Portfolio

This portfolio presents most of the projects completed during coursework for the Graduate Certificate in GIS for Archaeology at the University of West Florida.  These courses included Cartography, Introduction to GIS, Remote Sensing, GIS for Archaeology, and Special Topics in Archaeology.  It also includes a project begun during the final semester, Spring 2017, when I worked as an intern at the New Bedford Whaling Museum in Massachusetts.  I created an Esri Story Tour based on a logbook of a voyage taken by Horatio Hathaway from New York to China and back in 1850-1851. This project is ongoing.

A link to my portfolio can be found here, and a short video explaining my favorite project can be found here.  The Story Tour, although incomplete, can be viewed here.

Creating this portfolio required me to reflect back on how far I've come and how much I've learned over the past two years.  The maps show steady progress in my GIS skills and abilities, and the variety of topics and purposes for which these maps were made is quite impressive, looking back.  I feel like I have accomplished what I set out to learn and more.  I also realize where my strengths and weaknesses exist as they relate to GIS.  Overall, the process of creating the portfolio has left me feeling proud and hopeful that I can get meaningful, satisfying work in this field in the future.

Supervised Classification

This map shows a supervised classification to identify land use and land cover in Germantown, MD for the purposes of tracking changes in population and land consumption.  An image of the area was used to create signatures to be used in classification, and based on those signatures, the image was classified and the signatures merged into 8 classes.  The area for each class was calculated so that, when compared with images from previous or later years, the changes can be calculated.  

Creating signatures requires paying attention to what type of process you use for each signature.  Sometimes it is better to draw a polygon, other times Growing a Seed is the preferred method.  In general, the second method seemed to be best, although for the Fallow Field 1 signature, the polygon was a good choice.  By looking at histograms of each signature, it was possible to see if there was a good enough sample of pixels, and if there was separation in the spectral signature between classes.  By using the Mean Plot it was possible to determine the best band combination to create separation between each spectral signature.

With more time, even better signatures could have been created, but the distance image shows there were not too many poorly classified areas.

Unsupervised Classification

This is an example of unsupervised classification. Here a Landsat image was processed so that clusters of similar pixels were grouped as classes, and then those classes were matched to various feature types in the original image.  The result is a thematic map representing, in this case, 5 classes.  The classes can be analyzed to answer questions; in this case the question was what percentage of the image represented impermeable versus permeable surface.  The necessary information was contained in the Attribute table for the newly created classification image.  One issue that came up was that some pixels were classified in more than one group - for example there were green grass pixels on the roofs of some buildings.  The solution was to create a "Mixed" class that would allow for those situations where classification into a clearly identified group was not possible.

Thermal & Multispectral Analysis

The three images above all show different ways in which a remotely sensed image can be viewed in order to examine different features of interest.  In this image showing part of the coast of Ecuador, a feature that seems to be fields is visible to the lower right of the island. Each image shows the same feature, but what each band combination highlights is different.  In the true color image, the colors are what would be "expected" - vegetation is green, bare earth is grey, urban areas are tan.  The smaller rivers and streams show variation in depth based on the lightness or darkness of the brown and grey hues.  When the band combination is changed to the configuration in the upper right image, much of the vegetation stands out as bright green, but several areas, including the focus feature, show up as blue.  This highlights the moisture content of various elements in the image, including all rivers and streams, as well as areas of agriculture that are moist.  The third image shows Layer 6 - the Thermal band, in one color ramp.  The feature stands out as both an area of vegetation, and an area of relatively high moisture content because of the radiant energy produced by water in comparison to other substances.

Multispectral Analysis

This week's lab focused on learning how to read and understand histograms, and understanding how to identify spectral characteristics of various features..  We also explored various band combinations that would highlight certain features within an image.  Below are three examples.

The first feature we were to identify had a spike in Layer 4 between pixel value 12 and 18.  These low numbers indicate that the feature absorbs rather than reflects EMR, and this image clearly shows that characteristic, especially when the colors are manipulated so that the water is especially dark, contrasted with the light green vegetation and pink/red ground.

