Science Service System

Summary of Proposal LAN1830

TitleChange detection on damaged agriculturalarea by the Tohoku offshore earthquake and tsunami, Japan
Investigator YONEZAWA, Chinatsu - Tohoku University, Graduate School of Agricultural Sc., Faculty of Agriculture
Team Member
Prof YONEZAWA, Chinatsu - Tohoku University, Graduate School of Agricultural Science Faculty of Agriculture
Prof SAITO, GENYA - Tokyo Institute of Technology, Innovation Research Initiatives
Dr WATANABE , MANABU - Japan Aerospace Exploration Agency, Earth Observation Research Center
SummaryObjectives
Agricultural field on the pacific coast of the Eastern Japan were damaged by Tsunami caused by the Tohoku off-shore earthquake on March 11, 2011. Total 215km2 agricultural field were damaged, and 33 % of damaged field was reconstructed at one year after the earthquake. Land use has been changing from before the earthquake.
In this study, TerraSAR-X images on the damaged agricultural area is analyzed to extract the temporal change of flooding induced by Tsunami just after the earthquake and recovery process year by year. The change of flooding area just after the damage is extracted. Annual cropping pattern change before the earthquake is extracted and recovery process is evaluated.
Methods
TerraSAR-X data obtained before and just after the Tsunami are analyzed to extract flooding area and initial recovery process. Annual change of cropping pattern is extracted from the data obtained on rice growing season during 2011 and 2012. Agricultural parcel vector set is overlaid to TerraSAR-X image for high accuracy evaluation of land use change and statistical analysis.Terra/ASTER image with almost no clouds for extracting cropping pattern and ground truth data is used for evaluation of the analysis result of TerraSAR-X data.
Funding
The data analysis will cooperated with a project “Reconstruction process monitoring for agricultural field from tsunami damage using remote sensing image and parcel data“ supported by JSPS (Japan Society of Promotion of Science) KAKENHI Grant.
Data requirements
40 data observing the Sendai Plain is requested. Detail is described on detailed description.
Detailed report1.INTRODUCTION
An earthquake hit the eastern coast area of Japan at 14:46 (JST) on March 11, 2011. The magnitude as determined by the Japan Meteorological Agency was Mw 9.0. An enormous tsunami struck the coastal area after the earthquake, and more than 18,000 people either died or went missing. The tsunami damaged areas included a grain belt, the Sendai plain. A total of 215 km2 of agricultural fields were reportedly damaged, and several fields remained flooded until two months after the tsunami disaster. After the earthquake, many earth observation satellites acquired images of the damaged area. Analysis results of SAR data observing the damaged area have been reported.
We analyzed time series TerraSAR-X and TanDEM-X data observing the tsunami damaged agriculture area after the disaster occurrence to detect its drying process.
We also compare backscattering characteristics of TerraSAR-X, ALOS PALSAR and RADARSAT-2 data observing the damaged area for each agricultural parcel. TerraSAR-X is X-band, RADARSAT-2 is C-band, and ALOS PALSAR is L-band SAR. We compare analysis result of these dataset and discuss the effect of difference of wavelength for flooding area detection.

2. STUDY SITE AND DATA ANALYSIS
The study site consists of the agricultural fields on the Sendai plain, Japan. Paddy fields are expanded in this area. Sendai-shi is the largest municipality and consist five districts. Large parts of the agricultural fields in Wakabayashi-ku and Miyagino-ku and a part of the agricultural fields in Taihaku-ku were flooded by seawater following the tsunami. Wakabayshi-ku is selected for main target area for analysis. This area is flat and maximum elevation is 35 m. All analyzed data in this study contain whole area of the Wakabayashi-ku. Continuous observation was carried out by TerraSAR-X and TanDEM-X after the earthquake with HH polarization. All these data were obtained by stripmap mode and its spatial resolution is 1.2×3.3 m for slant range and azimuth on single look slant range complex data.
Backscattering coefficient (gamma naught) values were calculated for each data. The TerraSAR-X image was superimposed on map and mean backscattering coefficients were calculated for each agricultural parcel using vector boundary data. These agricultural parcel data were collected before the earthquake.
RADARSAT-2 HH image acquired on 12 March, 2011 were analyzed for comparison with TerraSAR-X image. Observation mode is wide fine and incidence angle is 37.7° with descending orbit.
Total fourteen HH-polarized ALOS PALSAR image acquired by FBS mode with range resolution is 7 to 44 m from March 13 to April 18 were analyzed. The smallest off nadir angle of these dataset is 14.0° and largest is 50.0°. Three off-nadir angles (21.5, 34.3, and 41.5) and fixed polarizations, i.e., HH, HH+HV, and full polarization, are assigned to the PALSAR regular-operation mode.

