供水系統的「乾旱」現象可視為系統供需間失衡的累積，在水文過程不確定下，其初期現象不易被發覺。也因此，適切的限水時機往往在乾旱被發覺前流失。台灣地區在經歷2002年及2003年的連續嚴重乾旱之後，經濟部水利署頒佈了「區域水資源調度機制」，明訂台灣各區域水資源局應依各水資源調度區之氣象、水文、水源及供需特性等，參考建議之主次指標因子，建立水資源預警指標(Water resources Early Warning Index, WEWI)，期以對未來的乾旱潛勢預作客觀評估，進而及早啟動相關因應措施，降低可能的衝擊。
水利署初步建議可對各指標因子綜合評分加權求出指標，現況建置之因子權重係由歷史乾旱紀錄以人為試誤調整而得，如是之預警方法除評分過程易流於主觀外，評估結果亦難以合理反映乾旱潛勢，故不利作為實務管理之客觀依據。鑑此，本研究在考量供水系統的特性及反映政府與水資源管理者對乾旱的認知下，以「平均缺水率」的概念建置水資源乾旱預警指標(water resources Drought Early Warning Index, DEWI)，並以供水系統的特性合併相關反應應變機制，建立合宜的乾旱分級標準，續就DEWI定義的預警期程建置其推論系統。
供水系統乾旱發生與否主決於兩項因素：(1)現況之水文及蓄水條件，及(2)未來預警期程內之氣候條件，因此現況水情與未來氣候推論為乾旱預警系統組成的最主要因素。在水文過程及未來氣候之不確定性下，為避免乾旱監測資訊對DEWI推論的誤差，與未來氣候的推論誤差累加於同一推論模式中，並考量實務上對未來氣候的推論多屬機率性預報，因此，本研究建置的乾旱預警系統同時包含定率及序率乾旱預警模式。
定率乾旱預警模式主要基於評估時刻的水情資訊(指標因子)對未來的缺水態勢根據歷史紀錄進行期望值推論。由於本文具體定義供水系統之乾旱特徵，故模式檢定結果可提供潛在的預測誤判機率。序率乾旱預警模式主要以水文歷程的序率性對預警期程內可能發生的乾旱等級作機率性預報，並結合中央氣象局未來三個月的天氣機率預報，賦予模式推論結果專業的未來氣候條件判斷成果。此種推論方式整合氣象、水文、水量供應、脆弱度等資訊進行模式的建置，符合一般對「乾旱預警系統」的預期與需求。實務應用上，本研究建置之水資源乾旱預警系統，可提供管理者定率預警模式的推論結果、乾旱情勢燈號跳脫機率，及序率預警模式推論的各等級乾旱發生風險，作為乾旱潛勢評估的依據。
本研究以台灣南部地區之曾文-烏山頭水庫系統為研究實例，以水庫的起始蓄水量及進水流量數列，進行至水文年結束(翌年五月下旬)的蓄水利用模擬，並以模擬期距內供水過程下的預期缺水率為DEWI值，參考行政院「旱災災害防救業務計畫」建議之旱災等級區分及該系統供水特性，設定乾旱指標之等級區間。
在模式建置部分，本研究選定部分水利署建議之指標因子並做適當修訂後，分別採用線性迴歸、倒傳遞類神經網路及適應性模糊推論系統，建置定率乾旱預警模式，結果顯示DEWI可較WEWI提供更適切的推論結果。序率乾旱預警模式以修正後的季天氣預報資料，對各水資源乾旱等級的發生機率進行推論，其整體推論趨勢與實際水資源乾旱情勢近似。整體而言，本研究建置的乾旱預警系統對預警期程內的乾旱潛勢提供客觀、完整的評估結果，可供決策者參考做為制訂抗旱因應策略的基準。

Signs of impending drought are often vague and the results generated from hydrologic conditions are uncertain. With regard to reservoir water supply systems, if it is known that a severe water shortage will occur in the future, then the water resources managers can adopt early water-limiting polices to reduce the effects of such shortages before the next wet season arrives. As an early drought prevention measure, the Water Resources Agency (WRA) of Taiwan decreed the 'Regional Water Resources Regulations', demanding that regional water resource offices establish an appropriate water resources early warning index (WEWI) according to the principles of meteorology, hydrology, water sources, and supply-demand traits for each water supply system. It defined seven factors based on the four principles mentioned above, and suggested a grading method using given weights (judged by water resource managers or a brains trust of the WEWI) to evaluate the likelihood of a drought occurring. However, this evaluation is subjective, and human error may affect the results when using trial and error to adjust the weights of the major and minor factors in the WEWI. In addition, the effectiveness of this system’s warnings has been doubted, because its early warning ability and warning risk cannot be explored. For decision makers, a water resources drought early warning index which dose no include the information evaluation period, drought characteristics, and potential risk of index values, it is not easy to use when formulating strategies for drought prevention. In view of these issues, this study proposes a water resources drought early warning index (DEWI) to avoid these disadvantages for decision makers. Since the droughts that concern engineers in Taiwan are related to the deficit of water resources, this study uses the expected-deficit-rate - average shortage rate (ASR) - to demonstrate the water shortage tendencies of a reservoir water supply system