Damage Assessment

This week, we looked at how to assess damage post-storms using aerial imagery. We compared raster files of a neighborhood in New Jersey from before and after Sandy. We had to assess damage done to properties based on these and rate it on a scale. We created a new feature class with points for each structure and including the information on damage and structure type in each point's tabular data. Next we looked at how close the structures were to the coastline, tallying the number of structures in each damage assessment based on location: 100m, 100-200m, 200-300m from the coastline. 

The pattern is very clear that the closer a structure was to the shoreline, the more likely it had major damage or was destroyed. The analysis shows that all structures within 300m of the coastline were at least affected by damage. The closer to the coastline a structure was, the more likely it was destroyed or had major damage. All destroyed structures were within 200m of the coastline and 83% of the major damage occurred within this distance as well. For those 200-300m from the coastline, over 90% of the damage was ranked as affected or minor. 

Structure Damage Category	Count of Structures 0-100 m from coastline	Count of Structures 100-200 m from coastline	Count of Structures 200-300 m from coastline
No Damage	0	0	0
Affected	0	9	10
Minor Damage	1	15	28
Major Damage	5	10	3
Destroyed	5	5	0
Totals	11	39	41

I think the aerial survey is a great method for a precursory look at what the damage could be, but I would hesitate to extrapolate overall damage without surveying a few more blocks using this method. Additionally, I think ground truthing is important to ensure quality control/quality assurance of the data. It was quite simple to see destroyed homes in the aerial, but judging minor vs major damage from sampling looking at if roofs are still intact is not going to give you the full picture of the impacts.

Coastal Flooding

This week we looked at coastal flooding through several projects focused on New Jersey and Southwest Florida. I was pretty excited to learn some of these techniques as this is something we consider in a lot of work I do on climate change impacts and cultural resources.

The first project involved comparing LiDAR data from pre- and post-Sandy on coastal New Jersey. We converted the point clouds into rasters and compared them to see how the storm affected the coastline. The result shows areas of erosion from the storm, including several areas where the ocean breached the island. 

The second portion of the assignment included calculating storm surge for Cape May County, NJ. This process involved using a DEM to find areas of elevation at 2 meters or lower and creating a polygon of this area. These areas represented where storm surge could flood the landscape. We had to do a simple math problem to figure what percentage of the county this area covered.

The last project was to calculate potential storm surge in Southwest Florida using both USGS and LiDAR derived DEMs. We used a similar process as in Cape May, selecting out pixels from a raster that were 1 meter or less and creating polygons for each data source. We used these polygons to select which buildings could be impacted by storm surge, comparing both methods to see how many structures were included in each model. The assignment assumed the LiDAR would be more accurate as it had a more nuanced representation of the landscape. It selected out less structures as it covered less area.

Visibility Analysis

This week, we worked through four Esri modules on Visibility Analysis. Here's a few observations, tips, and tricks I learned:

You can created some pretty advanced 3D landscapes in ArcGIS, with control over the textures and imagery. You can even set the sun to cast shadows as it would during a certain time of day in the global view!

It's important to consider where to set the ground surface when working in 3D scenes. Do you want the height based from the ground itself, the relationship to the ground, or an arbitrary point?

You can map things under the ground surface. Mind boggling and super important as an archaeologist!

You must first make the lines before conducting the line of site analysis. ArcGIS will color code the lines for you and it's easy to select out and delete lines that do not have intervisibility.

The viewshed analysis tools are pretty extensive and can be used for a lot of different things. I didn't realize you would end up with a raster depicting what is visible. To have this make a little more sense, especially when thinking about topography impact on viewsheds, you can drape the raster across a terrain in a 3D scene.

You can easily convert 2D layers and data to 3D layers and data in several different ways. The 2D features can be extruded to make them into 3D features. 

LiDAR

This week we played with LiDAR data to analyze the tree canopy in portion of the Shenandoah National Park. I was very excited to play with this data as I hope to use it as an important tool in some of my archaeological research. I've learned that LiDAR data is amazing but very slow and clunky in ArcGIS. I'm used to working with large datasets of 3D point clouds, but the manipulation of this data in GIS is very different than in the other programs I've used. And it takes time to process and even to draw the data for visualization. Important things to keep in mind as I start my LiDAR journey.

We used the data to look at a few different things for the area. The first was creating a digital elevation model (DEM) and a digital surface model (DSM). To do this, we filtered the points classed as ground surface (DEM) and those classified as non-ground (DSM). 

Next we looked at the height of the trees by comparing the vegetation height of the DSM to the ground heights of the DEM.

Finally, we calculated the canopy density by converting the ground and vegetation points into raster files and using several tools to categorize the pixels and compare the two datasets.

Crime Analysis with Hotspot Maps

 This week, we created hotspot analysis maps using three different techniques and compared the outcomes. The maps looked at homicides in Chicago in 2017 and compared the usefulness of their predictions against the 2018 data. The three methods we used included Grid-based Thematic Mapping, Kernel Density, and Local Moran's I. 

Grid-based thematic mapping involves classifying the homicides into small gridded areas across the city. This is accomplished through spatial joins of the homicides with the grid layer and filtering only the grids with crimes. We further sorted out only the top quintile of grids for consideration. 

Kernel Density involves creating a raster file based on the point data from the homicides. You use symbology to help classify and visualize the data, in this case focusing on areas with crime densities at three times the mean or higher. You can convert the raster into a polygon.

Local Moran's I classifies the homicides by census tracts, using a spatial join of these two layers. The Moran's tool in ArcGIS Pro will provide a statistical analysis of the area and you can out significant clusters of high crime density by choosing those that come back with a high-high rating (areas with high homicide rates near other similar areas). 

When crunching the numbers, the Local Moran's I seemed to have predicted the most homicides for 2018 based on the number of incidents as well as the percentage. However, when you take into account the density of the incidents per square mile, this method had the lowest number. I would say the Kernel Density seemed the best map as it had the second most predictions in number, the highest density predictions, and the size of the area fell in the middle - meaning it could be a more manageable area for police to cover.

Grid-Based Thematic Mapping

Kernel Density

Local Moran's I