Lab 5: Supervised Land Use Classification of Germantown, Maryland

This week’s lab focused on applying supervised image classification techniques using ERDAS Imagine to produce a current land-use map of Germantown, Maryland. The workflow required building a complete spectral signature set using AOIs at specific coordinates, refining signatures through histogram and mean plot evaluation, and running a maximum likelihood classification. I also created additional signatures for water and road surfaces to ensure the major land-cover types were accurately represented.

After generating the initial supervised classification, I recoded the output into eight final land-use classes—urban, road, grass, deciduous forest, mixed forest, fallow, agriculture, and water—and calculated the area for each category using the thematic attribute table. The final map includes the recoded classified image, an inset distance file showing areas of higher spectral uncertainty, and a legend with class names, colors, and area values.

Overall, this lab reinforced the importance of high-quality training signatures and careful class merging, and it demonstrated how supervised classification can support land-use monitoring and planning.

 

Lab 4: Spatial Enhancement & Multispectral Analysis – GIS 4035

This week’s lab introduced several important remote sensing skills that helped me better understand how satellite imagery is structured and how different filters, band combinations, and indices can bring out features that aren’t obvious at first glance. Module 4 focused on working in both ArcGIS Pro and ERDAS IMAGINE, and I definitely noticed that each program has its strengths depending on the task.

---

Getting Started: Image Enhancements

We began by learning about spatial enhancement techniques, which modify pixel values to highlight patterns, edges, or generalize areas of uniform texture. I applied both low pass and high pass filters to see how they changed the appearance of land cover. The low pass filter softened the image, while the high pass filter amplified edges—especially along roads, buildings, and ridgelines. These enhancements turned out to be important for understanding how different kernels and spatial frequencies affect interpretation.

Working With Multispectral Data

A major part of the lab involved exploring the individual spectral bands and learning what each wavelength reveals. Opening each layer in ERDAS IMAGINE helped me identify which bands were most useful for vegetation, soil, water, and built features.

I used grayscale views and the Inquire Cursor to evaluate DN values, which became very important once we reached the feature identification portion of the exercise.

Creating NDVI

Next, I created an NDVI raster using the built-in tool in ERDAS. NDVI’s formula (Band 4 – Band 3) / (Band 4 + Band 3) made it easy to separate healthy vegetation from clearcuts, soil, water, and other features.

Using the swipe tool to compare NDVI with the original image showed me how vegetation completely dominated the scene, which explained why so much of the NDVI surface looked “washed out.”

Exercise 7: Feature Identification and Subsetting

The final exercise brought together everything from earlier in the lab. I used the image histogram, grayscale views, multispectral combinations, and the Inquire Cursor to identify three distinct features in the scene. Once each feature was located, I created subsets and mapped them in ArcGIS Pro using different RGB band combinations.

Feature 1: River (Band Combo 4-3-2)

The river was identified by its very low NIR reflectance, visible as a spike between DN 12–18 in Layer 4. Water absorbs NIR strongly, so it appeared extremely dark in the grayscale display. The 4-3-2 composite (false color) made the river stand out sharply against bright red vegetation.

Feature 2: Barren Land (Band Combo 5-4-3)

This area showed a unique spectral signature—bright in visible bands (DN ~200) but very dark in SWIR (DN 9–11). This is typical of exposed soil or dry ground. Using the 5-4-3 combination highlighted these moisture differences and made barren areas easy to distinguish from vegetated surroundings.

Feature 3: Brackish or Shallow Water (Band Combo 3-2-1)

The last feature was a patch of water with unusually high reflectance in the visible bands but no change in SWIR. This indicated shallow or sediment-filled water. Using the 3-2-1 true-color display made this tonal variation in the water much more visible.

---

Reflection

This lab required a lot of trial and error, especially when navigating ERDAS IMAGINE’s tools and interpreting histograms. However, by the end I felt much more comfortable reading DN values, selecting effective band combinations, and understanding how different wavelengths help reveal different types of features. This was also my first time creating multiple image subsets for mapping, and it really helped reinforce how multispectral analysis supports real-world feature identification.

Exploring ERDAS Imagine: Navigating Lab 3a & 3b Challenges and Discoveries

 

Lab 3A – Intro to ERDAS IMAGINE and Digital Data

Overview

This week’s lab was my first deep dive into ERDAS IMAGINE, learning how to calculate properties of electromagnetic radiation (EMR) and how to view, manipulate, and subset raster imagery. I explored different datasets, including AVHRR and Landsat TM, and practiced preparing imagery for mapping in ArcGIS Pro.

I’ll admit, I struggled a bit at first — some of the tools in ERDAS IMAGINE have been renamed or rearranged since the lab instructions were written. It took some exploring to find the modern equivalents of certain commands, but once I got the hang of it, the workflow made a lot more sense

.

Key Takeaways

• Understanding EMR: I reviewed how wavelength, frequency, and energy relate through Maxwell’s and Planck’s formulas. It’s amazing how these invisible properties directly affect what we see in satellite images.

• Viewer Basics: I learned to open and display raster layers, use the Viewer to switch between True Color and Pseudo Color, and even set up multiple 2D views for side-by-side comparison.

• Navigation & Preferences: I customized the Viewer for easier zooming and set defaults like Fit to Frame and Background Transparent — which made a big difference when layering multiple images.

• AVHRR vs. Landsat: Comparing coarse-resolution AVHRR data with high-resolution Landsat TM imagery really highlighted how spatial resolution impacts what you can actually see.

• Subsetting Data: Using the Inquire Box and Subset Image tools, I clipped out a smaller area (tm_subset.img) and calculated the area for each land-cover class before exporting it for mapping in ArcGIS Pro

  
  

  .

Deliverable

I created a map of the TM Subset image in ArcGIS Pro with color-coded land-cover classes and area values in hectares.

Lab 3B – Intro to ERDAS IMAGINE and Digital Data Part 2

Overview

Lab 3B built on everything from 3A and introduced me to the analytical side of ERDAS IMAGINE. I learned to interpret image metadata, understand the different types of resolution, and analyze thematic data. Again, a few menu names didn’t match the lab sheet, so I spent some time experimenting to find where certain options had moved — but figuring it out gave me a better sense of how flexible the software really is

Key Takeaways

• Layer Info & Metadata: I examined image details like file type, projection, pixel size, and brightness statistics for subset_tm_00.img — useful context when deciding how to process or symbolize data.

• Spatial Resolution: I compared a series of Pensacola images (sra – srd) with pixel sizes from 2 m to 16 m. It was eye-opening to see how smaller pixels capture more ground detail (down to visible cars!) while larger ones blur those features together.

• Radiometric Resolution: Comparing 1-bit, 4-bit, and 8-bit images helped me understand how more bits allow finer distinctions in brightness — even if our eyes can’t always tell the difference.

• Spectral & Temporal Resolution: I reviewed how multiple bands improve spectral detail and how satellite revisit frequency defines temporal resolution.

• Thematic Raster Analysis: I opened soils_95.img and hydro_00.shp to calculate area and percent coverage of soil types. Then, I built a query to highlight soils that were fine-textured and humus-rich — indicators of higher erosion potential

  
  

  
  

🧭 Reflection

Together, Labs 3A and 3B gave me a solid foundation in ERDAS IMAGINE — from basic image viewing to raster analysis. While it was a little tricky adapting to tool names that didn’t quite match the instructions, the process really reinforced my understanding of image resolution and thematic mapping. These labs made me appreciate how much precision and preparation go into transforming satellite data into meaningful maps.