Human Activities Impact Prediction in Vegetation Diversity of Lar National Park in Iran Using Artificial Neural Network Model

• Published in: Integrated environmental assessment and management Vol. 17; no. 1; pp. 42 - 52
• Main Authors: Jahani, Ali, Saffariha, Maryam
• Format: Journal Article
• Language: English
• Published: United States Blackwell Publishing Ltd 01.01.2021
• Subjects: Artificial neural networks; Biodiversity; Biodiversity loss; Decision support system; Decision support systems; Graphical user interface; Human Activities; Human influences; Humans; Impact prediction; Iran; Landforms; Lar National Park; Livestock; Multilayer perceptron; Multilayer perceptrons; National parks; Neural networks; Neural Networks, Computer; Parks, Recreational; Plants; Sensitivity analysis; Soil; Soil moisture; Species extinction; Tourism; Training; Vegetation; Vegetation diversity; Iran; Multilayer perceptron; Decision support system; Vegetation diversity; Lar National Park
• ISSN: 1551-3777; 1551-3793
• DOI: 10.1002/ieam.4349

ABSTRACT The effects of livestock and tourism on vegetation include loss of biodiversity and in some cases species extinction. To evaluate these stressor–effect relationships and provide a tool for managing them in Iran's Lar National Park, we developed a multilayer perceptron (MLP) artificial neural network model to predict vegetation diversity related to human activities. Recreation and restricted zones were selected as sampling areas with maximum and minimum human impacts. Vegetation diversity was measured as the number of species in 210 sample plots. Twelve landform and soil variables were also recorded and used in model development. Sensitivity analyses identified human intensity class and soil moisture as the most significant inputs influencing the MLP. The MLP was strong with R2 values in training (0.91), validation (0.83), and test data sets (0.88). A graphical user interface was designed to make the MLP model accessible within an environmental decision support system tool for national park managers, thus enabling them to predict effects and develop proactive plans for managing human activities that influence vegetation diversity. Integr Environ Assess Manag 2021;17:42–52. © 2020 SETAC KEY POINTS Human activities endanger the sustainability of national parks, and the multilayer perceptron (MLP) model with R2 = 0.88 revealed the most accurate results in prediction of vegetation diversity. According to the results of the trained networks, the MLP model with 6 neurons in 2 hidden layers and logarithmic sigmoid (logsig) function results in the best prediction of vegetation diversity. Human activities class and soil moisture are the most influential factors affecting vegetation diversity in national parks. This research was designed with an applicable graphical user interface tool as an environmental decision support system, which is a shortcoming in most of the recent modeling for predicting changes in vegetation.