Laserfiche WebLink
<br /> . , ,. - <br />" <br />I <br />annual averagt;: tem~l1Itu:re did nol vary by more than ¡"Ffrom its average of 7(ì.3"F and <br />never had adifl'erence greater than J"P. AJthough greater differences ~experienced on <br />a month-to-month basis, they tend to average out over the year. This stability creates <br />stability inannwd water den¡¡Uld,. Ex1)ibit5 plots temperature against monthly water use <br />over 1975 to 2001. Waj:eruse and temperature arebigblyconelated. <br />Annual rainfall, in contrast, is much more variable; However, rain lends to fall in the <br />winter and eàrly sþringmonths when tetDpera,tures and inigationdemands are relatively <br />low. Hence, rainfiill va.riabiJitydoes n()uend to greatly impact annual water demlll1ds. <br />Rainfall has a much bigger impact on water supply (Hetch Hetcby) than demand. <br />Our model correlating water use with weather wasspecifit:>d. to measure the deviations in <br />water use (as shown in Exhibit 5) ftom deYíati()n$ in normal temperature and rainfall as <br />follows: <br /> " <br />WATERI '" Laix MONTHi,t + fJlxTEMPDEVI + ß2xRA.INDEVI <br /> ,., <br />where, <br />WATER: .. ratio of water USe in month tto a 12-month moving average of WATERI <br />MONTHi,t .. binary variable cre3ting one unique aìntercept for each calendar month <br />TEMPDEJlt .. temperature deviation in month trom the average format month fF) <br />RAJNDEJI! .. rainfall deviation in month 1Ìom the. average for that month (inches) <br />ai,fJl,fJ2 .. coefficients estimated using least squares regression <br />After es,timating the model, we used actual weather data from 1949 to 2001 to simulate <br />tbe maximum impacts on annual Wilier USe. Fi'lJm this ex.ercîse, we determine that water <br />use can be expected 10 vary +/- 2 percent from weather alone on annual basi$. <br /> Page 8 <br />