• Abdulla, F. A., , and Lettenmaier D. P. , 1997: Development of regional parameter estimation equations for a macroscale hydrologic model. J. Hydrol., 197, 230257.

    • Search Google Scholar
    • Export Citation
  • Anderson, E. A., 1973: National Weather Service River Forecast System—Snow accumulation and ablation model. NOAA Tech. Memo. NWS HYDRO-17, 217 pp.

    • Search Google Scholar
    • Export Citation
  • Anderson, R. M., , Koren V. I. , , and Reed S. M. , 2006: Using SSURGO data to improve Sacramento model a priori parameter estimates. J. Hydrol., 320, 103116.

    • Search Google Scholar
    • Export Citation
  • Arora, V. K., 2001: Streamflow simulations for continental-scale river basins in a global atmospheric general circulation model. Adv. Water Resour., 24, 775791.

    • Search Google Scholar
    • Export Citation
  • Baldocchi, D. D., 2003: Assessing the eddy-covariance technique for evaluating the carbon dioxide exchange rates of ecosystems: Past, present and future. Global Change Biol., 9, 479492.

    • Search Google Scholar
    • Export Citation
  • Baldocchi, D. D., , Finnigan J. , , Wilson K. , , Paw U K. T. , , and Falge E. , 2000: On measuring net ecosystem carbon exchange over tall vegetation on complex terrain. Bound.-Layer Meteor., 96, 257291.

    • Search Google Scholar
    • Export Citation
  • Bastidas, L. A., , Hogue T. S. , , Sorooshian S. , , Gupta H. V. , , and Shuttleworth W. J. , 2006: Parameter sensitivity analysis for different complexity land surface models using multicriteria methods. J. Geophys. Res., 111, D20101, doi:10.1029/2005JD006377.

    • Search Google Scholar
    • Export Citation
  • Betts, A. K., , Ball J. H. , , Beljaars A. C. M. , , Miller M. J. , , and Viterbo P. A. , 1996: The land surface-atmosphere interaction: A review based on observational and global modeling perspectives. J. Geophys. Res., 101 (D3), 72097225.

    • Search Google Scholar
    • Export Citation
  • Bohn, T. J., , Sonessa M. Y. , , and Lettenmaier D. P. , 2010: Seasonal hydrologic forecasting: Is there a role for multimodel ensemble methods? J. Hydrometeor., 11, 13581372.

    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., , Pollard D. , , and Thompson S. L. , 1992: Effects of boreal forest vegetation on global climate. Nature, 359, 716718.

  • Budyko, M. I., 1974: Climate and Life. Academic Press, 508 pp.

  • Burnash, R. J. C., , Ferral R. L. , , and McGuire R. A. , 1973: A generalized streamflow simulation system—Conceptual modeling for digital computers. U.S. National Weather Service and California Department of Water Resources Joint Federal and State River Forecast Center Tech. Rep., 204 pp.

    • Search Google Scholar
    • Export Citation
  • Chen, F., and Coauthors, 1996: Modeling of land-surface evaporation by four schemes and comparison with FIFE observations. J. Geophys. Res., 101, 72517268.

    • Search Google Scholar
    • Export Citation
  • Chen, F., , Janjic Z. , , and Mitchell K. E. , 1997: Impact of atmospheric surface-layer parameterizations in the new land-surface scheme of the NCEP mesoscale Eta model. Bound.-Layer Meteor., 85, 391421.

    • Search Google Scholar
    • Export Citation
  • Daly, C., , Neilson R. P. , , and Phillips D. L. , 1994: A statistical topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33, 140158.

    • Search Google Scholar
    • Export Citation
  • D’Odorico, P., , Caylor K. , , Okin G. S. , , and Scanlon T. M. , 2007: On soil moisture–vegetation feedbacks and their possible effects on the dynamics of dryland ecosystems. J. Geophys. Res., 112, G04010, doi:10.1029/2006JG000379.

    • Search Google Scholar
    • Export Citation
  • Duan, Q., , Sorooshian S. , , and Gupta V. K. , 1994: Optimal use of the SCE-UA global optimization method for calibrating watershed models. J. Hydrol., 158 (3–4), 265284.

    • Search Google Scholar
    • Export Citation
  • Ek, M. B., , and Holtslag A. A. M. , 2004: Influence of soil moisture on boundary layer cloud development. J. Hydrometeor., 5, 8699.

