GIS 5100 - Lab 7 - Coastal Flooding

This weeks lab focused on Coastal Flooding and predicting the impact of sea level rises or storm surges.  The two part lab covered Hawaii in Part A and Florida in Part B.  

Our task in Part A was to concentrate on the District of Honolulu and use 2010 Census data to report population density and demographics of the areas affected or not affected by a 3 foot and 6 foot rise in sea level.  The map below answers the question, what is the population density in areas with flooding due to a 6 foot rise in sea level.  The map shows the flood depth in the impacted census tracts.  Looking at the table one can see that the coastline of Honolulu where there are likely expensive high rise condominiums or hotels and therefore less owner-occupied housing units with higher non-native or white populations is most vulnerable to sea level rises.  

What is the population density in areas with flooding due to a 6 foot rise in sea level?

Variable                 Entire District   3 feet Scenario              6 feet Scenario

                                                      Flooded   Not-flooded   Flooded   Not-flooded 

Total Population      390,738             8,544        382,194         60,005     330,733

% White                     19.48              36.79        19.09             29.58       17.64

% owner-occupied  48.98                 32.22        49.36             36.97       51.16

% 65 and older        18.13                7.41         18.14              17.18       18.30

________________________________________________________________

This part of the lab was labor intensive with several table joins and statistical analysis.

Part B focused on Collier County, Florida and was a comparison of two different DEMs of the same area, a Lidar DEM and a USGS DEM.   I think naming this a USGS DEM is misleading as USGS has data from Lidar as well.  I think the USGS DEM is from the National Elevation Dataset with a 1 arc-second ~30 meter resolution.  Regardless, Part B's coastal flooding analysis showed how important image resolution is when selecting a DEM to predict sea level rise or storm surge impact.  The reliability of the USGS DEM was much weaker than the Lidar DEM as born out in our % omisson and % commission calculations.  The Lidar image had a cell size of 25 sq. ft and the USGS image had a 100 sq. ft. cell size.

Both Part A and Part B took me much more time than the 7 hours described in the assignment.  Partially because I'm interrupted a lot while working on assignments but this lab was definitely more time intensive for me.

It was also coincidental that in the news this week, both locally and nationally, there were articles regarding coastal flooding and rises in sea level.

http://gatehouseprojects.com/risingseas/home/site/nwfdailynews.com

http://www.sfgate.com/bayarea/article/Marin-San-Mateo-County-sue-big-oil-over-climate-11294549.php

GIS 5100 - Lab 6 - Crime Analysis

This weeks assignment had us learn about different crime analysis methods.   We focused on the Grid-based, Kernel Density, and Local Moran's I to analyze burglaries in 2007 in Albuquerque, New Mexico.  This was a very interesting and informative lab assignment. Here are my analysis steps for each method.

Hotspot Map 1 – Grid-based

1.    I performed a “Spatial Join” between the Grids and the Burglaries2007 shapefiles.

2.    I used an SQL Query to select only cells with a burglary count > 0 from Join_Grid_B2007 to create a new shapefile, J_GridB2007_O.shp

3.    Next, I sorted in descending order the attribute table on the Count field.   There were 507 total features and we were tasked to use quantile classification to create a top 20% quantile.  20% of 507 features was 101 features.   When I selected the 101 features with the highest count values I noticed that features 99-101 had a count value = 15.   There were other features that also had a count value = 15 but were not in the top 20%.  I created two new shapefiles.  The first, J_GB2007_20Pct contained 101 features.  I noticed that my new shapefile did not exactly match the sample image shown in the lab.  I used the Identify tool to see the count value of the three grid cells that were appearing in my image and not in the sample image.  My suspicions were correct, the three cells had count values = 15.  I created another new shapefile but this one had 98 features with count values >= 16.  The new shapefile, J_GB2007_19Pct was actually the top 19% grid cells.

4.    I added a field, DISSOLVE, to the attribute table and assigned a value of 1 to all records.

5.    I then used the Dissolve tool to create a single polygon of these grid cells.

My input feature was the J_GB2007_19Pct.shp and the output feature was GB2007_19Pct_D.  The dissolve field was my newly created DISSOLVE field and I made sure the option to “create multipart features” was checked.

Hotspot Map 2 – Kernel Density

1.    I set my environment to use the Grids shapefile as the extent boundary and Raster Analysis mask boundary.

2.    I used the Kernel Density tool under Spatial Analyst – Density.  My input feature was the Burglaries2007 features.  I left the Population field set to None.  I named my output raster KDBurg2007.  I set my output cell size to 100 feet and set the search radius to 2640 feet (.5 mile).  The area units was set to square miles.

3.    After running the tool under the Properties\Symbology tab I changed the classification to 3 classes using the Manual classification method and set Exclusions to 0 to exclude values = 0.  From the Classification Statistics box I recorded the non-zero mean value as well as the min and max values.  My three classifications were 0 – mean, mean – 3*mean, 3*mean – max

4.    The assignment said to select any area with a density of “at least 3 times the mean” so I used the raster tool Greater Than Equal to create a reclassified raster.  I then used the Raster to Polygon tool to create kdb07ge3mpy.shp  I used an SQL query to select records with a GRIDCODE = 1 and I had my Kernel Density polygon – KDB07GE_3M.shp

Hotspot Map 3 – Local Moran’s I

1.    I performed a “Spatial Join” between the Blockgroups2009Fixed.shp and the Burglaries2007.shp -  J_BG09F_B07.shp

2.    I added a new field in the attribute table called crime_rt

3.    I used the field calculator to populate the records with the number of burglaries per 1,000 housing units – crime_rt = Count/HSE_UNITS * 1000

4.    I ran the Cluster and Outlier Analysis under Spatial Statistics Tools\Mapping Clusters.  I used my joined block group (J_BG_09F_B07) as my input feature class and the crime_rt field as my input field.  My output feature was titled COA_BG09F_B07.shp

5.    I used a SQL query and selected only HH (high-high) clusters (areas with a high crime rate next to other areas with a high crime rate) and created COA_HH.shp

6.    I added a field, DISSOLVE, to the attribute table and assigned a value of 1 to all records.

I then used the Dissolve tool to create a single polygon of these grid cells.

Final Map – My final map displayed all three hotspot crime analysis methods.  I used the transparency setting to allow the user to see all shapefiles on top of each other.

Comparison of Different Analysis Methods for Crime Hotspot Mapping