Unit 3: Develop and Test Models of Landslide Susceptibility
How do geoscientists confidently create landslide susceptibility maps based on empirical data? Together with Unit 2: Examining the Distribution of Mass Wasting Events, this exercise helps students use GIS to gain quantitative landslide analysis skills. In this unit, students model and statistically validate landslide susceptibility maps using a combination of some, or all, of the landslide susceptibility values calculated in Unit 2. From their model validations, students then propose and defend a final landslide susceptibility map to be adopted for use in risk assessment. This unit uses ArcMap software.
Unit 3 Learning Outcomes
- Describe the role that predictive models play in the sociopolitical landscape, their intended use versus actual outcomes, and their limitations
- Differentiate between various mass movement susceptibility models and justify the use of each in small groups.
- Explore the effect of varying input data to model output by quantitatively evaluating model outputs and classifications with ROC curve analyses.
- Reflect on the unit exercises and reevaluate prior ideas or notions about modeling, its validity, and role in social change or policy making through a synthesis discussion.
Unit 3 Teaching Objectives
- Cognitive: Facilitate guided exploration of the development, analysis, and validation of model ouputs.
- Behavioral: Develop skills in computational experimentation and model validation.
- Affective: Encourage critical analysis of the relationship between predictive models and their application in the sociopolitical landscape.
Context for Use
This upper-level content is appropriate for junior- or senior-level courses in geohazards, geomorphology, GIS, remote sensing, and other geoscience courses where there may be an interest in developing skills in computational experimentation and model analysis and validation. This unit assumes that students have had significant exposure to GIS software/tools. The content in this unit is intended to follow Unit 2: Examining the Distribution of Mass Wasting Events, and it is necessary that the two units be completed together. This unit is best introduced after some ArcMap or QGIS skills have been practiced, so that students are familiar with GIS software, tools, and making maps. The content is designed to be executed as guided exploration: a combination of discussion, experimentation and analysis, and reevaluation. It is suggested that students work in small (4-5 person) groups. A very brief introductory discussion will help students engage in conversations about the sociopolitical space that susceptibility models can/do occupy.
This is a more advanced and quantitative approach than found in the introductory-level GETSI module Surface Processes Hazards: Living with Landslides - Unit 5: Mitigating future disasters: developing a mass-wasting hazard map
Description and Teaching Materials
Estimated time: Classroom (1 hour), Laboratory (3 hours), and External (3-4 hours). Additional time may be needed depending on students' experience with GIS software and tools. This unit may be shortened by having students only run one, or two additional iterations of the susceptibility model rather than the suggested three.
In Unit 2: Examining the Distribution of Mass Wasting Events, students were tasked with exploring single factors as an 'early stage landslide susceptibility map'. This module is inherently linked to Unit 2 by means of uniting single factor maps to generate multi-factor susceptibility maps. These maps are quantitatively and qualitatively assessed using receiver operating characteristic (ROC) curves and visual analysis, respectively. Susceptibility maps explored in this unit fit into the context of the larger module by encouraging students to critically evaluate the role and relevance of various factors towards prediction of landslide susceptibility.
Suggested reading in advance of this unit is a review of landslide susceptibility modeling include the peer-reviewed journal articles provided in the 'References and Resources' portion of Unit 2 materials (also provided in this unit).
Class discussion will help students recognize the role that predictive models play in the sociopolitical landscape. To facilitate such discussions, a presentation slide set with discussion prompts is provided. It focuses on the intended use versus outcomes of models, and model limitations. These discussions can be done as part of a think-pair-share rather than a whole class discussion.
- Unit 3 Role and Space of Modeling (PowerPoint 2007 (.pptx) 848kB Mar12 23) - this presentation includes goals for some of the discussion prompts in the notes portion.
Susceptibility Modeling & Analysis
Following the introduction, students should select a single region (Arizona or Puerto Rico) to continue working on for the duration of the module. Prior landslide susceptibility index (LSI) maps from Unit 2 should be shared with all the students, to encourage a deeper exploration of meaningful final susceptibility models. Towards developing a final susceptibility model, students will learn how to create susceptibility maps based on various factors and assess them using quantitative and qualitative methods. Additionally, students will analyze their maps, focusing on the distribution of final predicted risk classifications.
- Unit 3 Preparing a Final Susceptibility Map (Microsoft Word 2007 (.docx) 299kB Mar12 23) - this document includes steps on how to generate final susceptibility maps and assess them.
Quantitative and qualitative assessment of final susceptibility maps are done using ROC curves and visual analysis, respectively. The following presentation provides an introduction and step by step for how these two assessments can be done, and includes additional points for consideration and discussion. It is recommended that instructors present this material before the preparation of susceptibility maps, or between the classification of the maps and the assessment stage.
- Unit 3 Quantitative and Qualitative Models Assessment Presentation (PowerPoint 2007 (.pptx) 899kB Mar12 23) - this presentation includes background and methodology for quantitative and qualitative assessment of susceptibility models using ArcMap.
