Modeling Probability Distributions of Hydrologic Variables from NLDAS to Identify Water Cycle Extremes

Anticipating extreme events, whether potential for flooding or drought, becomes more urgent every year, with increased variability in weather and climate. Hydrologic processes are inherently spatiotemporal. Extreme conditions can be identified at a certain period of time in a specific geographic region. These extreme conditions occur when the values of a hydrologic variable are record low or high, or they approach those records. The calculation of the historic probability distributions is essential to understanding when values exceed the thresholds and become extreme. A dense data model in time and space must be used to properly estimate the historic distributions. The purpose of this research is to model the time-dependent probability distributions of hydrologic variables at a national scale. These historic probability distributions are modeled daily, using 35 years of data from the North American Land Data Assimilation System (NLDAS) Noah model, which is a land-surface model with a 1/8 degree grid and hourly values from 1979 to the present. Five hydrologic variables are selected: soil moisture, precipitation, runoff, evapotranspiration, and temperature. The probability distributions are used to compare with the latest results from NLDAS and identify areas where extreme hydrologic conditions are present. The identification of extreme values in hydrologic variables and their inter-correlation improve the assessment and characterization of natural disasters such as floods or droughts. This information is presented through a dynamic web application that shows the latest results from NLDAS and any anomalies.