Document Type : Research Article
Authors
1
Assistant Professor, Department of Climatology, Payame Noor University, Iran
2
Faculty Member, Iranian National Institute for Oceanography and Atmospheric Science (INIOAS), Tehran, Iran
Abstract
Global warming and its intensifying effects on arid urban regions highlight the necessity of integrated urban planning strategies. This study aims to forecast climate trends, evaluate the role of vegetation in mitigating environmental noise, identify drought-resistant plant species, and propose sustainable strategies for green space utilization in Birjand city. Annual temperature changes were predicted up to 2035 using the Long Short-Term Memory (LSTM) deep learning model, which indicated a gradual increase in mean annual temperature, posing potential challenges to urban livability. Based on these projections, the influence of vegetation cover on noise pollution was analyzed using remote sensing data. A vegetation index map was generated and reclassified into five categories, revealing that 47% of the study area lacked vegetation, 28% had sparse cover, 17% moderate, 6% good, and only 2% dense cover. Comparative analysis demonstrated that areas with denser vegetation experienced significantly lower noise levels, while sparsely vegetated or barren zones were exposed to higher acoustic pollution. Furthermore, tree species with high drought tolerance and effective sound absorption, such as Cupressus sempervirens var. pyramidalis, Robinia pseudoacacia, and Myrtus communis, were identified as suitable candidates for planting in traffic corridors and noise-prone urban areas. These findings emphasize the crucial role of urban green spaces in simultaneously mitigating thermal and acoustic stresses. Overall, the study underscores the importance of sustainable urban design through optimal distribution and species selection of vegetation, offering a practical pathway to counteract the adverse effects of climate change while enhancing environmental quality and residents’ quality of life.
Introduction
Rapid urbanization has led to increased noise pollution, primarily from transportation, which negatively affects quality of life, mental health, and physical well-being. Urban vegetation, particularly trees, offers a natural solution to mitigate noise pollution and improve mental health by enhancing green spaces. However, in arid regions, it is crucial to select drought-resistant species to balance the benefits of noise reduction with potential challenges such as allergenic pollen. This study first predicts annual temperature trends using a deep-learning LSTM model, then examines the role of vegetation in reducing noise pollution in Birjand. By integrating environmental data and modeling approaches, the research aims to provide sustainable urban design strategies that address both urban heat and noise pollution. The findings offer recommendations for urban planners on where to expand vegetation and which plant species are most effective in mitigating noise in arid environments.
Study Area
The city of Birjand is located at approximately 32°52′ north latitude and 59°13′ east longitude, covering an area of about 31,700 square kilometers in South Khorasan Province, eastern Iran. With a population of around 266,700, Birjand lies in a region characterized by a dry to semi-arid climate. The city has an average annual temperature of approximately 17.5 °C and receives about 134 mm of precipitation per year.
Material and Methods
In this study, the Long Short-Term Memory (LSTM) deep learning model was employed to predict temperature trends in the study area. Annual temperature data from 2000 to 2024 were collected, normalized using the Min-Max method, and divided into training (70%), validation (15%), and testing (15%) sets. The LSTM model, a type of advanced recurrent neural network (RNN), effectively captured complex temporal patterns and long-term dependencies within the data. Its architecture, comprising memory cells and three key gates (forget, input, and output), enabled robust modeling of long-term climate trends and seasonal fluctuations. Following training, the model was used to forecast annual temperatures through 2035. The results demonstrated high predictive accuracy, as confirmed by low RMSE values, and showed a clear warming trend aligned with observed climate patterns. Additionally, spatial analysis of noise pollution was conducted using GIS and the Inverse Distance Weighting (IDW) interpolation method. Field data on sound levels recorded at various times of day (morning, noon, and night) were used to generate detailed noise distribution maps for Birjand, providing insights into spatial variations in environmental noise exposure.
Results and Discussion
To predict temperature variations in the study area, a deep learning approach was adopted. Specifically, the Long Short-Term Memory (LSTM) neural network was employed to analyze the trend of mean temperature changes over the period 2021–2035. The results revealed a clear upward trend in average temperature, with a high coefficient of determination (R² = 0.84), indicating an estimated annual increase of approximately 0.08°C. The LSTM model was chosen for its proven capability to capture complex temporal dependencies in time-series climate data. Model outputs confirmed the efficiency of LSTM in accurately forecasting temperature trends, closely reflecting inter-annual fluctuations and highlighting a consistent warming trajectory for the region. These findings suggest that the study area is likely to experience noticeable impacts of climate change in the coming years. Throughout the model training process, the Mean Squared Error (MSE) consistently decreased across both training and validation datasets. As the number of training epochs increased, the model's prediction accuracy improved steadily. The sustained reduction in training error, along with a relatively stable error trend in the validation set, indicates effective learning without significant overfitting. Although a slight divergence between training and validation errors was observed in the final stages of training, the difference remained minor, suggesting that the LSTM network successfully captured the underlying temporal patterns in the data. Overall, the results highlight the robustness and high performance of the LSTM architecture in modeling climate-related time series and in providing reliable predictions of future temperature dynamics.
Spatial analyses indicated that commercial and industrial areas adjacent to busy roads exhibit the highest levels of noise pollution. In contrast, dense tree cover, particularly broad-leaved and coniferous species, significantly reduced unwanted noise levels. Due to their drought resistance and high efficiency in sound absorption, these species were identified as suitable options for planting along urban roads. Ultimately, the study emphasizes that selecting appropriate plant species and implementing smart green space design can simultaneously enhance urban livability, reduce noise pollution, and mitigate the impacts of global warming.
Conclusions
Urban green spaces, especially trees, play a vital role in mitigating noise pollution. This study highlights that proper green space design and the selection of suitable plant species around noisy areas such as streets and industrial sites can significantly reduce noise levels. Integrating green, low-noise land uses into urban development plans is essential for sustainable noise management. It is recommended that future studies leverage remote sensing and deep learning approaches to advance climate change monitoring and enhance urban planning strategies.
Keywords
Main Subjects
)
Send comment about this article