STATISTICAL EVALUATION OF CRITICAL DESIGN STORMS

1
STATISTICAL EVALUATION OF CRITICAL DESIGN STORMS

• F.N. Nnadi, Ph.D., P.E., Maria Rizou and Wendy
  Wert
• Civil and Environmental Engineering
• University of Central Florida
• 4000 Central Florida Blvd.
• Orlando, Florida 32816-2450

2
TABLE OF CONTENTS

• Background
• Scope and Objectives
• Data Source Weiss Factor
• Project Setup
• SMADA Modeling
• Statistical Analysis
• Results
• Conclusions and Recommendations

3
BACKGROUND

• Why this Study?
• Concern for drainage specifications in Florida
  due to
• large volume of rainfall (high risk from extreme
  events),
• flat topography, shallow GWT, and intensive
  urbanization.
• Controversy on the prediction efficiency of the
  various critical design storms.
• Inadequate testing of the historic and design
  storms over a range of conditions return
  periods, physiographic and climatological regions.

4
SCOPE AND OBJECTIVES
Scope Statistical Evaluation of Design Storms and
Actual Historical Storms over a range of Return
periods and a variety of Basins /Gage Stations
?
? temporal deviations spatial
deviations
5
SCOPE AND OBJECTIVES (cont.)

• Establish the watershed and pond combination
  which produces peak discharge for the defined
  duration at all gage stations .
• Design storms Run discrete storm simulations.
• Historical storms Run continuous simulation.
• Apply statistical tests to compare the results of
  (2) and (3) in order to identify the design storm
  that best approximates the actual storm.

6
Geographic variability

• Criteria for gage site selection
• Availability of historical rainfall data.
• Different rainfall patterns (IDF zones).
• Jurisdiction of different WMDs.
• 1 hr records Gainesville, Homestead, Panama
  City, Moore Haven, St. Leo.
• 15 min records Melbourne, Homestead, Panama
  City, St. Leo.

7
Locations of Watersheds
Gainesville
NWFWMD
SRWMD
SJWMD
Panama City
Melboerne
St. Leo
SWFWMD
Moore Haven
SFWMD
Homestead
8
Data Source

• Single-Event Rainfall Data
• Data used for Developing Standard IDF Curves with
  Weiss Factor Correction.
• Continuous Rainfall Data
• National Climatic Data Center, EarthInfo CD ROM
  with ASCII format data.

9
WEISS FACTOR
General concept
D 1hr
D
A
B
C
E
Dt recording
0.5
Schematic of a "non-inclusive" uniform 1-hr rain
falling in-between two adjacent 1-hr intervals.
10
Application of the Weiss Factor
11
Weiss Factor Effect on 1-hr Rainfall Data
12
Project Flowchart
13
Watershed Characteristics for Short Duration
14
Watershed Characteristics for Long Duration
15
SMADA Hydrologic Modeling

• SMADA
• Discrete modeling (FDOT, SCS II, SCS IIFL, SCS
  III, SFWMD72, SJRWMD96).
• Continuous simulation.
• Distrib
• Normal, 2 Parameter Log Normal, 3 Parameter Log
  Normal, Pearson Type III, Log Pearson Type III,
  and Gumbel Type I Extremal.

16
SMADA Modeling

• Multiple Rainfall Analysis (single-event
  simulation model)
• INPUT
• site specific rainfall depths
• watershed-pond parameters
• hydrograph generation method
• OUTPUT
• runoff values for various duration and return
  period combinations

17
SMADA Modeling

• Continuous Simulation Analysis
• INPUT
• Historic rainfall records (parsed with
  rparser.exe)
• watershed-pond parameters
• hydrograph generation method
• OUTPUT
• peak runoff values for each rainfall event.

18
SMADA Modeling

• Distribution Analysis (Distrib)
• INPUT
• maximum annual peak runoff values
• best fit distribution Normal, Gumbel, Pearson
  III, Log Pearson, 2 or 3 parameter Log Normal)
• OUTPUT
• peak runoff values corresponding to different
  return periods.

19
DISTRIB output
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
RESULTS ASSESSMENT

• Visual Comparison
• Statistical Analysis

24
VISUAL COMPARISON
Visual interpretation considers the minimum
absolute differences between the CS values and
the corresponding discrete method values at each
return period.
25
STATISTICAL ANALYSISUsing MiniTab Progam

• Normality Assessment
• Friedman two-way Analysis of Variance by ranks
• Friedman multiple-comparison test

26
Normality assessment

• Normal probability plot is a traditional
  technique.
• Types of tests Ryan-Joiner test, based on the
  Shapiro-Wilk test.
• Technique generalized least squares.

It is the linear regression of the ordered
observations with respect to their standardized
expected values.
It corrects the data if the observations were
ordered and thus are assumed to be correlated.
27
(No Transcript)
28
(No Transcript)
29

• Friedman two-way Analysis of Variance by ranks

Applicable when non-normality,
heterogeneity Subjects 9 test sites and 9
watersheds Data peak runoff values blocked by
return periods
The purpose of blocking is to sort experimental
units into groups of homogeneity with respect to
a dependent variable, such that the differences
between groups are as great as possible.
30
Advantage of blocking To remove data variation
associated with runoff changes along different
return periods Disadvantage one degree of freedom
is lost for each treatment after the previous
treatment
Hypotheses
31
Short Duration Analysis(1, 2, 4, and 8 Hour
Duration)
32
Friedman Test (Runoff Values, 15-min data), St.
Leo
33
Friedman Test (Runoff Values, 1-hr data), St. Leo
34
Summary of Approximations to CS values for Short
Duration
35
Long Duration Analysis(24, 72, 168, and 240 Hour
Duration)
36
Friedman Test (Runoff Values, 1-hr data), St. Leo
37
Summary of Approximations to CS values for Long
Duration
38
Remark Statistical evaluation considers that the
closest method to the CS is the distribution with
the minimum absolute difference of the rank sums
(Ti-Ti) between each discrete method and the CS
by blocking the return period.
39
(No Transcript)
40
CONCLUSIONS

• Visual and Statistical Analysis Agreement
• Regional Agreement for 1-hr Short Duration
• Obvious Regional variation for 15-min Short
  Duration
• Infiltration coefficient effect on the results of
  8-hr and 8k-hr watersheds
• Peak runoff results for long duration did not
  present same regional agreement