Bias Correction of Snow Water Equivalent Estimates in near real-time over the California Sierra Nevada
Abstract
Near real time monitoring the spatial distribution of snow water equivalent (SWE) is significant for the water resources management and water supply forecasting in California. However, the existing SWE estimation approaches contain large uncertainties in mountain areas due to the complex terrain, snow-vegetation interactions and diverse meteorological conditions. The Airborne Snow Observatory (ASO) has provided an unprecedentedly accurate SWE dataset in several basins of the California Sierra Nevada. It has been increasingly used as 'synthetic' ground truth information to evaluate the accuracy of other SWE products, whereas its utility in correcting the SWE residuals over large mountain regions has not been well examined. In this study, we leverage the ASO SWE in the Upper Tuolumne River Basin (UTRB) in California to correct the bias of a Near Real-Time Linear Regression SWE estimation Model (NRT-LRM). NRT-LRM relies on a generalized linear model (GLM) to simulate the effect of 18 independent physiographic variables and 1 historic reconstructed SWE on the dependent variable (i.e., snow pillow SWE observation). Three bias correction methods have been applied to reduce the residuals remaining in the NRT-LRM including linear scaling, variance scaling and Gaussian Process Regression (GPR). To explore the penitential utility of NRT-LRM in streamflow forecasting, we compare the correlation between pre- and post- bias-correction SWE data with Full Nature Flow (FNF) for each watershed (HUC 8) in the California Sierra Nevada. The results suggest that all three bias correction methods are capable of improving the accuracy of SWE estimation, while GPR shows the highest accuracy given the lowest remaining SWE residuals in the UTRB and the highest overall correlation with FNF. It compensates the limitations of the linear regression model by counting the non-linear correlations between SWE and the physiographic variables. This work integrates the existing satellite-, airborne- and ground-based snow observations to better estimate spatially distributed SWE in the California Sierra Nevada. The improved SWE information could largely benefit the current streamflow forecasting system and support the state-wide water management and water supply decision making.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.C33E1622Y
- Keywords:
-
- 0736 Snow;
- CRYOSPHERE;
- 0740 Snowmelt;
- CRYOSPHERE;
- 0758 Remote sensing;
- CRYOSPHERE;
- 1863 Snow and ice;
- HYDROLOGY