Dhruva Kathuria (618/UMBC) is a Co-Investigator on the recently funded NASA ROSES-Ecohydrology proposal “Enhancing flash drought prediction by quantifying vegetation responses to rainfall and heatwave pulses.” This three-year grant is led by Dr. Yanlan Liu (UCLA), with Drs. Kai Zhu (Univ. of Michigan), Alexey Shiklomanov (GSFC), and Kathuria, as the Co-Is.
The United States has been experiencing frequent flash droughts in recent decades, resulting in severe ecological and socioeconomic impacts. While normal droughts occur over two to three months with below-normal rainfall, flash droughts intensify over weeks, with a combination of little rain, high temperatures, and strong winds, resulting in water loss from evapotranspiration (ET). Current land surface models struggle to accurately predict flash droughts due to unclear ecohydrological mechanisms that drive the initiation and intensification of a flash drought. Mega-droughts typically are driven by climate oscillations, while flash droughts are substantially influenced by subseasonal-to-seasonal (S2S) land-atmosphere interactions mediated by vegetation. Following rainfall and heatwave pulses, soil and atmosphere dry-downs can trigger a cascade of ecohydrological feedback within the soil-plant-atmosphere system (SPAS). Under certain vegetation and atmospheric conditions, a positive feedback loop could emerge.
Dr. Kathuria explains, “The positive feedback can arise in some cases when a rainfall event is followed by a heatwave pulse. After rainfall pulses, large ET contributes to soil dry-downs, which can be further accelerated by the heatwave. When the soil dries out and the air heats up, plants can sometimes (though not always) respond by "shutting down" and releasing less water. This then backfires: less water released into the air means even hotter, drier conditions, which dries the soil further, triggering a vicious cycle that can rapidly spiral into a flash drought. We hypothesize that this happens more frequently in areas with higher biomass and in arid climates. To improve flash drought prediction, it is critical to mechanistically quantify the ecohydrological dynamics within the SPAS, which are co-regulated by vegetation and the physical environment, and to clarify their roles in driving flash droughts; hence, this forms the overarching goal of the project. This research will be among the first to mechanistically quantify how vegetation responses after rainfall and heatwave pulses regulate flash droughts.”
“The proposal has three stages, moving from data-driven insight toward improved forecasting. First, we will use a data-driven approach to explore the causal relationships between vegetation dynamics, climate conditions, and flash drought indicators across a wide variety of biomes. We plan to do this using causal machine learning methods applied to a broad mix of field and remote sensing data. Second, we will feed this understanding into a computationally efficient physical model, the Soil-Plant-Atmosphere System (SPAS), to test which physical mechanisms (soil drying, water movement within plants, and so on) contribute the most to accurately predicting flash droughts. We start with SPAS because it is computationally light and easy to run experiments on. Finally, we will use what we learn from SPAS to develop better parameterizations for the land surface model (Noah-MP) within NASA's Land Information System (LIS), improving its ability to forecast flash droughts. Noah-MP is computationally expensive, so the logic of the workflow is sequential: data-driven causal ML informs the efficient SPAS model, and SPAS in turn informs Noah-MP in LIS. Ultimately, the project will produce a large-scale model-data toolkit to improve flash drought prediction across the western U.S. The outcome will contribute to enhanced S2S hydroclimate forecasts and support mitigation strategies for future flash drought events.”