Premium
Model‐free statistical methods for water table prediction
Author(s) -
Yakowitz Sidney
Publication year - 1976
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/wr012i005p00836
Subject(s) - homogeneity (statistics) , series (stratigraphy) , cluster analysis , table (database) , regression , homogeneous , stochastic process , geology , statistics , computer science , data mining , mathematics , paleontology , combinatorics
In this study a new approach for predicting future values of well depths on the basis of regional water table records is presented. Basically, well water level depths are viewed as random sequences, and the assumption is made that the region to be analyzed can be partitioned into several subregions of unknown geographic shapes which are statistically homogeneous in the sense that the record of each well in a fixed subregion is a different realization of the same stochastic process. Methods from clustering and time series analysis are used to find first the subregions of stochastic homogeneity and then the statistical law for the time series of the wells in a given subregion. Forecasts are made and confidence bands constructed by using the methods espoused here (in conjunction with regression techniques) on Tucson basin data. The forecasts are compared to depths actually observed, and for many wells the agreement is sufficient to make these methods appear promising.