  • Ek, M. B., , Mitchell K. E. , , Lin Y. , , Rogers E. , , Grunmann P. , , Koren V. , , Gayno G. , , and Tarpley J. D. , 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 22, 8851, doi:10.1029/2002JD003296.

    • Search Google Scholar
    • Export Citation
  • Falge, E., and Coauthors, 2001: Gap filling strategies for long term energy flux data sets. Agric. For. Meteor., 107, 7177.

  • Finnerty, B. D., , Smith M. B. , , Seo D. J. , , Koren V. I. , , and Moglen G. , 1997: Sensitivity of the Sacramento Soil Moisture Accounting model to space–time scale precipitation inputs from NEXRAD. J. Hydrol., 203 (1–4), 2138.

    • Search Google Scholar
    • Export Citation
  • Gan, T. Y., , and Burges S. J. , 2006: Assessment of soil-based and calibrated parameters of the Sacramento model and parameter transferability. J. Hydrol., 320 (1–2), 117131.

    • Search Google Scholar
    • Export Citation
  • Goldstein, A. H., and Coauthors, 2000: Effects of climate variability on the carbon dioxide, water, and sensible heat fluxes above a ponderosa pine plantation in the Sierra Nevada (CA). Agric. For. Meteor., 101 (2–3), 113129.

    • Search Google Scholar
    • Export Citation
  • Gupta, H. V., , Sorooshian S. , , and Yapo P. O. , 1998: Toward improved calibration of hydrologic models: Multiple and noncommensurable measures of information. Water Resour. Res., 34, 751763.

    • Search Google Scholar
    • Export Citation
  • Gupta, H. V., , Bastidas L. A. , , Sorooshian S. , , Shuttleworth W. J. , , and Yang Z. L. , 1999: Parameter estimation of a land surface scheme using multicriteria methods. J. Geophys. Res., 104, 491503.

    • Search Google Scholar
    • Export Citation
  • Gutman, G., , and Ignatov A. , 1998: The derivation of the green vegetation fraction from NOAA/AVHRR data for use in numerical weather prediction models. Int. J. Remote Sens., 19, 15331543.

    • Search Google Scholar
    • Export Citation
  • Heuvelmans, G., , Muys B. , , and Feyen J. , 2004: Evaluation of hydrological model parameter transferability for simulating the impact of land use on catchment hydrology. Phys. Chem. Earth, 29 (11–12), 739747.

    • Search Google Scholar
    • Export Citation
  • Hollinger, D. Y., , Goltz S. M. , , Davidson E. A. , , Lee J. T. , , Tu K. , , and Valentine H. T. , 1999: Seasonal patterns and environmental control of carbon dioxide and water vapour exchange in an ecotonal boreal forest. Global Change Biol., 5, 891902.

    • Search Google Scholar
    • Export Citation
  • Hollinger, S. E., , and Isard S. A. , 1994: A soil moisture climatology of Illinois. J. Climate, 7, 822833.

  • Jarvis, P. G., 1976: The interpretation of leaf water potential and stomatal conductance found in canopies in the field. Philos. Trans. Roy. Soc. London, B273, 593610.

    • Search Google Scholar
    • Export Citation
  • Johansen, O., 1977: Thermal conductivity of soils (in Norwegian). Ph.D. thesis, University of Trondheim, 236 pp.

  • Koren, V. I., 2006: Parameterization of frozen ground effects: Sensitivity to soil properties. Predictions in Ungauged Basins: Promises and Progress, M. Sivapalan, Ed., IAHS, 125–133.

    • Search Google Scholar
    • Export Citation
  • Koren, V. I., , Schaake J. , , Mitchell K. , , Duan Q.-Y. , , Chen F. , , and Baker J. M. , 1999: A parameterization of snowpack and frozen ground intended for NCEP weather and climate models. J. Geophys. Res., 104 (D16), 19 56919 585.

    • Search Google Scholar
    • Export Citation
  • Koren, V. I., , Smith M. B. , , Wang D. , , and Zhang Z. , 2000: Use of soil property data in the derivation of conceptual rainfall-runoff model parameters. Preprints, 15th Conf. on Hydrology, Long Beach, CA, Amer. Meteor. Soc., 2.16. [Available online at http://ams.confex.com/ams/annual2000/techprogram/paper_6074.htm.]

