In Depth
For this project, our goal was to calculate the probability of different precipitation thresholds being met from certain points during the growing season through the end of August (as forward-looking windows). These probabilities were derived from historical climate data. Historical precipitation data was considered on a grid, for the years 1950-2022, with data gridded from Applied Climate Information System (ACIS) station records. For more information on the gridded ACIS precipitation data, please see the metadata at https://www.rcc-acis.org/docs_gridded.html. Multiple methods were considered for computing precipitation exceedance probabilities. For each grid cell, the sum of daily precipitation for a growing season from the date in question (every two weeks starting from April 1) through the end of August was considered for each year in the record. A basic probability was then derived from the historical record, where the number of years with precipitation from date_0 through August 31 that exceeded the threshold was expressed as a probability.
Mathematically, P = (N_years_exceed_threshold/N)*100, where N is the number of years in the record.
As a proof of concept, this process demonstrated that the probabilities can be calculated efficiently with the daily gridded precipitation dataset. In future work, we may reformulate this, such that a gamma distribution is assumed for the historical precipitation timeline in question. This way, the threshold amount may be added to the end of the time series, and its SPI calculated. From the SPI value of the threshold amount, the probability of that amount being exceeded may then be calculated.