Generating/Assessing Landslide Susceptibility Models
In this section of the unit, students (in groups of 4-5) will be responsible for generating at least three susceptibility maps (models) using LSI maps and data from Unit 2 on a single region of their choosing (Arizona or Puerto Rico). Students will then assess each model both quantitatively and qualitatively and select a final map they feel they can justify as the 'best' susceptibility model. Students will need to generate tables and maps to be utilized in a poster presentation (made through Microsoft Powerpoint). The final unit deliverable for summative assessment is a poster presentation that the instructor may choose to have students present to the class, or be graded independently.
If this unit is being preceded by Unit 2, then it is recommended that students share their data and maps from Unit 2 with the whole class to allow for a more robust experience in selecting not only different parameters, but different interpretations (classifications) of each parameter. Alternatively, prepared LSI maps are available for use (see Unit 3 Datasets below).
- Unit 3 Susceptibility Maps Student Poster Presentation Instructions (Microsoft Word 2007 (.docx) 202kB Mar12 23) - this document describes the final summative assignment, a scientific poster, for the students. Together with the summative assessment rubric for this unit, students should be able to develop a strong scientific poster. The final deliverable is meant to be a submitted assignment for each group, but instructors may choose to have students present their posters to the class.
End of Unit Discussion
At the end of the assignment and poster presentations, it is recommended that teachers guide their students through some discussion points that will help provide a close to the unit, but also provide an opportunity for metacognition and an opportunity to revisit the discussion offered in the introduction. Suggested discussion questions are below:
- What were some model limitations for your models? Why do these limitations exist?
- What role could your models play for individuals, in society, or in politics?
- Why might someone disregard your groups' final selected model?
- How could you communicate the worth of your model?
- This unit concludes predictive modeling (first introduced in Unit 2). How simple do you think predictive modeling is now? Does this differ from your response in Unit 2?
- What aspects of the modeling process do you think make it vulnerable when the results challenge social, political, or economic status quo?
- How has your opinion of modeling, its results, and its uses in guiding policy changed? Do you find yourself more, or less, critical? Explain.
A guide/key for the discussion prompts are provided here:
Unit 3 Datasets
- AZ_LSI (Zip Archive 75.3MB Jan9 19) - Arizona LSI factor raster data. This data is to be used if Unit 2 LSI factor datasets were unable to be generated, were corrupted, or if Unit 2 was skipped. The legend of classifications used is found below.
- PR_LSI (Zip Archive 18.9MB Jan10 19) - Puerto Rico LSI factor raster dataset. This data is to be used if Unit 2 LSI factor datasets were unable to be generated, were corrupted, or if Unit 2 was skipped. The legend of classifications used is found below.
- Unit 3 Classification Legend (Excel 2007 (.xlsx) 11kB Jan9 19) - legend of classifications used for Unit 3 data.
Teaching Notes and Tips
- This unit uses ArcMap software. See Unit 1 for related links.
- For the generation of LSI value field rasters and final susceptibility maps, it is recommended that a common server geodatabase, or folder be used. If this is not possible, make sure all outputs are saved into a single folder for the students. The final powerpoint scientific poster may be a large file, so delivery of the assignment by external drive or networked common space is recommended.
- Readings (see additional resources below) provide a good transition between Unit 2 and Unit 3. Have students (in particular) read the Chalkias et al. (2014) paper to better familiarize them with the concept of final susceptibility maps and ROC analysis.
Formative assessment for this unit is recommended as observation and feedback on student group discussions and questions. Additionally, further formative assessment can occur as student groups debate the combination factors to be used for their models; the teacher may choose to briefly touch base with each group to provide guidance and correct any misconceptions. Student metacognition is presented as part of suggested discussion guiding questions at the end of the unit (see above).
The powerpoint scientific poster is the summative assessment for Unit 3. Instructors may use this to evaluate how well student groups understand the modeling and assessment process. The rubric below can be modified to assign point values in a manner consistent with their course scheme.
- Unit 3 Poster Summative Assessment Rubric (Excel 2007 (.xlsx) 16kB Jan10 19) - this is the rubric for the scientific poster. It is recommended that teachers provide a copy to the students, or at least walk them through the grading criteria in advance of preparing their posters.
References and Resources
Additional Resources for Instructors:
Provided here are some additional readings on LSI/Frequency-Ratio methods. Teachers may choose to have students read these papers in advance of the class/lab to further motivate discussion. These resources are identical to those provided in Unit 2.
- Chalkias C., M. Ferentinou and C. Polykretis "GIS-Based Landslide Susceptibility Mapping on the Peloponnese Peninsula, Greece," Geosciences, 4, 176–190. (2014).
- Lee S. and B. Pradhan "Probabilistic landslide hazards and risk mapping on Penang Island, Malaysia," Journal of Earth System Science, 115, 661–672. (2006).
- Lepore C., S. A. Kamal, P. Shanahan and R. L. Bras "Rainfall-induced landslide susceptibility zonation of Puerto Rico," Environmental Earth Sciences, 66, 1667–1681. (2012).
- Li L., H. Lan, C. Guo, Y. Zhang, Q. Li and Y. Wu "A modified frequency ratio method for landslide susceptibility assessment," Landslides, 14, 727–741. (2017).