Module 3 - Electromagnetic Radiation (ERM) and Navigating Raster Imagery in ArcGIS Pro

In this lab assignment, we covered several topics involving electromagnetic radiation (EMR) and image processing in ArcGIS Pro. 

In Part A, we calculated EMR using Maxwell's wave theory expressed by the equation (C = λv) and the Planck relation (Q = hv). These foundational concepts allowed us to explore the behavior of electromagnetic waves and their energy quantization.

In Part B, we navigated to the raster layer tab in ArcGIS Pro, analyzing various groups such as Visibility Range, Effects, Compare, Rendering, Enhancement, and Rotation, utilizing a land cover thematic raster. We explored into raster layer properties and metadata, particularly focusing on multispectral (multiband) imagery, which is crucial for diverse land cover analyses.

In Part C, we transformed single-band images into composite multiband images, gaining insight into the imagery tab and the array of tools available for image analysis. We utilized Swipe and Flicker methods to compare and contrast two images effectively. For evaluation, we employed false color imagery band combinations: R:6, G:5, B:4 for images from Brownsdale, FL, and Pensacola, FL, along with R:7, G:5, B:4 to identify bright red/orange colors in the Blackwater River State Park, FL imagery.

The final step involved creating a map using a Land Use Land Cover raster of a forested area in Washington State. We focused on a specific area of interest and used the Clip Raster tool from the geoprocessing options to clip the image accurately. In the new clipped area's attribute table, we introduced a new field, SubHectares, calculating the Land Use Land Cover area in square meters using the formula (!Count!*900)/10000. Additionally, we created another field to quantify the area in acres via the equation Acres = !Count!* 900 / 4046.8564 . Finally creating a layout and incorporating all essential map elements

In Part D of the analysis, a comprehensive comparison was conducted, presenting four distinct images that varied in pixel (or raster cell) spatial resolution alongside another set of four images differing in radiometric resolution. This side-by-side evaluation facilitates an understanding of how these variations in resolution impact the visual and analytical outcomes of the images.

Module 2 - Land Use / Land Cover Classification - Ground Truthing & Accuracy Assessment

The lab assignment comprises two sections: 2a, focused on land use/land cover (LULC) classification, and 2b, involving ground truthing and accuracy assessment. In section 2a, a new map was created using aerial photography, establishing a polygon layer for LULC analysis, which included classifications such as Urban, Forest, Water, Wetland, and Barren Land. These Level I classifications were further refined into Level III categories for precision, detailing various types of land uses within each classification. 

In section 2b, ground truthing involved selecting 30 random sampling points using a GIS tool to ensure spatial distribution. The points were validated using Google Maps Street View, integrating classifications into a "Truthing" layer. The accuracy of the classifications was assessed using an Error Matrix to calculate user’s accuracy, producer's accuracy, and overall accuracy. A layout of the land use classification map with sample points was generated and exported in JPG format.

Module 1 - Visual Interpretation

In this laboratory exercise, the focus was on the interpretation of features observed in aerial photographs, examining both black and white and color imagery. 

In Exercise 1 of the lab I had to identify and interpret features present in black and white aerial photography, with a focus on the concepts of tone and texture. In this context, tone refers to the range of brightness and darkness within an image, contrasting with color photography, where tone is described using hue, saturation, and brightness. Texture indicates the variations in tone across a surface, resulting from the arrangement and repetition of these tones or colors. It can be classified into various categories, including very fine, fine, mottled, coarse, and very coarse. To achieve the exercise's objectives, I initially analyzed tone by mapping areas with distinct brightness levels, using polygons to outline each defined zone. Similarly, for texture assessment, I identified various textures and outlined areas with similar characteristics using polygons. This systematic approach facilitated a clear interpretation of the aerial imagery, highlighting the key characteristics of tone and texture.

In Exercise 2 of the lab we are required to identify features in black and white aerial imagery by employing four criteria: shape and size, shadow, pattern, and association. Each feature is marked with distinct points, utilizing different colors or symbols per criterion. The author successfully identified three objects based on shape and size characteristics: a commercial building, a road, and a swimming pool. Furthermore, three features were recognized through shadow analysis: a tree, a utility pole, and a water tower, identifiable by their shadows. Additionally, the association criterion was applied to identify three objects in relation to their surroundings, including a pier, a single-family house, and a hotel building. Finally, utilizing the pattern criterion, defined as the spatial distribution of objects across the terrain, the author identified two features: water and sand.

