WSP05706 - Laserfiche WebLink

Home Browse Search

, 172 WATERSHED MANAGEMENT ESTIMATING EROSION 173 TABLE 5.--MEASURED AND PREDICTED ANNUAL SEDIMENT YIELD (AC-FT/M12) FOR SELECT SEMIARID RANGELAMl WATERSHEDS . TABLE 6.--1lEGRESSION OF ACTUAL VERSUS PREDICTED SEDIMENT YIELDS: PREDICTED = a(actual)+b Tank Measured Predicted .yield number yield PSIAC Dendy/Bolton Fl axma n Renard 1 0.49 .0.29 201 B 0.83 -0.180 0.68 G 0.13 0.19 207 0.11 0.18 0.73 0.049 0.61 208 0.13 0.16 0.75 0.313 0.62 212 0.11 0.30 0.62 0.142 0.53 213 0.09 0.18 0.69 0.375 0.58 214 0.37 0.38 0.70 0.154 0.59 215 0.70 0.42 0.85 0.249 . 0.69 216 0.51 0.28 0.76 0.341 0..63 223 0.30 0.29 0.83 0.085 0.68 1 The B and G refer to brush and grass cover associated with the 1971 treatment of -th~ watershed. ----- METHOD N a b r2 PSIAC 10 .326 .172 0.63 Dendy-Bo lton 9 .226 .685 0.39 Fl axman 9 .077 .147 0.01 Renard 9 .158 .577 0.39 REPRESENTATIVENESS OF SHORT RECORDS The Flaxman method also agreed fairly well with the measured data because it was developed using data from a wide range of watershed con~ ditions in the western United States including three watersheds in Ari- zona. In contrast with two of the other methods, it does rot have a term reflecting drainage ar.ea. Historically, most sediment yield esti- mating equations include either runoff o~ watershed area_ or both. The Dendy-Bolton method overestimated sediment yield in all cases. PrediCtion might have been IOOre accurate if actual runoff data had been available to use in equation L Because the method involves only two parameters, it would oot be expected to explain as rrtJch of the observ~d variance as other methods. The Renard method also overestimated the sediment yield in all but one case. Predictions might improve if the techntque was used to simu- late the yield usi ng channel characteristics and observed runoff for each individual watershed rather than the average conditions with which the IOOdel was calibrated. For example some of the ponds had grass swails; in other locations, the channels 'were roore rectangular and con- tained large amounts of sand, which more nearly duplicated the comfi- tions of the large watersheds. Thus sediment accumulation in tanks with sand channe.ls (208, 214, 215, 216., .and 223) would be expected to be closer to the predicted, as observed on all but tank 208. A linear regression was computed between the measured and predict- ed sediment yields for each of the methods tested. The results are sum- marized in Table 6. It is 'not -surprising that the r2 values are as low as they are partly because of the relatively narrow range of sediment yield values for the data sets. At the same time, the very low value for the Flaxman method is surpriSing. This test indicated that the PSIAC method is statistically the best method to use. When relatively short records are used in developing and -testing predl ct i on schemes such as the sedi ment yi e 1 d method tested herei n, one immediately wonders whether the sample includes all extremes- of the cli- mate and if the-mean value indicates a long term mean. In the southwest- ern United States the coefficient of variation of annual precipitation is maximum for ~ny of the locations considered by Hershfield (6). Knisel et aT. (7) investigated methods to evaluate the length of record necessary for water resource data collection. One of the methods inves- tigated involved a cumulative surplus/deficit analysiS of the annual precipitation. The surplus/deficit analysis depicts trends ~hat may otherwise be obscure and is obtained by cumulating departures from a long term mean. Figl!re 2 illustrates the long-tenn a,~nual rainfall amounts. and cum- ulative surplus/deficit from the 13.66-1n. mean for the ralngage at Tontlstone within the Walnut Gulch Experimental Watershed. In only one year was ~ainfall above the long term mean for the period c~nsidered for most of the watersheds used in the evaluation. The negatlve slope..to the surplus/deficit graph for the periOd since 1957 illustrates the gen- eral dry trend during the study period, which since 1957. has been about 8% below normal. Thus. the vegetation cover would be poor and n.lfloff mi9ht be less than the long term mean. The importance of an unusual storm in affecting long. tenn trends has been well documented. Thus it is entirely poSSible that some of the observed yields are low because of low precipitation/runoff. Stock tanks 214, 215, and 216, on the other hand, have had some large storms during their short recordS, which may partly explain why the obs~rved yields for these ponds are larger and somewhat closer to the predlcted values. CONCLUSIONS 2912 1. Predicting sediment yields in the western United States is dif- ficult. Relatively large differences in the characteristics of the con- tributing watershed area over very short distance add to the prOblem. 2. Of the four methods investi9ated, the PSIAC method seems to provide the best results. Experienced people IIllst select values of the nine parameters, which can then prOduce results consistent with observa- tions. The PSIAC method is the only one tested which has factors which relate to management. Thus, it affords the opportunity to evaluate ,