Towards a better understanding of uncertainty sources in river temperature forecasting
Abstract
The potentially increasing pressure of high river temperatures on aquatic organisms calls for holistic river management strategies. Recent work highlighted the value of water temperature forecasting for better water resources allocation, but also stressed the need for a better understanding of the uncertainty associated with these forecasts. In this study, we quantify and disentangle the uncertainty induced by initial conditions and meteorological inputs in water temperature forecasts by producing daily ensemble hindcasts of water temperature. The hindcasts are executed for lead-times of 1 to 5 days using a semi-distributed rainfall-runoff/water temperature model. A particle filter data assimilation algorithm is implemented to create multiple sets of improved hydrological and thermal states as initial conditions to the forecasts. Ensemble meteorological forecasts are then fed to the model, using the aforementioned sets of improved initial conditions, to produce ensemble water temperature forecasts. Preliminary results show a larger contribution of the initial conditions (mean ensemble spread = 1.1°C) than that of the meteorological inputs (mean ensemble spread = 0.6°C) to the overall uncertainty of the system for 1-day ahead forecasts. The contribution of the meteorological inputs increases with lead-time to reach a mean ensemble spread of 1.8°C for 5-day ahead forecasts. For the same lead-time, the uncertainty of the initial conditions alone leads to a mean ensemble spread of only 0.5°C. These results suggest that the initial thermal conditions predominantly impact thermal forecasts for shorter lead times while meteorological inputs uncertainties take over for longer lead times. This study provides insights regarding the contribution of two major uncertainty sources to the overall uncertainty in water temperature forecasts. It also demonstrates the ability of the particle filter to properly update the state variables of a thermal model. These results can be used in strategies developed to target the most relevant uncertainty sources in a particular decision making context.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2016
- Bibcode:
- 2016AGUFM.H33B1535O
- Keywords:
-
- 1630 Impacts of global change;
- GLOBAL CHANGEDE: 1632 Land cover change;
- GLOBAL CHANGEDE: 1871 Surface water quality;
- HYDROLOGYDE: 1894 Instruments and techniques: modeling;
- HYDROLOGY