1. Introduction
Mesoscale regional climate models (RCMs) are recognized as an essential tool to address scientific issues concerning climate variability, changes, and impacts at regional to local scales. As the model resolution increases, land surface models (LSMs) coupled with RCMs need to incorporate more comprehensive physical processes and their nonlinear interactions. This has been the trend in recent developments (Stieglitz et al. 1997; Chen and Kumar 2001; Warrach et al. 2002; Niu and Yang 2003; Niu et al. 2005; Oleson et al. 2008; Choi and Liang 2010), but most LSMs still contain simplistic parameterizations that need improvements for terrestrial hydrologic processes. As a result, LSMs may produce nonlinear drifts in their dynamic responses to external forcings (e.g., Yuan and Liang 2011), which in turn feed back to the coupled climate system and ultimately lead to significant errors in predicting surface water and energy budgets. Improved parameterizations of key land surface processes, especially for the terrestrial hydrologic cycle, are needed to get better performance from RCMs.
Most current LSMs model only vertical moisture transport and therefore cannot capture spatial variability of soil water induced by topographic characteristics; thus, they are limited in predicting surface fluxes. Such one-dimensional models are unable to represent subsurface lateral transport induced by topography or moisture gradients. As one of the critical components in the terrestrial hydrologic cycle, runoff is estimated using the soil water budget without any explicit or parameterized routing treatment in most current models. As shown below, these models that disregard flow routing or runoff travel time over the basins predict surface runoff hydrographs with unrealistic sharp peaks and steep declining recessions. Moreover, ignoring the role of surface flow depth on the infiltration rate causes errors in both infiltration and surface flow calculations (Schmid 1989; Singh and Bhallamudi 1997; Wallach et al. 1997). Therefore, this study presents a numerical model based on a conjunctive solution of surface and subsurface flow interactions, typically applied for small-scale modeling, here specifically targeting for use in mesoscale land surface parameterizations over a continental scale to improve regional climate modeling, especially for the spatiotemporal distribution of surface and subsurface water that has significant influence on terrestrial water and energy budget.
To develop the improved runoff treatment in LSMs, this study has chosen the Common Land Model (CoLM) as the basic LSM that originally utilizes simplistic assumptions and crude parameterizations for the terrestrial hydrologic cycle, especially runoff processes as most LSMs do. The CoLM is an advanced soil–vegetation–atmosphere transfer model (Dai et al. 2003, 2004), which has been incorporated into the state-of-the-art mesoscale Climate–Weather Research and Forecasting (CWRF) model (Liang et al. 2005a,b,c,d, 2006, 2012) with numerous crucial updates and improvements for land processes (Choi 2006; Choi et al. 2007; Choi and Liang 2010; Yuan and Liang 2011). The original Community Land Model (CLM) and CoLM have been extensively evaluated for good performance against field measurements in a stand-alone mode as driven by the observational forcings (Dai et al. 2003; Niu and Yang 2003; Niu et al. 2005; Maxwell and Miller 2005; Qian et al. 2006; Niu and Yang 2006; Niu et al. 2007; Lawrence and Chase 2007; Lawrence et al. 2007; Oleson et al. 2008). Our own experience, however, has shown that a direct application of the CoLM for the CWRF at a 30-km grid spacing leads to serious problems in predicting the hydrologic cycle, especially runoff processes and basin discharge (Choi and Liang 2010; Yuan and Liang 2011).
A number of attempts have been made to couple explicit runoff schemes with LSMs. Walko et al. (2000) incorporated a modified form of Topography-based Hydrological Model (TOPMODEL) into the Land Ecosystem–Atmosphere Feedback model (LEAF-2) for land–surface processes in the Regional Atmospheric Modeling System (RAMS) in order to represent surface and subsurface downslope lateral transport of groundwater. Mölders and Rühaak (2002) coupled with an atmospheric model a hydrothermodynamic soil–vegetation scheme that incorporates surface and channel runoff. Since the runoff component was solved at a 1-km grid that differs from other terrestrial hydrologic schemes at a 5-km grid, the coupling using aggregation and disaggregation was required to introduce the runoff processes. Gochis and Chen (2003) developed a hydrologically enhanced form of the NOAH LSM incorporating a cell-to-cell surface flow routing scheme through disaggregated routing subgrids coupled to a quasi-steady state model for subsurface lateral flow. Maxwell and Miller (2005) added in the CoLM a variably saturated groundwater model ParFlow without explicit surface runoff scheme. As such, the model was mainly targeted to improve water table depth prediction, but it could not capture observed monthly runoff variations. Kollet and Maxwell (2006) incorporated an overland flow simulator into the ParFlow, followed by an evaluation against published data and an analytical solution for a V catchment (1.62 km × 1 km). Richter and Ebel (2006) developed a fully integrated atmospheric–ocean–hydrology model, the Baltic Integrated Model System (BALTIMOS), where the regional climate model REMO is coupled to the mesoscale hydrological model LARSIM for simulating the water balance of large river basins continuously. Fan and Miguez-Macho (2011) simulated lateral groundwater flow using estimates from three participants of the National Aeronautics and Space Administration's (NASA) Global Land Data Assimilation System (GLDAS) (Rodell et al. 2004) such as CLM, Mosaic, and Noah at 1-km spacing to improve water table depth prediction over North America.