during the evaluation period (i.e. the drought early warning period, from the evaluation time to the end of May) as the DEWI. The drought levels of DEWI are defined according to the drought level definitions of the ‘Drought Saving Program’ and water restriction standards of the reservoir operation rules: blue, green, yellow, orange, and red, as non-drought, normal drought, third-level drought, second-level drought and first-level drought, respectively. The definition of DEWI and its definitions of drought level are thus used to establish the water resources drought early warning system.
For the establishment of a water resources drought early warning system, because whether a drought occurs or not in a water supply system hinges on the immediate hydrologic situation and the climatic conditions during the period of early warning, the monitoring of the immediate hydrologic situation and the prediction of future climatic conditions are both very important. However, to date, most of the methods/models for drought estimation only or primarily depend on monitoring information related to the immediate hydrologic situation, and then utilize a deterministic model. On the other hand, the forecast results with regard to the future climate are usually given in probabilistic terms. To avoid errors of DEWI estimation and climate prediction are accumulated in an estimation model, the DEWI estimation in the proposed water resources drought early warning system is separated into deterministic and stochastic models according to the attributes of the information being evaluated. The deterministic DEWI estimation model is based on the immediate hydrologic information (index factors) to estimate the expect value of ASR (DEWI) according to the modeling data. Since the definition of DEWI is dependent on the characteristics of a water supply system, and not the experience of water resource managers, the results of model calibration can be provided to the decision makers as an indication of the potential risk of mistaken results. The stochastic DEWI estimation model is based on the stochastic characteristics of reservoir inflow and integrates the results of weather forecasting to estimate the probability of occurrence for classified levels of drought for DEWI from the start to the end of the evaluation period. The methodology proposed in this study fulfills the need for the development of a drought early warning system for water resources managers that integrates the related meteorological, hydrological, water supply, and vulnerability characteristics of a water supply system. The definition and the classification of DEWI are in agreement with the concepts of water resources management held by the decision makers, are produced by the operating characteristics of a water supply system, and can be directly connected with the related response mechanisms. For practical applications, the methodology proposed in this study can provide decision makers with the following information to formulate a suitable prevention policy for a potential drought: (i) a crisp value of ASR derived from the deterministic DEWI estimation model, (ii) the calibration information of each deterministic DEWI estimation model, and (iii) the probabilistic forecast results for the classified drought levels through the stochastic DEWI estimation model.
