Laserfiche WebLink
<br />., <br />003~~7 <br /> <br />. <br /> <br />5 <br /> <br />i <br />\, <br /> <br />the objective, samples should be taken from crees growing on flood plains. ~e- <br /> <br />riod of OV~TI'!P between cree-ring series and instrumented weat!terJ~cord~ i~ ~s~en: <br />tial for climatic reconstructions. A portion of the overlapping record is used to de- <br />velop mathematical relationships between the measured parameters and ring-width <br /> <br />indices, a procedure known as calihration. The remaining overlap is then used to <br /> <br />verify the established relationships; this is termed verification. <br /> <br />Multiple stepwise regression was used in earlier reconstructions to select a <br /> <br /> <br />small set of independent variables from a much larger group. This procedure was <br /> <br />flawed because of high intercorrelations between the independent variables. This <br /> <br />disadvantage has been overcome by using principal component analysis to develop <br /> <br />new sets of uncorrelated variables called empirical orthogonal functions, EOF's <br />(Lorenz, 1956; Kutzbach, 1967). EOFs are developed from a maoix of correlation <br /> <br />~ <br /> <br />coefficients among independent variables and the first few usually represent most of <br /> <br />the variability in the dendroclimatic data. Higher order EOFs are usually discarded <br /> <br />as they may well be mathematical artifactS of the technique and unrelated (0 the data. <br /> <br />The development of transfer functions in the calibration procedure is critical to the <br /> <br />extraction of climatic signals from cree-ring data Transfer functions come in many <br /> <br />forms with some utilizing single series and others sets of data. Since they are <br /> <br />empirically developed it is necessary to proof-test (verify) numerous transfer <br /> <br />functions. A portion of the data used for calibration is held back for proof-testing so <br /> <br />the transfer functions can be evaluated on data not used in their development. After <br /> <br />selecting an appropriate function, the final transfer function is developed from the <br /> <br />entire data set (Fritts, 1976) <br /> <br />l <br /> <br />Per;istence in tree-ring series, which may represent a biological memory of <br />climate coupled with a direct record of climatic per;istence, creates problems in cli- <br />matic reconsrructions. Earlier workers attempted to solve this problem by using <br />moving averages with arbicrarily chosen coefficients or weights. Although results <br />