H082-11
Development of a “nature run” for observation system simulation experiments (OSSE) for snow mission development
Thursday, 10 December 2020: 04:30
Virtual
Melissa Wrzesien1, Sujay V Kumar1, Carrie Vuyovich1, Ethan D Gutmann2, Rhae Sung Kim1, Barton A Forman3, Michael T Durand4, Mark S Raleigh5, Ryan Webb6 and Paul Houser7, (1)NASA Goddard Space Flight Center, Greenbelt, MD, United States, (2)National Center for Atmospheric Research, Boulder, CO, United States, (3)University of Maryland, College Park, MD, United States, (4)Byrd Polar Research Center, Columbus, OH, United States, (5)University Corporation for Atmospheric Research, Boulder, CO, United States, (6)University of New Mexico Main Campus, Department of Civil, Construction, and Environmental Engineering, Albuquerque, NM, United States, (7)George Mason University Fairfax, Fairfax, VA, United States
Abstract:
Observation System Simulation Experiments (OSSEs) are often conducted to evaluate the impact of data to be collected from proposed missions. These simulations are critical in the systematic assessment of competing mission designs and design choices. Further, these experiments also help quantify the utility of observations beyond the immediate variable of interest. Though snow is a critical component of the global water cycle, developing robust remote-sensing based snow measurements is a significant challenge. The use of OSSEs is useful in developing assessments of different snow observational methods. A “nature run” is a key component of the OSSE, where estimates of the true state are developed using a high-quality model and inputs. Using the nature run, different types of observations corresponding to different mission/sensor choices are then developed within the OSSE. The quality of the nature run significantly impacts the conclusions made from the OSSE.
Land surface models (LSMs) are used to develop spatially distributed estimates of snow water equivalent (SWE) and runoff. LSMs are limited by uncertainties in model physics and parameters, among other factors. In this study, we describe the use of model calibration tools to improve the snow simulations within the NoahMP LSM, for the purposes of developing a nature run for snow OSSEs. The NoahMP LSM is calibrated against the University of Arizona (UA) snow depth product over a Western Colorado domain for an average water year. We use a genetic algorithm within the Land Information System (LIS) for the calibration procedure. With calibrated parameters, we use LIS to produce calibrated and uncalibrated NoahMP simulations at ~1 km resolution for WYs 2010-2020. By evaluating both simulations against the UA dataset, we show that calibration decreases domain-averaged temporal RMSE for snow depth from 0.15 m to 0.13 m and improves the timing of snow ablation, with larger improvements in areas with higher SWE. Increased snow simulation performance resulted in improved estimates of model-simulated runoff, as compared with observations over small unmanaged basins. In one basin, the Nash-Sutcliffe Efficiency improves from 0.37 to 0.65. Results suggest that snow depth calibration can improve performance of other processes within the water cycle.
