climatic variables was first analyzed using Pearson correlation in SPSS software. Subsequently, the GMDH neural network model was implemented in MATLAB (R2014a) to predict radial growth based on climatic inputs. The data were randomly divided into training (70%), validation (15%), and test (15%) sets. Model performance was evaluated using statistical indices: Root Mean Square Error (RMSE), Mean Square Error (MSE), and Correlation Coefficient (R).
Results: Statistical analysis revealed a significant positive correlation between annual ring width and both annual precipitation (r = 0.188, p=0.002) and relative humidity (r = 0.173, p=0.004). In contrast, a significant negative correlation was found with average annual temperature (r = -0.185, p=0.002) and maximum annual temperature (r = -0.152, p=0.013). No significant correlation was observed with minimum annual temperature. The GMDH neural network model demonstrated high accuracy in predicting radial growth from climatic variables, with performance metrics on the total dataset as follows: RMSE = 3.86, MSE = 14.88, and R = 0.90. The model's predictions closely matched the observed growth trends, confirming its effectiveness.
Conclusion: The radial growth of Pistacia atlantica in the semi-arid Zagros forests is significantly influenced by climatic fluctuations. Increased precipitation and relative humidity positively enhance growth, while rising temperatures, particularly maximum temperatures, have a suppressive effect, likely due to increased evapotranspiration and water stress. The successful application of the GMDH neural network model, with its high predictive accuracy (R=0.90), establishes it as a powerful and reliable tool for modeling climate-growth relationships in complex forest ecosystems. These findings provide critical insights for developing climate-adaptive conservation and management strategies to enhance the resilience of P. atlantica stands against future climate variability.
