A publication of the Midwestern Regional Climate Center
March 11, 2015  




Go to the MRCC homepage

Follow us!
Follow us on Facebook!  Follow us on Twitter!



 High Resolution Drought Monitoring

High Resolution SPI Methodology (in brief)

The historical precipitation frequency distribution was determined using precipitation data from NWS COOP stations from across the contiguous US.  COOP stations meeting minimum data requirements were clustered into spatially discrete and acceptably homogeneous regions. Stations are clustered so that information from all the sites in a cluster can be used when estimating the distribution parameters for a particular site, yielding a more accurate estimate of the historical precipitation frequency distribution (Hosking and Wallis 1997).  Stations were clustered twice to reflect two different normals periods:  1971-2000 and 1981-2010.

Once final clusters were determined, the parameters of the Pearson Type III distribution were determined for each station, using information from each site in a given cluster, and data for each site’s period of record.  The distribution parameters were normalized by dividing by the historical mean at each station, allowing PRISM normals to be used as the location parameter.  Once parameters were normalized, they were interpolated to the same grid as the high-resolution precipitation estimates. 

To calculate SPI, gridded precipitation is summed for all the dates during a time period of interest.  The SPI generated here is calculated using both NWS AHPS precipitation as well as NCEP Stage IV precipitation estimates, resulting in two different SPI grids.  Both AHPS and NCEP Stage IV precipitation products have data over the contiguous US for 24-hour periods ending at 12Z (7am EST) each day.  Additionally, both products utilize a combination of surface gauge observations, radar-estimated precipitation, and, occasionally, satellite estimates of precipitation (Lin and Mitchell 2005).  The incorporation of radar-estimated precipitation means these products, and the SPIs generated using them, have the ability capture precipitation in locations where no or few surface gauges are located.

Once gridded precipitation has been summed for the given time period and ending date, it is divided by the corresponding PRISM normal precipitation to calculate the fraction of normal.  The fraction of normal precipitation is combined with the normalized distribution parameters for the given timescale and ending calendar day to determine the cumulative probability of the observed precipitation amount. The inverse normal function is then applied to calculate the SPI, which is normally distributed with a mean of zero and a standard deviation of one.

The result is four SPI “varieties” corresponding to combinations of gridded precipitation data (AHPS or NCEP Stage IV precipitation) and the cluster solutions and their associated PRISM normals (1971-2000 or 1981-2010).   The resulting SPI grids are generated over the contiguous US and updated daily.

SPI Blend
An experimental SPI calculation, known as SPI Blend has also been generated.  This is based on the assumption that more recent precipitation has a greater influence on the current severity of dryness (or wetness) than precipitation that fell farther in the past.  The SPI Blend accounts for this by linearly weighting precipitation, with the most recent precipitation receiving the greatest weight, and all other aspects of the calculation remaining the same as with conventional SPI.

Further Reading and Useful Links

AHPS Precipitation Estimates: http://water.weather.gov/precip/

NCEP Stage IV Precipitation Estimates: http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/stage4/

PRISM Climate Group: http://www.prism.oregonstate.edu/

SPI Blend Presentation: http://docs.lib.purdue.edu/ddad2011/15/


Hosking, J. R. M., and J. R. Wallis, 1997: Regional Frequency Analysis: An Approach Based on L-Moments. Cambridge University Press, 224 pp.

Lin, Y., and K. E. Mitchell, 2005: The NCEP stage II/IV hourly precipitation analyses: Development and applications. 19th Conf. on Hydrology, San Diego, CA, Amer. Meteor. Soc., 1.2.

D. Brent McRoberts and John W. Nielsen-Gammon, 2012: The use of a high-resolution standardized precipitation index for drought monitoring and assessment. J. Appl. Meteor. Climatol., 51, 68–83.

Daly, C., R. P. Neilson, and D. L. Phillips, 1994: A statistical–topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor.,33, 140–158. 

< Back to main article