Along with these attempts, we have been continuously seeking improved representations for the terrestrial hydrology in the CWRF. Choi (2006) and Choi et al. (2007) developed the 3D volume-averaged soil-moisture transport (VAST) model based on the Richards (1931) equation to incorporate the lateral flow and subgrid heterogeneity due to topographic characteristics introduced in the work of Kumar (2004). It was demonstrated that both the lateral flow and subgrid flux have important effects on total soil-moisture dynamics and spatial distribution. In general, soil-moisture moves from hillslopes to the lower regions by lateral and subgrid fluxes in the VAST model. Since the water would converge along topographic concave hollows, any water exceeding soil porosity needs to be transported to the vertical soil layers and then treated as lateral surface runoff. Hence, the improved terrestrial hydrologic scheme based on the VAST model also requires an explicit surface flow computation scheme. Choi (2006) incorporated a non-inertial diffusion wave model to account for the downstream backwater effect. This was an approximate form of the Saint-Venant (1871) equation, known for its efficiency in accuracy and computation (Ponce et al. 1978; Akan and Yen 1981; Hromadka et al. 1987; Morita and Yen 2002; Kazezyilmaz-Alhan et al. 2005). Choi and Liang (2010) incorporated the baseflow allocation scheme along with improved terrestrial hydrologic representations such as realistic bedrock depth, dynamic water table, exponential decay profile of the saturated hydraulic conductivity, minimum residual soil water, and maximum surface infiltration limit. This model, however, still simulates steep declining recession curves and relatively small total runoff, primarily because it underestimates the baseflow (Choi and Liang 2010). Therefore, this study has developed and implemented in the CoLM a new conjunctive surface–subsurface flow (CSSF) module that comprises an 1D diffusion wave model for the surface (overland) flow interacted with the 3D VAST model for the unsaturated subsurface flow and an 1D topographically controlled baseflow for the saturated subsurface flow, where all the components are designated at the 30-km grid scale. A new formulation is introduced for the baseflow to depict the effects of surface macropores and vertical hydraulic conductivity changes. The RCM grid-based overland flow routing process is introduced in this study for mesoscale LSMs to realistically predict the temporal variation of the spatial distribution of flow depth and runoff. Such coupling enables the CSSF to explicitly simulate surface runoff that results from both rainfall excess and moisture saturation in the whole soil column as well as its interactions with neighboring grids. This conjunctive flow model is solved by the mixed numerical implementation for each flow component and then substituted for the existing 1D hydrologic scheme in the CoLM.
Most hydrologic parameterization schemes in LSMs, including the CoLM and CLM, have been tested with field measurements at small catchment scales (Stieglitz et al. 1997; Warrach et al. 2002; Dai et al. 2003; Niu et al. 2005; Niu and Yang 2006), while directly applied in much coarser resolution climate models. Given the strong scale dependence, terrestrial hydrologic parameterizations, especially for runoff, must be tuned and evaluated at the same resolution as their host climate models. Therefore, we assess the new CSSF module as coupled with the current CoLM built in the CWRF for their designated application at the 30-km grid, focusing on the representation of surface and subsurface runoff. All schemes are explicitly implemented at the 30-km grids rather than hydrologic basins or catchments to avoid downscaling and upscaling exchanges between CWRF atmospheric forcings and land responses (Choi 2006; Choi et al. 2007; Choi and Liang 2010; Yuan and Liang 2011). This is an important advance to most previous LSMs that couple hydrologic and atmospheric schemes usually at different scales. Furthermore, the CSSF scheme with the scalable parameterization can be substituted for the terrestrial hydrologic scheme in most LSMs at any current and future finer spatial resolutions within the mesoscale range.
As demonstrated below, the CoLM+CSSF extension generates runoff variations much closer to observations. Section 2 presents a brief description of the original CoLM with our previous improvements. Section 3 elaborates the key features of the new CSSF parameterizations. Section 4 depicts the numerical implementation of the CSSF into the CWRF–CoLM at 30-km grids. Section 5 evaluates the CSSF skill enhancement in runoff processes and discharge predictions against daily observations at catchment-scaled study basins over the Ohio Valley. The final conclusions are given in section 6.
2. State of development of the CoLM
The original CoLM is well documented in Dai et al. (2003, 2004). Its major characteristics include a 10-layer prediction of soil temperature and moisture; a 5-layer prediction of snow processes (mass, cover, and age); an explicit treatment of liquid and ice water mass and their phase change in soil and snow; and a two-big-leaf model for canopy temperature, photosynthesis, and stomatal conductance. In coupling with the CWRF, the CoLM was consistently integrated with comprehensive surface boundary conditions (Liang et al. 2005a,b) and an advanced dynamic–statistical parameterization of land surface albedo (Liang et al. 2005c).
Recently, Choi and Liang (2010) identified several deficiencies in the CoLM formulations for terrestrial hydrologic processes and developed better solutions with a focus on stream discharge predictions. In particular, they have incorporated a realistic geographic distribution of bedrock depth to improve estimates of the actual soil water capacity, replaced an equilibrium approximation of the water table with a dynamic prediction to produce more reasonable variations of the saturated zone depth, used an exponential decay function with soil depth for the saturated hydraulic conductivity to consider the effect of macropores near the ground surface, formulated an effective hydraulic conductivity of the liquid part at the frozen soil interface and imposed a maximum surface infiltration limit to eliminate numerically generated negative or excessive soil moisture solution, and considered an additional contribution to subsurface runoff from saturation lateral runoff or baseflow controlled by topography.
