1. WATER RESOURCES 



Estimating Future Water Supply and Usage 



Most estimates of water supply 

 and usage have been couched in terms 

 of average annual water supply and 

 projected usage at some future date. 

 For small areas within the scope of 

 a single project or a system of proj- 

 ects, water supply is sometimes stated 

 as the mean flow available during 

 the most critical dry period in the 

 record. Such assessments have the 

 virtue of simplicity and are reason- 

 ably well understood by the layman. 



At the national level, a statement 

 of mean water supply and mean 

 usage is probably entirely adequate 

 because water-supply problems are 

 never solved at that level. At the 

 regional and local level, however, 

 use of the mean supply available 

 and a projected future usage deprives 

 the planner of the opportunity for 

 strategic evaluation of alternatives. 

 The planner is concerned with sup- 

 plying water for a specific period of 

 years into the future. It is virtually 

 certain that the actual streamflows 

 during this future period will not 

 duplicate those of the historic past 

 and that water usage at the end of 

 the period will not precisely equal the 

 forecast. Faced with such uncertainty, 

 the planner would be wise to treat 

 both variables in terms of probability. 

 Only through a probabilistic treat- 

 ment can he evaluate the risk of 

 expanding water-supply facilities too 

 fast, with consequent excessive costs 

 and risk of losing future technologi- 

 cal advantages, or of developing a 

 system so slowly as to threaten a 

 serious water shortage at some future 

 date. 



Estimates of Water Supply 



The data base for estimates of 

 water supply consist of approxi- 

 mately 10,000 gauging stations oper- 

 ated mostly by the U.S. Geological 



Survey; in addition, many thousands 

 of wells provide information on 

 groundwater levels. There may be 

 specific local deficiencies in this data 

 base, but on the whole it must be 

 judged reasonably adequate. It is 

 fortunate that this base exists, be- 

 cause only time can remedy deficien- 

 cies — from 30 to 100 years of record 

 are required to describe statistically 

 the characteristics of water supply. 



Qualifying Factors — Interpretation 

 of existing data on streamflow and 

 groundwater is complicated by the 

 fact that few stations record virgin 

 conditions. Regulation by reservoirs, 

 diversion from streams, pumpage 

 from groundwater, alteration of 

 stream channels, vegetation-manage- 

 ment practices, urbanization, and 

 many other factors render available 

 data series inhomogeneous over time. 

 In some cases, the effect of man's 

 activity is rather accurately known 

 and appropriate corrections can be 

 made. In most instances, however, 

 only the sign of the change can be 

 stated with accuracy. 



Synthetic Streamflow Records — 

 The last decade has seen the devel- 

 opment of hydrologic simulation us- 

 ing both digital and analogue com- 

 puters. Simulation is capable of 

 transforming precipitation data into 

 synthetic streamflow records. Simu- 

 lation brings many thousands of 

 precipitation stations operated by the 

 National Weather Service into the 

 data base and makes it possible to 

 make streamflow estimates at sites 

 where no gauging station exists. Be- 

 cause precipitation records are gen- 

 erally longer than streamflow records, 

 simulation permits the extension of 

 flow records at currently gauged sites. 



Similar development has taken 

 place with respect to simulation of 



groundwater basins primarily through 

 the use of analogue models. Al- 

 though these models cannot perfectly 

 reproduce historic streamflow or 

 groundwater basin performance be- 

 cause of errors in the data inputs 

 and deficiencies in the models them- 

 selves, errors in model outputs are 

 generally random and pose no serious 

 problem in probabilistic estimates of 

 water supply. Simulation models also 

 permit adjustment of observed flows 

 or groundwater levels to virgin or 

 natural conditions. It may be con- 

 cluded, therefore, that we are now 

 able to combine observed and syn- 

 thesized data into a data base cover- 

 ing a sufficient period of time to de- 

 fine the mean and variance of water 

 supplies with reasonable accuracy. 



Problems of Data Projection — The 

 historic data base, observed or simu- 

 lated, does not fully satisfy the need 

 for projections of future water sup- 

 ply, however. The water-supply 

 planner is concerned with possible 

 events over a specific period ranging 

 from 20 to 100 years in the future. 

 He is particularly concerned with 

 the sequences of annual flows, be- 

 cause a series of consecutive dry 

 years will impose a much greater 

 burden on his reservoir (surface or 

 subsurface) than the same number 

 of dry years dispersed over his plan- 

 ning horizon. To meet this problem, 

 the field of stochastic hydrology has 

 developed during the 1960's. 



Stochastic Hydrology — In stochas- 

 tic hydrology, generating functions 

 derived from the estimated statistical 

 characteristics of the historic record 

 are used in conjunction with random 

 numbers to generate many possible 

 flow sequences. Thus, a thousand 

 years of stochastic streamflow can 

 be broken into ten 100-year periods, 

 from which the planner can estimate 



197 