The second feature we were looking for had to satisfy two different criteria - a spike in pixel values around 200 in Layers 1-4, indicating high reflectivity, and a spike in pixel values between 9 - 11 in Layers 5 & 6.  The snow-capped mountains have this pattern.  A bright, nearly electric blue makes this feature stand out.

Variations in this water feature are highlighted by using a band pattern that allows the vegetation to remain muted.  Attention is then focused on the variations in the water, showing sediment buildup and a clear channel leading out to a larger body of water.  

Image Enhancement

This map shows an image that was enhanced using both Imagine and ArcGIS.  The first problem was to reduce the visibility of striping that occurred in the Landsat7 image.  Fourier Transform is a process that reduces the banding, although it can still be seen.  Further use of filters and adjustment of the histogram improved the contrast and detail, allowing the viewer to be able to see edges more clearly, and to be able to identify areas of vegetation, water, and urban or residential land use.

Intro to ERDAS Imagine and Digital Data

This is a subset of an image we processed using ERDAS Imagine.  First we had to learn a bit about how to use this new program, which seems to have a huge range of capabilities.  After loading a raster image into the Viewer and making sure the display was set to Pseudo Color, we spent some time learning how to navigate around the image and comparing one image at a lower level of detail (AVHRR) with one a LANDSAT Thematic Mapper satellite image.  The first was classified, while the second was continuous - raw data from the sensor.  We set the preferences to enable Fit to Frame and Background Transparent, making sure the Clear Display option was off, and experimented with options in the Multispectral tab to see what combination of Bands enhanced which types of features, such as vegetation, rivers, urban areas, etc.  The third part of this lab was the creation of a map which we began in ERDAS Imagine, where we created an attribute to show the area of each class in the image, and then exported a subset of the map to make a map in ArcMap.  This was done using the Inquire Box and then creating a subset image.  In ArcMap we made Class_Name the value field, with Unique Values chosen so that the hectares for each class would be listed.  I chose to edit the description so that the information would remain with the data rather than being temporary. 

Clearly the ERDAS Imagine program is very useful and has huge capabilities, but is limited by a few bugs which is unfortunate.

Truthing for Accuracy

The red and green dots represent 30 sample points that were randomly selected in order to check the accuracy of how they were classified in last week's lab.  I located them in a way to reduce bias, by creating and locating them in roughly equal proportions according to each classification type. Then, using Street View in Google Maps, I examined each site and recorded whether it was accurate or not, and if it was not, what classification it should have been.  Finally, I calculated the % accuracy of the sample points.  Because we generalized in making the polygons and classifying them originally, the accuracy was not very high.  Sometimes the point landed on a structure that was commercial rather than residential, although overall the polygon seemed to be the correct type.  In one case, what had been a sandy area when the photograph was taken had been built up with several houses.  What I had thought might have been an academic complex turned out to be several businesses and the Jackson State Fairgrounds.  Several areas that seemed forested were actually residential.  A closer reading of the classification descriptions also prompted me to change one or two classifications.

Level II LULC Classification

 

 

Classification of land use and land cover was the focus of this week's lab.  An aerial photograph of an area of Pascagoula, MS was digitized to create a land use/land cover map.  Features were identified using tone, pattern, shadows, size and shape, and associations, the subject of last week's lab, and then polygons were drawn around those features and classified.  The USGS Standard Land Use/Land Cover Classification System we used has several levels - we worked at Level II.  To begin, Level I classifications were identified, such as Urban or Built-up Land, Forest Land, and Water.  Within each of those, Level II categories were located, such as Residential areas, Lakes, and Deciduous Forest Land.  To create the map, the polygons were labeled, a logical color scheme was chosen to highlight the various categories, and the other essential map elements were added.

 

From this map, it is clear that Level II classification is quite general, and that by generalizing, many specific features and land types are incorrectly identified.  Level III classification provides far better identification, but is obviously more time consuming.

Visual Interpretation of Aerial Photographs

This week we produced 2 maps showing some of the ways that aerial images can be interpreted.  The first map shows variations in tone and texture.  There are 5 variations for each scale, with each polygon enclosing one variation.

 

 

The second map uses 4 criteria to identify features: Shape and Size, Shadows, Patterns, and Association.