The above improvements enable the CoLM to more realistically reproduce observations of terrestrial hydrologic quantities, where the skill enhancement is especially significant for runoff at peaks discharges under high-flow conditions. For convenience, the original model with these improvements is referred to as the CoLM in the subsequent sections, unless specifically noted otherwise. However, this model, using the current TOPMODEL equation, still underestimates the baseflow and thus results in steep declining recession curves and relatively small total runoff (Choi and Liang 2010). Therefore, we develop the CSSF approach below to further reduce these CoLM deficiencies so identified.
3. New CSSF parameterizations
Figure 1 illustrates the key terrestrial hydrologic processes represented in the current CoLM and the new CSSF module. Below we describe the new parameterizations introduced into the CSSF, including the treatments for various runoff components as shown in Fig. 2a and the surface flow routing scheme as illustrated in Fig. 2b. The CSSF also couples the VAST model to explicitly incorporate additive lateral and subgrid moisture transport fluxes due to local variations of topographic attributes.
Schematic diagram for the key terrestrial hydrologic processes represented in the current CoLM and the new CSSF parameterizations.
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
(a) Schematic diagram for surface runoff and subsurface runoff components at a single grid cell and (b) concept sketch for the developed conjunctive flow model for the four horizontal grid cells with multiple soil layers in the new CoLM+CSSF model.
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
a. Soil-moisture transport scheme
b. Subsurface runoff representation scheme
1) Baseflow runoff
The subsurface lateral flow mainly driven by complex terrain is collecting water along lower-valley regions. The lateral subsurface flow from the VAST model is limited to the vadose zone, the so-called interflow or throughflow, because it is modeled by the 
2) Bottom drainage runoff
3) Interflow runoff
4) Saturation excess runoff
c. Surface runoff representation scheme
1) Surface runoff
2) Surface flow routing
As delineated in Fig. 2b, in the hydrologically enhanced version of the CoLM, the surface flow equation in Eq. (23) relies on the exchange flux 
d. Total runoff representation scheme
4. Implementation of the new CoLM+CSSF model
The CSSF model is substituted for the existing terrestrial hydrologic representation in the CoLM. While the CoLM performs all computations independently at individual soil columns, the lateral subsurface flow and surface flow in the CSSF depend on neighboring grids and hence are calculated after the vertical subsurface water movement is computed using a time-splitting method.
Since the performance of the terrestrial hydrologic schemes depends strongly on the spatial scale, the CSSF skill enhancement in predicting runoff is evaluated over a relatively large catchment using the actual CWRF–CoLM 30-km grid mesh targeted for regional climate applications. To facilitate model comparison with observations, the experiments are conducted in a stand-alone mode, where the CoLM with or without the CSSF is driven by the most realistic surface boundary conditions (SBCs) and meteorological forcings. This avoids the complication from errors of atmospheric processes and surface–atmospheric feedbacks in the fully coupled CWRF. To implement the CSSF into the CoLM, a mixed numerical integration approach is adopted for different flow components. The 3D VAST is integrated using a time-splitting algorithm by separating the vertical and lateral components. An explicit method solves the lateral flow after a fully implicit method solves the vertical flow. The 1D diffusion wave model is solved by the MacCormack (1971) scheme with second-order accuracy in both space and time. The evaluation procedures on the CoLM+CSSF simulations are described below.
a. Study catchments
To appropriately evaluate the performance of the CSSF in the CoLM, we select three catchments around the Ohio Valley within the CWRF U.S. domain. These basins have observed records of streamflow discharges from the U.S. Geological Survey (USGS) National Water Information System (http://waterdata.usgs.gov/nwis/sw), and each contains the headwater of the stream. We choose one gauge station near each basin outlet: Kanawha River at Charleston, West Virginia (03198000), Kentucky River at Lock 4 at Frankfort, Kentucky (03287500), and Green River at Lock 2 at Calhoun, Kentucky (03320000). Figure 3 marks their locations while Table 1 gives more specifications. Figure 3 also illustrates the portion of the CWRF computational domain for U.S. climate applications (Liang et al. 2004) that covers the entire three study catchments. It contains a rectangular size of 690 km (23 grid cells) by 360 km (12 grid cells) at a 30-km spacing.
Plots of 30-km resolved flow directions (arrows), 1-km HYDRO1K stream network (curved lines), basin boundaries (black closed curves), model basin grid boundaries (white polygons), and the three USGS streamflow gauge stations (white circles) overlaid with spatial distributions of (a) the 30-km terrain elevation (black–gray gradation pixels) and (b) the 30-km USGS land cover type (gray-toned pixels), along with (c) a background map for study basin locations.
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
The selected USGS streamflow gauge stations for evaluation of the model performance. The drainage area for each station is documented by the USGS and the computational drainage grids are determined by 30-km flow directions for each of the three study basins.
b. Surface boundary conditions
The CoLM, as coupled with the CWRF, incorporates the most comprehensive SBCs based on the best observational data over North America constructed by Liang et al. (2005a,b). These include surface topography, bedrock depth, sand and clay fraction profiles, surface albedo localization factor, surface characteristic identification, land cover category, fractional vegetation cover, and leaf and stem area index for the 30-km grid scale constructed from raw data at the finest possible resolution. The spatial distributions of terrain elevation and land cover types are shown in Fig. 3, all at the CWRF 30-km grids, with major features summarized in Table 2. Figures 3a and 3b depict the distributions at the CWRF 30-km grid over the three study catchments for terrain elevation ranging from 140 to 991 m and land cover types consisting of cropland–woodland mosaic, deciduous broadleaf forest, and mixed forest. The bedrock depth distribution over these catchments, ranging from 0.87 to 4.97 m, is illustrated in Table 2.
