Philip's Geospatial Portfolio

Welcome

2008-09-28T19:22:00.000-07:00

Hello, and welcome to my portfolio. The goal of this blog is to give a little more information about myself and show some of the projects that I have completed while I was in school. Since I am currently engaged in a job search, I felt compelled create something that gives a little more insight about me and my experience beyond what a resume can provide.

Feel free to have a look around and send me an e-mail if you have any questions, or feedback.

Regards,

Philip

A bit about me

2008-09-27T19:41:00.000-07:00

I graduated in August 2008 from UCLA with a degree in Geography/Environmental Studies and a minor in Geospatial Information Systems and Technologies. Since graduating, I have moved to Portland with the intent to begin a career in the Geospatial industry.

Tioga Pass DEM

2008-09-01T19:02:00.000-07:00

For this DEM project, I chose to model a portion of Yosemite National Park in the central Sierra Nevada Mountains. I acquired the DEM data from the USGS seamless website. After acquiring the appropriate data, I decided to acquire GeoTiff raster data from the Global Land Cover Facility (GLCF) from the University of Maryland. I acquired ETM+ plus data for the same area and used ENVI to cut the Geotiff to an appropriate size and then combined layers three, two and one to form an RGB image. I then placed my TIF file into ArcGIS to place under the DEM for added realism. I then imported a land cover shape file that I acquired from the GAP Analysis website. I then combined the land cover types into eight types from over 30 based on type (Deciduous, Coniferous, Etc.). I then added some transparency to the land cover layer again in order to preserve realism. I added the highway layer, then used the 3D analyst tool to determine the elevation change of the Tioga Pass Road that moves from Yosemite Valley over Tioga Pass to the eastern side of the Sierras. The approximate elevations involved are from approximately 4500 feet at Yosemite Valley to over 10,00ft at Tioga Pass and then back down to around 6000 ft. near Mono Lake. This elevation change is accomplished at a distance of approximately 60 miles. As you can see from the elevation profile graph in the lower right corner of the map the there is quite a variety of climbs and descents along the way.

Mozambique Channel Cartography Project

2008-09-01T12:32:00.000-07:00

This historical map of the Mozambique Channel area is one of a series of maps created by me for Edward A. Alpers professor of History at UCLA. He gave me a list of place names for which I researched latitude and longitude coordinates and entered them into an excel spreadsheet. I acquired the appropriate shapefiles for ArcGIS and imported the XY coordinates from the spreadsheet. After the professor and I were reasonably pleased with the layout, the map was exported into Adobe illustrator for final element placement and cartographic detailing.

Change Detection

2008-08-30T21:44:00.000-07:00

The exersize was a change detection of the Dawa Co. Lake on the Tibetan Plateu. Using Landsat MSS, TM and ETM+ images band 4, a region of interest was selected the images subset and the TM and ETM images were resampled to the pixel count as the MSS for overlay purposes.
A band ratio adustment was performed on the TM image to negate the salt sediment that was showing up and all three images histograms were examined and the spectral threshold of the water was adusted accordingly. The image was stacked MSS = Red, TM = Green and ETM = Blue.
The change detection shows Dawa Co. is decreasing in size over time while it's small satellite lake is increasing.

Classification Comparison

2008-08-30T20:11:00.000-07:00

Images were subset calibrated and stacked bands 4,3,2 as RGB. The following land cover types were identified in the image.
1. Barren, 77 beaches, White, location Malibu and Santa Monica
2. Water, 53 Reservoirs, Black, location Santa Monica Mountains
3. Urban, 16 Mixed urban, Blue, location Downtown, Wilshire Blvd etc.
4. Forest Land, 43 Mixed Forest, Santa Monica Mountains
5. Water, Pacific Ocean, Black
6. Rangeland, 33 mixed rangeland, Greenish Brown, Santa Ynez mountains

NDVI was performed to evaluate vegetation densities throughout the region.

