Many current efforts aimed at climate change impact assessment and adaptation planning focus on water availability for both human populations and ecological systems (e.g. Trnka et al. 2012; Parmesan 2006). Projections of future climate scenarios from global climate models (GCMs) based on projected amounts and timing of precipitation and increases in air temperature are widely used in climate impact assessments (Girvetz et al. 2009). One of the goals of this study is to improve our understanding of the fate of precipitation in terrestrial ecosystems in the context of both historical and projected coupled climate-hydrology assessments. The three main pathways of precipitated water in a terrestrial system include the following: (1) returning to the air via evaporation and plant transpiration; (2) infiltrating subsurface into soils and potential recharge to aquifers; and (3) flowing “overland” to create runoff that feeds the flow of stream and river channel networks. These three terms represent the primary components of an all-purpose water balance that can be customized using site-specific data on topography, soils, and geology. Quantifying the relationships and tradeoffs between these pathways provides for much more detailed projections of the impacts of variability in water availability on ecosystems and their inhabitants. Although future climate change projections are variable due to uncertainties inherent to variable emissions scenarios and the range of available GCMs, mechanistic, process-based, hydrologic modeling informed by long-term empirical (measured) data sets can constrain the functional uncertainty of GCM-based future hydrology projections.
Reducing uncertainty in future climate-hydrology scenarios can be achieved by incorporating deterministic processes and empirically confirmed landscape characteristics into estimates of potential hydrologic outcomes. Validation of spatially explicit hydrologic models that quantify the water balance by comparing measured streamflow with model output is a promising approach to defining reasonable mechanistic relationships among climate, hydrology, and the landscape. These relationships can be calibrated using a historical baseline and then can be applied to assess future climate projections (Flint and Flint 2012a). The value of such a spatially validated mechanistic model is more robust projections for runoff and other components of the water balance under future climates. Effective ecological projections and planning in the face of climate change, especially in arid climates, now demand this level of hydrologic specificity (e.g. Marcarelli et al. 2010).
The scale of information needed by land and water managers is often finer than data generated by GCMs (Littell et al. 2012). Spatial downscaling resolves climate data to a spatial grain size that can be validated using watershed-based methods, applied to local landscapes, or analyzed across large regions. Downscaling is therefore a critical first step in developing estimates of water balance components for watersheds that are robust enough for use under current or future climates.
Fluctuations in runoff and recharge across multiple watersheds can be assumed to be monotypic or else variable in response to variable precipitation: in order to minimize uncertainty, there is a need for analysts to be able to model hydrologic cycles based on nearby conditions at the watershed scale. In addition, since relatively few watersheds are gaged, physically based models of hydrologic dynamics are often required to assess landscapes. For example, for the California hydrologic region, which includes all basins that drain into the state (Figure 1), there were approximately 1,700 streamgages in operation circa 2000, with periods of record ranging from 1 to 109 years; and only 1,400 with periods of record 5 years or greater. These streamgages represent less than a third of the 5,128 subwatersheds in California (http://www.cain.ice.ucdavis.edu/calwater/), which is presumably one of the better instrumented regions of the world.
California’s streamgage records display a wide variety of discharge dynamics, from flashy systems with high runoff peaks as a result of low permeability bedrock (such as granites in the Sierra Nevada) or large areas of impervious urban surfaces, to high baseflows with very permeable bedrock composition (e.g. volcanic rock; Flint and Flint 2007; Flint et al. 2011; Tromp-van Meerveld et al. 2007). The degree of climate aridity and soil type also affect potential hydrologic response to climate, with the deep unsaturated zones in arid regions or the deep soils of California’s Central Valley storing water when available from wet climate cycles that can be used as groundwater during dry periods (Flint and Flint 2007).
The objective of this paper is to document the development of a regional scale water-balance model that rigorously incorporates deterministic processes, and describe its application to the California hydrologic region at relatively fine spatial scales. The advantages of fine-scale application will be discussed. The calibration and validation of this model to measured streamflow provide confidence in the application of the model to both historical and future changes in hydrology as a result of climate that are described in a companion paper (Thorne et al. 2013). Further descriptions of the datasets discussed and the post-processing and availability of files are documented in Thorne et al. (2012). Previous versions of this model and applications to small regions or basins have been previously published and include Flint and Flint (2007, 2012a), Flint et al. (2011, 2012), and Micheli et al. (2012).
