Home Articles Construction of satellite image data set by multispectral and hyperspectral sensors

Construction of satellite image data set by multispectral and hyperspectral sensors

ACRS 2000

Poster Session 1

Construction of Satellite Image Data Set by Multispectral
and Hyperspectral Sensors

Yuzo Suga1, Kenji Takasaki2 and Shoji Takeuchi1


1Professor, 2Assistant, Hiroshima Institute of Technology,

2-1-1, Miyake, Saeki-ku, Hiroshima 731-5193, Japan

Tel & Fax : (81)-82-922-5204
E-mail:[email protected]


Keywords :
NOAA/AVHRR, SeaWiFS, MODIS, Landsat-7/ETM+, Regional Monitoring.


Abstract
Hiroshima Institute of Technology (HIT) has established the ground receiving station that has capabilities to capture and process the x-band down link data from earth observation satellite. This study deals with the construction of satellite image data set used to map the environmental information using several kinds of data from NOAA/AVHRR, SEASTAR/SeaWiFS, TERRA/MODIS, Landsat-7/ETM+, SPOT-1,2/HRV and SPOT-4/HRVIR captured in HIT ground station. This satellite image data set is constructed for the purpose of environmental information mapping from the wide area to the local area scale using the different spatial and spectral resolutions such as multispectral and hyperspectral sensor data. In this paper, some preliminary results are presented to demonstrate the capability for land and sea environmental monitoring using the data set produced by the initial operation of HIT ground station.


1. Introduction

Since 1980’s NOAA/AVHRR data have been widely used for monitoring of earth environment in global scale. For regional scale, AVHRR sensor still has been only one practical data source to be used to monitor land and sea environment, such as the distributions/conditions of vegetation, sea surface temperature (SST) and to monitor disaster phenomena such as forest fire or volcano eruption. On the other hand for local scale, Landsat-5/TM and SPOT/HRV have been main data sources for land cover mapping, land cover change detection, and disaster monitoring etc. Nowadays, some new sensors are going to be available in global and regional scale, i.e. SEASTAR/SeaWiFS and TERRA/MODIS. The spatial resolution of these sensors are moderate and compatible to that of AVHRR, however, the number of spectral band is significantly increased, especially for MODIS sensor. By these new sensors, the spectral information is significantly enhanced compared with that by conventional AVHRR sensor and therefore, they are highly expected to extract more plentiful and accurate environmental information in global or regional scale.

Recently Hiroshima Institute of Technology (HIT) has established the ground receiving station that has capabilities to capture and process the x-band down link data from various earth observation satellites, SeaWiFS, MODIS, Landsat-7/ETM+, and SPOT-1,2/HRV and SPOT-4/HRVIR. In this paper, we report the present status for receiving and processing of these satellite data and some preliminary results by data analysis of these satellite data to evaluate their capabilities for monitoring land and sea environment.


2. Sensor Specifications and Possible Sensor Combinations

The sensor specifications for AVHRR, SeaWiFS, MODIS and ETM+ are summarized in Table 1 to 4. As the wavelength characteristics of SeaWiFS and MODIS, it is pointed out that bandwidths are generally narrower than those of AVHRR and ETM+ as well as that number of band is much bigger. For example,
the band-1 of AVHRR corresponds to the band-6 of SeaWiFS as a wavelength region. However, the bandwidth for AVHRR is 0.1 μm, while that for SeaWiFS is 0.02 μm and five times narrower than AVHRR because the band-6 of SeaWiFS is designed to extract spectral signatures caused by absorption by chlorophyll. Therefore, the conventional vegetation index parameter like NDVI is still expected to be improved by replacing AVHRR band-1 with SeaWiFS band-6.


Table 1. Specification of NOAA-14,15/AVHRR.

Band No.
Wavelength Region(μm)
Spatial Resolution

1
2
3

4
5
0.58 – 0.68
0.725 – 1.1
3.55 – 3.93
or 1.58 – 1.64 (N-15 only)
10.3 – 11.3
11.5 – 12.5
1.1 Km
1.1 Km
1.1 Km

1.1 Km
1.1 Km

Table 2. Specification of SEASTAR/SeaWiFS.

Band No.
Wavelength Region(μm)
Spatial Resolution

1
2
3
4
5
6
7
8
0.402 – 0.422
0.433 – 0.453
0.480 – 0.500

0.500 – 0.520
0.545 – 0.565
0.660 – 0.680
0.745 – 0.785
0.845 – 0.885
1.1 Km
1.1 Km
1.1 Km
1.1 Km
1.1 Km
1.1 Km
1.1 Km
1.1 Km

Table 3. Specification of TERRA/MODIS.