The highest concentration of bright pixels represents areas of vegetation that are the most dense. Examples are golf courses, parks, and higher elevations of the santa monica mountains that have dense forest. The data range of these pixels is a positive value from 0 to 1. All of the pixels I observed were between 0 and .5
The highest concentration of dark pixels are in areas where there is no vegetation at all. Examples of this would be beaches, roads and dense urban areas such as downtown Los Angeles. The pixel value range would be between 0 and -1. The values I observed were all between 0 and -.5.

After the NDVI was performed, an unsupervised isodata classification was performed with 10 classes.

Dims: Full Scene (3,897,600 points)

Basic Stats Min Max Mean Stdev
Band 1 1 10 4.527976 2.901028

Histogram DN Npts Total Percent Acc Pct

Band 1 1 1070611 1070611 27.4685 27.4685
2 280327 1350938 7.1923 34.6608
3 204523 1555461 5.2474 39.9082
4 290595 1846056 7.4557 47.3639
5 326200 2172256 8.3693 55.7332
6 666843 2839099 17.1091 72.8422
7 554090 3393189 14.2162 87.0584
8 119768 3512957 3.0729 90.1313
9 74236 3587193 1.9047 92.0359
10 310407 3897600 7.9641 100.0000

1. Red- Water- Oceans and Lakes
2. Lt. Green – Mixed forest vegetation
3. Blue – Lower vegetation density natural grass lands, some agricultural crops.
4. Yellow- Grasslands, dead vegetation
5. Lt. Blue-Roads, Pavement
6. Pink – Urban structures Buildings
7. Maroon – urban mixed environment
8. Dk. Green – Bare Earth such as Beaches
9. Purple – flat vegetated areas, cemetery golf course
10. Mixed Terrain urban suburban (maybe Shadows?)

Classes 1,6,7 have a high percentage of pixels in the image, while 2,3,4,5,10 are in the middle range and class 8 and 9 are low. This would indicate that Water and Urban features are predominant in this Image.

New Orleans Elevation Analysis

2008-08-30T19:20:00.000-07:00

In the aftermath of Hurricane Katrina and the controversy surrounding the rebuilt levee system of New Orleans, I decided to do a project to examine elevation of the city to better understand the issues and challenges in keeping the city from flooding. This also gave me a chance to use some more detailed data sets, since Katrina the publicly available data for the area is unparalleled in it's accuracy. For the DEM image I was able to acquire a 1/9th arc second DEM as opposed to the typical 1/3rd arc second and allowed me to have accurate elevation information down to three meters instead of ten. This DEM image is coupled with a 1 meter/pixel digital orthophoto of the area to allow for a block by block analysis of the elevation. After overlaying the image on the DEM I was shocked to see how much of the City is far below sea level. With rising sea level and potential hurricane strikes an almost yearly threat, large scale flooding in New Orleans future is likely.

UCLA 2 Site Analysis

2008-08-30T19:17:00.000-07:00

The basis for this project was to complete a suitability analysis for a second UCLA satellite campus. Four factors were considered for the campus: Access to transportation, household income, earthquake fault lines and slope. Raster images were created from highway and earthquake fault line buffers, census data and DEM slope. The raster calculator was then used to calculate a site that was close to freeways, far from earthquake faults, in an area that lower income students had access to and was on level ground. A final city was chosen based on its raster calculator score and being suitable distance from the main campus in Westwood.

Precipitation Data - Spatial Interpolation

2008-08-30T19:05:00.000-07:00

In this project Interpolation was used from the spatial analyst extension was used to create a map precipitation in Los Angeles County. The data from each rain gauge station in the county was entered into an excel spreadsheet and the annual average and year to date totals were compared. The top image shows the difference between totals at each site with red being the most different and blue being most similar. Below the primary map are the two maps showing annual and year to date precipitation with red being lowest rainfall and blue highest rainfall.

Willow Fire Analysis

2008-08-30T18:46:00.000-07:00

This project began with the creation a geodatabase in ArcCatalog and importing the related shape files into it. I then used ArcToolbox to change the projection to UTM zone 11N, and use the Union and Dissolve tools to merge the separate fire daily burn areas. The Intersect tool was used to clip the vegetation type map within the border of the burn map. Excel was used to create a chart showing the vegetation types that were burned with the data coming from the attribute table of the intersected vegetation map.