Hydrologic modeling background
Many approaches to hydrologic modeling have been developed. The U.S. Geological Survey (USGS) Precipitation-Runoff Modeling System (PRMS) is used to simulate flows under future climate conditions at the watershed scale (Leavesley et al. 1992; Hay et al. 2011). This approach requires daily temperature and precipitation values that are applied to individual watersheds and used in a deterministic, distributed-parameter setting (Risley et al. 2011). The Variable Infiltration Capacity model (VIC) is a spatially explicit physical hydrology model, generally run regionally at coarse spatial scales, that balances energy and water budgets (Liang et al. 1994) and also runs using daily data (Wood et al. 2002). This model has also been applied to monthly climate in a model comparison study by Mauer et al. (2010), who found that model selection was less important for capturing high flow timing, but that for the low flows, the models tested varied, implying a need to vet model performance, particularly for aridifying regions. These rainfall-runoff models are specifically calibrated to streamgage data.
Other hydrologic modeling approaches have used streamgage data to validate the model projections using current or historical data. Alkama et al. (2011) developed the Interactions between Soil, Biosphere, and Atmosphere-Total Runoff Integrating Pathways (ISBA-TRIP) and looked at multi-decadal variability in continental runoff from 1960–1994 using 154 large rivers with different lengths of streamgage data for validation. Chiew et al. (2010) found that five different downscaling techniques all reproduced observed rainfall, and runoff models used were capable of reproducing observed streamflows for eight basins in Australia. These efforts point to the need to understand the capacity and limitations of hydrologic models that are used for future projections.
All these rainfall-runoff models rely on soil storage in some capacity yet do not incorporate bedrock properties; thus, they neglect the influence of spatially varying bedrock permeability in estimates of recharge. Experimental evaluations of hillslope processes include a few that have investigated the influence of bedrock permeability on hydrologic response to climate (Hutchinson and Moore 2000; Tromp-van Meerveld et al. 2007), while a few others numerically modeled watersheds including bedrock properties (Flint and Flint 2006; Jones et al. 2008; Hopp and McDonnell 2009). Generally, these models are two- or three-dimensional, finite-element models that explicitly incorporate bedrock but are computationally intensive and cover small areas. Historically, recharge estimates have relied on monthly water balance models that incorporate simulations of evapotranspiration (Alley 1984), inverse modeling (Sanford et al. 2001), or lysimetry and tracer tests (Gee and Hillel 1988). Water-balance modeling to assess both recharge and runoff has been done at the site scale (Flint et al. 2002a; Ragab 1996) and integrated with various measurements addressing different spatial scales (Flint et al. 2002b). Watershed-scale or regional-scale modeling to estimate recharge and runoff has been done using water-balance modeling by Hevesi et al. (2003), Flint et al. (2011), and Flint and Flint (2007).
Evaluating hydrologic response to climate in California
We used the Basin Characterization Model (BCM) to model the hydrologic cycle for the California hydrologic region [Figure 1; modified from Hickman (1993)]. This paper presents results for two 30-year periods from 1951–2010 for all watersheds and by ecoregion for precipitation, air temperature, April 1st snowpack, recharge, runoff, potential evapotranspiration (PET), actual evapotranspiration, and climatic water deficit, a parameter that is a function of PET and actual evapotranspiration (Stephenson 1998).
To develop confidence in the application of the model this paper also evaluates the reliability of hydrologic model performance by comparing basin discharge, a product of the runoff and recharge values generated by the BCM with streamgage data. Historical streamgage data were assembled from 138 mostly unimpaired basins (Figure 1), along with reconstructed unimpaired flows from 21 additional basins, and monthly and yearly summaries from streamgages were used to test how well the BCM model outputs perform on watersheds with varying bedrock permeability, soil properties, impermeable surfaces, and degrees of aridity. The results of this model testing permit hydrologic simulation performance within the study area due to influences of landscape variables.
Description of the Basin Characterization Model (BCM)
The Basin Characterization Model (BCM) is a regional water balance model (Flint and Flint 2007; Thorne et al. 2012). The BCM (Figure 2) mechanistically models the pathways of precipitation into evapotranspiration, infiltration into soils, runoff, or percolation below the root zone to recharge groundwater. The evapotranspiration component is derived through the use of PET equations (Priestley and Taylor 1972) that rely on the calculation of solar radiation using slope, aspect, topographic shading, and atmospheric parameters. For the purposes of comparison across watersheds (or other landscape units), PET in the BCM is not interactive with the other segments. In other words, potential water demand from plants is independent from other hydrodynamic components in the model. The soil storage component of the model uses soil properties to calculate how much soil moisture is available for plant evapotranspiration. Soil storage is also independent from the other hydrologic dynamics, except that groundwater recharge, calculated as infiltration below the zone of evapotranspiration, is calculated only from surplus, after soil moisture capacity has been filled. Groundwater recharge (recharge) is also tied to runoff, and the relationship between the two is driven by the level of permeability of bedrock.
