Volume 18, Issue 4 (1-2025)
مرتع 2025, 18(4): 561-581 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Beyranvand N, Hasanvand S, Sepahvand A, Tarnian F, Arjmand N. Modeling of Infiltration Rate in Different Vegetation Types by Various Soft Computing Techniques (Case Study: Alashtar Watershed, Lorestan Province). مرتع 2025; 18 (4) :561-581
URL: http://rangelandsrm.ir/article-1-1274-en.html
URL: http://rangelandsrm.ir/article-1-1274-en.html
Department of Range and Watershed Management, Faculty of Natural Resources, Lorestan University, Khorramabad
Abstract: (480 Views)
Background and objectives: During the rainy season, soil infiltration replenishes soil moisture in semiarid regions, influencing vegetation regeneration, erosion potential, and groundwater recharge. Infiltration of water into soil is a key process for providing water supply to plants and plays an essential role in controlling surface runoff and groundwater. This study investigated infiltration modeling in different vegetation types (Quercus brantii, Astragalus ecbatanus-Euphorbia denticulate, Grass, and Grass-Astragalus rhodoseminus) in the Alashtar watershed, Lorestan Province, Iran.
Materials and Methods: The study area, part of the Kashkan watershed in Lorestan Province, was selected for modeling infiltration rates using various soft computing techniques. The study area is located between 48°10′28″ - 48°23′29″ N latitudes and 33°45′17″ - 33°51′23″ E longitudes, covering approximately 112.54 km². Elevation varies from 1481 to 3613 meters above sea level. The area has a cold and semiarid climate with a mean annual rainfall of less than 570 mm. Five soft computing techniques—Support Vector Machine (SVM), Gaussian Process (GP), Multi-Layer Perceptron (MLP), and Random Forest (RF)—were used to model infiltration rates. The total dataset comprised physical soil characteristics, with 70% used for training and 30% for testing the models. The input data included time, sand, clay, silt, soil density, and soil moisture, while the output data were infiltration rates measured using a double-ring infiltrometer at 23 locations. Three statistical parameters—coefficient of correlation (C.C), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE)—were used to compare the efficiency of all models.
Results: The results indicated that the Astragalus ecbatanus-Euphorbia denticulate vegetation type had higher cumulative infiltration and average infiltration rates. Among the models, Random Forest (RF) and Gaussian Process with PUK kernel (GP-PUK) showed the least error in estimating infiltration rates with the input combination of time, sand, clay, silt, soil density, and soil moisture. The GP-PUK model demonstrated acceptable accuracy in Quercus brantii and Astragalus ecbatanus-Euphorbia denticulate vegetation types with correlation coefficients of 97.2% and 98.4%, respectively. The RF models performed better than other models in estimating infiltration rates for Grass and Grass-Astragalus rhodoseminus vegetation types, with correlation coefficients of 83.9% and 99.9%, respectively.
Conclusion: Predicting infiltration rates is crucial for hydrologic design, watershed management, irrigation, and agricultural studies. The results show that soft computing techniques have suitable capabilities to predict soil infiltration rates. These models can quantify infiltration amounts and estimate runoff in different vegetation types. The research findings can help local authorities manage and develop their areas systematically and effectively.
Materials and Methods: The study area, part of the Kashkan watershed in Lorestan Province, was selected for modeling infiltration rates using various soft computing techniques. The study area is located between 48°10′28″ - 48°23′29″ N latitudes and 33°45′17″ - 33°51′23″ E longitudes, covering approximately 112.54 km². Elevation varies from 1481 to 3613 meters above sea level. The area has a cold and semiarid climate with a mean annual rainfall of less than 570 mm. Five soft computing techniques—Support Vector Machine (SVM), Gaussian Process (GP), Multi-Layer Perceptron (MLP), and Random Forest (RF)—were used to model infiltration rates. The total dataset comprised physical soil characteristics, with 70% used for training and 30% for testing the models. The input data included time, sand, clay, silt, soil density, and soil moisture, while the output data were infiltration rates measured using a double-ring infiltrometer at 23 locations. Three statistical parameters—coefficient of correlation (C.C), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE)—were used to compare the efficiency of all models.
Results: The results indicated that the Astragalus ecbatanus-Euphorbia denticulate vegetation type had higher cumulative infiltration and average infiltration rates. Among the models, Random Forest (RF) and Gaussian Process with PUK kernel (GP-PUK) showed the least error in estimating infiltration rates with the input combination of time, sand, clay, silt, soil density, and soil moisture. The GP-PUK model demonstrated acceptable accuracy in Quercus brantii and Astragalus ecbatanus-Euphorbia denticulate vegetation types with correlation coefficients of 97.2% and 98.4%, respectively. The RF models performed better than other models in estimating infiltration rates for Grass and Grass-Astragalus rhodoseminus vegetation types, with correlation coefficients of 83.9% and 99.9%, respectively.
Conclusion: Predicting infiltration rates is crucial for hydrologic design, watershed management, irrigation, and agricultural studies. The results show that soft computing techniques have suitable capabilities to predict soil infiltration rates. These models can quantify infiltration amounts and estimate runoff in different vegetation types. The research findings can help local authorities manage and develop their areas systematically and effectively.
Keywords: Lorestan province, Alashtar watershed, Infiltration, vegetation types, Learning algorithms.
Type of Study: Research |
Subject:
Special
Received: 2024/07/11 | Accepted: 2025/02/15 | Published: 2025/01/29
Received: 2024/07/11 | Accepted: 2025/02/15 | Published: 2025/01/29
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
