Thursday, October 19, 2017

Lab 7: Multispectral Analysis

Multispectral analysis involves using filters to view a multispectral image at different bands, or wavelengths, of radiation. Understanding how the different wavelengths interact with the environment allows me to better identify what is shown in the image. The task for this lab was to be given a criteria about a group in one or more bands for an image and for me to locate the area the criteria refers too. To do this I used various tools in ERDAS Imagine, like the images histogram or the inquire tool to inspect the pixels that make up the raster under various bands. I made a separate map for each feature I was to identify and explained on each how I identified them.

Project 3 - Statistical Analysis Prepare Week

Project number three for Special Topics in GIS is to use statistical analysis to draw a correlation between known methamphetamine lab locations an the social data provided the 2010 census. Since the first week is prepare week most of what I did was to prepare data for the analysis. Census data is divided up into areas called census tracts, which are designed to break the population up into chunks of roughly similar population. The problem with census data is that it's not really suitable for analysis in its unaltered state. One statistic will be for the number of single men living in a household by themselves. While that's a useful thing to know just having the number of single men living alone isn't useful for analysis because there is no greater context. In order for me to use it in analysis it needs to be normalized, which usually means to turn it into a percentage so that the data is meaningful. By finding the total number of households in that census tract and then dividing the number of single male households by that number we get the percentage of households in each census tract that belongs to a single male occupant. This is a meaningful statistic because it doesn't rely on ancillary information to make sense. The original number was useless without the context of how many households there were. The majority of the prepare week work was to manually, and then with a python script, turn a select number of census statistics into data that's meaningful for analysis.

The second problem was to create meaningful data to compare it against. I had a shapefile providing locations of meth labs busted around Charleston, WV between 2004 and 2008. Preparing the data wasn't as simple as locating the number of meth labs in each census tract because that data is similarly meaningless for analysis. Instead I needed to normalize that data spatially. I used a join tool to detect each meth lab within 100 meters of each census tract then divided that number by each tracts area, creating a density statistic for each tract. Next week I can use both sets of data I prepared to perform the final statistical analysis.

Saturday, October 14, 2017

Lab 6: Spatial Enhancement

Most of this lab was an explanation of how various filters affect imagery. There are different effects and transformations you can use to alter your data for easier analysis. The specific example we used in this labs deliverable was an image from the damaged platform Landsat 7. The imagery from Landsat 7 has an effect called striping, where diagonal bands of null data cross each image at regular intervals. I used a technique called Fourier Transformation to reduce the impact of these null bands by warping the data on either side of the bands closer together. Then I used other filters to attempt to sharpen and clarify the undamaged data that was left.

After the fourier transformation I ran my raster image through a sharpening filter. The 3x3 sharpening kernal filter works by analyzing the cells in groups of nine, in a 3x3 square. The filter then emphasizes the difference in the each cell compared to the ones adjacent to it, making the image appear sharper. With the newly sharpened image I then ran it through a standard deviation focal tool. This uses a similar technique, it examines cells in a larger group and reduces the impact of cells that have the same value. The reason I did this is because the remaining bands were all the same cell value, and the standard focal tool helped eliminate them further. The resulting image can be seen below.

Wednesday, October 11, 2017

Mountain Top Removal - Appalachian Mountains

Report week for the Mountain Top Removal (MTR) project. This week is the culmination of my previous efforts along with my fabulous Group 1 to analyse and present a story map showing that analysis of MTR in West Virginia. This link leads to a web map showing my groups combined polygon layer of our MTR analysis. The final week involved using spectral data of the area I was assigned in ERDAS Imagine and comparing it to a landsat image from 2010 in order to reclassify each band as belonging to an MTR site or a non-MTR site. Then I used that reclassified raster to create a polygon layer that I then removed the rivers and streams from, to prevent false positives with the spectral data from the stream beds. The final polygon was then accuracy checked with random points and presented to my group leader so that he could merge it with the rest of my groups polygon layers.

Here I'll show you an example of our polygon layer over a background elevation layer and then the same area with a historical elevation map behind it, showing the changes made by MTR over the years.

Finally, my journey of learning about, analyzing, and presenting my findings about MTR have all been collected in a single place, my MTR Map Journal.

Thursday, September 28, 2017

Intro to Electromagnetic Radiation

The title sounds like this is going to be way more fantastical than it probably is. This week I went over the basics of the visual and non-visual spectrum for Electromagnetic Radiation (EMR) and how it effects remote sensing. Different methods of remote sensing use different EMR bands and this changes what data they collect and how they collect it. The lab this week used ERDAS Imagine to examine different types of raster imagery of the United States and some specifically of Mount Olympus in Washington State. Much of the lab dealt with an introduction to the math behind EMR and as an introduction to ERDAS Imagine, which is a software engine designed to analyse and modify raster imagery. The final part of the lab involved selecting an area of the Mt. Olympus thematic raster to display. The raster imagery was rendered in pseudo color, which displays the thematic image by associating each identified class with a specific color for easier examination.

Friday, September 22, 2017

Special Topics - Mountain Top Removal Analyze week

The second week of the Mountain Top Removal project was the analysis week. I was assigned one of my groups LandSat images, taken in 2010, for the purpose of discovering the spread of Mountain Top Removal (MTR) from the 2005 images. The first step in the analysis is to take the landsat images and clip them down to just the area my group is assigned to. Once my area has been isolated the image was opened in ERDAS imagine which broke the pixels of the image apart by their elevation data, classifying them into fifty different classes. Then I manually identified which of those classes were part of MTR practices based on my previous examination of the 2005 MTR polygon and raster files. With the elevations classified between MTR areas and normal mountain range the resulting raster file was brought into arcmap and reclassified to exclude everything but the potential MTR areas. The result is this Raster image showing the areas which I suspect to be expanded MTR, and also some rivers.

Wednesday, September 20, 2017

Photo Interpretation - Truthing

Truthing is the process of examining your Land Use and Land Cover classifications using supplementary data or on site examination. This week I took the same map I had classified last week and used supplementary data, also known at google streetview, to confirm 30 semi-random points placed on my map. I spread the points fairly evenly across the polygons on my map but they were placed randomly inside the polygons. Then I located the point on google maps and used street view to confirm that the point fit the classification I gave it. Even though the area the point is in would be classified as "residential" if the point itself lands on top of a road it would be flagged as inaccurate. Once all the points are truthed the map is graded based on it's classification accuracy, mine was 73% accurate.