3. RESULTS AND DISCUSSION
3.1 Extraction of recovery process
Figure 1 shows some of the TerraSAR-X images obtained after the disaster for Sendai plain. The distribution of the mean backscattering coefficient for each agricultural parcel is shown using a color ramp. On March 12, it is obvious that the areas reached by the tsunami show smaller backscattering coefficients than the areas that were not reached. On March 17, overall backscattering coefficients in tsunami reached area increased. The image on April 11 was obtained from ascending orbit, and this image still show that several fields show low backscattering. On September 4, approximately 6 month after the disaster, it seems that the difference of backscattering coefficient between tsunami reached area and out of the tsunami reached area is obscure.
Nine data were obtained with same orbit direction and incidence angle in analyzed data set. Figure 2 shows the images for Wakabayashi-ku from these data obtained descending orbit with 37° of incidence angle. The tsunami reached area show smaller backscattering coefficients than out of the area in the image obtained on 12 March. On 23 March, the low backscattering area is found on coastal side (denoted ‘A’ in figure 2). However, backscattering coefficients on mountainous side agricultural field in tsunami reached area was increased. Drying from tsunami damage in these area is suggested. From March 23 to May 6, the low backscattering area in coastal side becomes small. Low backscattering area (denoted ‘B’ in figure 2) appears on out of the tsunami reached area on May 17. This area is estimated as flooding area for rice planting. Early May is rice planting season in this region. The low backscattering fields is found just out of the tsunami reached area (denoted ‘C’ in figure 2) on 28 May. This area still shows smaller backscattering on 8 June. It is estimated as rice planting on area ‘C’ is later than the area ‘B’. The low backscattering field in area ’A’ becomes small on 19 June, however, it is still be recognizable. On September 4, backscattering coefficients of area A, B, and C are similar to surrounding area, more than -11dB. This suggests rice growing on area B,C and drying from flooding by tsunami on area A.
We investigate average of the mean backscattering coefficients on agricultural fields reached by the tsunami and on the fields out of the tsunami reached area in Wakabayashi-ku (Figure 3). The tsunami reached fields show smaller backscattering coefficients than the not reached fields except of 28 May to 16 June, flooding for rice planting and early growing season. It is found that the backscattering coefficients affected by incidence angle, it means generally less than 30 degree incidence angle data show large backscattering coefficients, and more than 30 degree incidence angle data show small backscattering coefficients. It is difficult to discuss drying process from tsunami damage from just temporal change of backscattering coefficients change with various incidence angle dataset.

3.2 Comparison with RADARAST-2 and ALOS PALSAR data
3.2.1 RADARSAT-2
Figure 4shows images obtained by RADARSAT-2 on 12 March, 2011. TerraSAR-X observed same area with ca. 5 min of observation time deference and its image is shown on Figure 1 (a). Distribution pattern of backscattering coefficients are similar between TerraSAR-X and RADARSAT-2 data. Tsunami reached area areas show smaller backscattering coefficients than the non-reached area both TerraSAR-X and RADARSAT-2 image.
Wakabayashi-ku contains large flooded agricultural fields. Mean backscattering coefficients of TerraSAR-X image is -16.6 in tsunami reached area and it is -9.5 in the fields where tsunami not reached. It is -16.3 in tsunami reached area and -9.8 in tsunami not reached area on RADARSAT-2 image. Backscattering coefficients value are similar between TerraSAR-X and RADARSAT-2 in spite of difference of observation wavelength.
A scatter plot to show relationship backscattering coefficients in Wakabayashi-ku between TerraSAR-X and RADARSAT-2 is shown in Figure 5. Correlation coefficients between TerraSAR-X and RADARSAT-2 is 0.77.