The practicality of the proposed water resources drought early warning system is demonstrated through a case study in the Tsengwen-Wushantou reservoir water supply system. The results indicate that the deterministic DEWI estimation model can provide better estimation results with regard to potential droughts than the WEWI estimation model. In addition, the stochastic DEWI estimation model estimates the probability of occurrence for each water resources drought level with the modified seasonal weather forecast result, and the estimation results are in good agreement with the real water resources drought situation. The estimation results from the deterministic and stochastic DEWI estimation models complement each other, and can provide more useful and objective evaluations for decision makers who need to devise strategies to face the challenges of an impending drought.

Akaike, H., “Information theory and an extension of the maximum likelihood principle”, Proceedings of 2nd International Symposium on Information Theory, Akademiai Kiado, Budapest (1973).
Bacanli, U., Firat, M. and Dikbas, F., “Adaptive neuro-fuzzy inference system for drought forecasting”, Stochastic Environmental Research and Risk Assessment, Vol. 23, No. 8, pp. 1143-1154 (2009).
Bowden, G.J., Maier, H.R. and Dandy, G.C., “Optimal division of data for neural network models in water resources applications”, Water Resour. Res., Vol 38, No. 2, pp. 2-1-2-11 (2002).
Cancelliere, A. and Salas, J.D., “Drought length properties for periodic-stochastic hydrologic data”, Water Resour. Res., Vol. 40, W02503 (2004).
Chang, Y.T., Chang, L.C. and Chang, F.J., “Intelligent control for modeling of real-time reservoir operation, part II: artificial neural network with operating rule curves”, Hydrological Processes, Vol. 19, No. 7, pp. 1431-1444 (2005).
Chidthong, Y., Tanaka, H. and Supharatid, S., “Developing a hybrid multi-model for peak flood forecasting”, Hydrological Processes, Vol. 23, No. 12, pp. 1725-1738 (2009).
Cigizoglu, H.K., “Application of generalized regression neural networks to intermittent flow forecasting and estimation”, Journal of Hydrologic Engineering, Vol. 10, No. 4, pp. 336-341 (2005).
Cuddy, S., Letcher, R., Chiew, F.H.S., Nancarrow, B.E. and Jakeman, T., “A role for streamflow forecasting in managing risk associated with drought and other water crises”, Drought and water crises: science, technology and management issues, Taylor & Francis, Boca Raton, pp. 345-365 (2005).
CWB, Seasonal weather forecast report, Central Weather Bureau, Ministry of Transportation and Communications, Taiwan (2008).
Fowler, H.J. and Kilsby, C.G., “A weather-type approach to analysing water resource drought in the Yorkshire region from 1881 to 1998”, Journal of Hydrology, Vol. 262, No. 1-4, pp. 177-192 (2002).
Frick, D.M., Bode, D. and Salas, J.D., “Effect of drought on urban water supplies. II: Water-supply analysis”, Journal of Hydraulic Engineering, Vol. 116, No. 6, pp. 754-764 (1990).
Hald, A., A history of mathematical statistics from 1750 to 1930, Wiley series in probability and statistics, Wiley, New York (1998).
Hayes, M.J., Drought indices, What is drought?, National Drought Mitigation Center (2006).
Haykin, S.S., Neural networks: a comprehensive foundation, Prentice Hall, Hamilton, Ontario, Canada (1999).
Heim, R.R., “A review of twentieth-century drought indices used in the United States”, Bulletin of the American Meteorological Society, Vol. 83, No. 8, pp. 1149-1165 (2002).
Huang, W.C. and Yuan, L.C., “A drought early warning system on real-time multireservoir operations”, Water Resour. Res., Vol. 40, W06401 (2004).
Huang, W.C. and Chou, C.C., “Drought early warning system in reservoir operation: theory and practice”, Water Resour. Res., Vol. 41, W11406 (2005).
Iglesias, A., Garrote, L. and Cancelliere, A., “Guidelines to develop drought management plans”, Coping with drought risk in agriculture and water supply systems, Springer Netherlands, Dordrecht, pp. 55-65 (2009).
Jain, S.K. and Das, A., “Application of ANN for reservoir inflow prediction and operation”, Journal of Water Resources Planning & Management, Vol. 125, No. 5, pp. 263-271 (1999).
