عنوان مقاله [English]
Terrain morphology, provides a lot information for researchers in the field of environmental science. One of the goals in geomorphology is identification, and analyzing terrain landforms. In the past, the identification of landforms was performing field-based or using topographical maps. Manually, which was time-consuming and difficult, and in vast areas, it was facing many problems. In this article we had attempted to identify glacial and sub glacial landforms including: glacial cirque, glacial sinkhole (Tarn Lake), summit, saddle, ridgeline, drainage line, and sedimentary fans. For recognizing these landforms, two modelling approaches have followed. First is a conceptual modeling, which uses kernel pattern modeling. This modeling level, provides the condition in which, terrain morphology compares to the reference pattern. Second is an object-based modeling, which uses reference object to recognize landforms. Above mentioned landforms, considered in this research, recognizes, using both modeling approaches, and the results represents in the form of maps. Some typical landforms considered to make accuracy assessment and performance control, for each model output possible. To automate modeling procedures, Python programming language used widely. Eventually, all codes and scripts prepared in the build-in Graphical User Interface (GUI) programming environment of python (Tkinter), and the software, named: Landform Detector V.1 prepared. Accuracy assessment, shows that landform recognition process had performed with 60% in average, which respect to the landform complexity, is acceptable. Average accuracy of the considered models are equal to 51.58 % and 50.60 % for conceptual and object-based approaches, respectively. In result, object-based approach had a better performance overall.
Terrain Landform Recognition Based on Kernel Pattern Modeling (with Focus on Glacial and Subglacial Landforms), in Alborz Mountainous belt
Terrain morphology, provides a lot information for researchers in the field of environmental science. One of the goals and subjects in geomorphology is identification, classification and analyzing terrain landforms. In the past, the identification of landforms was performing field-based or using topographical maps (Usually in the form of contour maps) manually, which was time-consuming and difficult, and in vast areas, it was facing many problems. In this article we had attempted to identify glacial and subglacial landforms including: glacial cirque, glacial sinkhole (Tarn Lake), summit, saddle, ridgeline, drainage line, and sedimentary fans (Alluvial fan, colluvial fan, and glacial outwash fan). For recognizing these landforms, two modelling approaches have followed. First is a conceptual modeling, which uses kernel pattern analyzing and modeling. This modeling level, provides the condition in which, terrain morphology compares to the reference pattern and then its similarity calculates. Second is an object-based modeling, which uses reference object to recognize landforms by calculating the deviation of terrain morphology and reference object.
Materials and methods
Digital Surface Model (DSM) was used as an elevational source of the terrain’s surface morphology. In this research, this DSM dataset was used to model landform recognition and classification. To identify landforms, the Alborz mountainous belt located in northern part of Iran was used. for recognizing typical glacial and subglacial landforms, Alamkuh and Takht-E-Soleiman glacial site, which is in the western part of Alborz mountainous belt, has considered. To meet research’s goals, seven most frequent and distinct glacial and subglacial landforms which are important in the field of glacial geomorphology, considered. These landforms are included: Glacial cirques, glacial saddle and sinkhole (Potential tarn lakes), summits, ridgelines, drainage lines and valleys floor, and glacial outwash fans (Alluvial fans, Colluvial fans). Recognition modeling of all these landforms, are performed in two modelling approaches. The study modeling environment is covering digital terrain modeling. Raster analysis concepts which are dealing with gridded data structures, are considered here principally. Landform classification, recognition and detection are the sub-category in the digital terrain modeling environment. Glacial and subglacial landforms are the matter of subject in the field of landform recognition.
Results and discussion
In this paper, we have two different approaches against landform recognition modeling. Two new and innovative models were developed for the detection and classification of some glacial and subglacial landforms. First approach was according to a kernel pattern analysis and the second one was object-based. In the field of kernel pattern analysis, some key factors should have been taken into consideration. First, was the pattern or the morphology of the kernel. To design the templates, we used a conceptual model representing the local morphology of the selected landform. Preparing this pattern could also be done, with regarding to the allometrical and morphometrical attributes of the landforms, which of course, is our suggestion for the future studies. The next issue in this modeling level was, the scale challenges. We suggested two concepts related to the scale, including geometrical and morphological scales. We don’t work on the automated algorithms to change morphological scales of the landform kernel patterns, but we automated algorithms to change geometrical scales with regarding to the concepts of the linear interpolation. We tried to codify and automate all algorithms and formulas using Python programming language and introduced a Graphical User Interface (GUI) for them. We put all the code components together and provided a software called Landform Detector V.1 for this purpose. This software would be able to run all models and algorithms introducing here
In this study two different approaches provided in the field of glacial and subglacial landform recognition. First is object-based modeling approach, which compares landform reference object to the terrain surface to calculate deviation percentage. Because this model, uses the average of the real samples of the terrain landform, the morphometric, allometric, and geometric ratios between the various components of a landform are well maintained. As a result, this model can well overcome the scale challenges. Less deviation from reference object in this model would increase the possibility occurrence of that specific landform. Second is kernel pattern analysis and modeling approach, which uses a conceptual space to model terrain landform. This model type, analyzes, resemblance levels of the terrain surface with the object pattern to calculate similarity percentage. In this area, geometrical and morphological scales, provided to satisfy scale dependencies. In result, the more similarity with the reference pattern, the more possibility occurrence of that specific landform. These two modeling approaches, would be able researchers to recognize and classify complex landforms. Also make it possible to deal with scale challenges and dependencies. Our suggestion for future studies is to focus on algorithms that can automatically change geometric and morphological scales. Pattern analysis algorithms would be also useful concepts in this area of interest.
Keywords: Kernel Pattern modeling, landform recognition, Terrain Landform, Object-based modeling, Glacial and subglacial landform.