WSP00066

• File Number: 8143.400
• Description: John Martin Reservoir
• State: CO
• Basin: Arkansas
• Water Division: 2
• Date: 8/16/1966
• Title: Suggestions for mathematical analysis of historical records
• Water Supply Pro - Doc Type: Report/Study

,

ERNAL USE ONLY

~ I' \It 1\0 \?

Arkansas Valley in Colorado

<::>
o
N
~
I-"
-.l

Suggestions for Q~t~"':-l~tical analysis
oi histoi~~Jl records

Introduction

Two appro,::,::,c:12S seem appropriate in the c..n&lysis or historical'records.

(1) River llgainll and (2)"system 10ss.1T The accompanying d:.agram~ic sketch

may illustrate my idea. The sketch is very greatly over-s~:;l/~ified, needless

to say, but may be adequate to demonstrate the difference b2cween the two

approaches.. Both approaches \ViII be handica.pped by the complete lack of ,some

data, and in<lCCU1~ac:~,..s and incomplct;12SS oE ",".v.s.ilable data~. For example, records

of histori.::al chat:2;2s in gr"und-wate:r storage are virtually nonexistent; pre-

cipit~tion reco~ds a~e not detailed enough to be more than a crude measure of

amounts and s?atisl distribution] a very few of items B, C, and G have been

contin~ously ffie.:lSUrcJ since only about 1940; item F can only be estimated;

there ~re no h~~:orical records of underflow. However, preliminary analyses

have encou:r.::ged'me to the extent thai.:. I would recommend fu:cc:her investigations~

r::..:., inv<2st:.g.:.rte river "gain", the dependent variable shoul0 ~e H+D+E-A-C-G,

durir: ~Jeri.o~':'s ;,:hat they were all measured. The 0ependent var iable for system

Hloss;/ should 0<=. A+B-F-H. The independent variables in both investigations

shoulc De: suriace water applied to lands within the systero; water pumped

from c~a valley fill; and precipitation. Logically, some measure of solar

radia~~on also should be considered, but similar investigations I have made in

the p~_oc r0~ch have indicated that; first, there is undue correlation between,

for e~~~ple, ?an evaporation at John Martin, and precipitation; and second,

that the ra~lge - from year to y-.;;ar - in total radiation is too small to warrant

its i~clusi0n as a variable.

,

',:",

T:'12 pitfalls - and pratfalls - in using multiple regression analysis are

too Hc..:~_ kr.own for me to elaborate, except to say that we will not obtain

absolu::..:: proof of anything. I am almost certain, though, that c;:rc::ul considera-

tion of ~he results of mathematical analysis of historical' datn w~:: give us a

much better understanding of the system - and of data deficiencies v

',',

1

,'. -~,.<

.-.

'"''