The science of data assimilation (DA) is the key to extract the knowledge of satellite data for improving predictability of landatmosphere interactions and reducing uncertainties in historical…

This federal grant, awarded by the National Aeronautics and Space Administration (NASA) to the Regents of the University of Minnesota, Office of Sponsored Projects Administration, supports research on data assimilation techniques to enhance the predictability of extreme weather events and soil moisture dynamics. The $299,467 grant aims to develop new variational data assimilation methods that can better account for non-Gaussian errors, observational outliers, and systematic biases in weather and land surface models. The research will prototype these techniques within a coupled Weather Research and Forecasting (WRF) and Noah Land Surface Model system to improve the assimilation of precipitation and soil moisture data from NASA's Global Precipitation Measurement (GPM) and Soil Moisture Active Passive (SMAP) satellite missions. The resulting open-source software tool is intended to accelerate further scientific discoveries and potential operational applications in NASA's Earth system models. This grant does not have a set-aside designation.

The science of data assimilation (DA) is the key to extract the knowledge of satellite data for improving predictability of landatmosphere interactions and reducing uncertainties in historical reanalysis of the hydrologic cycle. Currently, most of the operational systems still use decades-old variational approaches that prevent US to take full advantage of new satellite data and go beyond our state-of-the-art predictive skills. In particular, classic variational methods (1) are unable to accurately forecast the dynamics of hydrologic variables with isolated extremes and jump discontinuities (e.g., precipitation supercells and squall lines); (2) cannot properly account for large differences between model forecasts and satellite observations (e.g., small-scale convective precipitation cells that are observed by satellite but are not captured in model forecasts); (3) are incapable to rigorously treat systematic biases in model forecasts (e.g., underestimation of soil moisture over a specific soil/vegetation type); and (4) lack observability (e.g., information of surface soil moisture retrievals can not directly propagate to lower soil layers). This research has two main hypotheses. Firstly, IT hypothesize that the key atmospheric states and fluxes exhibit a specific prior structure in an appropriately chosen transform domain (e.g., fourier), which could be used to significantly improve predictability of extreme precipitation. Secondly, IT hypothesizes that new techniques#that not only minimize the error variance but also directly penalize the mismatch between two probability distributions#can be used to markedly reduce uncertainties in forecasts of soil moisture dynamics. This research aims to develop a new class of variational DA techniques that enables to enhance predictability of extreme events, formally accounts for systematic forecast biases, and also enhances observability in weather-land prediction systems. This involves a series of complementary directions to answer some key questions. How can we develop new and efficient techniques to conduct non-gaussian DA in transform domains (e.g., fourier, wavelet) for removing geophysical errors without significant smoothing effects on extremes? How can we make variational DA robust to observational outliers? What is the best way for direct penalization of the distance between the probability distributions of the model forecast and the analysis state for rigorous treatment of systemic biases? Is there any way to increase observability in DA systems by accounting for cross-correlation between the model and observation errors? As a result of this research, the hope is that we can reduce uncertainties in prediction of extreme precipitation; take full advantage of satellite soil moisture retrievals when land surface model outputs are systematically biased, and increase root-zone soil moisture observability through precipitation and soil moisture data assimilation. To test the hypotheses, this research will prototype a new variational DA module to be used in a coupled system of the weather research and forecasting (WRF) model and the noah land surface model for joint assimilation of the active/passive retrievals by the global precipitation measurement (GPM) and the soil moisture active passive (smap) satellites. Using ground-based reference observations, the performance of the new DA techniques will be validated and assessed based on the specified objectives over the contiguous united states. An interoperable open source software tool will be made publicly available to accelerate further scientific discoveries and foster potential operational applications in NASA's earth system models.