Effect of Atmospheric Correction on Satellite Image Data by Lowtran Model
Kiyoshi Toril*, tomoyuki Mase** and Takashi Hoshi***
*Dept. of Agric. Eng. Kyoto univ, Kyoto, Japan
** Ministry of Agric., Forestry & Fisheries, Tokyo, Japan
***Dept. of Comp. & Information Sci . Ibaragi Univ, Hitachi, Japan
Fax (Int'l+81)+773-64-3617,
E-Mail cxk02755@niftyserve.or.jp
Abstract
Remote sensing technique is one of the most expected tools to be used in environmental measurements. However, distortions and noises contained in satellite images become serious impediments for quantitative analysis.
In the present study, we used Lowtran Model to eliminate atmospheric effects with respect to thermal infrared, visible and near infrared bands due to technical differences and examined the result. Data used in the study are Landsat 5 TM data on Chaophraya plain, Thailand, and Kojima bay Okayama Pref., Japan We obtained interesting data derived from the difference in atmosphere in the tropics and temperate zone.
1. Introduction
Accuracy in earth survey by satellites has been raised with the advancement in space technology and measurement techniques and objects of the survey have also been variegated, Observations of vegetation, soil water, and chlorophyll content and earth surface temperature are included in the related fields. With the progress, progress, programs for satellite data analysis has also been developed and excellent soft wears such as arc/info, IDRISI and ERDAS have become popular. Analytical method has shifted from visual and qualitative ones at the beginning to quantitative one noises and they form a great hindrance on quantitative analysis. These errors and noises are roughly classified into 3 categories; noises derived from ability of a sensor, geometrical distortions, and atmospheric effects. Here, we discuss about elimination of atmospheric effects, which is the main theme of correction.
Presented at the 16th Asian Conference on Remote Sensing, Surananree University of Technology, Nakhon Ratchasima, Thailand, November 20-24, 1995.
2. Concepts of Atmospheric Correction
Radiation entering a sensor is classified as in Fig. 1. Atmospheric correction is the processing to eliminate S2 and S4 leaving S1 and S2 which are the objects of observation as well as to eliminates S5 in Fig. 2 contaminating from the vicinities of the observed pixels. However, techniques for atmospheric correction have not been established and many studies are still going on at present. A general formula for transmission of radiation in a multiple atmospheric model is as follows,
Figure 1 Incxident radiation to satellite senser
Figure 2 In the case of visible, and near-infrared band
where I
v: spectral radiance in frequency range v, v::optical thickness,
m: cosine of the zenith angle,
f: azimuth, :
eearth : emission rate of the ground,
eair emission rate of the medium, Bn: Planck function,
Planck constant, n: frequency, c: velocity of light, k: Boltzmann constant, temperature of the medium, temperature of the ground, av: albedo of the primary scattering of the medium, and Pn: probability function of the scattering angle. The left hand side of the equation is the radiance observed by a satellite the first term of the right hand side is the surface radiation, the second term is the atmospheric radiation and the third term is the scattered light including that by the earth surface. Optical thickness, , is the integrate of the volumetric dispersion coefficient, K
v, from the upper boundary of the atmosphere
(z=
µ)
to the altitude z and is defined as follows,
To solve the equation (1), parameters including atmospheric temperature and optical thickness (scattering coefficient) are needed. Formerly, these parameters were decided on the basis of the actual measurement values obtained from a radiosonde in the vicinity of the site of analysis or correlation in observation intensities in multiple bands. However, in recent years, excellent atmospheric models such as LOWTRAN and FASCODE have been published facilitating easy access to these parameters. In the present study, we used LOWTRAN 7 as an atmospheric model in our calculation. LOWTRAN 7 is a program source written in FORTRAN and, in a simple case, it can provide optical thickness, atmospheric temperature and altitude distribution of aerosol content in each model selected from Table 1.
Table 1 Standard input parameters in LOWTRAN 7
| Parameters related to temperature, humidity, gaseous concentration |
Parameters related to aerosol |
| Tropical Type |
No aerosol |
Summer type |
Stratospheric type |
| Midlatitude summer type |
Rural type (23k km visual range) |
Winter type |
Aged volcanic type |
| Midlatitude winter type |
Rural type (5km visual range) |
|
Fresh volcanic type |
| Subarctic summer type |
Marine type |
|
|
| Subarctic winter type |
Urban type |
|
|
| 1976 U.S. type |
Tropospheric type |
|
|
| User definition |
Fog type |
|
|
| |
Desert type |
|
|
| |
User definition |
The image data used in the present study is the one observed by LANDSAT-Thematic Mapper (TM) and it has 7 bands ranging from visual to thermal infrared wavelengths (0.45.12.5
mm). The parameters used in Equation (1) differ greatly depending on the wavelength and the method of atmospheric correction differs accordingly. In the present study, we discuss about atmospheric corrections in 2 wavelength zones, visual and near infrared zone and thermal infrared zones.
2.1 Data correction in thermal infrared zone
LANDSAT-TM consists of 7 channels and only one channel is designated to the thermal infrared zone(10.4
mm-12.5
mm). Calculation for correction of thermal infrared band data concerns with accurate calculation of the earth surface temperature. While the model equations to convert the observed values to the surface temperature have been published by NASA,etc. they are either uniform totally ignoring longitude and atmospheric effects of the observation site of localized concerning specific data. So in the present study, we modified the correction technique to be applicable to all sorts of atmosphere or earth surface can be ignored and, therefore, radiation laws can be simplified greatly. In other words, upward radiance, Iobs from the earth surface to the satellite is expressed by the following equation obtained by multiplying each term of Equation (1) by the satellite response function Rn, and integration it by the instantaneous angle of view and frequency. (3)
In the equation u1u2 the range of instantaneous field of view and v1 v2 : the range of frequencies observed by the satellite. In the calculation, atmosphere up to 100-km altitude was horizontally divided into 32 uniform layers and the atmospheric up to 100-km altitude was horizontally divided in to 32 uniform layers and the atmospheric temperature and optical thickness determined by LAWTRAN 7 as well as the response function of the sensor were substituted into Equation (3). Then. was changed in the temperature range corresponding to the atmospheric model and the observed radiance, I
abs was determined .While LOWTRON presents a mean data of a long period including fine and rainy weather, observation by the thermal infrared remote sensing is limited to fine weather and therefore, humidity which is an important factor in the present problem becomes excessive in comparison to that of the atmosphere at the time of observation. Because of this, only humidity was modified to the respective winter types in the mid latitude and subarctic summer models in the present study, The maximum input radiance of the sensor was set as
I
max =B
v (340) and the minimum input radiance as
I
min =B
v (200) and the digital output value, D, is recorded as follows.
By the above equation the supplied satellite data, D, and the ground temperature,
T
earth are correlated as follows, Here, aerosol models were all designated to the rural model (5 km visual range).
Figure 3 Correlation between quantitized radiance and surface temperature
The ratio of atmospheric radiation included in the observed radiance,
I
abs becomes as Fig. 4 The temperature of the observed radiance,
T
abs of Fig. 4 is defined by Bn (T
abs)=I
abs
Figure 4 The ratio of atmospheric radiance to observed value
