臺灣地區降雨型態分佈不均，並且地形不易留存地表水資源，在此情形下，地下水資源作為廉價且易於開發的補充性水資源被大量開發，然而過度的開發地下水資源卻易釀成地層下陷的禍端。本研究選擇台灣九大地下水分區中，地層下陷發生情況最為嚴重的濁水溪沖積扇作為研究區域，利用自2009年至2019共計11年的各地層下陷GPS固定站觀測資料，藉由參考標準化雨量指數法(SPI)以及標準化地下水位指數法(SGI)之計算流程，建立能廣泛應用於各種不同狀況的地層下陷時間序列分析法-標準化地層下陷指數法(SSI)；並且在成功確立標準化地層下陷指數地計算流程與數據處理步驟後，本研究將進一步嘗試利用長短期記憶類神經網路之輔助，將標準化地層下陷指數的應用由地層下陷時間序列變化分析之領域推展至地層下陷時間序列趨勢變化預測的領域。
經計算之後所得之標準化地層下陷指數(SSI)在時間序列的變動趨勢上與地層下陷GPS固定站觀測資料完全相同並且其資料數值經過標準化處理，對於應用在分析地層下陷長期變動趨勢以及數值化評估地層下陷狀況之領域具有相當之實用性。
而長短期記憶網路的訓練結果顯示，標準化指數資料組的趨勢模擬預測結果曲線在與歷史資料序列曲線的趨勢吻合程度表現上與原始數據資料組相差無幾，然而由於應用標準化指數來進行趨勢模擬預測相較於使用原始觀測資料進行趨勢模擬預測尚有諸如無因次值統一化等的優勢，因此本研究認為將標準化指數法應用於地層下陷趨勢變化預測領域此一想法是可行且具有發展性的，唯其具體理論建立與發展以及技術的實際應用，未來還需後續研究資料補充。
關鍵詞：濁水溪沖積扇、地層下陷、標準化地層下陷指數法(SSI)、長短期記憶類神經網路、地層下陷地表沉陷量趨勢預測

The distribution of rainfall patterns in Taiwan has been uneven since long ago, and the terrain could not easily retain much surface water resources. Under such circumstances, groundwater is widely developed as a cheap and easy-developed supplementary water resource. However, excessive extraction of groundwater have caused some land subsidence problems in Taiwan. In order to study the behaviors of land subsidence, Zhuoshui River Alluvial Fan in Taiwan was selected as the research area. First, a new index for standardizing time series and characterizing land subsidence situations, the Standardized Subsidence Index (SSI), was described. The SSI was built referring to the calculation process of the Standardized precipitation index (SPI) and Standardized groundwater index (SGI). Then these Standardized indexes has been calculated for land subsidence monitoring data recorded between 2009 and 2019 from 5 GPS land subsidence observation stations. Finally, after finishing the calculation of the Standardized subsidence index (SSI); we utilized a long-short term memory (LSTM) model which provides a deep-learning-based time-series processing method for modeling the land subsidence prediction. The monitoring data of groundwater level observation stations and rainfall observation stations from 2009 to 2019 were collected as two input variables of all datasets. All of the collected datasets would be divided into two groups, the original datasets group and the standardized index datasets group. Original datasets group included the monitoring data of rainfall observation station, groundwater level observation stations, and GPS land subsidence observation station. The Standardized index datasets group was included the SPI, SGI and SSI. The results show that the prediction of land subsidence is not much different between both groups of LSTM training results, but the standardized index datasets group has an advantage in the other aspect such as the standardized value of the data. Generally speaking, this study considers that utilizing standardized index method for modeling the land subsidence prediction is feasible, but more research was needed for the theory to be completely built.
Key Words：Zhuoshui River Alluvial Fan, Land subsidence, Standardized Subsidence Index (SSI), Long-short term memory algorithm

1. Abdulrazzaq, Z. T., Hasan, R. H., & Aziz, N. A. (2019). Integrated TRMM data and standardized precipitation index to monitor the meteorological drought. Civil Engineering Journal, 5(7), 1590-1598.
2. Abiodun, O. I., Jantan, A., Omolara, A. E., Dada, K. V., Mohamed, N. A., & Arshad, H. (2018). State-of-the-art in artificial neural network applications: A survey. Heliyon, 4(11), e00938.
3. Bazrafshan, J., Hejabi, S., & Rahimi, J. (2014). Drought monitoring using the multivariate standardized precipitation index (MSPI). Water resources management, 28(4), 1045-1060.
4. Bloomfield, J. P., & Marchant, B. P. (2013). Analysis of groundwater drought building on the standardised precipitation index approach. Hydrology and Earth System Sciences, 17(12), 4769-4787.
