Extreme Precipitation
An extreme weather event is an occurrence that is significantly different from typical weather at a specific location and time of year. Extreme precipitation events can lead to flooding, mudslides and other damaging events. In a changing climate the intensity and frequency of such events will likely change across California. This tool visualizes how climate models predict extreme precipitation events will change over the 21st century.
- Source: Cal-Adapt. Data: LOCA Downscaled CMIP5 Projections (Scripps Institution of Oceanography), Gridded Observed Meteorological Data (University of Colorado, Boulder).
-
Four models have been selected by California’s Climate Action Team as priority models for research contributing to California’s Fourth Climate Change Assessment (Pierce et al., 2018). Projected future climate from these four models can be described as producing:
- A warm/dry simulation (HadGEM2-ES)
- A cooler/wetter simulation (CNRM-CM5)
- An average simulation (CanESM2)
- The model simulation that is most unlike the first three for the best coverage of different possibilities (MIROC5)
- Source: Cal-Adapt. Data: LOCA Downscaled CMIP5 Projections (Scripps Institution of Oceanography), Gridded Observed Meteorological Data (University of Colorado, Boulder).
-
Four models have been selected by California’s Climate Action Team as priority models for research contributing to California’s Fourth Climate Change Assessment (Pierce et al., 2018). Projected future climate from these four models can be described as producing:
- A warm/dry simulation (HadGEM2-ES)
- A cooler/wetter simulation (CNRM-CM5)
- An average simulation (CanESM2)
- The model simulation that is most unlike the first three for the best coverage of different possibilities (MIROC5)
- Source: Cal-Adapt. Data: LOCA Downscaled CMIP5 Projections (Scripps Institution of Oceanography), Gridded Observed Meteorological Data (University of Colorado, Boulder).
-
Four models have been selected by California’s Climate Action Team as priority models for research contributing to California’s Fourth Climate Change Assessment (Pierce et al., 2018). Projected future climate from these four models can be described as producing:
- A warm/dry simulation (HadGEM2-ES)
- A cooler/wetter simulation (CNRM-CM5)
- An average simulation (CanESM2)
- The model simulation that is most unlike the first three for the best coverage of different possibilities (MIROC5)
- Source: Cal-Adapt. Data: LOCA Downscaled CMIP5 Projections (Scripps Institution of Oceanography), Gridded Observed Meteorological Data (University of Colorado, Boulder).
-
Four models have been selected by California’s Climate Action Team as priority models for research contributing to California’s Fourth Climate Change Assessment (Pierce et al., 2018). Projected future climate from these four models can be described as producing:
- A warm/dry simulation (HadGEM2-ES)
- A cooler/wetter simulation (CNRM-CM5)
- An average simulation (CanESM2)
- The model simulation that is most unlike the first three for the best coverage of different possibilities (MIROC5)
About
Cal-Adapt’s Extreme Precipitation Tool describes what an extreme precipitation event looks like by providing estimates of intensity and frequency of extreme precipitation events. The tools and visualizations allow you to examine how extreme precipitation events are likely to change in a warming climate over locations of interest to you.
By default, Cal-Adapt calculates extreme values of precipitation over a 2-day period, and defines an extreme event as the lowest value from Annual Maximum values in the historical period (1961–1990). Users can override these defaults by selecting a new “event duration” (number of days over which precipitation accumulates), or by selecting a different “threshold“ value that corresponds to either the 90th, 95th or 99th percentiles. The tool then displays the extreme events that exceed the threshold in different ways. The Frequency chart shows the estimated intensity of precipitation events (Return Level) for a selected period (Return Period) and how it changes over the historical period (1961–1990), mid-century (2035–2064) and end-century (2071–2099). The other charts display the total number of events, the timing of these events and the longest stretch of consecutive extreme events.
What is a Threshold value?
