Long Term Prediction of Drought Events in Southwestern US

We developed and applied an integrated approach to construct predictive tools with lead times of 1 to 12 months to forecast monthly drought indices over three pilot study areas (i.e. Sacramento County, CA, Kings County, CA and Clark County, NV). The following steps were conducted: (1) acquire, and generate historical (1950 - 2016) precipitation spatial data averages for the three counties using over 40 rain gauges, (2) acquire and pre-process monthly values for 34 climatic indices influencing the regional and global climatic patterns (e.g., Northern Atlantic Oscillation [NOI], Southern Oscillation Index [SOI], and Tropical North Atlantic Index [TNA]) over the period ranging from January 1950 to December 2016, (3) construct a neural network model to extract relationships between the observed precipitation and the controlling factors (i.e. climatic indices with multiple lead-time periods), (4) construct a neural network stepwise algorithm to determine the most dominant and predictive climatic indices in the region and (5) use the predicted values to determine long term drought indices for the region. These methodology could lead the way to the development and implementation of long-term water and agricultural management scenarios for southwestern US.