DEVELOPMENT AND APPLICATION OF A PARSIMONIOUS WATER BALANCE MODEL FOR MEDITERRANEAN CLIMATE CONDITIONS
- Author(s): Moran, Thomas C.
- Advisor(s): Hunt, James R
- et al.
The annual water balance is an important indicator of the hydrologic function and utility of a watershed, and yet there has been relatively sparse research of the special considerations that control the yearly partition of precipitation in a Mediterranean climate (MC) like that of California. In particular, there is a gap in empirical characterization of the annual water balance over a broad collection of watersheds spanning the diverse climate and landscape conditions of the state. This research develops and applies a top-down, parsimonious, physically interpretable water balance model that explicitly accounts for seasonality, a critical climate factor for MC regions.
The research was motivated by the observation of a straightforward, linear relationship between total annual precipitation and streamflow for watersheds in the Russian River Basin of northern California. A dataset of monthly water balance variables was developed to meet the criteria of accurate estimations, geographic contiguity, and temporal longevity, continuity, and consistency. Inspection of the long-term water balance for 159 watersheds in the state led to a more general form of the precipitation-streamflow relationship, a segmented linear model. Model parameters were estimated for each watershed via regression of water balance observations using a structural probabilistic model that was resilient to uncertainties in the input data.
Model parameter estimates displayed aggregate clustering by prevailing wetness conditions, as well as geographic regionalization. The average predictive uncertainty for gaged watersheds ranged from 50 to 125 mm per year in terms of area-normalized streamflow. Modeled streamflow residuals were used to evaluate historical changes in the water balance, revealing a decreasing trend in the streamflow of most California watersheds during the onset of the climate change era, controlling for precipitation. Sensitivity analysis showed that changes in the seasonality of precipitation and potential evapotranspiration have an order-of-magnitude larger impact on the water balance relative to other climate drivers. Spatial proximity correlation and watershed feature regression both showed promise as methods for estimation of model parameters in ungaged watersheds. The model was also contextualized with regards to the influential Budyko curve.
This research demonstrated that a parsimonious and interpretable model was capable of describing the annual water balance for the diverse hydrologic conditions across California. By focusing on the analysis of many watersheds over long timeframes it was possible to characterize and interpret broad trends and patterns that influence the water balance in MC regions.