Satellite Remote Sensing, Resolution Characteristics, and Image Classification

Remote Sensing Terminology and Satellite Missions

• Foundational Review: Previous lectures established the definition of remote sensing and provided an overview of the process, specifically focusing on electromagnetic radiation and the differences between passive and active sensors.
• Modern Applications of Satellite Remote Sensing: Satellites are currently used for weather monitoring, assessing ground conditions following natural disasters (e.g., floods and hurricanes), classifying land use or land cover, and providing imagery for inaccessible areas such as high mountains and polar regions.
• Technological Lifecycle: Older satellites are regularly decommissioned, while newer ones integrated with advanced technology are launched annually.
• Nadir/Nadir: This refers to the specific point on the ground located directly below a remote sensor.
• Satellite Mission Characteristics: Selecting appropriate sensors and platforms requires considering resolution characteristics, which often involve trade-offs (increasing one type of resolution frequently necessitates a decrease in another).

The Landsat Program: History and Evolution

• Significance: Landsat represents the world’s largest and longest running continuously acquired collection of space-based, moderate-resolution land remote sensing data.
• Primary Users: It is essential for professionals in forestry, agriculture, natural resource management, and conservation.
• Change Detection: With data spanning four decades, Landsat is fundamental for analyzing land use change and development over time.
• Historical Timeline of Launches:     * Inception: Inspired by 1960s weather satellites.     * Landsat 1: Launched by NASA in $1972$ to test the feasibility of multispectral earth observation from an unmanned platform.     * Landsat 2: Launched in $1975$.     * Landsat 3: Launched in $1978$.     * Landsat 4: Launched in $1982$.     * Landsat 5: Launched in $1984$; it remains the longest-operating earth observation satellite, providing imagery for nearly $29$ years.     * Landsat 6: Failed to orbit in $1993$ and provided no data.     * Landsat 7: Successfully launched in $1999$ and continues to provide daily global data.     * Landsat 8: Launched in $2013$ and remains operational for daily global data.     * Landsat 9: Launched in September $2021$, with imagery currently available.
• Management: Originally managed by NASA, it is now a joint effort between NASA and the United States Geological Survey (USGS).
• Orbital and Swath Characteristics:     * Orbits: All Landsat satellites are in near-polar sun-synchronous orbits.     * Altitude (Landsat 1–3): Approximately $900\,km$ with a revisit period of $18$ days.     * Altitude (Landsat 4–9): Approximately $700\,km$ with a revisit period of $16$ days.     * Swath Width: Sensors collect data across a swath that is $185\,km$ wide.     * Full Scene Dimensions: Defined as $185 \times 185\,km$.

Remote Sensing Scanning Technologies

• Cross-Track Scanners (Across-Track):     * Utilize rotating or oscillating mirrors to scan the earth in a series of lines from one side of the sensor to the other.     * A bank of internal detectors measures energy in specific wavelength ranges/spectral bands.     * This energy is converted into electrical signals and then digital data for computer processing.
• Push Broom Scanners (Along-Track):     * Contain no moving parts.     * Utilize a linear array of sensors, typically Charge Coupled Devices (CCDs).     * Each sensor corresponds to a specific area sampled on the ground across a line perpendicular to the flight direction.     * Forward motion is provided by the flight of the satellite platform itself.

Evolution of Landsat Sensors

• Multispectral Scanner (MSS):     * Found on Landsat $1$ through $4$.     * Detects radiation in $4$ spectral bands: Green, Red, and two Near Infrared (NIR) bands.     * Spatial Resolution: Approximately $60 \times 80\,m$.     * Radiometric Resolution: $6$ bits ($64$ digital numbers).     * Technology: Cross-track oscillating mirror. Routine collection ceased in $1992$.
• Thematic Mapper (TM):     * Superseded the MSS with $7$ spectral bands.     * Spatial Resolution: $30\,m$ for most bands; thermal band resolution is $120\,m$.     * Utilizes oscillating mirror technology.
• Enhanced Thematic Mapper Plus (ETM+):     * Standard on Landsat $7$.     * Added a panchromatic (black and white) band with a resolution of $15\,m$.     * Uses oscillating mirror technology.
• Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS):     * Featured on Landsat $8$ (as OLI/TIRS) and Landsat $9$ (as OLI-2/TIRS-2).     * Technology: Push broom instruments.     * New Spectral Bands (OLI): Deep blue visible channel (Band 1) for water/coastal investigation; Short Wave Infrared (SWIR) channel (Band 9) for cirrus cloud detection.     * Band Characteristics: Bands have become narrower in newer iterations, enabling better characterization of land cover condition.