Summary of terrain elevation, bedrock depth ranges, and major land use types for the three study basins.
The new CSSF model requires additional SBCs for each 30-km grid, such as standard deviations of subgrid local terrain slopes in the two horizontal coordinate directions, and eight surface flow directions to incorporate the topographic effects on soil moisture transport as well as lateral surface and subsurface flows. These fields were also constructed from the same 1-km DEM data. Following Lear et al. (2000), the double maximum algorithm (DMA) based on the eight-direction pour-point model is used to determine the flow direction that most realistically represents the dominant direction of the river network within the 30-km grid box. The maximum flow accumulation is first calculated from the 1-km DEM, and then the DMA uses a unique division of the 30-km grid box into four subsections generated from two offset 30-km meshes overlaid with the 1-km grids. Finally, the DMA extracts river networks from the 1-km resolution and upscales them to the CWRF 30-km computation grid resolution. Figure 3 shows that the so-derived 30-km flow directions represent well the feature of the Global Hydrological 1 kilometer database (HYDRO1K) stream network. See Liang et al. (2005a,b) and Choi (2006) for the details of data sources and construction methods for SBCs.
c. Meteorological forcings and initial conditions
The stand-alone simulations of the CoLM with and without the CSSF model are driven by the same meteorological forcings constructed from the best available observational North American Regional Reanalysis (NARR; Mesinger et al. 2006). The NARR adopts a 32-km grid, close to that of CWRF, and provides 3-hourly atmospheric and land data over an extensive area that completely includes our U.S. computational domain. The outcome represents a major improvement upon the earlier global reanalysis datasets in both resolution and accuracy. The required atmospheric variables to drive the CoLM stand-alone simulations are pressures at the lowest atmospheric layer and the surface (Pa), temperature at the lowest atmospheric layer (K), specific humidity at the lowest atmospheric layer (kg kg−1), zonal and meridional winds at the lowest atmospheric layer (m s−1), the lowest atmospheric layer height (m), convective and resolved rainfalls [mm (3 h)−1], snow [mm (3 h)−1], planetary boundary layer height (m), and downward longwave and shortwave radiations onto the surface (W m−2). They are remapped onto the CWRF 30-km grid by linear spatial interpolation. Note that the NARR data for soil temperature and moisture are given only in four layers of 0–10, 10–40, 40–100, and 100–200 cm below the surface. Mass conservative vertical interpolation for the 11 soil layers in the CoLM along with a conventional LSM spinup strategy is adopted to initialize the CoLM. Specifically, the CoLM integration is started at 0000 UTC on 1 January 1995 and run continuously throughout the whole year of 1995 as driven by the NARR data. This is repeated for five cycles with the same 3-hourly NARR forcings of 1995. The resulting conditions at the end of the fifth cycle are considered to be fully consistent with the atmospheric forcings and hence used as the initial conditions for the subsequent CoLM simulation to be evaluated against observations.
5. Runoff simulation results
We have first examined the sensitivity of the existing CoLM modified by Choi and Liang (2010) to two calibration parameters, the decay factor f (2–10 m−1) and the maximum baseflow coefficient 
Comparison of (top) the benchmark efficiency and (bottom) correlation coefficient with the decay factor f and the maximum baseflow coefficient 
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
The sensitivity analysis of the new CoLM coupled with the CSSF is performed for the two cases, the CoLM+CSSF simulations with and without the new baseflow scheme. When the new baseflow scheme is coupled, calibration is done for the two parameters: the decay factor f (2–10 m−1) and the anisotropic ratio 
Comparison of the (top panels for each 
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
Table 3 summarizes the BE and R scores for total runoff from the CoLM and CLM+CSSF models with each selected calibration parameter set. When the CSSF is coupled, the results are significantly improved where much higher values are obtained for both scores than the CoLM results. Figure 6 compares the time series of daily specific discharges (per unit drainage area) during 1995 observed and simulated by the baseline CoLM (Dai et al. 2003) and the improved CoLM (Choi and Liang 2010) with–without coupling the CSSF at the three gauge stations in the study domain. The original and current CoLM without CSSF schemes produce discharges as pulse fluctuations as a result of quick response to rainfall events, causing no recession time and overall underestimation of runoff, whereas the hydrologically enhanced version of the CoLM with the CSSF captures the seasonal variability of streamflow quite realistically.
Comparison of the model performance measured by the benchmark efficiency BE and the correlation coefficient R of the modified CoLM and the new CoLM+CSSF models with each selected calibration parameter set for the three study river basins based on the simulation result in 1995.
Comparison of daily time series of model simulated specific discharges of the total runoff from the baseline model (the original CoLM) by Dai et al. (2003), the CoLM modified by Choi and Liang (2010), and the new CoLM+CSSF in this study, along with the daily observations from the three USGS gauge stations in 1995. The hyetographs of the observed total precipitation are plotted along the right-hand vertical axis.
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
With each set of the selected calibration parameters, the CoLM uncoupled and coupled with the CSSF were run continuously for 5 yr from January 1995 to December 1999. As shown in Fig. 7, the runoff simulated by the CoLM alone shows an overestimate for the peaks but an underestimate in recession periods, while the coupling with the CSSF results in a significant improvement by capturing observations much more closely over the three study catchments. It is also clear that the coupling with the CSSF significantly improves the CoLM performance in representing the basin flow dynamics. Figure 7 and Table 4 demonstrate that the CoLM+CSSF incorporating the role of surface flow depth contribution to infiltration results in shallower water table depth and enhanced baseflow generation. In the CoLM+CSSF the routed surface flow depth and the large f value allow more water recharged to soil layers to raise the water table and increase subsurface flow.
