Final Project: Landslide Susceptibility in MN

Goals and Research Question
The goal of this project was to map landslide susceptibility throughout the state of Minnesota using multiple variables.  The research question pursued was "Which regions within the state of Minnesota are most prone to landslide hazards?'  In the end, the hope would be that a correlation would be revealed between landslide prone areas which had association with previous landslide occurrences.

Data & Methods
Data was acquired from Minnesota GIS Clearinghouse and Multi-Resolution Land Characteristics Consortium.  The DEM and soil data were acquired from the clearing house, and land cover was acquired from MRLC.  Slope, aspect, land cover and soil were all found to have significant roles in landslide occurrences.  Therefore, out of the DEM, slope and aspect maps were created using the slope and aspect spatial analyst tools.  Figure 1 below shows a slope map of the state of Minnesota.

Figure 1.  A slope map of Minnesota

Slope was reclassified into five categories.  The lowest slope represented lowest landslide risk, highest represented highest landslide risk (very low, low, moderate, high, very high). 

Aspect was reclassified, having highest landslide densities in the W, then N, E, S.  South was ranked lowest risk (south facing slopes typically have the lowest landslide densities).  East was moderate, north was high, and west was very high.  A flat value was assigned as very low risk.  

Land cover was reclassified and ranked from very low to very high as well.  Open water being very low, woody wetlands low risk, forest and shrub lands moderate risk, urban high risk, barren lands very high risk.


Soil erodibility, represented in K values, was reclassified and broken up into 5 classes from .02 to .55.  Figure 2 below shows the K Factor scale. .02 was considered very low, being composed of fine textures with high clay content, and .55 being ranked high.  This consisted of silt sized particles which were prone to crusting and were easily detachable. 

Figure 2.  A K Factor soil scale

After all the values were reclassified, they were inputted into the weighted overlay tool and a map was produced out of the combined variables.  Figure 3 below shows the weighted overlay inputs, and figure 4 shows the workflow.

Figure 3.  All four variables inputted into weighted overlay tool

Figure 4. A workflow of the data processing

Results
Figure 5 below shows the resulting susceptibility map compared alongside a historical map of landslides in the state of Minnesota.  The results revealed that there is a fair amount of relevance among the four weighted factors and previous landslide occurrences.  It's most prominent in the southwestern part of the state, particularly in Winona county.  The Hennepin County area appears rather moderate for the susceptibility map compared with the high concentration of landslide occurrences in the historic map .  A recommendation for future mapping could consider putting more weight on urbanization or population density as a landslide risk. 

Figure 5.  the resulting susceptibility map (left) compared alongside to a historic landslide map of Minnesota

Conclusions
The motive of this project was to model how GIS can be a useful tool in delineating areas of high or low risk for landslide hazards.  The results revealed a moderate correlation between the four chosen factors and historic landslides.  Future work could involve more a detailed analysis of landslide prone neighborhoods for urban planning.   


Sources
Chalkias, Christos, Maria Ferentinou, and Christos Polykretis. "GIS-Based Landslide Susceptibility Mapping on the Peloponnese Peninsula, Greece." Geosciences. August 20, 2014. www.mdpi.com/journal/geosciences.

Jennnings, C.E., M. Presnail, E. Kurak, R. Meier, C. Schmidt, J. Palazzolo, S. Jiwani, E. Waage, and J.M. Feinberg. "Historical Landslide Inventory for the Twin Cities Metropolitan Area." Minnesota Department of Natural Resources, 2016, 1-34.

https://files.dnr.state.mn.us/waters/watermgmt_section/shoreland/landslide-inventory.pdf.
Keller, Edward A. “Introduction to Environmental Geology.” 3rd ed. Upper Saddle River, NJ: Pearson Prentice Hall, 2005.

Lee, Saro , and Touch Sambath. "Landslide susceptibility mapping in the Damrei Romel area, Cambodia using frequency ratio and logistic regression models." Korea Institute of Geoscience & Mineral Resources 50 (April 2006): 847-55. Accessed March 3, 2018. https://geo.mnsu.edu/geo/Landslides-LCCMR/Susceptibility%20Mapping/Lee%20and%20Sambath%202006.pdf.

DEM and Geology data: http://www.mngeo.state.mn.us/chouse/metalong.html#elevation

Land cover data: https://www.mrlc.gov/nlcd11_data.php

Lab 3: Watershed Analysis

Goals and Background
The purpose of this lab was to have students gain an understanding of how to properly delineate watersheds.  The study area used for this particular watershed analysis was in Adirondack Park, New York.  A watershed is an area of land which surface waters meet in a concentrated area.  This generally forms into a lake or stream.  Different hydrology tools within ArcGIS were used to delineate the provided watershed features based on attributes such as topography.

Methods 

Part 1:
Adirondack Park data was initially gathered from a New York GIS governmental website, and a hydrology shapefile was downloaded from Cornell University's CUGIR site.  The digital elevation model of North America was gathered from ArcGIS online.  The park boundary and hydrology shapefiles were initially under different coordinate systems, so the project tool was used to turn it into the correct projection, as well as changing the cell size to 60 by 60 meters.  After that, it was clipped to the buffer layer.  


