كاربرد الگوريتم هوش جمعي كلوني زنبورهاي مصنوعي در بهينه‌سازي تخمين و پهنه‌بندي شدّت فرسايش باد با استفاده از داده‌هاي ژئومورفولوژي (مطالعه موردي: حوضه آبريز دشت بيرجند، استان خراسان جنوبي)

Application of Collective Intelligence of Artificial Bee Colony Algorithm in Optimization of Estimation and Zoning of Wind Erosion Intensity Using Geomorphological data (Case Study: Birjand Plain’s Drainage Basin, South Khorasan Province)

براي تخمين شدّت فرسايش باد كه وسيع ترين قلمرو فعّاليّت را دارد، استفاده از مدل‌هاي منطقه‌اي اجتناب‌ناپذير است. اين مدل‌ها، تجربي بوده و مبتني بر مؤلّفه‌هايي هستند كه از طريق امتيازدهي در آستانه‌هاي تعيين‌شده به تخمين فرسايش منجر مي‌شوند. تفاوت تجربيات و متغيّر بودن مؤلّفه‌هاي ورودي اين مدل ها، باعث بروز ناسازگاري و كاهش قابليّت اعتماد تخمين‌ها مي‌شود. هدف از اين مطالعه، بهينه‌سازي تخمين فرسايش باد در حوضه آبريز دشت بيرجند از طريق كمينه نمودن ناسازگاري امتيازات مؤلّفه‌هاي ورودي مدل مورد استفاده است. براي اين منظور، مدل تجربي پهنه‌بندي شدّت فرسايش باد سازمان جنگل‌ها و مراتع ايران  با استفاده از الگوريتم هوش جمعي كلوني زنبورهاي مصنوعي، بهينه‌سازي شده است. پس از تعيين امتيازات مؤلّفه‌هاي مدل اوّليه، منطقه مورد مطالعه به پيكسل‌هاي 200×200 مترمربعي تبديل و با استفاده از دستگاه مختصات قطبي به 82 ناحيه تقسيم شد. سپس الگوريتم بهينه‌سازي كلوني زنبورها در سه مرحله شامل: 1 تخصيص، 2 پايش و 3 نتيجه‌گيري توسّط زنبورها، اجرا و پيكسل‌هاي با بيشترين پتانسيل فرسايش شناسايي شده‌اند. نتايج، نشانگر انتقال حدود 49% از مساحت طبقات فرسايش باد در مدل سازمان جنگل‌ها و مراتع ايران به طبقات فرسايشي بالاتر در روش الگوريتم كلوني زنبورها است. بر اين اساس، الگوريتم كلوني زنبورها حسّاسيت بسيار زيادي در ارائه طبقه‌بندي كلاس‌هاي فرسايش باد دارد. آزمون واريانس مساحت كلاس‌هاي فرسايشي حاصل از دو روش، حاكي از قابليّت اطمينان بيشتر به نتايج روش الگوريتم كلوني زنبورها است. بيشترين نرخ فرسايش در لندفرم مخروط افكنه‌ها رخ داده و بيش از 90% كانون‌هاي فرسايش در واحد ژئومورفولوژيك دشت‌سر واقع شده است.

The domain of wind erosion is wider than the other erosion processes, so the use of regional models is inevitable to estimate its intensity. The experimental models depend on components rated in defined ranges evaluating the amount of erosion.  Different experiences and also variety of input components of the model lead to some inconsistency in the results, and decline the reliability of estimation. The aim of this study is to optimize the estimation of wind erosion in Birjand plain through removal and mitigation of the effects of different rating experiences.  In this paper, the data obtained from the experimental model of Iranian Research Institute of Forest and Rangeland (IRIFR) are optimized using collective intelligence artificial bee colony algorithm. To achieve this purpose, after calculating the components of Iranian research institute of forest and rangeland model, the investigated area was divided into pixels of 200×200m. The pixels were located into 82 subdomains by using polar coordinates in order to decrease the computational time. Then optimization of bee colony algorithm was implemented in three steps: (1) the allocation process, (2) the investigation process and (3) conclusion process by the bees. Finally the pixels with greatest potential erosion were identified. About 49% of the area of wind erosion classes in IRIFR model moved to higher erosion classes in bee colony algorithm. Therefore bee colony algorithm is highly sensitive in the classification of wind erosion. The variance test of the erosion classes obtained by the two methods showed more reliability of bee colony results. The results showed the highest erosion rates occurred in the alluvial fan landforms and more than 90 percent of erosion centers are located in the pediment of geomorphologic unit. Extended Abstract 1Introduction  Approximately 28% of the habitable lands of the earth are under the influence of human activities that has led to loss of soil fertility. Wind erosion lead to loss of biodiversity and production capacity of the ground known as desertification. The domain of wind erosion is wider than the other erosion processes, so the use of logical models is inevitable to estimate its intensity. The experimental models depend on components rated in defined ranges, and then evaluate the amount of erosion. Experimental models’ components are rated using field information and expert opinions; due to different expert experiences and also variety of input components, there is no consensus. So one of the fundamental disadvantages of the models is related to rate of input components which leads to different results. The aim of this study is to optimize the estimation of wind erosion in Birjand plain through removal and mitigation of the effects of different rating experiences. The case study in this article is drainage basin of Birjand plain which is located in the South Khorasan province. It covers an area of 3425 square kilometer, of which 980 square kilometer is the plain extent and is very susceptible to wind erosion and the rest is mountainous. The average annual precipitation in the plain over a period of 50 years equals to 177 mm and its climate is generally dry. 2 Materials and Methods  In order to study wind erosion, erosion models with different goals, such as Successive approximation of erosion and designing of erosion control tools, are used. In this paper, the data obtained from the experimental model of forests and rangelands of Iran (IRIFR) is optimized using collective intelligence artificial bee colony algorithm. To achieve this purpose, after calculating the components of Iranian research institute of forest and rangeland model, the investigated area was divided into pixels of 500 and in order to decrease the computational time, the pixels were located into 82 subdomains by using polar coordinates. Then optimization of bee colony algorithm was implemented in three steps: (1) the allocation process, (2) the investigation process and (3) conclusion process by the bees. Finally pixels with greatest potential erosion were identified. The center of erosion in the investigated area, that is the critical part of the region in terms of erosion, was also identified. Finally using multivariable statistical methods, the centers and spatial pattern of zones which were obtained by both models were compered. 3 Results and Discussion  About 47% of the area of wind erosion classes in IRIFR model moved to higher erosion classes in bee colony algorithm. Therefore bee colony algorithm is highly sensitive in the classification of wind erosion. Although the area of output zones obtained by the two models was very similar in both of them (about 78%). But there is significant difference in terms of the spatial pattern of erosion classes. The variance test of the erosion classes obtained by the two methods showed significant difference. The lower variance amount of classes obtained by bee colony algorithm showed that the results of this method have more reliability in wind erosion classification. 4 Conclusion Identification of 82 critical wind erosion centers in the investigated area and its agreement with real evidence shows the importance of this method in investigating the effective wind erosion components. The results showed the highest erosion rates occurred in the alluvial fan landforms and more than 90 percent of erosion centers are located in the pediment of geomorphologic unit.