Comparison of daily time series of the specific discharges and water table depth (
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
Comparison of 5-yr-averaged simulation results from the CoLM alone and the new CoLM+CSSF models along with observations for the three study river basins. that the quantity 
The CoLM predicts very little subsurface runoff, even after incorporating more advanced representations, including realistic bedrock depth, dynamic water table, exponential decay profile of the saturated hydraulic conductivity, minimum residual soil water, and maximum surface infiltration limit (Choi and Liang 2010). In general, when total runoff is dominated by its surface component, LSMs tend to overestimate runoff peaks and underestimate runoff recession. This may partially explain why the CoLM simulates extremely weak seasonal–interannual soil moisture variability (Yuan and Liang 2011). On the other hand, the CoLM coupled with the CSSF produces much larger subsurface runoff by incorporating the effects of surface flow depth and surface macropores on the baseflow generation. The ratios of subsurface to total runoff increase from 62.0%, 33.9%, and 50.3% to 89.2%, 65.5%, and 57.4%, respectively, for the Kanawha River, the Kentucky River, and the Green River basins (Table 4). The study demonstrates that the baseflow generation is extremely important for capturing streamflow observations. Note that baseflow runoff is comparable to total subsurface runoff in the CoLM with the CSSF because other subsurface components are confined by the use of the bedrock depth and the supersaturation prevention scheme. Figure 8 illustrates that the baseflow is predominant in low-flow seasons while surface runoff is more important during the high-flow season, especially May–June. One exception is for the Kanawha River basin, where no considerable storm events occurred over that period. Therefore, the new CSSF parameterizations that enforce interactions between the routed surface flow and the subsurface flow with topographic subgrid soil-moisture variation and baseflow generation simulate total runoff more realistically, especially in the recession part of the hydrograph. Relative to observations for the Kanawha River, the Kentucky River, and the Green River basins, the 5-yr-averaged runoff is only 20.4%, 58.4%, and 48.1%, respectively, in the CoLM, but increased to 81.0%, 84.2%, and 106.9% in the CoLM+CSSF (Table 4 and Fig. 9). However, this hydrologically enhanced version of the CoLM still cannot capture the recessions in the early spring flooding events. As shown in Fig. 9, the simulated monthly flow volumes for each of the study basins are in general less than observations during February–April. One possible reason is that the new model does not yet include aquifer recharge, regulation storage, deep aquifer groundwater flow, and channel flow routing. In addition, the snowmelt scheme may also contribute to the model–observation discrepancy (Yuan and Liang 2011). These factors warrant further investigation.
Modeled 5-yr-averaged climatology of monthly total model runoff consisting of surface runoff and baseflow components in the new CoLM+CSSF model simulation for (top to bottom) the three study river basins.
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
Comparison of 5-yr-averaged climatology of monthly total runoff from the CoLM and the new CoLM+CSSF models along with observations for (top to bottom) the three study river basins.
Citation: Journal of Hydrometeorology 14, 5; 10.1175/JHM-D-12-0168.1
6. Conclusions and summary
Most existing LSMs predict soil moisture transport only in the vertical direction and estimate surface runoff from local net water flux (precipitation minus surface evapotranspiration and soil-moisture storage). However, we found that subsurface subgrid and lateral fluxes have significant impacts on soil-moisture spatial variability (Choi et al. 2007), and an explicit surface flow routing scheme is required for the comprehensive terrestrial hydrologic cycling in LSMs (Choi 2006; Choi et al. 2007; Choi and Liang 2010). They should be incorporated in a fully interactive manner to affect the hydrologic cycle both locally and in adjacent areas. To this end, we have developed the CSSF module that comprises the 1D diffusion wave surface flow model coupled with the 3D VAST subsurface flow model and the 1D topographically controlled baseflow to substitute the existing hydrologic module in the CoLM. The CSSF is implemented into the CoLM by a time-split mixed numerical approach, where the subsurface flow model is separated for the vertical component by an implicit differencing method and the lateral component by an explicit method, and the surface flow model is solved by the MacCormack scheme. The implementation and verification of the CSSF in coupling with the CoLM is unique as compared with the conventional approach in that the hydrological and atmospheric components interact directly at the identical mesoscale grid mesh without aggregation or disaggregation.
The model performance is evaluated using stand-alone simulations of the CoLM with and without the CSSF driven by realistic SBCs and reliable NARR climate forcing data around the Ohio Valley on the same 30-km grid over the U.S. domain of the coupling CWRF to be applied. The CSSF incorporates advanced representations for the lateral and subgrid soil-moisture transport, the surface flow routing and interaction with subsurface flow, and the topographically controlled baseflow. As a result the coupling with the CSSF simulates the total runoff much closer to observations than the CoLM alone, especially for the declining recession curves of hydrographs. The CSSF so developed demonstrates a significant surface flow depth contribution to infiltration causing enhanced baseflow generation. Although the soil-moisture simulation performance is not directly evaluated due to lack of observations in the study basins, Yuan and Liang (2011) have also demonstrated the superiority of the CSSF in simulating the observed Illinois soil moisture variations from an offline test against the CoLM and CLM. The terrestrial water dynamics by full interactions between surface and subsurface flow in the hydrologically enhanced version of the CoLM with the CSSF represents an important advance to the simple soil water budget used in the original and our modified CoLM (Choi and Liang 2010). Ignoring these new processes in the CoLM can cause significant model errors and, consequently, unrealistic model parameters targeted for calibration. The redistribution of surface and subsurface water by the new model may also have a large impact on the prediction of the surface energy balance as well. Note that the original CoLM has been demonstrated by numerous studies for its good performance in global climate models (see the introduction in section 1). We contend that a resolution increase must couple with advanced model physics representation at that scale to realize improved predictions. The newly developed CSSF model provides a suite of improved modeling capability for the CoLM to better characterize surface water and energy fluxes crucial to climate variability and change studies at regional–local scales.