The extreme threshold sets the conditions for which a precipitation event is considered “extreme“. By default, the threshold is set to the lowest annual maximum precipitation accumulation in the historical record (1961 to 1990). Other alternative threshold values (90th, 95th and 99th percentiles) are based on commonly used quantiles over the historical record. Selecting too high a threshold (in arid locations) or too low a threshold can decrease the reliability of the estimates.
What is an Event Duration?
Event duration is the number of days over which precipitation falls that contribute to a single event. Changing this value will change the extreme threshold.
What is a Return Period?
The return period estimates the average time between extreme events. This is sometimes worded as a “1 in x years” event.
What is a Return Level?
The return level is the estimated amount of precipitation that would be expected to be exceeded once every return period. Effectively it is the inverse of the return period. Instead of wondering how often an extreme precipitation event will occur, we are instead considering once in any given time period what would extreme precipitation event look like? The return level is similar to the accumulated precipitation threshold, but is estimated from the underlying statistical distribution of modeled precipitation data in future climate scenarios. By contrast, accumulated precipitation threshold are calculated from historical observed values.
Technical Approach
Extreme Value Theory (EVT) is a statistical methodology used for describing rare events. There are several ways to apply EVT to precipitation data inlcuding fitting a Generalized Extreme Value distribution (GEV) over block maxima (annual maximum value) and the Peaks-Over-Threshold (POT) approach where probability distribution of exceedances over a pre-defined threshold are modeled using a generalized Pareto distribution (GPD). This tool explores extreme events in California using a POT approach.
Data values that exceed a high predefiend threshold, by default the lowest value from Annual Maximum values in the historical period (1961–1990), are extracted from a 30 year daily time series. If there are any back-to-back events only the largest such event is included. A generalized Pareto distribution is applied to this partial duration time series. Shape and scale parameters for the distribution are estimated using the Maximum Likelihood method. Return levels for selected Return Periods are estimated from the fitted model. Confidence intervals at the 95% level for each return level are estimated using the Profile Likelihood method, where sufficient (n > 100) events exist.
User Admonishment
The Extreme Precipitation Tool is designed to broadly inform potential changes in extreme precipitation intensity and frequency across a wide range of environments and climate zones in California. On a local scale different statistical assumptions (i.e. using annual maximal values rather than partial duration time-series, fitting techniques for distribution parameters and choice of extreme value distribution) may be more appropriate. We encourage users to ensure the empirical fit of the applied distribution is acceptable to their end use before using estimates produced from this tool for planning purposes.
References
- Wilks, D. (2011). Statistical methods in the atmospheric sciences (3rd ed.). Oxford ; Waltham, MA: Academic Press.
- Gilleland, E. (2015). Introduction to Extreme Value Theorem Analysis. National Center for Atmospheric Research.
- Coles, S. (2001). An introduction to statistical modeling of extreme values. London: Springer-Verlag. ISBN: 1-85233-459-2.
Data Sources
LOCA Downscaled CMIP5 Projections
Scripps Institution Of Oceanography - University of California, San Diego
Daily climate projections for California at a resolution of 1/16° (about 6 km, or 3.7 miles) generated to support climate change impact studies for California’s Fourth Climate Change Assessment. The data, derived from 32 coarse-resolution (~100 km) global climate models from the CMIP5 archive, were bias corrected and downscaled using the Localized Constructed Analogues (LOCA) statistical method. The data cover 1950-2005 for the historical period and 2006-2100 (some models stop in 2099) for two future climate projections. Details are described in Pierce et al., 2018.
Gridded Observed Meteorological Data
University of Colorado, Boulder
Historical observed daily temperature data from approximately 20,000 NOAA Cooperative Observer (COOP) stations form the basis of this gridded dataset from 1950–2013 at a spatial resolution of 1/16º (approximately 6 km). Observation-based meteorological data sets offer insights into changes to the hydro-climatic system by diagnosing spatio-temporal characteristics and providing a historical baseline for future projections. Details are described in Livneh et al., 2015.