In exercise 3 of the lab we interpret color information by using a true color TIFF file and subsequently comparing the identified colors with those in an infrared image of the same geographical area. The process involves selecting five specific features from the true color image, denoting each feature with a point, and documenting their corresponding colors in a predefined table. The features identified for this exercise include a river, a forest located to the southwest of the river, marshes situated in the northeast portion of the image, grass found in a golf course, and a water basin. Following this, a color comparison of the same features in the infrared images is conducted, with results also recorded in the same tabulated format. This exercise aims to enhance the understanding of how different imaging techniques can represent environmental features across various spectral ranges.

Color-infrared (CIR) aerial photography, or "false color" photography, is utilized for natural resource interpretation, unaffected by atmospheric haze. Live vegetation typically displays red tones, with intense reds indicating healthy growth and lighter shades signaling reduced vigor. Bare soils are depicted in shades of white, blue, or green, generally darker indicating moisture levels. Man-made features reflect their materials, with asphalt appearing dark, gravel and dirt roads lighter, and concrete light-toned. Water is represented in varying blues, from nearly black for clean water to pale blue for sediment-rich areas, with shallow stream colors influenced by bottom materials like sand.

 

Module 3.1- Scale Effect and Spatial Data Aggregation

In the final lab of the semester, we explored several key concepts related to geospatial data analysis. First, we investigated the impacts of scale on vector data, which involves understanding how changes in scale can affect data representation and interpretation. Next, we examined the influence of resolution on raster data, focusing on how the spatial resolution of raster datasets can impact the detail and accuracy of the analysis. Additionally, we delved into the Modifiable Area Unit Problem (MAUP) by employing Ordinary Least Squares (OLS) regression analysis to showcase how aggregation levels can distort statistical outcomes. The lab also provided insights into identifying multipart features within spatial datasets, enhancing our skills in data classification and analysis. Lastly, we measured gerrymandering to assess its implications on electoral boundaries and representation, illustrating the practical applications of spatial analysis in political contexts. This multifaceted approach allowed us to engage with important theoretical and practical aspects of geographic information systems (GIS).

Scale in cartography denotes the relationship between a distance on a map and its corresponding distance in reality, commonly presented as a representative fraction, such as 1:1,200. Maps are categorized based on scale into large-scale and small-scale maps. Large-scale maps have a smaller denominator, enabling them to represent smaller areas with a high level of detail and reduced geometric error. Conversely, small-scale maps possess a larger denominator, allowing them to cover broader areas but resulting in lower detail and an increased likelihood of geometric inaccuracies. This distinction is crucial for map interpretation and usage in various applications.

Resolution in raster data is characterized by the cell size, which is the smallest spatial unit that can be represented and measured. A finer resolution, indicated by a smaller cell size, captures more detail and allows for the representation of smaller features; however, it also results in increased storage demands and computational requirements. Conversely, a coarser resolution, denoted by a larger cell size, diminishes detail and may lead to the "mixed pixel" problem, where a single pixel encompasses multiple land cover types, thereby reducing classification accuracy. A general guideline states that to be discernible, a real-world feature should be at least as large as one cell.

Gerrymandering is the manipulation of an electoral constituency's boundaries so as to favor one party or class. The Polsby–Popper score is a mathematical compactness measure of a shape developed to quantify the degree of gerrymandering of political districts. This score ranges from 0 to 1, where a score of 0 signifies a complete lack of compactness, indicating a highly irregular shape, while a score of 1 denotes maximal compactness, reflecting a shape that is highly regular.

Screenshot of the worst offender a highly irregular shape

Module 2.2 - Surface Interpolation

Interpolation is a Geographic Information Systems (GIS) technique that estimates values at unmeasured locations by utilizing known sample points, based on the principle that spatially proximate points exhibit similar values. This method converts point data into continuous rasters, enhancing the visualization and analysis of spatial patterns. It is applicable across numerous measurements, including elevation, rainfall, and noise levels.

