Hamidreza Keshtkar; Hassan Yeganeh; Omid Kavoosi
Abstract
Ferula gummosa is one of the rare and valuable species in Iran's rangelands, which is exploited by local stakeholders due to its high economic value. Protecting this species can help maintain the biodiversity and stability of mountainous areas. This study was conducted to compare the performance of six ...
Read More
Ferula gummosa is one of the rare and valuable species in Iran's rangelands, which is exploited by local stakeholders due to its high economic value. Protecting this species can help maintain the biodiversity and stability of mountainous areas. This study was conducted to compare the performance of six predictive models: Artificial Neural Networks, Random Forest, Classification Tree Analysis, Surface Range Envelope, Generalized Boosting Machines, and Generalized Linear Models. To evaluate the interactions between topographic factors and other variables, two environmental datasets were quantified and used for model calibration. The first dataset includes eleven factors covering topographic, climatic, edaphic, and remote sensing variables. Meanwhile, the second dataset contains six factors, focusing on climatic, edaphic, and remote-sensing variables. Model accuracy was evaluated using the True Skill Statistic (TSS), the area under the curve of the Receiver Operating Characteristics (ROC), and the Accuracy Index. The evaluation indices indicate that the Generalized Boosting Machine (GBM) model predicted the ecological niche of F. gummosa more accurately than the other methods. Additionally, the results showed that removing topographical variables reduced the model accuracy by 11 to 25%. The slope, NDVI, wetness, and soil groups were found to be the most important factors in mapping potentially suitable habitats for the target plant. According to the results obtained from the GBM model, approximately 45% of the Ghorkhoud area is in excellent condition. This knowledge can aid in the selection of predictors for practical Species Distribution Model (SDM) applications and provide information on which modeling techniques are most useful for a group of species.
Marziyeh Haji Mohammadi; Aliakbar Nazari Samani; Arash Zare Garizi; Hamidreza Keshtkar; Mahmood Arabkhedri; Amir Sadoddin
Abstract
The SWAT model is widely used to simulate watersheds and evaluate the impact of conservation watershed management practices. In this model, the simulation of the watershed processes is based on hydrological response units (HRUs) which are created by overlaying land use /land cover, soil and slope maps. ...
Read More
The SWAT model is widely used to simulate watersheds and evaluate the impact of conservation watershed management practices. In this model, the simulation of the watershed processes is based on hydrological response units (HRUs) which are created by overlaying land use /land cover, soil and slope maps. Meanwhile, in the HRUs definition steps, these units become conceptual and lose their spatial location and continuously. This approach is a useful and often inevitable way to simulate large and heterogeneous watersheds in terms of computational efficiency. However, if the aim is spatializing and evaluating the effectiveness of management methods on runoff, sediment and other pollutants in medium to small basins, it is necessary to know the exact location of HRUs. The purpose of conducting this study was present a new approach to defining spatial and independent HRUs and compare the simulation results based on this method with the standard form of the model. In the new approach, independent and spatial HRUs are defined through pre-processing procedures in GIS and uniquely named soil units. The model results of both approaches were very similar and no significant difference was observed in the model outputs in Taleghan watershed. The Nash-Sutcliffe coefficient of the simulated runoff and sediment at the outlet with the standard approach was 0.75 and 0.64, respectively. While, it was obtained 0.74 and 0.61, respectively for the new approach. The definition of spatial HRUs by applying the proposed method provides more tangible and practical outputs, which is more beneficial for identifying the critical areas as well as locating conservation practices compared to the conceptual HRUs approach.
Serveh Darvand; Hassan Khosravi; Hamidreza Keshtkar; Gholamreza Zehtabian; Omid Rahmati
Abstract
The purpose of this study was to compare machine learning models including Support Vector Machine, Classification and Regression Tree, Random Forest, and Multivariate Discriminate Analysis to prioritize susceptible areas to dust production. To determine the dust days, hourly meteorological data of Alborz ...
Read More
The purpose of this study was to compare machine learning models including Support Vector Machine, Classification and Regression Tree, Random Forest, and Multivariate Discriminate Analysis to prioritize susceptible areas to dust production. To determine the dust days, hourly meteorological data of Alborz and Qazvin provinces and satellite images of the same days for the period 2000 to 2019 were used. 420 dust collection points were identified and the map of their distribution was prepared. The maps of factors affecting the occurrence of dust, including landuse map, soil orders map, slope map, slope aspect map, elevation map, vegetation map, topographic surface moisture, topographic surface ratio, and geology mam were prepared. Using the mentioned models, the impact of each of the effective factors of dust was determined and prioritization maps of dust harvesting areas were prepared. Models were evaluated using the ROC curve. According to the results, the elevation factor is more important in all models than the other parameters used in the model. The modeling results also showed that the Random Forest )RF( and Multivariate Discriminate Analysis (MDA) models had the highest values of accuracy (0.96), precision (0.94), Probability of Detection (POD) (0.98), and False Alarm Ratio (FAR) (0.051) compared to the others. The performance of the RF and MDA models is better than the other models, followed by the Support Vector Machine (SVM) and Classification and Regression Tree (CART) models, respectively. Also, in evaluating the models using Receiver Operating Characteristic (ROC), the RF model was selected as the best model.