Therefore, the BCM can model the response of any given watershed to climate as driven by its energy balance (based on latitude, longitude, elevation, slope, and aspect), soil moisture storage capacity, and the characteristics of the materials that are deeper than the rooting zone, including deep alluvial valleys or bedrock that can permit percolation into groundwater. The BCM calculates hydrologic variables on a grid cell basis and can be run at any spatial resolution, generally limited by data resolution, computing power, or file storage capabilities. Grid cell values can be summarized for any spatial pattern, such as watersheds. A post-model calculation for basin discharge can be performed.
The BCM has several subroutines or modules: the calculation of potential and actual evapotranspiration and climatic water deficit; snow accumulation and melt; available water; and recharge and runoff (Figure 2). The model begins with climatic inputs of precipitation and air temperature. This is followed by the calculation of PET, which relies on an hourly energy-balance calculation, based on solar radiation, air temperature, and the Priestley-Taylor equation (Flint and Childs 1991). Clear sky PET is calculated using a solar radiation model that incorporates seasonal atmospheric transmissivity with site parameters of slope, aspect, and topographic shading (to define the percentage of sky seen for every grid cell) (Flint and Childs 1987). Hourly PET is aggregated into monthly time series, and cloudiness corrections are made on the basis of calibrations using cloudiness data from National Renewable Energy Laboratory (NREL; http://www.nrel.gov/; Flint and Flint 2008). Modeled PET for the southwest United States has been calibrated to measured PET from California Irrigation Management Information System (CIMIS) and Arizona Meteorological Network (AZMET) stations (Flint and Flint 2007).
Using PET and gridded precipitation, maximum and minimum air temperature, and the approach of the National Weather Service Snow-17 model (Anderson 1976), the snow module accumulates, sublimates, and melts snow to produce available water (Figure 2). These inputs to the water balance have been calibrated regionally to solar radiation and PET data, and snow cover estimates have been compared to Moderate Resolution Imaging Spectroradiometer (MODIS) snow cover maps (Flint and Flint 2007). This paper presents further snow module calibration work.
The BCM’s available water calculation quantifies water that is available for use in the remaining parts of the BCM, which balance watershed hydrologic components (Figure 2). Available water occupies the soil profile, where it will become actual evapotranspiration (AET), and may also result in runoff or recharge, depending on the soil storage and permeability of the underlying bedrock. Total soil-water storage is calculated as porosity multiplied by soil depth. Field capacity [soil water volume at −0.03 megapascals (MPa)] is the soil water volume below which gravity drainage is negligible, and wilting point (soil water volume at −1.5 MPa) is the soil water volume below which actual evapotranspiration does not occur (Hillel 1980). Once available water is calculated, it may exceed total soil storage and become runoff, or it may be less than total soil storage but greater than field capacity and become recharge. Anything less than field capacity is calculated as AET, at the rate of PET for that month, until it reaches wilting point. This permits the subsequent calculation of climatic water deficit (CWD).
When soil water is less than total soil storage and greater than field capacity, soil water greater than field capacity equals recharge. If recharge is greater than bedrock permeability (K), then recharge = K and excess becomes runoff, else it will recharge at K until field capacity is reached. Runoff and recharge are combined to calculate basin discharge, and actual evapotranspiration is subtracted from PET to calculate CWD.
The BCM can be used to identify locations and climatic conditions that generate excess water by quantifying the amount of water available either as runoff generated throughout a basin or as in-place recharge (Flint and Flint 2007). Because of the grid-based, simplified nature of the model, with no internal streamflow routing, long time series for very large areas can be simulated easily. However, if local unimpaired streamflow data are available, estimated recharge and runoff from each grid cell can be used to calculate basin discharge that can be extrapolated through time for varying climates. In addition, the application of the model across landscapes allows for grid-based comparisons between different areas. Because of the modular and mechanistic approach used by the BCM, it is flexible with respect to incorporating new input data or updating of algorithms should better calculations be derived. All input files necessary to operate the BCM, and the output files resulting from the simulations, are shown in Thorne et al. (2012; Appendix A). A complete list of all input and output variables and definitions is included in Thorne et al. (2012; Appendix B).