Fundamental Types of Resolution

• Spatial Resolution:     * Defines the amount of detail visible and the minimum size of a feature discernable in an image.     * Determined by pixel size in a raster grid.     * High Resolution: $< 2\,m$ pixel size (e.g., Ikonos, QuickBird satellites). Used for topographic/thematic mapping and city planning.     * Medium Resolution: Between $2\,m$ and $30\,m$ (e.g., Landsat $7$ and $8$). Used for global observations and deforestation monitoring.     * Low/Coarse Resolution: $> 30\,m$. Used for global/weather monitoring (e.g., MODIS - Moderate Resolution Imaging Spectroradiometer, used for wildfire burn areas). Often coincides with high repetition rates.
• Spectral Resolution:     * Refers to the number, spacing, and width of spectral bands detected.     * Band: A continuous range of wavelengths detected (e.g., Blue: $0.4\text{ to }0.5\,\mu m$; Red: $0.6\text{ to }0.7\,\mu m$).     * Panchromatic (Pan): A wide band (e.g., $0.5\text{ to }0.7\,\mu m$) that records all visible light in one black/white band.     * Hyperspectral Sensors: Feature hundreds of narrow channels providing a near-continuous spectrum (e.g., AVIRIS - Airborne Visible Infrared Imaging Spectroradiometer with $224$ channels from $0.4\text{ to }2.5\,\mu m$; Hyperion).
• Radiometric Resolution:     * Describes the ability to discriminate slight differences in energy; specified by grayscale values.     * $1$-bit: $2^1 = 2$ values (pure black and white).     * $2$-bit: $2^{2} = 4$ values of gray.     * $8$-bit: $2^{8} = 256$ shades of gray ($0$ for black, $255$ for white). Higher radiometric resolution requires more storage space.
• Temporal Resolution:     * The time required for a satellite to return to the same spot (revisit period or repetition rate).     * Factors: Altitude, orbit, sensor capability.     * Standard Revisit periods: Landsat $7/8$ is $16$ days; Ikonos is $14$ days.

Spectral Reflectance and Band Combinations

• Spectral Signatures: Different features reflect/absorb radiation differently across wavelengths.     * Vegetation: Chlorophyll absorbs red light but reflects most Near Infrared (NIR) radiation.     * Water: Absorbs most Infrared radiation; appears black in longer wavelengths.     * Soil: Reflects significant amounts of Mid-Infrared radiation.
• Band Combinations in RGB Display:     * True Color: Red channel = Red band, Green channel = Green band, Blue channel = Blue band. (Landsat $7$ bands $3, 2, 1$; Landsat $8$ bands $4, 3, 2$).     * Color Infrared (False Color): Red channel = NIR, Green channel = Red, Blue channel = Green. This highlights vegetation as bright red.     * SWIR False Color: Uses NIR and Short Wave Infrared bands to make water and rivers stand out prominently.     * Data depth: RGB monitors display $256 \times 256 \times 256$ combinations, totaling approximately $16,800,000$ distinct colors.

Image Classification in GIS

• Definition: Computer-assisted interpretation reducing image areas into categories based primarily on spectral signatures.
• Classification Criteria: Success depends on the presence of distinctive signatures for classes and the ability to distinguish them from other patterns.
• Types of Classification:     * Supervised: Uses training sets or sample areas delineated by the user to guide software.     * Unsupervised: The computer automatically groups pixels based on statistical patterns without user-defined training sites.
• ArcGIS Pro Implementation: The "Imagery Tab" becomes active when a raster dataset is added.     * Classification Wizard: A guided tool composed of best practices covering: Pre-processing, Segmentation, Training sample selection, Training, Classifying, and Accuracy Assessment.     * Applications: Quantifying land cover changes over time (e.g., forest converted to urban) and creating land use maps (commercial, industrial, residential, pasture).

Questions & Discussion

• Question: Which image has the highest resolution among the samples? (Left: Europe low-res; Middle: Paris medium-res; Right: Paris high-res).
• Answer: The image on the right is the highest resolution, where the Eiffel Tower is clearly discernable.
• Student Resource: The USGS "Earth Now" website allows users to see near-real-time recordings of what Landsat satellites are currently capturing.