In this lab assignment, three interpolation techniques, Thiessen, Inverse Distance Weighted (IDW), and Spline interpolation, were employed to analyze water quality conditions in Tampa Bay, specifically focusing on the Biochemical Oxygen Demand (BOD) measured in milligrams per liter. 

1. Thiessen Interpolation: This method is a straightforward nearest-neighbor approach that divides geographic space into polygonal regions defined by the proximity of known data points. Each polygon represents the area nearest to a particular sample point, and values are assigned throughout the polygon based on the value of the nearest known point.

2. Inverse Distance Weighted (IDW): IDW calculates values for unsampled areas by averaging values from neighboring sample points, giving more weight to closer points. This method is particularly effective with datasets that have a high density of sample points and is typically used where spatial data shows gradual changes.

3. Spline: This technique uses mathematical functions fitted to the data to minimize surface curvature. The output is a smooth surface representation ideal for depicting phenomena that show gradual variations, such as topographic gradients. Interpolation thus serves as an essential tool in GIS for analyzing and visualizing spatial data distributions effectively.

First, we started with utilizing a dataset included of samples collected from Tampa Bay over a short timeframe. An initial analysis was conducted to determine the minimum, maximum, average, and standard deviation of the BOD parameter through a non-spatial technique, which involved examining the statistics of the point data layer.

Subsequently, the Thiessen polygon tool was employed to create polygons based on the proximity to known data points. This analysis also included checking the properties of the polygons and recording the relevant statistics, including the minimum, maximum, average, and standard deviation.

Following that, IDW Spatial Analyst tool was utilized to generate a new interpolation raster, which depicted the concentration of BOD across Tampa Bay. The results were further analyzed using the Zonal Statistics as Table tool to calculate and record statistics from the newly created raster.

Lastly, the Spline tool was applied to create an alternative raster utilizing the Regularized spline method. This final raster revealed an anomaly in the northern portion of the study area, where there was a very high concentration of BOD, despite the absence of sampling points in that specific region. This observation raises questions about the accuracy and representation of the data in that area. Two nearby sampling locations showed biochemical oxygen demand (BOD) levels of 1.157 mg/L and 2.2 mg/L, with the higher reading identified as an outlier. To address this discrepancy and to avoid distorting the smooth curve, we can either remove the outlier or replace it with an interpolated value. In this case, the point corresponding to the elevated BOD reading was removed, and spline interpolation analysis was subsequently re-executed.

Map showing Spline Interpolations using Tension type

Module 2.1 - Surfaces: TINs and DEMs

In this week's lab assignment, we began by previewing a Triangulated Irregular Network (TIN) file in ArcGIS Pro. A TIN is a data model used to represent terrain heights by entering the x, y, and z coordinates of measured points. These points can connect to form the smallest triangles possible between any three adjacent points, creating a network of triangles. Few points are sufficient for large, flat areas, while a higher density of points is necessary to accurately capture small-scale variations (Bolstad & Manson, 2022, p. 61).

We started by created a new scene, incorporated the TIN file, and designated it as an Elevation Source. Subsequently, we exaggerated the terrain by increasing the vertical exaggeration and examined the data.

Following that in Part B, we utilized a DEM to create a Ski Run Suitability Map. To achieve this, we first established a new local scene and added the DEM raster. We then employed the Raster to TIN tool to generate a TIN file from the DEM and set it as the elevation source. Next, we applied the reclassify tool to reclassify the elevation from the DEM file, then generated a slope raster from the DEM, and reclassifying it, as well as creating and reclassifying an aspect raster. Ultimately, we merged the three reclassified rasters to produce the final suitability map.

In Part C, we opened a new local scene and added a TIN file, setting it as the elevation source and applying a vertical exaggeration of 2.00. We utilized the symbology options under TIN Layer > Drawing to visualize the TIN file, incorporating slope, aspect, edges, and contour lines, and examining the elevations and points distribution given the nature of the terrain to determine if the distribution is appropriate.

When overlaying el_points shapefile onto the TIN, the density and distribution of points must correspond to the spatial variability of the terrain. The point distribution indicates that the higher uniform area between elevations of 1898.11-2022 meters is flat, resulting in fewer collected points. In contrast, the lower elevations of 637-790.89 meters, which can be viewed as not overly steep, exhibit a high point density. While increased point density in flat regions can enhance accuracy, in smooth terrains, additional points may not significantly improve the surface representation. In these areas, fewer points are sufficient due to the gradual changes in the surface.

