應用MODIS時間序列影像探討東日本大海嘯對稻田受損及復育之影響

Analysis of Damaged Rice Fields and Rice Restoration after he Great East Japan Tsunami Using Time-Series MODIS Data

蕭雅勻(Ya-Yun Hsiao)

指導教授 ： 陳繼藩

國立中央大學/工學院/土木工程學系/碩士(2015年)

摘要

2011年3月11日，日本東部海上發生九級地震引發海嘯，造成東北沿岸地區農業上巨大的損失，共23,600公頃的農地受到損害。單在宮城縣和福島縣的海嘯受災區，稻田就占受災區60%以上，稻作為主要作物，因此，了解日本災後稻田受損和復育的情況對於農政單位將會是一項重要的議題。遙測影像提供多時序、大範圍的資料，而稻作在光譜上的反應比其他土地利用更為明顯，故本研究使用2010年(海嘯前一年)、2011年(海嘯發生年)、2012年(海嘯後一年)、2013年(海嘯後二年)、2014年(海嘯後三年)的MODIS影像，探討受災區稻田於海嘯發生前、後的分布情形以及災後的復育情況。本研究主要分為四個步驟：第一步驟，使用MODIS影像建立常態化差分植生指標(Normalized Difference Index，NDVI)的時間序列資料，並以小波轉換法(wavelet transform)濾除時序資料的雜訊，保留時序資料上有關稻作的資訊。第二步驟，使用支援向量機(Support Vector Machine, SVM)進行影像分類。第三步驟為精度評估。第四步驟，分析2010年至2014年間稻田受損和復育的情況。研究成果顯示，2010年至2014年的總體精度(overall accuracy)分別為88.0%、91.7%、89.5%、92.2%和90.8%，Kappa值分別為0.76、0.83、0.78、0.84和0.81。2011年海嘯過後，受災區有84.9%的稻田受損而無法耕種，2012年日本農民在沿岸地區開始陸續種植稻作，整體復育率為27.4%，2013年整體復育率為38.5%，2014年整體復育率為33.1%。將2010年至2014年的分類成果和統計資料相比，RMSE皆僅占總面積3%以下，顯示本研究能應用於災後稻田受損和復育之相關研究。

關鍵字

MODIS ； 水稻復育 ； 常態化植生指標 ； 東日本大海嘯

並列摘要

The 2011 Great East Japan Tsunami,happened on 11 March, 2011, destroyed 23,600 hectares of cultivation areas in northeastern coastal area of Japan, especially in Fukushima Prefecture and Miyagi Prefecture. More than 60% of inundation area is the area of rice fields. Therefore, monitoring the damaged rice fields and restoration of rice area after the tsunami is critical to provide agronomic planners with valuable information for the effective crop management strategies. Satellite imagery can provide multi-temporal and wide region data, and the spectral profile of rice is different from other land use types. This study used Moderate Resolution Imaging Spectroradiometer (MODIS) data from 2010 to 2014 to investigate the damaged rice fields and restoration of rice areas after the tsunami. The procedure of data processing consists of four steps: (1) data pre-processing to produce the time-series Normalized Difference Vegetation Index (NDVI) data and data filtering of the NDVI time-series data by wavelet transform; (2) paddy rice mapping using support vector machine (SVM); (3) accuracy assessment, and (4) area estimation of damaged rice fields and their restoration. The mapping results indicated the overall accuracy is 88.0%, 91.7%, 89.5%, 92.2% and 90.8% and the Kappa coefficient is 0.76, 0.83, 0.78, 0.84 and 0.81 from 2010 to 2014, respectively. After the tsunami (2011), 84.9 % of rice fields damaged by the tsunami is estimated. In 2012, the restoration rate is 27.4%. In 2013, the restoration rate is 38.5%. In 2014, the restoration rate is 33.1%. The comparison of RMSE divided by the total area in classification results and statistic data from 2010 to 2014 are all lower than 3%. It indicated this study can apply to the analysis of damaged rice fields and their restoration.

並列關鍵字

MODIS ； rice fields restoration ； Normalized Difference Index ； the Great East Japan Tsunami

參考文獻

林世峻、莊智瑋、何世華、林昭遠，2008，"植生指標對影像分類準確度影響之研究"，水土保持學報，40(2)：181-193。

Boschetti, M., Stroppiana, D., Brivio, P. A., et al. (2009). Multi-year monitoring of rice crop phenology through time series analysis of MODIS images. International Journal of Remote Sensing,30(18), 4643-4662.

Chagué-Goff, C., Przemyslaw, N., Henri, K. Y. W., et al. (2012). Environmental impact assessment of the 2011 Tohoku-oki tsunami on theSendai Plain. Sedimentary Geology, 282, 175-187.

Chen, C. F., Huang, S. W., Son, N. T., et al. (2011). Mapping double-cropped irrigated rice fields in Taiwanusing time-series Satellite PourI’Observation de la Terre data. Journal of Applied Remote Sensing, 5, 1-18.

Dewi, K. S., Ishak, H. I., Widyo, N. S., et al. (2010). Detecting Rice Phenology in Paddy Fields with Complex Cropping Pattern Using Time Series MODIS Data:A Case study of Northern Part of West Java-Indonesia.ITB Journal of Science, 42A(2), 91-106.