A series of thunderstorms moving along a stalled frontal boundary dropped extremely heavy rain on much of southern Minnesota on
August 18, 19, and 20, 2007. The most intense precipitation rates occurred during the afternoon
and evening hours of Saturday, August 18, and the early morning hours of Sunday, August 19. Over the course of the event, all or portions of
28 counties received at least four inches of rain. Six-inch totals were common across the region, and portions of southeastern Minnesota reported
astounding rainfall amounts ranging from 8 to 20 inches. The heaviest rainfall reports came from Winona, Fillmore, and Houston counties, where 36-hour totals
exceeded 14 inches. The largest multi-day rainfall total reported (through Monday, August 20) was 20.85 inches
observed near the town of Houston in northern Houston County. An official National Weather Service climate observer near Hokah in
Houston County reported a storm total of 16.27 inches. Of the 16.27 inches, 15.10 inches fell within the observer's 24-hour
observation cycle ending at 8:00 AM on Sunday, August 19. This is the largest 24-hour rainfall total
ever recorded by an official National Weather Service reporting location in Minnesota. The previous Minnesota record was 10.84 inches,
measured at Fort Ripley in Crow Wing County on July 22, 1972.
The deluge produced flooding tied to seven fatalities. Major flood damage occurred in many southeastern Minnesota communities. Hundreds
of homes and businesses were impacted. Reports of stream flooding, urban flooding, mudslides, and road closures were numerous throughout
The combination of huge rainfall totals and a very large geographic extent, make this episode one of the most significant rainfall events
in Minnesota's climate history. A six-inch rainfall total for a given location in this region over a 24-hour period is said to
be a "100-year" (1% probability) storm. The area receiving six or more inches during a 24-hour period in the midst of this torrent encompassed thousands of
square miles. Other heavy rainfall events during this decade of comparable magnitude and spatial coverage include extraordinary
rainfalls in northwestern Minnesota on June 9-10, 2002, and in
southern Minnesota on September 14-15, 2004.
Further Discussion of Rainfall Return Period:
When addressing the relative rarity of this rainfall event, it is useful to focus on a single location to use as a case study. As detailed above,
an official National Weather Service Cooperative observer located one mile south of Hokah (Houston County) recorded 15.10 inches of
rain for the 24-hour period from 8:00 AM, Saturday, August 18 to 8:00 AM, Sunday, August 19. The 3-day storm total for this location was 16.27 inches.
This was not the largest rainfall total reported from the storm event, however the value has the endorsement of a multi-agency "climate extremes committee"
and is easily defensible. Other "non-official", nevertheless quite valid, rainfall reports coming from other locations in southern Winona and northern
Houston counties support the conclusion that the Hokah observation represents the rainfall amounts for at least two hundred square miles in that vicinity.
The 15.10 inch 24-hour total is the largest 24-hour rainfall total ever recorded by an official National Weather Service reporting location in Minnesota.
This is a significant detail. The National Weather Service has maintained a network of 150 to 200 volunteer weather observers for 116 years. The previous
Minnesota record was 10.84 inches. For the Hokah report to top this record by over four inches is truly remarkable.
The probability of receiving six or more inches of rain in a 24-hour period, in a given year, for a given location in southeastern Minnesota
is approximately 1% (100-year storm). One can logically infer that the probability of receiving 15 inches of rain in a 24-hour period is far, far
less than 1%. The question becomes ... how much less? Is it 0.2% (500-year), 0.1% (1000-year), 0.01% (10,000-year), etc.?
As mentioned, there are over 100 years of rainfall data available for many Minnesota locales. The challenge is to extrapolate to recurrence
intervals that reach beyond our recorded experience. Mathematical techniques are available to make those extrapolations. The technique utilized, along
with the data sets used for input, can create a variety of answers, none of which can be declared as "truth". The Minnesota State Climatology Office
has determined that a "generalized logistic distribution" used on two long-term data sets from southeastern Minnesota offers the best visual
representation of the extreme value behavior. Using this distribution and these data sets, the State Climatology Office estimates that the
15.10-inch 24-hour total reported at Hokah has a return period of somewhat over 2000 years.
The Probable Maximum Precipitation (PMP) value for a 24-hour period
in southeastern Minnesota is approximately 31 inches. The PMP is a hypothetical upper limit reflecting the atmosphere's capacity for producing a perfect
rainmaker in a given area.
The return period discussion above revolves around extreme value statistics for a single location. If one takes into account the considerable area receiving
the extraordinary rainfall totals, the August 18-20 event becomes even more statistically improbable.
When asked to provide a recurrence interval estimate for this event, an investigator should note the enormous gap between the 100-year 24-hour
threshold (six inches), and the fifteen or more inches that fell during a 24-hour period on that ill-fated weekend. Although a precise recurrence interval
cannot be determined, we can safely state that the probability of being subjected to such an occurrence was considerably below one percent. What are, from a risk
management perspective, the meaningful distinctions between a 1.0% (100-year) probability, a 0.1% (1000-year) probability, and a 0.01% (10,000-year) probability
of occurrence? That is to the judgment of the investigator.
Location by location rainfall totals are available as comma-separated text files. The files provide location coordinates that are followed by a multi-day
precipitation total in inches. The files include rainfall data for the entire state of Minnesota for the August 18-20, 2007 time period.
Imagery from the La Crosse National Weather Service Doppler radar site is available without charge from the National
Climatic Data Center (NCDC). Radar-based precipitation estimates for the flood event can be obtained from
this archive. One-hour precipitation total estimates were calculated by the La Crosse radar facility at five-minute
increments during the storm. The imagery can be viewed using free visualization and analysis software
available from NCDC. In addition to viewing the radar data, the software allows the user to export the data
to common formats such Shapefile, Arc/Info ASCII Grid, and more. Level III one-hour precipitation data are provided
at a resolution of two kilometers in radial distance, by one degree of azimuth arc. The La Crosse office of the National
Weather Service points out to users of the radar imagery that a
radar "fault" is evident
southwest of the facility during the event. Data found in this arc should be treated carefully.
The State Climatology Office thanks the Soil and Water Conservation Districts, the National Weather Service, and
all of the diligent volunteer precipitation observers who make analyses of these events possible.