You are here:
Use of hydrologic landscape classification to diagnose streamflow predictability in Oregon
Citation:
Patil, S., J. Wigington, S. Leibowitz, AND R. Comeleo. Use of hydrologic landscape classification to diagnose streamflow predictability in Oregon. JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION. American Water Resources Association, Middleburg, VA, 50(3):762-776, (2014).
Impact/Purpose:
The ability to predict when, where, and how much water will be available for use is critical to the decision making of water resource managers. Increasingly, water resources planning and allocation decisions will have to be made in places where little or no past data exists, and the future effects of climate and land-use change on water availability are highly uncertain. Physics-based models of hydrology play a critical role in our understanding of how river basins function and provide a sound scientific basis for decision making under uncertainty. These hydrologic models are primarily used for future predictions of streamflow in river basins. However, research has shown that the prediction skill of hydrologic models differs from region to region. Unfortunately, the underlying causes for these differences are still poorly understood. This research investigates whether the new hydrologic classification system of landscapes, developed by scientists at EPA, can improve our understanding of where and why models fail to predict. Oregon Hydrologic Landscape (OHL) classification was used in conjunction with a spatially lumped hydrologic model to analyze the spatial patterns of streamflow predictability. We found that the model shows a tendency to perform poorly in regions classified with spring snowmelt-dominated flow regime, high aquifer permeability, and medium to high soil permeability. Our results indicate that poor characterization of the snow processes and difficulty in estimating external gains and losses of deep groundwater are the primary reasons for poor model predictability in Oregon. We will use these results to (1) develop a framework for improving the scientific soundness of hydrologic models, (2) provide estimates of the prediction skill at river basins with little or no past data, and (3) assess the potential impacts of climate and land-use change on streamflow predictability.
Description:
We implement a spatially lumped rainfall-runoff model to predict daily streamflow at 88 catchments within Oregon, USA and analyze its performance within the context of Oregon Hydrologic Landscapes (OHL) classification. OHL classification is used to characterize the physio-climatic conditions that potentially favor high (or low) streamflow predictability. Results show that high prediction catchments (Nash-Sutcliffe efficiency of (NS) > 0.75) are predominantly classified as rain dominated with very wet climate, low aquifer permeability, and low to medium soil permeability. Most of these catchments are located in western Oregon, west of the Cascades Mountain Range. Conversely, most low prediction catchments (NS < 0.6) are classified as snow dominated with high aquifer permeability and medium to high soil permeability. They are mainly located in the volcano-influenced regions near the High Cascades. Using a subset of 36 catchments, we further test whether class-specific model parameters can be developed for prediction at ungauged catchments. In most catchments, OHL class-specific parameters provide predictions that are on par with individually calibrated parameters (performance decline < 10%). However, large performance declines are observed in OHL classes where hydrologic predictability is not high enough to begin with. Results suggest that higher uncertainty in the rain-to-snow transition of precipitation phase and the difficulty in estimating external gains/losses of deep groundwater are major concerns for modeling at lower prediction catchments of Oregon. Moreover, regionalized estimation of model parameters appears to be more useful in regions where physio-climatic conditions favor good hydrologic predictability.