Show:
Principles
Image Processing
Vegetation Indices
Classification
100

What is being detected by remote sensing instruments?

Electromagnetic Radiation 

100

Why, when comparing images from different years, is it important to try and have the images coincide on the same day and month of acquisition as closely as possible?  

Seasonal variability – especially for vegetation

100

Define spectral signatures, with examples.

the pattern (graph) of different levels of reflectance of veg type throughout the EMR spectrum. 

100

What are the different types of change?

  1. Short term (weather) 

  1. Cyclic (seasonal) 

  1. Directional (Urban developments) 

  1. Multidirectional (deforestation and regression) 

  1. Events (natural hazards, disasters and fires)  

200

Name and define the two Sensor Types.

  1. Passive: Measuring Natural radiation (source= sun) 

  1. Active: Measuring radiation transmitted from sensor (LiDAR) 

 

200

What regions of the spectrum show the largest reflectance of vegetation/soil/water

  1. Green/NIR for Vegetation 

  1. Red for Soil  

  1. Depends on oceanic factors (phyto, salinity, sediment) Blue  

200

Describe three ways remote sensing is used to monitor the Ocean.

  1. Sea Surface Temperature: Measures thermal radiation 

  1. Ocean Salinity: Measures microwaves (change and amount received) 

  1. Sea Surface Height: Measures the time radar takes to return to sensor  

200

What are the different approaches for change detection?

  1. Two most simple/common 

  1. Spectral change detection – identifies changes on the spectral characteristics of the spatial units (i.e. pixels), includes techniques that account for the spatial arrangement of pixels (i.e. context) 

  1. Class change detection – consists of a post-classification comparison (i.e. GIS overlay) 

300

What are the two types of satellite orbits? Give an example of each.

  1. Near polar orbit: high inclination angles, altitude range 400-1000 km, orbital period of about 100 minutes, SUN SYNCHRONOUS, e.g. Landsat – repeat coverage every 16 days 

  1. Geostationary orbit: 35,786 km altitude exactly, same location relative to earth, coverage does not change, e.g. Geostationary Operational Environmental Satellites (GOES) 

300

Most RS systems can collect data in both a panchromatic and a multispectral mode. What is one advantage of each mode?

  1. Panchromatic = better spatial resolution (Detailed Maps) 

  1. Multispectral = better spectral resolution (Change detection and indices)  

300

How can RS be used in agriculture and forestry? Give an example of each application and explain which data would be used for that application.  

Agriculture 

  1. Mapping out crop/vegetation types 

  1. Precision agriculture 

  1. Crop yield estimation  

Forestry 

  1. Land use change/ analysis 

  1. Deforestation tracking 

  1. Wildfire monitoring  

300

Describe unsupervised image classification.

  1. Algorithms work by grouping ‘similar’ pixels into the same class, while ensuring that each class is sufficiently ‘dissimilar’ to all others as to be considered separate 

  1. Different algorithms used, most common are K-means and ISODATA 

400

List and define three types of atmospheric scattering.

  1. Rayleigh: radiation interacts with particles/atmospheric molecules that are smaller in diameter 
  2. Mie: radiation interacts with particles similar size, occurs lower parts of atmosphere, influences longer wavelengths
  3. Non-selective: Radiation interacts with particles larger than, scatters wavelengths equally resulting in clouds appearing white/grey 


400

What types of preprocessing steps may occur prior to main data analysis of a remotely sensed image?

  1. Radiometric correction – scene illumination and reducing image noise 

  1. Atmospheric correction – statistical and complex radiative transfer-based methods 

  1. Geometric correction – accounting for systemic and random distortions, e.g. curvature of the earth 

400

Explain the equations for NDVI, NDWI and NDDI.

  1. NDVI (Normalized Difference Vegetation Index) = NIR-RED/ NIR+RED 

  1. NDWI (Normalized Difference Water Index) = NIR-SWIR/NIR+SWIR 

  1. NDDI (Normalized Difference Drought Index) = NDVI-NDWI/ NDVI+NDWI 

400

How can classification accuracy be assessed?

  1. Accuracy assessment is essential for both supervised and unsupervised; compares a classification with ground-truth data to evaluate how well the classification represents the real world 

  1. Most common method for assessment is a confusion/error matrix, which considers producer's accuracy (omission error), user's accuracy (commission error), and overall accuracy 

  1. Kappa coefficient as measure of agreement between classification map and reference data (highly used, highly debated) 

500

Where are the atmospheric windows that we can effectively use in remote sensing?

  1. Atmospheric windows: the wavelength regions outside the main absorption bands of the atmospheric gases ( H20, CO2,O3) 

  1. Four principal windows: 1. visible –NIR (0.4-2.5 micrometers), 2. mid-IR (3-5 micrometers), 3. thermal IR (8-14 micrometers), 4. microwave (1-30 cm) 

  1. Think of diagram

500

What are the 4 types of image resolution that we are concerned about when interpreting remote sensing data? Give examples of each.

  1. Spatial resolution: ability to separate between objects (think pixel size) 

  1.  Temporal resolution: time between observations (LandSat 16 day, MODIS 1-2) 

  1. Spectral resolution: location, width, and sensitivity of chosen wavelength bands (think number of bands, Landsat8: 11, Sentinel 2: 13, MODIS: 36),  

  1. Radiometric resolution: precision of observations (think number of grey levels: LandSat 12bit or 4,096) 

500

Describe what a vegetation index is, how can they be used for distinguishing between veg types and conditions?

  1. Calculation of reflectance of vegetation at different wavelengths 

  1. -1 to 1 value 

  1. NDVI vs NDWI vs NDDI 

500

Describe supervised image classification.

  1. Different methods for supervised classification: maximum likelihood, nearest neighbor, artificial neural networks, decision trees, support vector machines, spectral mixture analysis 
  2. Classification uses training classes/ data samples: garbage in, garbage out 
  3. Parametric: assumption based, pixels classified based on model 
  4. Non-parametric: knowledge based (Random Forest) decision trees, classified based on rules, characteristics, and distance  
  5. Soft/ Hard Pixels
  6. Supervised classification has higher accuracy, takes more time