5 Remote Sensing and Imagery

How remote sensing works

Remote sensing is the collection of information about a place or object without direct physical contact. Earth-observation sensors on satellites, aircraft, and drones measure energy reflected or emitted by land, water, and the atmosphere. Those measurements can be processed into images and other datasets for analysis in a GIS.

Sensors are either passive or active. Passive sensors measure naturally available energy, usually sunlight reflected from Earth or heat emitted by its surface. Active sensors send out energy and measure its return.

Sensors, energy, and spectral information

Different materials interact with energy in different ways. A surface's spectral signature is its pattern of reflection or emission across wavelengths. Sensors record measurements in wavelength ranges called bands. A multispectral image contains several bands, whereas hyperspectral imagery records many narrow, closely spaced bands.

Passive optical imagery generally needs daylight, and clouds can obscure the ground. Radar is an active sensor that sends microwave pulses; it can operate without sunlight and often observes through clouds, so it can complement optical imagery. Lidar sends laser pulses and can measure surface height and structure.

Raster imagery and pixel values

A raster is a grid of rows and columns. Each cell, or pixel, represents a ground area and stores a measured or calculated value. A single-band image has one value per pixel; a multiband image stores a value for each band at each pixel. For example, a multispectral image might have separate blue, green, red, near-infrared, and shortwave-infrared bands.

Raster data are not limited to photographs. A digital elevation model (DEM) stores elevation values for cells. Other raster examples include temperature, precipitation, and other climate grids containing measured or modeled values; land-cover maps and vegetation indices that are categories or calculations derived from imagery; and radar or lidar products describing surface characteristics, height, or structure.

A raster's cell size, coordinate reference system, and georeferencing determine how it aligns with other GIS layers. Pixel values may be calibrated measurements, scaled digital numbers, categories, or derived values, and are not necessarily expressed in familiar units. Check the metadata and processing level before analysis.

Four kinds of resolution

Resolution describes different properties of a dataset, so “higher resolution” does not mean the same thing in every context.

• Spatial resolution is the ground area represented by a pixel. A 10-meter pixel represents a smaller area and can show finer features than a 30-meter pixel. However, a pixel does not guarantee that every feature of that size can be distinguished; sensor design, contrast, and processing matter too.

• Spectral resolution describes which wavelengths are measured and how finely they are divided into bands. More or narrower bands can help distinguish materials with different spectral signatures.

• Radiometric resolution is the sensor's ability to record differences in energy and relates to how measurements are encoded. Greater radiometric resolution can represent smaller differences in signal, but does not by itself guarantee more accurate measurements.

• Temporal resolution is how often a sensor observes the same area. Frequent observations are useful for tracking change, but clouds, acquisition schedules, and data availability affect which observations can actually be used.

These qualities involve trade-offs. A dataset with very fine spatial detail may cover a smaller area or have fewer bands or observation dates. Choose imagery based on the feature, scale, timing, and measurement needed.

Viewing and interpreting image colors

A true-color composite assigns visible red, green, and blue bands to their corresponding display colors, approximating how a scene might appear to human vision. A false-color composite assigns other bands to visible display colors to make patterns easier to see. Displayed colors therefore do not always match the real-world colors of objects.

Healthy vegetation reflects strongly in near-infrared wavelengths, which people cannot see. If near-infrared is assigned to the display's red channel, vegetation may appear red or pink. Water generally absorbs near-infrared energy and often appears dark in that band. Other band combinations can highlight moisture, burned areas, or other features. Before interpreting color, check which bands were assigned to the displayed red, green, and blue channels.

Several visual clues can support interpretation:

• Tone or color: relative brightness or displayed color, influenced by measured bands and image processing.

• Texture and pattern: for example, the regular pattern of fields or the varied texture of woodland.

• Shape, size, and context: a feature's outline and scale, and its relationship to nearby features.

• Shadows: useful for judging structure, but capable of hiding surface details.

Limits and uncertainty in interpretation

Visual clues support interpretation but do not prove what a feature is. Similar materials can look alike, and the same material can look different as lighting, moisture, season, viewing geometry, or processing changes. Clouds, haze, and shadows can obscure or alter the signal. When reliable identification matters, compare the image with other dates, bands, reference data, or field observations.

A workflow for interpretation and NDVI

For a satellite image of a watershed, use a consistent sequence to investigate vegetation and water:

1. Check the image date, sensor, band descriptions, pixel size, coordinate system, and processing information.

2. View a true-color composite as an overall visual reference.

3. Create or inspect a false-color composite that includes near-infrared. Vegetated land may contrast strongly with water, which often appears dark in near-infrared.

4. Use a consistent color stretch and inspect clouds, shadows, and image edges that could affect the view.

5. Compare apparent patterns with another date or a reference layer before drawing conclusions.

A common numerical measure is the Normalized Difference Vegetation Index (NDVI), calculated from near-infrared reflectance, denoted by $\mathrm{NIR}$, and red reflectance, denoted by $\mathrm{Red}$:

NDVI emphasizes the contrast between vegetation's typically strong near-infrared reflectance and its absorption of red light. It is an indicator, not a direct measurement of plant health; soil, clouds, shadows, season, and sensor differences can affect the result.