Band No.
Wavelength Region(μm)
Spatial Resolution

1
2
0.620 -0.670
0.841 – 0.876
250 m
250 m

3
4
5
6
7
0.459 – 0.479
0.545 – 0.565
1.230 -1.250
1.628 – 1.652
2.105 – 2.155
500 m
500 m
500 m
500 m
500 m

8
9
10
11
12
13
14
15
16
17
18
19
26
0.405 -0.420
0.438 – 0.448
0.483 – 0.493

0.526 – 0.536
0.546 – 0.556
0.662 – 0.672
0.673 – 0.683

0.743 – 0.753
0.862 – 0.877
0.890 – 0.920
0.931 – 0.941

0.915 – 0.965
1.360 – 1.390
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km

20
21
22
23
24
25
27
28
29
30
31
32
33
34
35
36
3.660 – 3.840
3.929 – 3.989
3.929 – 3.989*
4.020 – 4.080
4.433 – 4.498
4.482 – 4.549
6.535 – 6.895
7.175 – 7.475
8.400 – 8.700
9.580 – 9.880
10.780 -11.280
11.770 – 12.270
13.185 – 13.485
13.485 – 13.785
13.785 – 14.085
14.085 -14.385
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km
1 Km

* Saturation level is smaller than previous band.

Table 4. Specification of Landsat-7/ETM+.

Band No.
Wavelength Region(μm)
Spatial Resolution

1
2
3
4
5
6
7
8
0.45 – 0.52
0.53 – 0.61
0.63 – 0.69
0.78 – 0.90
1.55 – 1.75
10.4 – 12.5
2.09 – 2.35
0.52 – 0.90
30 m
30 m
30 m
30 m
30 m
60 m
30 m
15 m

Among the four sensors listed in Table-1 to 4, ETM+ is special due to its much higher resolution and narrower swath, which has been used mainly for monitoring local areas. However, as the first seven bands of MODIS almost corresponds to the wavelength regions by ETM+, there might be some possibility to use both sensor data as the complement data in spatial resolution and swath. For example, the up-scaling of land cover information using Landsat-5/TM and NOAA/AVHRR was attempted (i.e. Takeuchi and Inanaga, 2000) and proved to be effective for monitoring wide areas more accurately even if both spectral information is not compatible each other. Therefore, up-scaling using TM/ETM+ and MODIS band-1 to 7 is expected to bring more accurate monitoring capability by the combined utilization of multiple sensor data.

SeaWiFS is originally designed to extract some physical parameters for oceanic applications, like water quality and chlorophyll content using visible band information. Therefore, one of the standard ways for sensor combinations will be the combined use of visible bands from SeaWiFS with the thermal bands from AVHRR, the former brings water quality and chlorophyll content and the latter sea surface temperature (SST). In addition, due to its capability for non-saturated observation of land areas, the similar combination might be effective for land environmental monitoring. For example, NDVI by SeaWiFS might be better than that by AVHRR as described previously, or, the enhanced vegetation index (EVI) is possibly extracted from band-1/2, 6, and 8 of SeaWiFS, while it is impossible by AVHRR data. The latter approach is also possible for MODIS band-1 to 3 (Justice et al, 2000 and Huete et al, 2000). These multiple sensor combinations are expected to enable to achieve more accurate land environmental monitoring in regional scale.

ACRS 2000

Poster Session 1

Construction of Satellite Image Data Set by Multispectral
and Hyperspectral Sensors


3. Data Reception and Processing in Hit

At a moment, HIT has two antenna systems, a 7.3 meters Transportable Ground Station (TGS) antenna and a 13 meters antenna, the latter was just established in March this year. Currently SeaWiFS and MODIS are received by the 7.3 meters antenna and Landsat-7/ETM+ and SPOT-1,2, and 4 are received by the 13 meters antenna. In addition, NOAA/AVHRR is received by a 1.2 meters antenna system.

Landsat-7/ETM+ data are processed soon after reception in the 13 meters antenna system. As the products, Raw, Level-0, Level-0R, Level-1R and Level-1G are archived. The format of Level-1R is HDF only and those for Level-1G are HDF, FAST and GeoTIFF. SPOT-1, 2 and 4 data are also processed in the 13meter antenna system and Level-1B with CAP format is archived as the SPOT data product. For NOAA/AVHRR, raw AVHRR data are converted to NOAA/NESDIS Level-1B format. For SeaWiFS and MODIS, all data are archived in HDF format.


4. Some Preliminary Results

We preliminary evaluated the spectral information by AVHRR, SeaWiFS, MODIS, and ETM+ data. AVHRR, SeaWiFS, and MODIS data were acquired on the same day (Aug. 4, 2000). ETM+ was acquired on July 22, 2000, almost same season as that for the formers. Figure 1 shows AVHRR false-color composite image, in which red and blue are assigned to band-1 and green for band-2. As there are only two bands for visible and near-infrared regions, there is no other way to enhance spectral information except NDVI, which will be discussed later.