Part 2:
The flow directions were then calculated by utilizing the Flow Direction tool.  The sinks were removed by using the fill tool.  After this is completed, the flow direction tool is used again to calculate the flow direction.  Once completed, the area where the water collects must be assessed by using the flow accumulation tool. The watersheds were then delineated by choosing a threshold of 50,000, 100,000, and 500,000.  Unique identifiers were then used for each stream with the help of the stream link tool.  The watershed tool was then used in order to be able to clip the park boundary.

Results
Figure 1 shows the delineation of watersheds of the park using a threshold of 50,000.  This resulted in 174 unique watersheds.  Figure 2 shows the same thing, other than the fact that the cell size was 120 by 120 meters rather than 60 by 60.  The total watersheds resulted in 16.  Figure 3 is a map which shows the vector streams of Adirondack in comparison to the original hydrology feature.  The original, having a cell size of 60 by 60, had considerably more watersheds than the vector, being 120 by 120.

Figure 1. A map which delineates watersheds of Adirondack Park with a threshold of 50,000

Figure 2. A map displaying delineated watersheds of Adirondack Park, with a cell size of 120 rather than 60m.  

Figure 3.  A map comparing the original hydrology streams vs the vector streams

Sources 
(1) Adirondack Park boundary shapefile from the New York State GIS Clearinghouse at
http://gis.ny.gov/
(2) A hydrology shapefile from Cornell University's Geospatial Information
Repository (CUGIR) site at http://cugir.mannlib.cornell.edu/index.jsp
(3) A 30-arc-second digital elevation model (DEM) of North America accessed from ArcGIS online

Lab 2: Georeferencing and Data Creation

Goals and background
The purpose of this lab was to give students a review of georeferencing skills as well as digitizing and creating new feature classes.  This was done through georeferencing a city map of Eau Claire from 1878 and comparing it to a current map to visualize the change in water features over the last 140 years.

Methods 
The first part of this lab requires students to georeference an 1878 map of Eau Claire (shown in figure 1) and a 2018 map by using the georeferencing toolbar provided in ArcMap.  The old map is aligned to the current map by adding ground control points in the middle of street intersections.  The aim was to get the EMS error as low as possible, managing to bring it to approximately 1.36 with 29 GCPs as shown in figure 2.

Figure 1: 1878 map of the city of Eau Claire

Figure 2: A list of ground control points as well as the total RMS error

Part two of the lab had students create a feature class, digitizing polygons within it. All of the major water features in the 1878 image were to be digitized within the study area.  The same process was done for the 2018 map of Eau Claire.  This was conducted by utilizing the editor tool and selecting "create polygons". 

Results
Figure 3 below displays a georeferenced map of Eau Claire placed over a current map.  The river between 1878 and now do not completely line up, but georeferencing them still did an impressive job with overall alignment. 

Figure 3: A georeferenced map of Eau Claire (1878-2018)

Figure 4 below shows a map of the two water features.   As you can observe, the river is generally more narrow now than it was 140 years ago.  The difference in the amount of water in the study area is over 1,000,000 square feet less than what it was in the 1878 map. 

Figure 4: A comparison map of water features from 1878-2018

Sources
1. David Rumsey Map Collection https://www.davidrumsey.com/luna/servlet/detail/RUMSEY~8~1~4181~480085#

2. Master_Centerlines feature class (clipped), Eau Claire County

3. World Topographic Map, ESRI, 2018: https://www.arcgis.com/home/item.html?id=30e5fe3149c34df1ba922e6f5bbf808f

Lab 1: Review of ArcGIS Basics

Goal and Background

The purpose of this blog was to refresh students with using ArcGIS software, particularly with ArcCatalog and ArcMap.  Students were reminded of foundational file types such as points, lines, and polygons.  They were then reintroduced to utilizing ArcCatalog for previewing and organizing map layers. They also were asked to identify various features of a map document via the identify tool with objects such as streets.  Attribute tables were also observed to examine different layers.  Lastly, students were asked to make a couple of basic maps consisting of different variables in order to compare/contrast.  

Methods 

The first question of this lab required students to identify what the geographic coordinate system, projected coordinate system and projection of a shapefile provided in the given data.  This was simply done by pulling up layer properties and viewing the data source.  Another question asked students to identify associated attributes with a provided donut shop layer.  This was executed by right clicking the layer and opening the attribute table.  The next question asked to identify the street names of the "Major Highways" crossing the Redlands map.  Opening up the attribute table, the two major highway classes were State 30 and Orange St, as shown in figure 1.  

Figure 1 - Major Highway Class

Lastly, students were asked to create two maps with different variables of the city of Erie. The first variable chosen was population density.  The variable for the other map was median rent.  For population density, number of persons was the first variable and was normalized with square miles per tract in order to properly represent spatial patterns.  This wasn't necessary for median rent, as it was displayed as a single variable and already standardized.
  

Results
When comparing the two maps, it was surprising to see there was not a correlation between higher concentrations of people and higher median rent.  The original assumption was that city centres would be the highest for rent, but the maps for the most part proved opposite of that, as shown in figures 2 and 3 below. 

Figure 2: Population Density per sq mile

Figure 3: Median Rent (USD)

Sources
Dr. Caitlin Curtis, University of Wisconsin-Eau Claire, Dept of Geography and Anthropology