Moreover, the steep regions with elevations ranging from 944.78-1252.56 meters have a lower point count. Insufficient sampling in uneven or variable terrains can lead to distortions in the resulting surface. If the TIN shows a hilly landscape characterized by sharp ridges and valleys, we should observe a greater concentration of points along these features.

In the final Part D, we introduced a new scene and incorporated two layers: a points layer and a study area. We utilized the Create TIN tool to generate a TIN layer from the points layer. The symbology was modified to display only contour lines. The next step involves generating contour lines from the DEM based on the points layer using the spline tool, followed by creating contours from the DEM with the same intervals used for the TIN contours. After generating both sets of contours, we compared the contour lines produced by the two distinct methods.

Comparing TIN and DEM contours reveals key differences: TIN contours are jagged and sharp, while DEM contours are smoother and more rounded, particularly evident in steep areas with low elevation point density. In flat regions, both models yield similar contour patterns due to effective interpolation. TINs, based on a Triangulated Irregular Network, offer greater accuracy in complex terrain by using non-overlapping triangles for interpolation. Conversely, DEMs use a raster grid. Thus, the choice between TIN and DEM should align with specific application requirements and data needs.

Screen shot showing TIN and DEM contour lines

Bolstad, P., & Manson, S. (2022). GIS Fundamentals A First Text on Geographic Information Systems. Elder Press.

Module 1.3 - Data Quality : Assessment

The goal of accuracy assessment in mapping involves the verification of data reliability and quality. In the context of road networks, this assessment evaluates positional accuracy, attribute accuracy, completeness, and consistency. Given that road networks are essential to applications such as address geocoding and routing, it is important that the attribute data is both complete and precise.

There are various methods to ensure data accuracy. One effective approach is ground truthing, which utilizes high-precision GPS units to validate data accuracy. Additionally, data can be compared against high-resolution imagery or existing high-accuracy datasets to further determine its reliability.

Both positional accuracy and completeness are critical in the evaluation of data accuracy. Positional accuracy is determined by comparing the data against a higher quality dataset. Completeness, on the other hand, assesses the degree to which real-world objects are adequately represented. This involves evaluating the number of objects that should exist within the database but are absent, as well as identifying data that may need to be excluded.

Thematic accuracy and thematic completeness represent two additional measures of data quality. Thematic accuracy is measured by the extent to which attribute values correspond with reference data, such as the correct recording of street names or speed limits. On the other hand, thematic completeness is assessed by the number of missing attributes.

In the previous lab assignment 1.2 Data Quality – Standards, we evaluated the quality of two road networks: the streets of Albuquerque, derived from a shapefile provided by the city, and the Street Map USA data obtained from the TeleAtlas product. The aim was to evaluate positional accuracy using metrics defined by the National Standards for Spatial Data Accuracy (NSSDA).

In this lab assignment 1.3 Data Quality – Assessment, we initially calculated the total length of roads from two distinct layers: Jackson County Street centerlines and TIGER Roads, in order to evaluate which dataset exhibits greater completeness. 

Comparing the two shape files Street_Centerlines and TIGER_Roads we find that the total length of roads represented in the TIGER_Roads shapefile exceeds that of the street_centerline shapefile by around 509.4 kilometers. This suggests that the TIGER_Roads dataset includes a larger quantity of digitized streets. Typically, TIGER data incorporates unpaved roads and private driveways, which are generally absent from local street centerlines. It can be concluded that although TIGER data may not possess the positional accuracy of local street centerlines, it often provides greater completeness.

In the second step, we employed a 5km x 5km grid to determine the total length of roads contained within each grid for both networks. As a result, we identified the number of grid polygons where the county centerlines are more complete than the TIGER roads, as well as the number of polygons where the TIGER roads surpass the county centerlines in completeness. 

In the final step, we calculated the percentage difference in total length between the two road datasets for each grid polygon, utilizing the centerlines as the reference and applying the formula: 

% difference = (total length of centerlines - total length of TIGER roads) x 100%                                                          (total length of centerlines)

Ultimately, we produced a choropleth map that illustrates the percentage differences between the two datasets.

Map layout showing the percentage difference between the two data sets for each grid polygon