5. Chu, H. J. (2018). Drought detection of regional nonparametric standardized groundwater index. Water Resources Management, 32(9), 3119-3134.
6. Demirci, M., Üneş, F., & Körlü, S. (2019). Modeling of groundwater level using artificial intelligence techniques: A case study of Reyhanli region in Turkey.
7. Dibike, Y. B., & Solomatine, D. P. (2001). River flow forecasting using artificial neural networks. Physics and Chemistry of the Earth, Part B: Hydrology, Oceans and Atmosphere, 26(1), 1-7.
8. Everitt, B. S.: The Cambridge Dictionary of Statistics, 2nd Edn.,Cambridge University Press, Cambridge, 2002.
9. Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. MIT press.
10. Guenang, G. M., & Kamga, F. M. (2014). Computation of the standardized precipitation index (SPI) and its use to assess drought occurrences in Cameroon over recent decades. Journal of Applied Meteorology and Climatology, 53(10), 2310-2324.
11. Helm, D. C. (1975). One‐dimensional simulation of aquifer system compaction near Pixley, California: 1. Constant parameters. Water Resources Research, 11(3), 465-478.
12. Hinton, G. E., Osindero, S., & Teh, Y. W. (2006). A fast learning algorithm for deep belief nets. Neural computation, 18(7), 1527-1554.
13. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
14. Khorrami, M., Abrishami, S., Maghsoudi, Y., Alizadeh, B., & Perissin, D. (2020). Extreme subsidence in a populated city (Mashhad) detected by PSInSAR considering groundwater withdrawal and geotechnical properties. Scientific reports, 10(1), 1-16.
15. Li, H., Zhu, L., Gong, H., Sun, H., & Yu, J. (2020). Land subsidence modelling using a long short-term memory algorithm based on time-series datasets. Proceedings of the International Association of Hydrological Sciences, 382, 505-510.
16. McKee, T. B., Doesken, N. J., & Kleist, J. (1993, January). The relationship of drought frequency and duration to time scales. In Proceedings of the 8th Conference on Applied Climatology (Vol. 17, No. 22, pp. 179-183).
17. Osti, A. L., Lambert, M. F., and Metcalfe, A. V.: On spatiotemporal drought classification in New South Wales: Development and evaluation of alternative techniques, Aust. J. Water Resour., 12, 21–34, 2008.
18. Pathak, A. A., & Dodamani, B. M. (2019). Trend analysis of groundwater levels and assessment of regional groundwater drought: Ghataprabha River Basin, India. Natural Resources Research, 28(3), 631-643.
19. Python, fitter, https://fitter.readthedocs.io/en/latest/
20. Python, Keras, https://github.com/keras-team/keras
21. Python, sklearn.preprocessing-QuantileTransformer, https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.QuantileTransformer.html
22. Rahmati, O., Falah, F., Naghibi, S. A., Biggs, T., Soltani, M., Deo, R. C., ... & Bui, D. T. (2019). Land subsidence modelling using tree-based machine learning algorithms. Science of the Total Environment, 672, 239-252.
23. Staudemeyer, R. C., & Morris, E. R. (2019). Understanding LSTM--a tutorial into long short-term memory recurrent neural networks. arXiv preprint arXiv:1909.09586.
24. Uddameri, V., Singaraju, S., & Hernandez, E. A. (2019). Is standardized precipitation index (SPI) a useful indicator to forecast groundwater droughts?—Insights from a Karst aquifer. JAWRA Journal of the American Water Resources Association, 55(1), 70-88.
25. Varsamopoulos, S., Bertels, K., & Almudever, C. G. (2018). Designing neural network based decoders for surface codes. arXiv preprint arXiv:1811.12456.
26. Vicente-Serrano, S. M., Beguería, S., & López-Moreno, J. I. (2010). A multiscalar drought index sensitive to global warming: the standardized precipitation evapotranspiration index. Journal of climate, 23(7), 1696-1718.
27. Vidal, J.-P., Martin, E., Kitova, N., Najac, J., and Soubeyroux, J.M.: Evolution of spatio-temporal drought characteristics: validation, projections and effect of adaptation scenarios, Hydrol. Earth Syst. Sci., 16, 2935–2955, doi:10.5194/hess-16-2935-2012, 2012.
28. Water, U. N. (2020). 2020 UN World Water Development Report, Nature-based Solutions for Water.
29. Wu, H., Hayes, M. J., Wilhite, D. A., & Svoboda, M. D. (2005). The effect of the length of record on the standardized precipitation index calculation. International Journal of Climatology: A Journal of the Royal Meteorological Society, 25(4), 505-520.
