The LOUZIE project started in 2003 as a collaborative effort amongst Hydrometeorological Analysis and Support (HAS) Forecasters at the Lower Mississippi River Forecast Center. As a joint project with Joshua Palmer (now Service Hydrologist in Greer, SC), the acronym LOUZIE stands for Light, Over-Estimate, Under-Estimate, Zero, Intermittent, and Extreme which are descriptors for how observed gage data compares to radar estimated precipitation in quantitative precipitation estimates (QPE) used in hydrologic modeling efforts for real-time forecasting at the National Weather Service.

The simplistic approach began on paper, tracking gauges for errors, and developed into a regional QA-QC and gauge maintenance program for all the Weather Forecast Offices across the Southeast U.S. In addition, it received national acclaim through implementation in algorithms for quality control at the National Centers for Environmental Information (NCEI), the Storm Prediction Center, National Centers for Atmospheric Research, and National Severe Storms Laboratory – for example, the MRMS product included the logic within LOUZIE for quality assurance. Weather & Water has implemented similar procedures in DAART and CRAFTEA.


El Vado Dam Corrective Action Study involved stochastic flood modeling and, thus, required consideration of snowmelt processes due to its location in the Sangre de Cristo Mountains in northern New Mexico. Unfortunately, as if often the case in remote areas, the observations in the region were severely limited both historically (only one long term reporting site in the mountainous terrain inside the watershed of interest). Nearby sites were tens to nearly a hundred miles away on different aspect terrain.

In response, Jason developed a mechanism to back-cast (reanalyze) historical time series for the individual watersheds and elevation bands within the hydrologic model using the 4 sites as predictors for estimating the values. The overlapping period of SNODAS (a gridded snow water equivalent product) served as the training dataset (a machine learning process) to develop highly-skilled correlations. That same method was then applied to estimate historical time series for a period of over 30+ years using an 8-year overlap. This advancement is applied at Weather & Water to generate time series of precipitation, temperature, wind, and other variables of interest – along with extraction of modeled data from reanalysis and reforecast products from NOAA and other well-respected data sources. Learn more on our page for the DAART and CRAFTEA products.