Automatic approach for deriving fuzzy slope positions

中文版

Overview

Spatial gradation information of slope positions (e.g., ridge, backslope, and footslope) is important for terrain-related geographical or ecological modeling. The quantification of such gradation information is the so-called fuzzy slope positions.

Among existing methods for deriving fuzzy slope positions, the prototype-based inference method is more reasonable because of not only its inference in both spatial and attribute domain but also its use of the typical locations as prototypes, which inherently contain the characteristics of the slope position distribution in a study area. However, its practicability is currently limited due to the extensive manual operations and parameter-settings, such as preparing topographic attributes as input, finding prototypes of slope positions, and setting parameters for fuzzy inference.

This study proposed an approach to automate the whole workflow of the prototype-based method. Instead of being determined totally by users in the original method, in the proposed approach the typical locations and the fuzzy inference parameters for each slope position type can be automatically determined based on the common expert knowledge and data mining. Furthermore, the preparation of necessary topographic attributes is automated, which means that the proposed automatic approach needs only one necessary input, i.e. the gridded DEM of the study area. The proposed approach is implemented as a configurable Python script to organize the workflow, in which all of the compute-intensive procedures are speeded up by parallel computing based on message passing interface (MPI). Case studies shows that this approach can derive fuzzy slope positions reasonably and efficiently.

Similar to the prototype-based method for deriving fuzzy slope positions, many other geospatial analysis methods currently also need manual operations and parameter setting processes for carrying out a workflow in practice. These processes are tedious for users and selecting reasonable parameter values is not only subjective but also error-prone. This study shows an example of designing automatic approach for geospatial analysis method, in which the parameter-settings can be automatically determined by the combination of expert knowledge (e.g., the curve shapes of FMFs for individual topographic attributes, and parameter settings related to RPI [Relative Position Index]) and data mining techniques (e.g., FMF parameter estimation based on the frequency of topographic attribute values in candidate areas of typical locations). The parallel computing technique can be used to achieve high computational efficiency for such automatic, but compute-intensive workflow. The basic idea in the proposed approach is potentially useful for automation of other similar geospatial analysis methods.

Software availablility

• Open source at github.com/lreis2415/AutoFuzSlpPos, also integrated into the watershed modeling framework namely SEIMS.

Representation

• An automatic approach of prototype-based fuzzy slope positions. AAG Annual Meeting, Apr. 5–9, 2017. Boston, USA. Download PPT (alternative download link)

Course

• (in Chinese)国科大研究生普及课《流域分析与建模》课程实验——流域空间参数提取： Installation for environment and Brief manual, Program and demo data, PPT.

Review history

• Submission: 2017-01-16

• With Editor: Dr. Takashi Oguchi (oguchi (at) csis.u-tokyo.ac.jp)

• Major revision: 2017-09-17

  • Reviewer: This paper presents an interesting approach to automatize the overall workflow of the prototype-based inference method developed by Qin et al., (2009) for deriving slope position typologies (i.e., ridge, shoulder slope, backslope, footslope, and valley). The basic idea is to overcome the limitations of the prototype-based inference method for deriving fuzzy membership values that are mainly due to the “subjective” assignment of the user of a set of explicit rules for each slope position type. The methodological approach is quite original and its implementation, freely available on the web, surely represents an interesting tool for the scientific community. However, I have some concerns about the structure of the manuscript. Several parts of the text are dedicated the description of the methodology but a chapter dealing with the comparison with other methods of classification and related discussion citing international literature is completely missing. Despite the long text describing the approach some essential information is poorly presented (e.g. which algorithms are used to derive slope and profile curvature?). Moreover, in my opinion in its present form, the manuscript seems more suitable for a journal focused on computer science in geoscience (e.g., chapter 4 dedicated to the description of the implementation, the computational efficiency used as evaluation methods) than for a journal as Geomorphology. This could be addressed by reducing the text related to the technical issues as computational efficiency, strengthening the case study part maybe comparing the results of the approach with an expert classification and drafting a discussion chapter as mentioned above.

• Resubmit after major revision: 2017-10-24

• Minor revision: 2017-12-03

  • Reviewer: The authors have made a great effort to address most of the issues raised in my previous reviews and the manuscript has been improved. In particular, I appreciate the authors’ effort in revising and reorganizing the manuscript in order to highlight the geomorphometric approach devised in their work, reducing the information on the implementation and, thus, making the manuscript more suitable for a journal as Geomorphology and its readers.

• Resubmit after minor revision: 2017-12-03

• Accepted: 2017-12-15

Citation

Zhu, L.J., Zhu, A.X., Qin, C.Z., and Liu, J.Z. 2018. Automatic approach for deriving fuzzy slope positions. Geomorphology, 304: 173–183. doi:10.1016/j.geomorph.2017.12.024