Jang, J.S.R., “ANFIS: adaptive-network-based fuzzy inference system. Systems”, Systems, Man and Cybernetics, Vol 23, No. 3, pp. 665-685 (1993).
Jenkins, M.W. and Lund, J.R., “Integrating yield and shortage management under multiple uncertainties”, Journal of Water Resources Planning & Management, Vol. 126, No. 5, pp. 288-297 (2000).
Kallis, G., “Droughts”, Annual review of environment and resources, Vol. 33, No. 1, pp. 85-118 (2008).
Karamouz, M. and Vasiliadis, H.V., “Bayesian stochastic optimization of reservoir operation using uncertain forecasts”, Water Resour. Res., Vol. 28, No. (5), pp. 1221-1232 (1992).
Keyantash, J. and Dracup, J.A., “The quantification of drought: an evaluation of drought indices”, Bulletin of the American meteorological society, Vol. 83, No. 8, pp. 1167-1180 (2002).
Kim, T.W. and Valdes, J.B., “Nonlinear model for drought forecasting based on a conjunction of wavelet transforms and neural networks”, Journal of Hydrologic Engineering, Vol. 8, No. 6, pp. 319-328 (2003).
Maier, H.R. and Dandy, G.C., “Neural networks for the prediction and forecasting of water resources variables: a review of modelling issues and applications”, Environmental Modelling and Software, Vol. 15, No. 1, pp. 101-124 (2000).
Murphy, A.H., “ What is a good forecast? An essay on the nature of goodness in weather forecasting”, Weather and Forecasting, Vol .8, No. 2, pp. 281-293 (1993).
Nicholls, N., Coughlan, M.J. and Monnik, K., “The challenge of climate prediction in mitigating drought impacts”, Drought and water crises: science, technology, and management issues, Taylor & Francis, Boca Raton (2005).
O'Brien, K.L. and Leichenko, R.M., “Double exposure: assessing the impacts of climate change within the context of economic globalization”, Global Environmental Change, Vol. 10, No. 3, pp. 221-232 (2000).
Saidy, D., “Early warning and response”, International Journal for African Studies 2, http://www.brad.ac.uk/research/ijas/ijasno2/saidy.html (1997).
Salas, J.D., Applied modeling of hydrologic time series, Water Resources Publications, Littleton, Colo. (1980).
Shibata, R., “Selection of the order of an autoregressive model by Akaike's information criterion”, Biometrika, Vol 63, No. 1, pp. 117-126 (1976).
Shin, H.-S. and Salas, J.D., “Regional drought analysis based on neural networks”, Journal of Hydrologic Engineering, Vol. 5, No. 2, pp. 145-155 (2000).
Simon, H., Neural networks: a comprehensive foundation, Prentice Hall International, New Jersey (1999).
Steinemann, A.C., Hayes, M.J. and Cavalcanti, L.F.N., “Drought indicators and triggers”, Drought and water crises: science, technology, and management issues, Taylor & Francis, Boca Raton (2005).
Stigler, S.M., The history of statistics: the measurement of uncertainty before 1900, Belknap Press of Harvard University Press, Cambridge, Mass (1986).
Stigler, S.M., Statistics on the table: the history of statistical concepts and methods, Harvard University Press, Cambridge, Mass (1999).
Sudheer, K.P. and Jain, A., “Explaining the internal behaviour of artificial neural network river flow models”, Hydrological Processes, Vol. 18, No. 4, pp. 833-844 (2004).
Tannehill, I.R., Drought: its causes and effects, Princeton University Press, Princeton, New Jersey (1947).
Tayfur, G., Moramarco, T. and Singh, V.P., “Predicting and forecasting flow discharge at sites receiving significant lateral inflow”, Hydrological Processes, Vol. 21, No. 14, pp. 1848-1859 (2007).
Trenberth, K.E. and Caron, J.M., “Estimates of meridional atmosphere and ocean heat transports”, Journal of Climate, Vol. 14, No. 16, pp. 3433-3443 (2001).
