Please use this identifier to cite or link to this item:
|Title:||ATMOSPHERIC CORRECTION OF THE IRS HIGH RESOLUTION IMAGES|
|Authors:||Srivastava, Shailendra Shanker|
|Keywords:||PHYSICS;ATMOSPHERIC CORRECTION;IRS HIGH RESOLUTION IMAGES;REMOTE SENSING|
|Abstract:||For ideal remote sensing, a non-interfering atmosphere is needed, which would not modify the radiation from the source. But, it is not the reality, scattering due to the molecules and aerosols in the atmosphere and absorption due to the gaseous constituents of the atmosphere, all modulate surface reflectances as well as exo-atmospheric photons, which propagate through the atmosphere to the satellite sensor. These photons interact with the atmospheric constituents, and undergo change in their frequencies or direction of propagation or both, the degree of which depends on the abundance and distribution of the absorbers, emitters and scatterers present in the atmosphere. Under some conditions (e.g. over oceans), satellite measured radiances comprise of more than 90% contribution from the atmosphere, but even much smaller effects would degrade the quantitative use of this data. Therefore it becomes necessary to understand fully the various processes like scattering, absorption as well as their radiative effect on the satellite imagery to take full advantage of the sensors onboard the spacecrafts. Removal of this atmospheric effect is called atmospheric correction. Physical principle behind the atmospheric correction is to distinguish between the signal and the noise (let us call the atmospheric effect as noise, though it is a part of the sensed signal) introduced by the atmosphere. Atmospheric correction is one of the most difficult tasks facing the remote sensing community. Over the past 30-35 years, a number of algorithms have been developed to correct for variations in atmospheric transmission. There are radiative transfer models, e.g. MODTRAN, 6S, FLAASH etc., which serve as cornerstones for satellite remote sensing. They are used in both forward and reverse problems. But, there does not exist any single approach, which is universally accepted and being used operationally. One of the major limitations is need of supply of data relating to the atmospheric conditions at the time of imaging, which can be fulfilled either by in situ measurements at the time of imaging or retrieved through some mathematical model or some appropriate assumptions. Linear array of discrete detectors, CCDs (Charged coupled devices) are being used in Indian remote sensing satellites (IRS) space-borne opto-electronic telescopes. IRS Satellite systems use push broom technique to collect the data, utilizing the satellite motion. It has Multispectral high-resolution imager like LiSS-3/4 (Linear Imaging Self Scanner, 23.5m/5.8m spatial resolution), and AWiFS (Advanced Wide Field Sensor, Abstract 56.5m) for the remote sensing applications in the field of agriculture, ecology, geology etc. Satellite images have to undergo a chain of radiometric and geometric preprocessing before they are ready to be used for some application. Atmospheric correction is one of the most important components of a typical radiometric correction procedure for the satellite images. Optical depth (deciding the visibility) calculation is one of the components of the radiative transfer problems. Rayleigh optical depth is an important component of the total optical depth. It is a function of several optical parameters. Two different approaches are found in its calculations, first, step-by-step use of Rayleigh scattering theory and the second is the use of the best-fit equations (mostly with only one independent variable, i.e. the wavelength). Hoyt (1976), Frohlich and Shaw (1980), Teillet (1990), Bucholtz (1995), and Bodhaine et al. (1999) have treated the subject almost entirely, but in isolated segments. They have evaluated and compared different inputs and different methods. However, uncertainties are observed in the calculations, which are more in the case of the use of the best-fit equations or in the use of interpolation of the Rayleigh optical depths or due to different depolarization factors. In this thesis, we have worked to evolve a method to determine Rayleigh optical depth, which uses more realistic refractive indices (scaled for moisture level and CO2 concentrations in the atmosphere), and depolarization factor (including the rotational Raman scattering) adjusted for different constituents of the atmosphere instead of fixed values for them. We computed Rayleigh Optical Depths (ROD) quantum mechanically (semi-classically) for the 0.4-0.9-gm range (at an interval of 0.05 gm), which is used frequently in remote sensing. We have compared the ROD values with the well-established Elterman's model calculations for trend analysis purpose. Elterman's calculations have already proved their credentials as they are being used in several models. The calculated Rayleigh optical depth is coming 3.4% lower than the values generated by previous researchers. This difference is due to the fact that for the first time, we have incorporated the vector coupling coefficients (Clebsch-Gordan coefficients), which insure the conservation of angular momentum during photon-particle interaction. We have calculated it for the lower rotational quantum numbers, valid for atmosphere above the troposphere. We incorporated this method in a radiative transfer model 6S (Second Simulation of Satellite Signals in Solar Spectrum), which we will