Remote Sensing Data Resolution

Exam Demonstrate understanding of the capabilities and strengths of various types of remote sensing Demonstrate an understanding of limitations and uncertainty in the techniques Show that you can perform and justify the use of analyses such as data interpolation Apply the skills that you have developed to solve problems Demonstrate an understanding of the wide-ranging applications of RS methods for business, government, and science Lecture 1 - Remote Sensing data resolution Advantages of RS; ● Better coverage in time and space ● Better access to data - fast and guaranteed ● Better quality - consistent and controlled ● New measurements from radar sensors ● Synergy - diverse payloads with many instruments simultaneously taking measurements ● However, still need in situ data to validate and calibrate RS data Passive approach; uses natural radiation form the sun or other emitted signal Active approach; illuminates the subject from an artificial source Imaging sensor; creates a picture by scanning across linear array of detectors while array moves through space Non-imaging sensor; makes individual point measurements Synthetic Aperture Radar = active imaging LiDAR = active non-imaging Sonar = active non-imaging Multi-spectral imager = passive imaging Aerial photography = passive imaging Passive microwave radiometer = passive non-imaging Earth observation from satellites; ● Orbital parameters; ○ Altitude ○ Eccentricity (describes deviation from circular orbit) ○ Inclination (angle at which it crosses the equator, thus defines coverage) ● Different orbits have different characteristics ○ Low Earth Orbit; 400-1000km, low field of view, high spatial resolution, polar orbit, coverage attained in strips ○ Geostationary Orbit; 36000km, large field of view, poor spatial resolution, weather and communication satellites Resolution; ● Spatial; minimum separation of two discrete objects for them to be resolved individually, grid size of image raster is proxy for spatial resolution ● Temporal; ability to pinpoint events/changes in time, determined by satellite orbit, velocity related to orbit radius and eccentricity, inclination determines coverage and repeat period ● Radiometric; determines how finely a system can represent or distinguish differences of intensity, expressed as a number of levels or pixel depth, high radiometric resolution the better subtle differences can be represented ○ Signed integer means it can have negative values ○ Unsigned integer means all values will be positive ○ Floating point numbers store numbers after the decimal point ● Spectral; represents its ability to resolve features in the electromagnetic spectrum Landsat programme; ● USGS ● Launched in 1972 - longest satellite record ● Multispectral imager ● 30m resolution ● Landsat 8; operational land imager (OLI) and thermal infrared sensor (TIRS), more sensitive to variations in radiance than L7 as has higher pixel depth, 16-day repeat cycle, launched 2011, scene size of 170km N-S and 183km E-W Sentinel; ● ESA ● Sentinel 1; polar orbiting, all weather, day and night, radar imaging ● Sentinel 2; polar orbiting, multispectral high resolution imaging ● Sentinel 3; multi-instrument mission to measure sea surface topography, sea and land surface temperature, ocean colour and land colour with high accuracy and reliability ● 10m resolution, 12-bit ● 5 day repeat cycle (at equator) - two satellite configuration ● 290km swath SRTM; ● 11-day mission flow by space shuttle in 2000 ● Mapped elevation of 80% of Earth's land area ● ● Used InSAR 30m resolution Lecture 2 - The Electromagnetic Spectrum Matter exists electromagnetic radiation when it has a temperature above absolute zero Wavelength of radiation is a function of temperature (Wien’s law) Wave theory; ● An electrical field which varies in magnitude in a direction perpendicular to the direction in which the radiation is travelling ● A magnetic field orientated at right angles to the electrical field ● Both these fields travel at the speed of light ● The speed of the electromagnetic wave = frequency x wavelength Frequency = number of waveforms passing point at one time in Hertz Quantum theory; ● Suggests that EM radiation is composed of many discrete units or quanta (photons) ● The energy of the quantum is given by Planck-Einstein relation E=hf ● Combining wave and particle theories for EM radiation you get E proportional to constant/wavelength ● Longer the wavelength of EM radiation the lower its energy content Rayleigh scattering; ● Atmospheric particles have diameters that are very small relative to wavelength of the radiation eg dust particles ● Scattering increases as wavelength becomes shorter Non-selective scattering; ● Caused by particles much larger than the wavelength of the scattered radiation eg water droplets ● Scattering is non wavelength-dependent Incoming energy from the sun depends on latitude (distance/path length) and varies with season Interaction with