The CSSF model, albeit with many new advances, has yet to incorporate a more complete list of important factors for comprehensive terrestrial hydrologic simulations over larger basins, including aquifer recharge, regulation storage, consumptive use, and channel and groundwater flow routing across the basin boundaries. In the current CSSF mode, the overland-based surface flow scheme cannot fully capture the storage effect of real streams, and the topographically controlled baseflow scheme associated with the water table depth neglects the possible contribution from the deeper aquifer underlying the bottom of the LSM soil column. These areas will be our future targets to improve.
Acknowledgments
The research was supported by the NOAA Education Partnership Program (EPP) COM Howard 00073421000037534, Climate Prediction Program for the Americas (CPPA) NA11OAR4310194 and NA11OAR4310195, Environmental Protection Agency RD83418902, National Science Foundation ATM-0628687, and the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (NRF-2013R1A2A2A01008881).
APPENDIX
Soil Hydraulic Conductivity
To improve estimates of the actual soil water capacity, we adopt the geographically distributed bedrock depth profiles as constructed at the CWRF 30-km grid by Liang et al. (2005b). To approximate the water drainage through bedrocks, we assume the hydraulic properties of bedrocks whose porosity is 0.05 and saturated hydraulic conductivity is 1% of that in the soil layer right above, as similarly introduced in many LSMs (Abramopoulos et al. 1988; Xue et al. 1991; Boone and Wetzel 1996; Sellers et al. 1996).
REFERENCES
Abramopoulos, F., Rosenzweig C. , and Choudhury B. , 1988: Improved ground hydrology calculations for Global Climate Models (GCMs): Soil water movement and evapotranspiration. J. Climate, 1, 921–941.
Akan, A. O., and Yen B. C. , 1981: Diffusion-wave flood routing in channel networks. J. Hydraul. Eng., 107, 719–732.
Beven, K. J., 1982a: Macropores and water flow in soils. Water Resour. Res., 18, 1311–1325.
Beven, K. J., 1982b: On subsurface stormflow: An analysis of response times. Hydrol. Sci. J., 27, 505–521.
Beven, K. J., 1984: Infiltration into a class of vertically non-uniform soils. Hydrol. Sci. J., 29, 425–434.
Beven, K. J., and Kirkby M. J. , 1979: A physically based, variable contributing area model of basin hydrology. Hydrol. Sci. Bull., 24, 43–69.
Bonan, G. B., 1996: A land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and user's guide. NCAR Tech. Note NCAR/TN-417+STR, 150 pp. [Available online at ftp://ftp.daac.ornl.gov/data/model_archive/LSM/lsm_1.0/comp/NCAR_LSM_Users_Guide.pdf.]
Boone, A., and Wetzel P. J. , 1996: Issues related to low resolution modeling of soil moisture: Experience with the PLACE model. Global Planet. Change, 13, 161–181.
Brooks, R. H., and Corey A. T. , 1964: Hydraulic properties in porous media. Hydrology Paper No. 3, Colorado State University, 27 pp. [Available online at http://www.wipp.energy.gov/library/cra/2009_cra/references/Others/Brooks_Corey_1964_Hydraulic_Properties_ERMS241117.pdf.]
Chen, J., and Kumar P. , 2001: Topographic influence of the seasonal and interannual variation of water and energy balance of basins in North America. J. Climate, 14, 1989–2014.
Choi, G. W., and Molinas A. , 1993: Simultaneous solution algorithm for channel network modeling. Water Resour. Res., 29, 321–328.
Choi, H. I., 2006: 3-D volume averaged soil-moisture transport model: A scalable scheme for representing subgrid topographic control in land-atmosphere interactions. Ph.D. dissertation, University of Illinois at Urbana-Champaign, 189 pp.
Choi, H. I., and Liang X.-Z. , 2010: Improved terrestrial hydrologic representation in mesoscale land surface models. J. Hydrometeor., 11, 797–809.
Choi, H. I., Kumar P. , and Liang X.-Z. , 2007: Three-dimensional volume-averaged soil moisture transport model with a scalable parameterization of subgrid topographic variability. Water Resour. Res., 43, W04414, doi:10.1029/2006WR005134.
Cosby, B. J., Hornberger G. M. , Clapp R. B. , and Ginn T. R. , 1984: A statistical exploration of the relationships of soil moisture characteristics to the physical properties of soils. Water Resour. Res., 20, 682–690.
Dai, Y., and Coauthors, 2003: The common land model. Bull. Amer. Meteor. Soc., 84, 1013–1023.
Dai, Y., Dickinson R. E. , and Wang Y.-P. , 2004: A two-big-leaf model for canopy temperature, photosynthesis, and stomatal conductance. J. Climate, 17, 2281–2299.
