International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
  
‚Choose study area 
Synchronous 
. monitoring — 
  
| | radiometric correction | 
Y 
— Geometric correction | 
  
“Image data 
pretreatment 
| | atmospheric correction | 
| 
| Empirical regression - | 
Construct | | model 
L ie. [ 
Analysis of = 
results 
Figure l. The Frame of Research Plan 
“Artificial neural 
network model abel 
  
  
  
2. DATA COLLECTION AND PROCESSING 
2.1 Synchronous Monitoring 
Water quality monitoring experiment synchronized with 
Landsat satellite was performed in PoYang Lake in July 8, 2001. 
The locations of sampling points were set as Figure 2. 
  
T / / T * 
Aj CM s 
Vi = x 
y NaniCuang S 4 
f * > x P ER ár 
Cu f Sr LA 
C j ' 34 
Jah « x 
] 1 ^ "n. 3 
Ï e X © 
1 d > 1 fux 4 
| - f gio i03 
TH % ÿ ey el Qt" 
gt Á aes NT 4 
  
  
  
5 Ha 
a £T 5 : 
et "€ Sampling Poinfs — "a. rà 
va CU Poyangiane /-PHéPre Lp 
E SS 1 t es MPa e d A 
o Tm 
  
Figure 2. Distribution of sampling points 
There were ten sampling points, at which SS, chl-a, TN, TP, 
CODmn, DO, temperature and pH were monitored. DO, 
temperature and pH were measured in situ, and the others in 
laboratory. Data of longitude and latitude were obtained by 
GPS at each point in situ. 
2.2 Remote Sensing Data 
Remote sensing data adopted synchronous Landsat 7 ETM+ 
data for its good spatial resolution, which path/row numbers 
were 121/40. The satellite image was clear and cloud-free. 
2.2.1  Pre-processing of Remote Sensing Data 
The remote sensing data needs several steps of pre-processing 
before an inversing model is applied, which include radiometric 
correction, geometric correction and atmospheric correction. 
Commonly the purchased image has been processed by 
radiometric correction and original geometric correction, so 
jobs that users need to do are accurate geometric correction and 
atmospheric correction. The accurate geometric correction in 
this study was accomplished by ground control points (GCPs), 
678 
whose precision was better than one pixel. The following part 
discussed the procedure of atmospheric correction. 
2.2.2 Atmospheric Correction 
PCI, a commercial image processing software package, 
provides a set of atmospheric correction tools for sensors of TM, 
MSS and SPOT, such as ATCORO, ATCORI and ATCOR2. 
The flow chart of atmospheric correction is showed as Figure 3. 
Original Data: TM/MSS or SPOT3/4 
* 
ATCORO Define aerosol optical depth: VISIBILITY | 
  
  
   
: x. : 
. Create reflectance image: 
cATCOR | Using ATCORO to give VISIBILITY. 
Y 
Reflectance Image Without Adjacency Effect 
FAV Lowpass filter 
En Create lowpass reflectance image 
Y 
Lowpass Reflectance Image 
ATCOR2 C reate improved reflectance image with 
.adjacency effect — —— 
Y 
Reflectance Image With Adjacency Effect 
Flow chart of atmospheric correction provided by 
PCI 
Figure 3. 
The theory of this method is given by Richter (Richter, 1990; 
Richter, 1996), whose main idea is that according to standard 
atmospheric categories, atmospheric dispersion has been 
calculated in different aerosol types, different sun zenith angles, 
different altitudes and different atmospheric visibilities, the 
results of which are stored in a directory, like as look-up table. 
In actual application atmospheric correction is performed 
according to this table. Categorizing basis of Richter method 
comes from middle resolution atmospheric transmission model 
- MODTRAN. Its arithmetic also considers and corrects the 
adjacent effect of ground reflection. 
3. EMPIRICAL REGRESSION MODEL 
Empirical Regression Model often sets remote sensing data 
(atmospheric correction or not) as independent variables and 
concentration of water quality components as dependent 
variable(s) to construct their relative equations. In order to 
review effects of empirical regression models, this study 
designed following combinations, and detailed description was 
showed in the previous study (Kuang, 2002) . 
€ Atmospheric correction: yes or no. two cases. 
€ Independent variables: 89 kinds of remote sensing band 
combination, such as RI, RI*R3, (R3*R5)/In(R1*R2) 
and so on (Kuang, 2002). 
€ Dependent variables: SS and chl-a. 
3.1 Results of Regression Model 
The datasheet of water quality monitoring and remote sensing 
data were showed in Table 2. Calculating the relative 
coefficient of every above combination (the total mounts of 
combination were 356), results that had good correlation were 
listed in Table |. 
Int 
  
3.2 
Ta 
Th 
Th 
at 
ap] 
res 
ser 
the 
rer 
otl 
fur 
Th 
Col 
to 
4.1 
Fig 
str 
bai 
cry 
inp 
of 
*