An Alternative Temporal
Rainfall Distribution for
Hydrologic Analysis and
Design
Joseph P. Wilson, PE, PHWilson Hydro, LLCPhD CandidateMissouri University of Science TechnologyRolla, Missouri
Sorry Folks, Missouri is CLOSED due to high water
MoDOTRoad Conditions
I think we can all agree that this
Is not the same as this
Session Topics
Why is the temporal distribution of rainfall important?
Rainfall temporal distributions currently used in hydrologic modeling.
Some of the issues with currently used rainfall distributions.
An updated temporal distribution.
Design Storms
Most jurisdictions in the United States use the design storm approach to hydrologic modeling for design and analysis.
A design storm is a way of representing storm rainfall with respect to depth, duration, temporal distribution, frequency and spatial distribution.
Based on the assumption that, under average conditions, a rainfall of a given frequency will produce runoff of the same frequency.
Rainfall Temporal Distributions
The selected design storm for hydrologic modeling has a significant impact on the computed results.
Rainfall Temporal Distributions are generally categorized as:
Simple geometrical shapes.
Nested intensity.
Statistical analysis of a gage network.
Simulation based on stochastic rainfall models
Based on an observed storm.
NRCS Type I, Ia, II, and III
Huff’s Quartile
Atlas 14 Distributions
New Frequency Modified Pilgrim-Cordery
Temporal Distributions for Discussion
NRCS 24-Hour Distributions
Nested Intensity.
Developed by SCS using rural rain gages, expanded on work by Hershfield.
NRCS proposing to
discontinue use.
Huff’s Quartile Distributions
Median CurvesPoint Rainfall
• Derived from 49 gage network in Illinois
• Events > 0.5”• 11 year period of
record, 1955-1966• Revisited in 1990 to
add 5 urban gages andResults included inBulletin 71.
Atlas 14 Distributions
Median Curves24-Hour Duration
• Published with Atlas 14 Region 8, 2013
• Based on precipitation depth > 2-year recurrence interval
• Distribution by duration
• 6-hour curves not based on true quartiles
They are Somewhat Similar
Frequency Modified Pilgrim-Cordery Distribution
Based on the method developed by Pilgrim and Cordery, 1975.
PC Ranking method based on arbitrary number of observed rainfall events.
Annual exceedance series used.
Depths non-dimensionalized.
Original method average period used for each rank. FMPC uses most frequent rank.
Average depth per rank assigned.
Frequency Modified Pilgrim-Cordery Method
This analysis uses the NOAA hourly rainfall for 48 stations in Missouri defined as Hydrologic Region 4 in Atlas 14, Volume 8.
Period of record ranges from 30-67 years.
The annual exceedance series was used.
Durations of 6-, 12-, and 24-hours.
48 Stations Studied
Data is Messy!
Data Processing
Period of record >30 years.
Missing data does not exceed 5%.
Accumulated totals excluded.
Consecutive totals summed for consecutive 6-, 12-, and 24-hour periods. Precipitation must occur in at least 80% of the time periods.
Overlapping time periods eliminated.
Data Processing, cont’d
The WILRAIN program was developed to process over 27 million lines of rainfall data.
A separate MODPC program was developed to derive the Frequency Modified Pilgrim-Cordery Distribution and compute the RMSE.
The programs can be applied to data from any NOAA hourly or 15-minute rainfall station or can be customized for other data formats.
FMPC Distribution - 48 Stations
Computed Error
St. Louis Lambert InternationalCumulative Mass Diagram, Hourly Data
St. Louis Regional Analysis
St. Louis Regional Analysis
St. Louis County Flood Study, MSD, FEMA
St. Louis FMPC compared to Huff’s 2nd Quartile
St. Louis FMPC compared to Huff’s 3rd Quartile
St. Louis FMPC compared to NRCS Type II
SummaryFrequency Modified Pilgrim-CorderyDistribution has a smaller RMSE than Huff’s or the NRCS Type II, and more closely represents observed data.
The method can be applied to any NOAA station recording hourly or 15-minute data.
Can be used at a single station or as a regional analysis.
It is important to use the best available data.
RecommendationsSelect the distribution that best fits your data.
Prefer a statistical based distribution.
Rainfall distributions in order of preference:
1. Frequency Modified Pilgrim-Cordery.
2. Pilgrim-Cordery.
3. Huff’s quartile/Atlas 14
4. As a last resort:
Nested intensity from local data.