Impact of Rescaling Anomaly and Seasonal Components of Soil Moisture on Hydrologic Data Assimilation

Abstract:

In hydrological sciences many observations and model simulations have moderate linear association due to the noise in the datasets and/or the systematic differences between their seasonality components. This degrades the performance of model-observation integration algorithms, such as the Kalman Filter, that result in more accurate analysis when there is higher degree linear association between the observation and the model. To alleviate the impacts of the systematic differences, observations are linearly transformed via various rescaling techniques from observation to model space before they are assimilated into the models. In general such techniques consider rescaling either the entire timeseries or only the anomaly components of observations. In this study the impacts of rescaling seasonality and anomaly components of observations separately or together and the seasonality of these rescaling coefficients are investigated. Experiments are performed for both synthetic and real data cases. Both synthetic and real data assimilation experiments use an Antecedent Precipitation Index (API) model and assimilate observations in Kalman Filter framework, while real case assimilate LPRM observations over four USDA-ARS watersheds where the station-based observations are used to validate the analysis. Results show rescaling observations more strongly to the model is favorable when the model is more skillful than observations, while rescaling observations strongly to the model degrades the analysis if observations are relatively more accurate.