Earth’s surface; radiation that is not absorbed or scattered in the atmosphere can reach and interact with the surface, three forms of interaction take place when energy strikes ● Absorption ● Transmission ● Reflection (specular, diffuse) Specular reflection; ● Smooth surfaces with respect to incoming wavelength ● Almost all radiation is redirected ● In a perfect specular reflector angle incidence = angle reflection Diffuse reflection; ● Rough surfaces relative to incoming wavelength ● Energy scattered in all directions ● Back scatter and forward scatter Spectral reflectance; ● The amount of solar radiation, absorbed, transmitted and reflected varies with wavelength ● Variation of wavelength differs between different substances ● Substances / land cover types can be distinguished by their spectral reflectance, thus we need to view them using a range of wavelength bands Pan sharpening; ● Process of merging higher-resolution panchromatic and lower resolution multispectral imaging to create a single high-resolution colour image Lecture 3 - Spectral signatures and NDVI Spectral reflectance (colour); ● For a given substance the amount of solar radiation absorbed transmitted or reflected varies with wavelength (colour) ● This variation with wavelengths differs between substances ● Substances (land cover) can be distinguished by their spectral reflectance, we need to view them using a range of wavelength bands Vegetation index and other indices; ● Spectral signatures in leaves of vegetation are a product of the physical, chemical and biological characteristics; red light 400-700nm is useful for photosynthesis so it is absorbed, whereas infrared photons do not carry enough energy for photosynthesis so infrared light is reflected → we can use the difference between red and infrared to indicate photosynthesis and thus vegetation ● Normalised Difference Vegetation Index (NDVI) = (NearInfraRed-Red)/ (NearInfraRed+Red) ● Normalised Difference Snow Index helps to distinguish snow vs clouds ● Normalised Difference Water Index; for vegetation = (NearInfraRed-SWInfraRed)/(NearInfraRed+SWInfraRed), for water bodies = (Green-NearInfraRed)/(Green+NearInfraRed) Feature space; ● Use different spectral response of different materials to separate them ● Some variation in spectral signatures of different land use ● Pixel to pixel comparison based on Euclidean spectral distance Image classification; ● Makes image easier to interpret qualitatively ● May reveal hidden information ● Spectral analysis and image classification vital for natural resource mapping at large scales ● Classification can be unsupervised or supervised: ● Unsupervised; ○ Use spectral information to group pixels with similar spectral properties into clusters ○ No prior knowledge required ○ Low opportunity for human error ○ Unique classes are recognised as distinct units ○ Spectrally homogenous classes may not correspond to informational categories of interest ○ Limited control over menu of classes ○ Spectral properties of specific informational classes will change over time → problematic to compare classes between images ○ K-means clustering; changing position of centroid and assigns closest class to centroid based on distance to centroid ● Supervised; ○ Identify classes based on prior knowledge ○ Create training sites for each class, telling software what forest land cover looks like, derive spectral signature for each class and select rules for group membership ○ Areas of interest delineated by user ○ Spectral information on the cover types is gathered for these areas ○ Parallelepiped or box classifier; Each pixel compared to each class block, if it fits in overlapping boxes it uses minimum distance from mean to decide the class, easy to programme, fast in operation, Min and max values may not be truly representative, Shape is not ideal for nature of the data, Does not assign all pixels ○ Minimum distance; find closest cluster mean for each pixel, assigns all the pixels but ignores class variability ○ Maximum likelihood; location shape and size of clusters is determined from statistical properties of training data, based on probability of belonging to a class, usually provides the most accurate results, high computational cost, input data must be normally distributed, overclassfies classes with large distribution Lecture 4 - Active remote sensing; radar and sonar Basic active RS principles ● Active microwave section of EM spectrum, wavelengths 1mm to 1m ● The ‘radio window’ - atmosphere absorbs certain parts of the EM spectrum ● Radar ranging concept; send out as radio wave, hits object, time for return wave after hitting the object ● Radar altimeter; provides elevation information but no imaging capability ● Radar systems defined by the range of depression angles ● Side-looking airborne radar allows images to be formed ● Geometry leads to geometrical distortions ● Geometric errors; ○ Foreshortening (making something look nearer than it is) ○ Layover ○ Shadowing ● Imaging radar resolution; beam width (function of wavelength and antenna length) ● Synthetic aperture radar; records a ‘history’ of reflections, allows synthesis of a much larger antenna for increased spatial resolution → uses Doppler effect ● Phase information ● Interferometric SAR; utilises phase shift information in radar signals, phase differences contain topography and surface deformation ● Sonar has the same principles as radar but measures water depth from a point on the surface ● Sonar spatial resolution depends on water depth and instrument beam angle Data interpolation ● Interpolation is the process of taking known data points to estimate the values of unknown points ● Inverse Distance Weighted; assumes that each input point has a local influence that diminishes with euclidean distance, it weighs the points closer to the processing cell greater than those further away, assumes the variable being mapped decreases in influence with distance from its sampled location ○ Advantages; can estimate extreme changes in terrain such as cliffs or fault lines, dense evenly space points are well interpolated (flat areas with cliffs) and can increase or decrease amount of sample points to influence cell values ○ Disadvantages; cannot estimate above maximum or below minimum values, not very good for peaks and mountainous terrains ● ● ● Spline interpolation; estimates values using a mathematical function that minimizes overall surface curvature, resulting in a smooth surface that passes exactly through the input points ○ Advantages; useful for estimating above maximum and minimum points, creates a smooth surface effect ○ Disadvantages; cliffs and fault lines are not well presented because of smoothing effect, when sample points are close together and have extreme differences in value it does not work well because it uses slope calculations to figure out the shape Kriging; considers both distance and the degree of variation between known data points, iterative process including exploratory statistical analysis of the data, variogram modelling creating the surface, used when there is a spatially correlated distance or directional bias in the data (soil science and geology) ○ Advantages; directional influences can be accounted for (lava flows and winds), can exceed the minimum and maximum point values ○ Disadvantages; does not pass through any of the point values and causes interpolated values to be higher or lower than real values, computationally expensive Natural neighbour interpolation; basic equation, uses areas of overlapping polygons to define weights, local coordinates define the amount of influence any scatter point will have on output cells ○ Advantages; exact as the resulting function retains the value at each data point, continuous everywhere except at the data points, performs well even when the data is highly anisotropic, handles large datasets efficiently Coasts guest lecture Funded by DEFRA Wave riders; wave height, wave period, wave direction, sea temperature LiDAR and fieldwork briefly LiDAR principles; ● Airborne systems ● Satellite systems ● Terrestrial Laser Scanners ● All techniques use lasers (light amplification by stimulated emission of radiation, monochromatic and waves are phase coherent, therefore can use phase shift to increase precision of measure, can measure fractions of wavelengths displacement, very fine spatial resolution) LiDAR types; ● Profiling lasers; downward pointing, closely spaced measurements along-track, measures range from aircraft or satellite to surface, in sensor elevation and orientation is known surface elevation can be calculated ● Imaging lasers; integrated scanning mechanism, differing characteristics for each mechanism, increased swath width, intensive post-processing required LiDAR data characteristics; ● Imaging LiDARs produce point cloud data → vector based with XYZ coordinates, plus time and intensity, several different rasters can be made from a point cloud ● Generated from return laser pulses in wave form LiDAR processing; ● Primary and secondary returns (eg first return from tree top canopy and second return for woody trunk then 3rd 4th 5th could be ground return) ● Conversion from point cloud to raster required Terrestrial Laser Scanning; ● •Satellite & airborne are less useful at the coast; rapid changes in topography and aspect introduce errors, Low point density in areas of high slope, TLS is a useful tool to fill this gap ● Safer - enables accurate measurement of complex features from a safe distance ● Accuracy ~ ± 10 cm - an order of magnitude greater than that achieved by airborne LiDAR (~ 1m) ● Challenging – TLS often sited on a soft, sandy beach that may result in subsidence of the instrument as it operates a scan (10-40 minutes) ● Varies with the complexity and reflectivity of the cliff surface, weather conditions, and ground conditions ● Information from the TLS surveys allows accurate digital 3D models of the cliff surface to be generated