Lidar DEM Rendering

2008-08-27T20:15:00.000-07:00

This DEM was created in ArcScene from a DBF(Dbase 4) file of LiDAR returns. The second series of returns were used in order to represent the bare earth data. After the returns were added into ArcMap using add XY Data. A TIN of the data was created using the 3D Analyst toolbar and then Hillshade was produced from the TIN. The HillShade was brought into ArcScene for the 3D rendering. The DEM is not a true bare earth DEM as you can see some man-made structures in the rendering.

Classifcations Part 2

2008-08-02T21:49:00.000-07:00

Salmon Crisis Map

2008-08-01T13:10:00.002-07:00

Click on image for larger view.

This map was created to show the rapid decline of pacific chinook salmon populations between 2002 and 2007. The basic state shapes were imported from ArcGIS into illustrator where it was maniputlated and the table and population graphics were created from scratch.

Soil Map Project

2008-08-01T12:38:00.000-07:00

This map was created in conjunction with an exercise in a soil and water conservation class, and was created as a planning map that would show potential soil related development hazards in the area. The soil map was digitized from a NRCS soil map book of Fresno County California and was geo-referenced using Township and Range coordinates. Polygons were created to lay on top of certain area in order to identify soil hazards but had some transparency so as not to obscure the soil map.

Rio Purus Project

2008-08-01T12:26:00.000-07:00

This project was completed for Dr. Susanna Hecht, Professor of the school of Public Policy UCLA. The project dealt with two historic maps of the Purus river in the Amazon Basin. The maps were digitized then geo-referenced. Afterwards the river channels of each map were traced, and shape files were created, in order to compare the maps with satellite imagery. Mr. Sid files of the area were downloaded from the JPL Zulu website and regions of interest were identified for closer analysis of geomorphological changes. Unfortunately although the maps were highly accurate for their time, they did not have the accuracy necessary to complete a change detection analysis against the satellite imagery.

GIS Programming Project

2008-08-01T12:22:00.000-07:00

The GIS programming project was designed to automate site analysis. This was done by creating a custom user interface and then modifying several scripts downloaded from the ESRI development network to automate such processes as buffer conversion vector to raster, raster calculator and raster to vector. Over 1000 lines of code were used to create this project.

Assignment 5 - Introduction to Cybercartography: Building a Google Map Mash‐up

2008-07-30T21:40:00.000-07:00

View Larger Map

I treated my google mashup as more of a "proof of concept" than a final product. The idea behind my mashup was to compare apartment rentals throughout the city. What I did was to find apartments on craigslist, which on this map I limited to 1 bedroom apartments in the central LA area. I put about 20 push pins on the map and when you click on each push pin the title is rental price and there is a link in the push pin back to the original craigslist posting. I envision using this in a couple of ways, on a smaller scale you could create a personal "shopping list" that you could easily take with you on a mobile device such as an iphone to go apartment hunting. On a larger scale with many many more entries you could really get an idea of comparible rentals in a single neighborhood, which could be useful for a renter or landlord as a way to get "comps". You could also compare neighorhoods against each other, or compare against demographic data for research purposes. If there was a way to mine the data from craigslist to automate the process you could probably create a site like zillow.com for renters, which to my knowledge doesn't exist any real capacity.

Assignment 3 - Making a better map

2008-07-17T13:19:00.000-07:00

Click on a map for a larger image.

Part 1

The equal interval classification is a very basic classification that is derived from taking the lowest and highest value and creating breaks at equal intervals between those values. For my map I chose to make 5 classes in order to give as detailed a picture as possible but try to preserve clear color definition between the classes. I chose to use a single hue in the classification for this map as I felt it would give a good sense of low to high by going light to dark, no need to make it too complex. I felt that this classification did not convey as much information about the population distribution over the state as I would like, due to the range in values from low to high being so vast and the distribution being skewed, basically 95% of the county's are in one class. I would not choose this classification for this particular data. If the range were narrower and the data more evenly distributed this classification could work.