3.2.2 ALOS PALSAR
We compared TerraSAR–X and ALOS PALSAR data obtained on 16 and 23 March, 2011. The observation orbit of TerraSAR-X on 16 March is Asceing and that of ALOS PALSAR is Descending. On 23 March, both satellite orbit are descending.
Figure 6 shows distribution of mean backscattering coefficient for each agricultural parcel with color gradation on coastal area of Sendai-shi for these data, respectively. Generally ALOS PALSAR data show lower backscattering coefficient than the TerraSAR-X image. It is necessary to consider about the ALOS PALSAR images were acquired by irregular off nadir angle, it means insufficient calibration. On 16 March image, tsunami reached area shows lower backscattering coefficients than the no flooded area on TerraSAR-X image (Figure 6 (a)). This distribution pattern is found from ALOS PALSAR image (Figure 6 (b)). Especially low backscattering area in tsunami reached area appeared on the image obtained on 23 March (Figure 6 (c) and (d)). These area seems to be correspond with flooded area at eight days after the tsunami confirmed by ALOS AVNIR-2 image (Figure 7). Especially, TerraSAR-X image shows good agreement.
Correlation of backscattering coefficients between TerraSAR-X and ALOS PALSAR is appeared (Figure 9). Correlation coefficients are 0.42 on 16 March and 0.55 on 23 March. Correlation between TerraSAR-X and ALOS PALSAR is smaller than that between TerraSAR-X and RADARSAT-2. We found that backscattering coefficient is affected by not only presence of water, but also other factors such as surface roughness, soil moisture and so on from analysis of time series ALOS PALSAR data [4]. A strong scatter has effect to increase backscattering intensity for neighbouring area in the case of ALOS PALSAR because its spatial resolution is 7 to 44 m.

CONCLUSION
We analyzed TerraSAR-X images of agricultural field damaged by Tsunami caused by the 2011 Tohoku earthquake using individual agricultural parcels, and obtained following results:

1. Flooded agricultural field show lower backscattering coefficients than the surrounding area. Lower backscattering coefficients area were found on the image obtained on 3 month after the disaster occurrence, however, these area disappeared on 6 month after the disaster. This suggests the Tsunami flooded area was dried until 6 month after the Tsunami.
2. Backscattering coefficients is affected by incidence angle. Generally, large incidence angle data show smaller backscattering coefficients than the smaller incidence angle data. 3. Flooding for preparing rice planting appear as small backscattering coefficients. The agricultural parcel of which backscattering coefficients decrease from early May to June corresponds to rice planting area.
4. The difference of average for the mean backscattering coefficients of agricultural field parcels between the tsunami reached area and the not reached area is the largest on 12 March, the next day of the disaster and it generally decrease with time. This difference becomes negative from late May to June, rice planting and early growing season.
5. Two data obtained different sensors on same day were compared. TerraSAR-X and RADARSAT-2 data obtained on 12 March, the next day of the tsunami occurrence show similar behavior in spite of different wavelength. Correlation coefficients are 0.77 on agricultural fields in Wakabayashi-ku distinct.
6. TerraSAR-X and ALOS PALSAR data obtained 16 and 23 March were also analyzed and compared. Correlation coefficients on Wakabayashi-ku were 0.42 and 0.55 on March 16 and 23, respectively. Both data set shows smaller correlation coefficients than the TerraSAR-X and RADASAT-2.

The result of this study suggests that time series of TerraSAR-X data generally reveals flooding and drying process from tsunami damage. However, it is necessary to consider about observation direction and incidence angle to discuss the change of backscattering coefficients. Management and reconstruction of tsunami damaged agricultural field is still important matter. The analysis result of TerraSAR-X data provides relevant information.

ACKNOWLEDGMENTS
RADARSAT-2 data are provided by GEO’s Tohoku-oki Event Supersite Website. ALOS PALSAR data are provided under the agreement of JAXA Research Announcement. This research is supported by by JSPS (Japan Society of Promotion of Science) KAKENHI Grant Numbers 24510252,30221774. Agricultural parcel data is provided by land improvement association of Miyagi prefecture.

Back to list of proposals

© DLR 2004-2016