Peak discharge prediction, recession analysis, and evaluation of streamflow response to rainfall for watersheds in the Luquillo Experimental Forest, Puerto Rico

Department of Natural Resources Management and Engineering, University of Connecticut, Storrs, CT 06269, coronaex@hotmail.com

Abstract

Peak Discharge Estimation in the Luquillo Experimental Forest, Puerto Rico, using Multiple Regression Analysis and G.I.S.

The Luquillo Experimental Forest (LEF) is located in the Sierra de Luquillo, in the northeast corner of Puerto Rico. Instantaneous peak discharge data from 10 USGS gaging stations located in the forest were used to perform a peak discharge frequency analysis. These stations had records for different periods, ranging from 13 to 38 years. Multiple regression equations were developed to relate peak discharge values to drainage basin and climatic characteristics. Digital Elevation Models, stream networks, geology, drainage basin boundaries, and rainfall data were integrated in a Geographic Information System (GIS) to calculate the desired parameters. Regression equation models are provided to estimate the 2, 10, 25, 50, and 100-year peak discharge at ungaged sites whose characteristics are similar to the study area. In this study, basin area was the most significant parameter (alpha = 0.01) to predict peak discharge for all models. In addition, the 2-year 24hr rainfall intensity was significant to predict the 2-year peak discharge. The coefficient of determination (R²) and its adjusted version (Adj. R²) for the regression equations range from 0.70 to 0.90. Considering the small sample size, these statistics show that the model predictability is highly acceptable.

Recession Analysis for watersheds in the Luquillo Experimental Forest to predict streamflow recessions during rainless periods

An analysis of hydrograph recessions and rainfall data was performed to develop Master Recession Curves (MRC), for the Río Fajardo and Río Espíritu Santo watersheds. These rivers are part of the drainage network of the Luquillo Experimental Forest in Puerto Rico. The slope of the MRC is considered as the recession constant k, and can be used to predict baseflow recessions during periods of low rainfall. To account for seasonal rainfall patterns, the data were grouped into dry and wet month recessions. Sets of three MRC’s per season for each watershed were developed: one using the Matching Strip Method (MS) and two using variations of the Correlation Method (CM). These two variations of CM were the Envelope line procedure (CME) and the least squares Regression (CMR). The observed vs. predicted recessions for each method were compared in order to find the method with the best fit to the selected data. The selected method was the CMR, which predicted most of the recessions within a range of ± 2-cfs. This method proved to be faster, easier, and less subjective than the others. The recession constant from this method was used to estimate the time it would take the discharge to reach the Q99 flow duration, which was considered as a critical flow. Based on this study, water managers have between 8 to 12 days warning before streamflow goes from Q90 to Q99. In addition, it was found that Río Fajardo has fewer and shorter periods of low rainfall and tends to recede faster than Río Espíritu Santo, especially during the wet season.

Streamflow response to rainfall in the Luquillo Experimental Forest

Daily average discharge from the long-term gage on the Río Espíritu Santo (5006380) and rainfall records from the El Verde weather station were examined to determine river response to rainfall events. The Río Espíritu Santo watershed is located at the Luquillo Experimental Forest (LEF) in Puerto Rico. The resulting database was divided into different ranges of initial flow conditions. Each range was analyzed visually and then statistically using simple and multiple regression techniques. The visual analysis and the simple regression equations were used to estimate the amount of rainfall necessary to produce certain streamflow response, and to assess the influence of the initial flow conditions in such estimation. The goal of the multiple regression equations was to improve the simple regression equations that used total rainfall as the only predictor, by adding other parameters like the watershed storage and antecedent rainfall conditions. Two types of multiple regression models were developed. The Daily based (D) models seek to predict streamflow response following one or two days of rainfall. The Event based (E) models were used to predict the total streamflow increase from the beginning of the rainfall event to the date where the peak discharge was recorded. Although daily-based variables are easier to use, the E models had higher R²and smaller standard errors (%SE) than the D models. The final models developed with the full data set had better R², Adj. R², and %SE than any other model that used a specific range of initial flow. However, the multiple regressions do not show a significant improvement compared to the simple regression equations.

• WEBB Project Bibliography
• WEBB Project Chief Home Page
• Caribbean Water Science Center home page