On the other hand, as SeaWiFS has total eight bands in the same wavelength regions, there might be various band combinations to enhance spectral information in land areas as well as in water areas. Figure 2 shows an example for such an enhanced false-color composite image. In Figure 2, red is assigned to the enhanced image by the computation of (B6-B1)/(B6+B1), where B1 and B6 are data values for band-1 and band-6 respectively. This image enhances the difference of reflectance in band-1 (blue-color region) and band-6 (red-color region). The other two color components, green and blue, are assigned to band-8 (near-infrared) and band-4 (green-color region) respectively. Therefore the color image in Figure 2 includes all information from blue, green, red and near-infrared regions. The result clearly indicates the differences between forested land covers and urban/agricultural land covers, while in AVHRR image these differences are not identified so clearly.

Figure 3 shows an example of MODIS false-color composite image acquired on the same day as AVHRR and SeaWiFS. The color assignment is that red is assigned to band-6, green to band-2, and blue to band-1. Band-6, 2 and 1 correspond to band-5, 4 and 3 of ETM+ respectively. Figure 4 shows the false-color of ETM+ in the color assignment similar to MODIS, namely red is assigned to band-5, green to band-4 and blue to band-3. Both images indicate similar color tones for similar land covers. This result suggests that there might be a possibility to use both data as a complement data set as described in section 2.

Figure 5 shows an enhanced color image using principal component analysis (P.C.A.) for eight bands data only in water areas. The colors are assigned to the first three principal components, red to PC-1, green to PC-2 and blue to PC-3. The image clearly shows the differences in water quality between the water in Pacific Ocean and the water in coastal regions or Seto Inland Sea. Figure 6 shows the color image for sea surface temperature (SST) derived from AVHRR on the same day. It is clearly seen from both images that the boundary for two different water quality in Figure 5 corresponds to the boundary for two different temperature regions in Figure 6, namely between inner or coastal water areas with lower temperature and the warm ocean current (Kuroshio) with higher temperature. This result suggests that the combined use of visible information from SeaWiFS and thermal information from AVHRR is very effective for monitoring sea surface environmental conditions.

Figure 7 and 8 show the comparison of NDVI images derived from AVHRR and SeaWiFS data respectively. As described in section 2, there might be a possibility to improve NDVI parameter by using SeaWiFS band-6 as a visible band. The comparison of Figure 7 and 8 supports this improvement. In Figure 7 it is rather difficult to discriminate the differences between forested land covers and agricultural land covers (see Kyushu region in Figure 7). On the other hand, in Figure 8, it is very easy to discriminate these differences. The observation seasons of both images are summer and rice is planted in almost agricultural areas and it is almost a peak condition for rice-growth in this season. Therefore, the result of Figure 7 and 8 clearly supports that NDVI derived from SeaWiFS is more effective than that by AVHRR for monitoring vegetative conditions.

ACRS 2000

Poster Session 1

Construction of Satellite Image Data Set by Multispectral
and Hyperspectral Sensors

5. Conclusion

Some preliminary results on the evaluation of spectral information for regional land and sea environmental monitoring using NOAA/AVHRR, SeaWiFS, MODIS and Landsat-7/ETM+ were presented in this paper. These preliminary results support that the data from above new sensors are capable to monitor land and sea environmental conditions more accurately in regional scale if the sensors are combined each other. From now we will develop the combination methods for above sensors to establish effective data sets based on HIT ground receiving facilities, and also conduct various kinds of feasible studies for land and sea environmental monitoring around the Japan Islands.


References

  • Takeuchi, S. and A. Inanaga, 2000. Scaling-Up of Land Cover Information by Using Multiple Resolution Satellite Data. Adv. Space Res. Vol.26, No.7, 1127.
  • Justice, C., J.Townshend, E.Vermote, et al. 2000. Preliminary Land Surface Products from the NASA Moderate Resolution Imaging Specroradiometer (MODIS). Proceedings of IGARSS’2000 Symposium.
  • Huete, A., K.Didan, Y.Shimabokuro, L.Ferreira, and E.Rodriguez. 2000. Regional Amazon Basin and Global Analysis of MODIS Vegetation Indices: Early Results and Comparisons with AVHRR. Proceedings of IGARSS’2000 Symposium.


Fig.1.NOAA-14/AVHRR False-Color.
(Red:B1, Green:B2, Blue:B1)

Fig.2. SeaWiFS Enhanced False-Color.
( Red:(B6-B1)/(B6+B1), Green:B8, Blue:B4)

Fig.3. MODIS False-Color.
(Red:B6, Green:B2, Blue:B1)

Fig.4. Landsat-7/ETM+ False-Color (Path:112).
(Red:B5, Green:B4, Blue:B3)

Fig.5. SeaWifs Enhanced False-Color by P.C.A.
(Red:PC-1, Green:PC-2, Blue:PC-3)

Fig.6. NOAA-14/AVHRR Sea Surface
Temperature (SST) – Color.

Fig.7. NOAA-14/AVHRR NDVI-Color.
[ NDVI = (B2-B1)/(B2+B1) ]

Fig.8. SeaWiFS NDVI-Color.
[ NDVI = (B8-B6)/(B8+B6) ]