Abstract
Analyses of Total Quarterly Rainfall (TQR) depth time series (1964-2014) at four gauges surrounding one potential rainwater harvesting site in this region show strong temporal and spatial variability. Using Box-Jenkins Methodology, three best-fit seasonal ARIMA models for 10-year predictions of TQR were selected; among more than 20 models, based on criteria of minimum values of variance, AIC and SBC. Comparisons between the forecasted and the actual TQR show good agreement. Data analyses show that spring records have the heaviest rainfall. Accuracy assessment results of the models indicate that for the most favorable model 58% of the residuals are within +/- 10 mm, and 74% are within +/- 20 mm. Making use of these results, it was possible to obtain seasonal predictions of runoff volumes for water harvesting at the potential site. The results indicate that the largest runoff volumes are expected to occur during the springs of 2016, 2019 and 2022 with an estimate of approximately 3.764 Mm(3). Such results are useful for the design and maintenance of the storage facilities of the harvested water.