Regularized joint inverse estimation of extreme rainfall amounts in ungauged coastal basins of El Salvador |
| |
| Authors: | Michael J Friedel |
| |
| Institution: | (1) Crustal Imaging and Characterization Team, United States Geological Survey, Denver Federal Center, MS 964, Box 25046, Lakewood, CO 80225, USA |
| |
| Abstract: | A regularized joint inverse procedure is presented and used to estimate the magnitude of extreme rainfall events in ungauged
coastal river basins of El Salvador: Paz, Jiboa, Grande de San Miguel, and Goascoran. Since streamflow measurements reflect
temporal and spatial rainfall information, peak-flow discharge is hypothesized to represent a similarity measure suitable
for regionalization. To test this hypothesis, peak-flow discharge values determined from streamflow recurrence information
(10-year, 25-year, and 100-year) collected outside the study basins are used to develop regional (country-wide) regression
equations. Peak-flow discharge derived from these equations together with preferred spatial parameter relations as soft prior
information are used to constrain the simultaneous calibration of 20 tributary basin models. The nonlinear range of uncertainty
in estimated parameter values (1 curve number and 3 recurrent rainfall amounts for each model) is determined using an inverse
calibration-constrained Monte Carlo approach. Cumulative probability distributions for rainfall amounts indicate differences
among basins for a given return period and an increase in magnitude and range among basins with increasing return interval.
Comparison of the estimated median rainfall amounts for all return periods were reasonable but larger (3.2–26%) than rainfall
estimates computed using the frequency-duration (traditional) approach and individual rain gauge data. The observed 25-year
recurrence rainfall amount at La Hachadura in the Paz River basin during Hurricane Mitch (1998) is similar in value to, but
outside and slightly less than, the estimated rainfall confidence limits. The similarity in joint inverse and traditionally
computed rainfall events, however, suggests that the rainfall observation may likely be due to under-catch and not model bias. |
| |
| Keywords: | El Salvador Monte Carlo Nonuniqueness Prediction Prior information Extreme rainfall Regionalization Return periods Recurrence intervals Regularization Uncertainty Ungauged |
| 本文献已被 SpringerLink 等数据库收录！ |
|
