Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Characterization of Within-Day Beginning Times of Storms for Stochastic Simulation

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  Transactions of the ASABE. 55(4): 1179-1192. (doi: 10.13031/2013.42242) @2012
Authors:   J. V. Bonta, S. P. Hardegree, J. Cho
Keywords:   Beginning times of storms, Kernel density estimation, Mixed doubly truncated normal distributions, Storm simulation, Weather generator

The beginning times of storms within a day are often required for stochastic modeling purposes and for studies on plant growth. This study investigated the variation in frequency distributions of storm initiation time (SI time) within a day due to elevation changes and month. Actual storms without 24 h constraints were used, as opposed to simply bursts of precipitation within a 24 h period. Two methods of characterizing and quantifying these distributions were investigated: kernel density estimation (KDE), and a mixed doubly truncated normal (MDTN) distribution method using nonlinear curve fitting subject to bounds on the parameters. Parameter estimation methods were also investigated. Data came from the raingauge network maintained by the USDA-ARS at the Reynolds Creek Experimental Watershed in southwest Idaho over a 982 m elevation gradient. There was no difference between frequency distributions of SI time with elevation or precipitation type over the 147 km2 study area. There was a significant shift in SI-time distribution from earlier in the morning in late fall and winter to early afternoon during the spring and summer. Both the KDE and MDTN methods accurately characterized the observed histograms, which included near-uniform, single-mode, and bimodal distributions. The MDTN method worked well most of the time (~97%) but can have mathematical convergence problems. An SI-time analysis based on a 24 h cycle starting at 2100 h yielded a better fit to the data than a standard day defined to start at midnight using the MDTN method. Exploratory regressions between the four MDTN parameters and several readily available independent variables did not yield consistent or significant predictive relationships. Cumulative distributions for either the KDE or MDTN methods are suggested for stochastic modeling purposes on a monthly basis, as they represent well observed histograms of SI times. The KDE method is suggested for use because of its simplicity in ungauged areas as long as neighboring data are available. The methods have utility for characterizing time variation of other weather elements.

(Download PDF)    (Export to EndNotes)