during the third period but a differing daily amount distribution could cause highly differing delta inundations 

 in one period versus the other period. One approach to study this possibility is the cross-tabulation of daily 

 precipitation. The Corpus Christi WSO AP daily precipitation was selected for analysis because of better 

 quality, despite the shorter record period that initiated in 1948. 



Given in table 3 are the frequencies from cross-tabulation of the Corpus Christi daily precipitation in 

 approximately one-inch size categories. Inspection of the frequency patterns over time su^ests some 

 differences in shorter time periods of about 10 years (particularly the first 10 years). The number of years per 

 each of the 3 Rincon Project periods in succession was 18, 24 and 18 years. Available Corpus Christi WSO AP 

 data limited the initial study period to 10 years (1948-99). Cross-tabulations present outcomes such that pattern 

 comparisons for selected periods can often be made using muUtvariate analysis. A distribution-free 

 permutation test was selected and applied in exploratory analysis to explore whether the three distribution 

 patterns differed. The statistical test applied is one of a family of tests known as multi-response permutation 

 procedures (MRPP) (Mielke et al., 1976, 1982). These tests avoid making the unjustified assumption that the 

 joint distribution of the response measurements is multivariate normal or some other specific distribution. 



Application of the MRPP yielded a P-value of 0.08 in comparing frequency patterns of all three periods. In the 

 exploratory sense, the first 10 years of frequencies were separately compared with those of each of the 24 and 

 18 year periods using MRPP, and ia both applications found to differ at better than the 5 percent level. 

 Application of MRPP in a comparison of frequencies of the 24 versus 1 8 year periods was not significant at the 

 5 percent level. In this exploratory analysis, no attempt was made to correct for experimentwise error rate 

 under a partial null hypothesis in which some comparisons are equal but others differ (Miller, 1981). The 

 results of these comparisons are not surprising given the extreme level of drought in the 1950s. Additionally, 

 shorter comparison periods, such as the 10-year period in this data, likely increase the probability of finding 

 significant differences, given the dependence in rime in precipitation data. Multivariate analyses of 

 crosstabvJated daily precipitation from the Corpus Christi WSO AP gauge express differences in the 

 precipitation distribution of the 1950s drought years versus distributions from the other study periods of the 

 60-year record under consideration. The second study period precipitation distribution did not differ from that 

 of the third and most recent period. 



Inspection of event frequencies of large precipitation amounts (=> 4 inches per day) revealed 25 events in the 

 Corpus Christi WSO AP data. There were 14 years of the record with tropical cyclone occurrences (8-La Niiia, 

 3-El Nino, 3-baseline). Ten large events (40 percent) occurred in La Niiia years of which seven events (70 

 percent) were during five tropical cyclone years. Seven large events (28 percent) occurred in El Niiio years but 

 none in tropical cyclone years. Eight events (32 percent) occurred in baseline years of which three events (38 

 percent) were in tropical cyclone years. Seven of 21 years with large events involved tropical cyclones. The 

 three largest category events were La Nina events in years without tropical cyclones. Two of the three largest 

 events occurred during the drought years of the 1950s. These results suggest somewhat over one-third of large 

 precipitation events occur during the La Niiia phase and often in years (La Niiia) with tropical cyclones. A 

 modest number of large events seem to occur during the El Nifio phase but seldom involving years with 

 tropical cyclones. Large events occur in baseline years, some of which may be tropical cyclone years. These 

 results are generally consistent with those of the computed precipitation anomalies. An additional multivariate 

 comparison of interest that was not conducted is the comparison of distributions between ENSO types and 

 baseline years. 



The crosstabulated data revealed 21 (40 percent) of 52 years included large precipitation events. Of the 21 

 years, 6 were El Niiio, 8 were La Niiia, and 7 were baseline years. Tropical cyclones occurred in seven high- 

 event years: five were La Niiia years and two were baseline years. Annual precipitation exceeded 40 inches in 7 

 years of the record. Five years were baseline years, one year was La Niiia, and one year was El Nino and the 

 greatest amoimt (48.07 inches) of the record. Two years of the high seven years did not have a high daily- 

 amount event. These results express a role by ENSO but also other mechanisms in causing Nueces River 

 watershed precipitation. 



yippendixD ♦ D-9 