The mean-standard deviation classification system was another classification system that I felt generally did not fit well with this type of data. Again if the population data fit within a normal bell shaped curve, it would likely create a more accurate representation of the population. I did like that the standard deviation classification system in Arcgis forces you to use a 2 hue color classification, as it does a good job of illustrating what data is above and below the mean, but that could pose a limitation on where you could use that classification (ie. black and white would get a little messy or difficult to read). The 2 hues allowed me to have more classes without complicating the map too much and improve precision without sacrificing readability. Overall, I would only use this classification for data that is in a normal distribution.

The natural breaks classification system was the classification that I felt was the most successful and gave an accurate result in a short time without having to spend the extra time coming out with a manual classification system. For simplicity I chose a single color classification that progressively darkened as the population increased. I chose six class breaks in order to give as much detail as possible but still give the reader the ability to distinguish between the classes on the map easily. The colors in the legend for the second and third classes are close but in the map itself they are easily distinguishable. I found myself occasionally spending to much time looking at the color gradient in the legend without actually taking the time to look see how the colors were working in the map itself. Overall I would likely try the natural breaks first as a classification scheme as it is a more intelligent algorithm to determine class breaks, unless I was trying to measure change over time, then it would not be appropriate since it will break at different points each time the data changed.

The quantile classification system breaks data up into classes of equal numbers from low to high with the closest value, so for example if you had 100 total data points and 5 classes there would 20 data points in each all in a certain range. This would work well if the data values were distributed relatively evenly across the distribution, but again with this data that is highly skewed with a large range it does not work as well as I would have liked. The key problem comes on the higher end of the distribution with the last classification break ranging from about 750,000 to over 9 million. This makes the map appear to have many counties of unbelievably high population. Because there are so few counties at the top of the range, adding more classes would not help unless you had 20 or more which would make the map totally unreadable. The 2 color classification that was used was mostly as a test for the population change map in part 2, but I think it does a solid job of distinguishing between the classes.

Part 2

For the population change map I chose to use a manual classification as the only one of the previous classification that gave results that I was happy with was the natural breaks, which does not work well when trying to show change over time. The other classification systems were not working well at year 1970 and also did not work well to demonstrate population change. When creating maps with the the 3 systems besides that natural breaks, the maps looked too similar over time. This is likely because the shape of the distribution did not change much over time, and although the population values increased across the board, this really only shifted the values over. This means the mean-standard deviation distribution was very similar regardless of time and not viable. The other two classification systems the equal interval and quantile, also did not work for the same reasons they weren't appalling in the original maps. I learned throughout this exercise that examining the distribution is an important step in determining the what classification will work or is the "best fit" for a particular data-set. In the end I determined that to best show population change over time specific classes with hard numbers should manually be created. In the end I chose seven classes that were a hybrid approach of grouping the smaller population counties into similar sized classes and then creating population classes on what would be considered population milestones such as 100,000 or 1 million. My intention was to give the map user not only a sense that growth was occurring but also an easy quantification of it.

For the design of the map I chose a 2 hue classification of a sand color and a poinsettia red. I felt these colors were pleasing to the eye, did not detract from each other or were too similar and helped to communicate the wide range of county populations. I also wanted communicate the change in populations as the primary focus so I decluttered the map by removing the county names and created a thicker line between the counties to improve definition but made the line grey as to not be too bold. For each frame I also included the date in large easily visible font so that you can clearly see when the date changes. I put a drop shadow under the state to give the appearance of it popping above the other states which are grey with grey boundaries to set them back from the California figure. With regards to legend I put the population numbers in italics to improve readability of all the zeros particularly at that particular font size. Overall I feel that these elements have created a successful map that easily communicates to the reader.

A full size version of the map can be found at the link below.
http://pschlatter.bol.ucla.edu/calmap_popchange.swf