30. Zhao, K., & Chen, S. N. (2011). Study on artificial neural network method for ground subsidence prediction of metal mine. Procedia Earth and Planetary Science, 2, 177-182.
31. 方熙廷. (2018). 平原與沿海地區地下水預測之類神經時空分析與理論解析.
32. 王詩婷, 陳宇文, 黃韋儒, 張良正, 江崇榮, 王詠絢, ... & 陸挽中. (2019). 濁水溪沖積扇之地下水文特徵研究-地下水位退水線探勘分析. 農業工程學報, 65(2), 12-22.
33. 左正民, & 張良正. (1996). 遺傳演算法於地層下陷模式參數檢定之應用 (Doctoral dissertation).
34. 江衍銘， 張麗秋， 張斐章 (2002) 回饋式類神經網路於二階段即時流量預測. 台灣水利, 50, 2: 15-21.
35. 吳國儒, 高慧珊, 鍾昌翰, 何宜樺, & 張斐章. (2008). 以類神經網路推估蒸發量. 農業工程學報, 54, (3), 1-13.
36. 呂珍謀, 朱木壽, & 陳厚元. (2010). 小波及灰色理論應用於地層下陷預警之研究-以濁水溪沖積扇為例. 農業工程學報, 56(4), 70-82.
37. 周秉誼. (2016). 淺談Deep Learning原理及應用.https://www.cc.ntu.edu.tw/chinese/epaper/0038/20160920_3805.html
38. 林崇瑋，2010，應用類神經網路推估雲林、嘉義地區之地層下陷。國立成功大學資源工程研究所碩士論文。
39. 邱正鈞, 萬昱廷, & 王士榮. (2020). 以倒傳遞類神經網路進行地下水位與抽水量推估. Journal of Taiwan Agricultural Engineering, 66(2).
40. 洪致文, & 施明甫. (2017). 台灣氣象乾旱指數的建立與嚴重乾旱事件分析. 大氣科學, 45(2), 145-165.
41. 翁叔平,陽承道. (2013).應用標準化降水蒸發散指數分析臺灣百年來 乾溼變化的低頻特徵與遙地相關
42. 國家災害防救科技中心，(2016)，臺灣氣候變遷災害衝擊風險評估報告，國家災害防救科技中心，新北市。
43. 康家桂，2006，以時間序列預測麥寮地區地層下陷之研究。國立交通大學土木研究所碩士論文。
44. 楊亞欣，2013，濁水溪沖積扇地下水與地層下陷聯合運用模擬與分析。國立成功大學資源工程研究所碩士論文。
45. 楊苡絜. (2018).以Sentinel-1A SAR及SBAS方法監測雲林地層下陷.
46. 經濟部，(2016)，氣候變遷進行式，氣候變遷水環境知識庫。
47. 經濟部水利署，2019，｢中華民國108年臺灣地區水文年報｣，經濟部水利署。
48. 經濟部水利署，2020，｢地下水保育管理暨地層下陷防治計畫｣，經濟部水利署。
49. 經濟部水利署，2020，｢雲彰地區地層下陷具體解決方案暨行動計畫｣，經濟部水利署。
50. 經濟部水利署，(2010)，北部旱災潛勢資料繪製示範計畫。經濟部水利署，執行單位：國立台灣大學。
51. 經濟部水利署，(2011)，中南部旱災潛勢資料繪製示範計畫。經濟部水利署，執行單位：國立台灣大學。
52. 經濟部水利署全球資訊網. (2019).https://www.wra.gov.tw/Default.aspx
53. 經濟部水利署地理資訊平台，https://gic.wra.gov.tw/gis/
54. 葉信富, 張家富, 李哲瑋, & 李振誥. (2016). 以標準化地下水與降雨指數法評估高屏溪流域之乾旱特性. Journal of Chinese Soil and Water Conservation, 47(1), 45-52.
55. 葉信富, 張家富, 李哲瑋, & 李振誥. (2016). 以標準化地下水與降雨指數法評估高屏溪流域之乾旱特性. Journal of Chinese Soil and Water Conservation, 47(1), 45-52.
56. 蔡., Tung-Lin T., 楊., 黃., Jinn-Chuang Y., & Liang-Hsiung H. (2000). 區域性地下水超抽導致地層下陷模式之發展與應用. http://hdl.handle.net/11536/66469.
57. 蘇苗彬、陳旺志(1999)，序率式時間序列分析應用於地下水位與地表沉陷量之模擬，第八屆大地工程學術研究討論會論文集，pp818-832。