Data mining for long term prediction of hydrologic variables based on pattern fitting for advisory services

Abstract

Water resource systems are in need of advisories for irrigation newlineplanning and water system operation. Increase in population, changes in newlineclimate pattern and changes in land use land cover dynamics has resulted in newlinedeficit rainfall in certain areas. Hence, sustaining life in rainfall deficit areas newlinerequire better management of the water resource system.The success and the reliability of the system will depend on the continuous simulation spanning multiple years, the archives of previous newlineprediction is needed to forecast critical events as such events are of low newlinefrequency. Expert knowledge and opinion need to be used for the continuous newlineimprovement of the advisory. This kind of forecast system is needed where newlinethe dynamics of the water system is getting affected with land use land cover newlinechanges taking place and places affected by the chaotic nature of the global newlinewind circulation for predicting weather.This dissertation is focused on exploring possibilities to better newlinemanage the water distribution systems at micro level by developing simple newlinetools that can suggest advance prediction of the water stress, using newlinecomputational simulation models making use of the historical cycle and newlinetrends in the hydrologic variable data series.The study attempt to predict hydrological variables newlineevapotranspiration, precipitation and groundwater level, as these variables are newlineindicators of water stress in a region. The prediction is attempted for different newlinelong time horizons that can help in continuous monitoring and can help in newlineoperational management of the water resources system, the study is conducted newlineusing data collected for Vellore, Tamil Nadu an arid area with deficit rainfall newline newline