Elsenbeer, H., Cassel D. K. , and Castro J. , 1992: Spatial analysis of soil hydraulic conductivity in a tropical rainforest catchment. Water Resour. Res., 28, 3201–3214.
Entekhabi, D., and Eagleson P. S. , 1989: Land surface hydrology parameterization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate, 2, 816–831.
Fan, Y., and Miguez-Macho G. , 2011: A simple hydrologic framework for simulating wetlands in climate and earth system models. Climate Dyn., 37, 253–278.
Freeze, R. A., and Cherry J. A. , 1979: Groundwater. Prentice-Hall, 604 pp.
Gochis, D. J., and Chen F. , 2003: Hydrological enhancements to the community Noah Land Surface Model. NCAR Scientific Tech. Rep., 77 pp. [Available online at http://nldr.library.ucar.edu/collections/technotes/asset-000-000-000-516.pdf.]
Hromadka, T. V., II, McCuen R. H. , and Yen B. C. , 1987: A comparison of overland flow hydrograph models. J. Hydraul. Eng., 113, 1422–1440.
Jha, R., Herath S. , and Musiake K. , 2000: River network solution for a distributed hydrological model and applications. J. Hydrol. Processes, 14, 575–592.
Kazezyilmaz-Alhan, C. M., Medina C. C. Jr., and Rao P. , 2005: On numerical modeling of overland flow. Appl. Math. Comput., 166, 724–740.
Kollet, S. J., and Maxwell R. M. , 2006: Integrated surface-groundwater flow modeling: a free-surface overland flow boundary condition in a parallel groundwater flow model. Adv. Water Resour., 29, 945–958.
Kumar, P., 2004: Layer averaged Richard's equation with lateral flow. Adv. Water Resour., 27, 521–531.
Kumar, P., Verdin K. L. , and Greenlee S. K. , 2000: Basin level statistical properties of topographic index for North America. Adv. Water Resour., 23, 571–578.
Lawrence, D. M., Thornton P. E. , Oleson K. W. , and Bonan G. B. , 2007: The partitioning of evapotranspiration into transpiration, soil evaporation, and canopy evaporation in a GCM: Impacts on land–atmosphere interaction. J. Hydrometeor., 8, 862–880.
Lawrence, P. J., and Chase T. N. , 2007: Representing a new MODIS consistent land surface in the Community Land Model (CLM3.0). J. Geophys. Res., 112, G01023, doi:10.1029/2006JG000168.
Lear, M., Olivera F. , Famiglietti J. S. , and Maidment D. R. , 2000: Upscaling river network extractions from global digital elevation models. CRWR Online Rep. 00-6, University of Texas at Austin, 110 pp. [Available online at http://www.crwr.utexas.edu/reports/pdf/2000/rpt00-6.pdf.]
Liang, X.-Z., Li L. , Kunkel K. E. , Ting M. , and Wang J. X. L. , 2004: Regional climate model simulation of U.S. precipitation during 1982–2002. Part I: Annual cycle. J. Climate, 17, 3510–3528.
Liang, X.-Z., Choi H. I. , Kunkel K. E. , Dai Y. , Joseph E. , Wang J. X. L. , and Kumar P. , 2005a: Development of the regional climate-weather research and forecasting model (CWRF): Surface boundary conditions. ISWS Scientific Rep. 2005-01, Illinois State Water Survey, 32 pp. [Available online at http://www.isws.illinois.edu/pubdoc/sr/iswssr2005-01.pdf.]
Liang, X.-Z., Choi H. I. , Kunkel K. E. , Dai Y. , Joseph E. , Wang J. X. L. , and Kumar P. , 2005b: Surface boundary conditions for mesoscale regional climate models. Earth Interact., 9, doi:10.1175/EI151.1.
Liang, X.-Z., and Coauthors, 2005c: Development of land surface albedo parameterization bases on Moderate Resolution Imaging Spectroradiometer (MODIS) data. J. Geophys. Res., 110, D11107, doi:10.1029/2004JD005579.
Liang, X.-Z., Xu M. , Zhu J. , Kunkel K. E. , and Wang J. X. L. , 2005d: Development of the regional climate-weather research and forecasting model (CWRF): Treatment of topography. Proc. 2005 WRF/MM5 User's Workshop, Boulder, CO, NCAR, 9.3. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2005/abstracts/Session9/3-Liang.pdf.]
Liang, X.-Z., and Coauthors, 2006: Development of the regional Climate-Weather Research and Forecasting model (CWRF): Treatment of subgrid topography effects. In Proc. 7th Annual WRF User's Workshop, Boulder, CO, NCAR, 7.3. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2006/abstracts/Session07/7_3_Liang.pdf.]
Liang, X.-Z., and Coauthors, 2012: Regional Climate–Weather Research and Forecasting Model (CWRF). Bull. Amer. Meteor. Soc., 93, 1363–1387.
MacCormack, R. W., 1971: Numerical solution of the interaction of a shock wave with a laminar boundary layer. Proceedings of the Second International Conference on Numerical Methods in Fluid Dynamics, Lecture Notes in Physics, Vol. 8, Springer-Verlag, 151–163.
Mahrt, L., and Pan H. , 1984: A two-layer model of soil hydrology. Bound.-Layer Meteor., 29, 1–20.
Maxwell, R. M., and Miller N. L. , 2005: Development of a coupled land surface and groundwater model. J. Hydrometeor., 6, 233–247.
Mesinger, F., and Coauthors, 2006: North American regional reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360.