refer to as "modified 6S", henceforth in this thesis. ii Abstract Aerosols have heterogeneous spatial and temporal distribution with a lifetime of a day (in the lower tropospheric region) to several days or even months in the upper atmosphere. The varying spatial and temporal patterns of aerosol, their size and number distribution, different composition and absorption characteristics make its monitoring more elusive than molecular species, whose characteristics are relatively better understood. We have also worked on the aerosol characterization and its retrieval for their use in atmospheric correction. We made use of spectral Aerosol Optical Thickness (AOT) to decipher the particle size (rough estimation) of the involved aerosol. AOT was collected for the month of January for the consecutive two years (2008 & 2009). It was to prepare a theoretical base for a large-scale study for the Ahmedabad region. Diurnal variations as well as large day-to-day variations (up to 20-40%) in AOT are observed at 380, 440, 500, 675, and 870nm wavelengths. Data measured onboard Oceanic Research Vessel (ORV) "Sagar-Kanya" during a sea truth collection campaign in Arabian Sea (13° 8' N 73° 28' E on 21st April 2009 9.00am) and data from Aerosol Robotic Network (AERONET) site at Kanpur (26°30'46" N, 80°13'55"E) are also utilized to see the curvature (Spectral variation of Angstrom exponent) of the aerosols. Curvature and Angstrom exponent values in 0.380-0.440 IAM and 0.675-0.870 inn intervals indicate coarse particle abundance (a positive curvature) over Ahmedabad (due to desert dust) and Ocean (due to sea salt), while fine particle abundance (negative curvature) is observed over the Kanpur site (due to industrial combustion). We have generated a prototype of lookup table for aerosol optical thickness (AOT) retrieval with modified 6S, though it requires more rigorous modeling. We, at Space applications centre (SAC) have our ground calibration test site at "Chharodi". It is around 35kms away from Ahmedabad. Artificial targets (of asbestos, concrete, mirror & cloth) of different shapes (rectangular, radial, circular etc. for the determination of required geometric/radiometric parameter from it), sizes (according to the spatial resolution of the sensor) and colour (black, white and different gray shades) are used. These targets are well characterized in terms of their geometries, location and spectral properties. It is extensively used for the in-flight calibration and validation purposes for various IRS satellite sensors. We have corrected a few LiSS-4 (of IRS-P6) and PAN (of Cartosat-2A) images over the Chharodi test site and validated the results with the synchronous measurements of reflectances/radiances of the targets with the satellite over pass. AOT, water content and ozone content values are taken from the EOS-MODIS (Earth Observation Satellite- iii Abstract Moderate Resolution Imaging Spectro-radiometer) global geophysical parameter images. We have chosen continental aerosol model, which consists of dustlike, water-soluble particles and soot particles. The targets used for validation at the test site are black cloth (reflectance 3.1%), bare soil (reflectance 20.2%), and white cloth (reflectance 80.2%). After correction with modified 6S, the radiance (atmospherically corrected) of the white cloth (bright target) is enhanced by 86.48% (value increased to 28.93 mW/cm2-pm-sr from 15.51 mW/cm2-pm-sr), which is closer to the ground measured radiance (24.0 mW/cm2-vim-sr). The radiance of bare soil is enhanced by 44.82% (value increased to 8.80 mW/cm2-um-sr from 6.28 mW/cm2-um-sr), which is greater than ground-measured radiance (6.0 mW/cm2-pm-sr). It has definitely improved the quality of the images. In this work, we also corrected (for atmospheric effects) hyper-spectral imagery (one set) of Hyperion data of Earth Observing-I (ED-I) satellite of NASA (National Aeronautics and Space Administration) using FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes). Spectral signatures of the natural ground targets (Viz. forest, soil and water body) of the atmospherically corrected data is compared with in situ measured spectrum of the same targets, but in situ measurements were neither of the same area nor at the same day and time (due to nonavailability of data). It shows that corrected spectrum is better related to the ground measured reflectances. Reflectance features such as linear slope in the Visible and NIR for soil spectra and steep rise in reflectance from VIS-Red to IR in vegetation spectra are better resolved after correction. In this work, we have brought out a new procedure to calculate the Rayleigh optical depth, which is related to the actual atmospheric conditions more realistically. We incorporated this method into the well-known radiative transfer code "6S", and then corrected the satellite images with the "modified 6S". It has shown a significant improvement in the corrections with respect to the original model. We successfully corrected several IRS images. We worked on the aerosol characterization and retrieval. We generated a prototype of lookup table for AOT retrieval. We also corrected one hyper spectral data set. The scope for future work lies in the development of the inversion techniques to retrieve the aerosol size distribution, which can be used in atmospheric correction.|
|Research Supervisor/ Guide:||Vyas, N. K.|
|Appears in Collections:||DOCTORAL THESES (Physics)|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.