Remotely sensed evapotranspiration to calibrate a lumped conceptual model: Pitfalls and opportunities

- New satellite ET data offers opportunities for model parameter and structure identifiability.
- Model performance reduces, adding MODIS ET as a calibration set, due to conceptual model structure.
- Streamflow and ET based model calibration requires understanding of water balance uncertainties.

SummaryPhysically representative hydrological models are essential for water resource management. New satellite evapotranspiration (ETobs) data might offer opportunities to improve model structure and parameter identifiability, if used as an independent calibration set. This study used a modelling experiment on 4 catchments in New South Wales, Australia, to investigate whether MODIS (16A3) ETobs can be used to improve parameter calibration for low parameter conceptual models. The catchment moisture deficit and exponential routing form of the model IHACRES was used to test calibration against streamflow, MODIS ETobs or a combination setoff the two. Results were compared against a regionalized parameter model and a model using MODIS ETobs directly as input. Firstly, the results indicated that the observed water balance of the catchments has, currently unexplained, large positive differences which impact the calibrated parameters. More generally, using MODIS ETobs as a calibration set, results in a reduction of the model performance as all residuals of the local water balance and timing differences between the water balance and the outflow need to be resolved by the routing component of the model. This is further complicated by variations in land cover affecting the MODIS ETobs. Finally this study confirms that the calibration of models using multiple environmental timeseries (such as MODIS ETobs and Q) can be used to identify structural model issues.