Mölders, N., and Rühaak W. , 2002: On the impact of explicitly predicted runoff on the simulated atmospheric response to small-scale land-use changes—An integrated modeling approach. Atmos. Res., 63, 3–38.
Morita, M., and Yen B. C. , 2002: Modeling of conjunctive two-dimensional surface-three-dimensional subsurface flows. J. Hydraul. Eng.,128, 184–200.
Niu, G.-Y., and Yang Z.-L. , 2003: The Versatile Integrator of Surface and Atmosphere processes (VISA) Part 2: Evaluation of three topography based runoff schemes. Global Planet. Change, 38, 191–208.
Niu, G.-Y., and Yang Z.-L. , 2006: Effects of frozen soil on snowmelt runoff and soil water storage at a continental scale. J. Hydrometeor., 7, 937–952.
Niu, G.-Y., Yang Z.-L. , Dickinson R. E. , and Gulden L. E. , 2005: A simple TOPMODEL-based runoff parameterization (SIMTOP) for use in GCMs. J. Geophys. Res., 110, D21106, doi:10.1029/2005JD006111.
Niu, G.-Y., Yang Z.-L. , Dickinson R. E. , Gulden L. E. , and Su H. , 2007: Development of a simple groundwater model for use in climate models and evaluation with Gravity Recovery and Climate Experiment data. J. Geophys. Res., 112, D07103, doi:10.1029/2006JD007522.
Oleson, K. W., and Coauthors, 2008: Improvements to the Community Land Model and their impact on the hydrological cycle. J. Geophys. Res., 113, G01021, doi:10.1029/2007JG000563.
Ponce, V. M., Li R.-M. , and Simons D. B. , 1978: Applicability of kinematic and diffusion models. J. Hydraul. Eng., 104, 353–360.
Qian, T., Dai A. , Trenberth K. E. , and Oleson K. W. , 2006: Simulation of global land surface conditions from 1948 to 2004: Part I: Forcing data and evaluations. J. Hydrometeor., 7, 953–975.
Richards, L., 1931: Capillary conduction of liquids through porous mediums. Physics, 1, 318–333.
Richter, K.-G., and Ebel M. , 2006: Analysis of runoff for the Baltic basin with and integrated Atmospheric-Ocean-Hydrology Model. Adv. Geosci., 9, 31–37.
Rodell, M., and Coauthors, 2004: The global land data assimilation system. Bull. Amer. Meteor. Soc., 85, 381–394.
Saint-Venant, B., 1871: Theorie du mouvement non-permanent des eaux avec application aux crues des rivieres at a l'introdution des marees dans leur lit. Acad. Sci. Paris C. R., 73, 147–154.
Schaefli, B., and Gupta H. , 2007: Do Nash values have value? Hydrol. Processes, 21, 2075–2080, doi:10.1002/hyp.6825.
Schmid, B. H., 1989: On overland flow modelling: Can rainfall excess be treated as independent of flow depth? J. Hydrol., 107, 1–8.
Sellers, P. J., Los S. O. , Tucker C. J. , Justice C. O. , Dazlich D. A. , Collatz G. J. , and Randall D. A. , 1996: A revised land surface parameterization (SiB2) for atmospheric GCMs. Part II: The generation of global fields of terrestrial biophysical parameters from satellite data. J. Climate, 9, 706–737.
Sevuk, A. S., and Yen B. C. , 1973: Comparison of four approaches in routing flood wave through junctions. Proc. 15th Int. Assoc. of Hydraulic Engineering and Research World Congress, Istanbul, Turkey, IAHR, 169–172.
Singh, V., and Bhallamudi S. M. , 1997: A complete hydrodynamic border-strip irrigation model. J. Irrig. Drain. Eng.,122, 189–197.
Sivapalan, M., Beven K. J. , and Wood E. F. , 1987: On hydrologic similarity: 2. A scaled model of storm runoff production. Water Resour. Res., 23, 2266–2278.
Stieglitz, M., Rind D. , Famiglietti J. , and Rosenzweig C. , 1997: An efficient approach to modeling the topographic control of surface hydrology for regional modeling. J. Climate, 10, 118–137.
Tsai, C. W.-S., and Yen B. C. , 2001: On noninertia wave versus diffusion wave in flood routing. J. Hydrol., 244, 97–104.
Walko, R. L., and Coauthors, 2000: Coupled atmosphere-biophysics-hydrology models for environmental modeling. J. Appl. Meteor., 39, 931–944.
Wallach, R., Grigorin G. , and Rivlin J. , 1997: The errors in surface runoff prediction by neglecting the relationship between infiltration rate and overland flow depth. J. Hydrol., 200, 243–259.
Warrach, K., Stieglitz M. , Mengelkamp H.-T. , and Raschke E. , 2002: Advantages of a topographically controlled runoff simulation in a soil–vegetation–atmosphere transfer model. J. Hydrometeor., 3, 131–148.
Woods, R. A., and Sivapalan M. , 1997: A connection between topographically driven runoff generation and network structure. Water Resour. Res., 33, 2939–2950.
Xue, Y., Sellers P. J. , Kinter J. L. III, and Shukla J. , 1991: A simplified biosphere model for global climate studies. J. Climate, 4, 345–364.
Yuan, X., and Liang X.-Z. , 2011: Evaluation of a Conjunctive Surface–Subsurface Process Model (CSSP) over the contiguous United States at regional–local scales. J. Hydrometeor., 12, 579–599.