USISDR, Drought risk reduction framework and practices: contributing to the implementation of the hyogo framework for action, International Strategy for Disaster Reduction (2007).
Vlachos, E. and James, L.D., “Drought impacys”, Coping with droughts, Water Resources Publications, Littleton, Colo. (1983).
Vogel, R.M., “Reliability indices for water supply systems”, Journal of Water Resources Planning and Management, Vol. 113, No. 4, pp. 563-579 (1987).
Wei, W.W.S., Time series analysis: univariate and multivariate methods, Pearson Addison Wesley, Boston (2006).
Wilhite, D., Svoboda, M. and Hayes, M., “Understanding the complex impacts of drought: a key to enhancing drought mitigation and preparedness”, Water Resources Management, Vol. 21, No. 5, pp. 763-774 (2007).
Wilhite, D.A., “Planning for drought: a process for state government”, IDIC Technical Report Series 90-1, NE: International Drought Information Center, University of Nebraska, Lincoln (1990).
Wilhite, D.A., Drought and water crises: science, technology, and management issues. Taylor & Francis, Boca Raton (2005).
Wilhite, D.A. and Glantz, M.H., “Understanding the drought phenomenon: the role of definitions”, Water International, Vol. 10, No. 3, pp. 111-120 (1985).
中興工程顧問股份有限公司，「臺灣地區水資源水情決策支援系統建置計畫總報告」，經濟部水利署委辦計畫，臺中 (2007)。
吳柏林，「時間數列分析導論」，華泰出版社，臺北 (1995)。
李彥璋、周乃昉，「貝氏序率動態規劃在水庫即時操作之應用」，國立成功大學水利及海洋工程學系碩士論文，臺南 (1993)。
周乃昉、林政浩，2008，「水庫運用規線之設計方法」，中華民國專利證書，發明第I300813號。
林俊全，「台灣的天然災害」，臺灣地理百科，遠足文化，新店 (2004)。
社團法人台灣農業工程學會，「旱災潛勢定義及其分析方法之建立」，經濟部水利署委辦計畫，臺北 (2009)。
徐享崑、尚新民，「河川流量繁衍模式之比較研究」，台灣水利，第33卷，第4期)，第50~81頁 (1985)。
張斐章，「乾旱時期河川流量之預測」，台灣水利，第37卷，第4期，第50~63頁 (1989)。
莊光明、甘俊二，「灌溉節水技術手冊」，農委會水利特刊，行政院農業委員會，台北市七星農田水利研究發展基金會，臺北 (1995)。
許晃雄，「建立乾旱監測與預警整合系統之構想」，全球變遷通訊雜誌，第37期，第22~27頁 (2003)。
童慶斌、劉子明、李明旭，「應用季節性氣候預報於水庫蓄水量預測」，農業工程學報，第52卷，第2期，第51~66頁 (2006)。
溫漢章、郭振泰，「利用合成流量進行水資源系統分析-以濁水溪流域為例」，國立臺灣大學土木工程學系碩士論文，臺北 (2001)。
萬象、廖元熙、張文綺，「應用乾旱指標於乾旱管理之研究」，第十一屆水利工程研討會，台北，第O67~O72頁 (2000)。
經濟部水利署，「旱災災害防救業務計畫」，臺北 (2003a)。
經濟部水利署，「區域水資源調度機制」，2003.10.08行政院核定版，臺北 (2003b)。
經濟部水利署，「區域水資源調度機制-修正案(稿)」，臺北 (2005)。
經濟部水利署，「曾文水庫運用要點」，臺北 (2009)。
監察院，「第○九一二二○○二八五號糾正案文」，臺北 (2002)。
顏君凌、周乃昉，「翡翠水庫不同乾旱限水標準之評比」，國立成功大學水利及海洋工程學系碩士論文，臺南 (2003)。
羅樹孝，「水文學辭典」，茂昌，臺北市 (1995)。