    • Search Google Scholar
    • Export Citation
  • Koren, V. I., , Smith M. , , and Duan Q. , 2003: Use of a priori parameter estimates in the derivation of spatially consistent parameter sets of rainfall-runoff models. Calibration of Watershed Models, Q. Duan et al., Eds., Vol. 6, Water Science and Applications, Amer. Geophys. Union, 239–254.

    • Search Google Scholar
    • Export Citation
  • Koster, R. D., , Suarez M. J. , , Ducharne A. , , Stieglitz M. , , and Kumar P. , 2000: A catchment-based approach to modeling land surface processes in a general circulation model 1. Model structure. J. Geophys. Res., 105 (D20), 24 80924 822.

    • Search Google Scholar
    • Export Citation
  • Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305, 11381140.

  • Livneh, B., , Xia Y. , , Mitchell K. E. , , Ek M. B. , , and Lettenmaier D. P. , 2010: Noah LSM snow model diagnostics and enhancements. J. Hydrometeor., 11, 721738.

    • Search Google Scholar
    • Export Citation
  • Loescher, H. W., , Law B. E. , , Mahrt L. , , Hollinger D. , , Campbell J. L. , , and Wofsy S. E. , 2006: Uncertainties in, and interpretation of, carbon flux estimates using the eddy covariance technique. J. Geophys. Res., 111, D21S90, doi:10.1029/2005JD006932.

    • Search Google Scholar
    • Export Citation
  • Mahmood, R., , and Hubbard K. G. , 2003: Simulating sensitivity of soil moisture and evapotranspiration under heterogeneous soils and land uses. J. Hydrol., 280, 7290.

    • Search Google Scholar
    • Export Citation
  • Mahrt, L., , and Ek M. , 1984: The influence of atmospheric stability on potential evaporation. J. Climate Appl. Meteor., 23, 222234.

  • Mahrt, L., , and Pan H. L. , 1984: A two-layer model of soil hydrology. Bound.-Layer Meteor., 29, 120.

  • Maurer, E. P., , Wood A. W. , , Adam J. C. , , Lettenmaier D. P. , , and Nijssen B. , 2002: A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States. J. Climate, 15, 32373251.

    • Search Google Scholar
    • Export Citation
  • McCumber, M., , and Pielke R. A. , 1981: Simulation of effects of surface fluxes of heat and moisture in a mesoscale numerical model. Part 1: Soil layer. J. Geophys. Res., 86, 99299938.

    • Search Google Scholar
    • Export Citation
  • Milly, P. C. D., , Betancourt J. , , Falkenmark M. , , Hirsch R. M. , , Kundzewicz Z. W. , , Lettenmaier D. P. , , and Stouffer R. J. , 2008: Stationarity is dead: Whither water management? Science, 319, 573574.

    • Search Google Scholar
    • Export Citation
  • Mishra, V., , Cherkauer K. A. , , and Shukla S. , 2010: Assessment of drought due to historic climate variability and projected future climate change in the midwestern United States. J. Hydrometeor., 11, 4668.

    • Search Google Scholar
    • Export Citation
  • Nash, J. E., , and Sutcliffe J. V. , 1970: River flow forecasting through conceptual models part I — A discussion of principles. J. Hydrol., 10, 282290.

    • Search Google Scholar
    • Export Citation
  • Niyogi, D., , Alfieri J. G. , , Blanken P. D. , , Chen F. , , LeMone M. A. , , Mitchell K. E. , , and Ek M. B. , 2008: Estimation of the minimum canopy resistance for croplands and grasslands using data from the 2002 International H20 Project. Mon. Wea. Rev., 136, 44524469.

    • Search Google Scholar
    • Export Citation
  • Pan, H. L., , and Mahrt L. , 1987: Interaction between soil hydrology and boundary layer development. Bound.-Layer Meteor., 38, 185202.

  • Pielke, R. A., , Avissar R. , , Raupach M. , , Dolman A. J. , , Zeng X. , , and Denning A. S. , 1998: Interactions between the atmosphere and terrestrial ecosystems: Influence on weather and climate. Global Change Biol., 4, 461475.

    • Search Google Scholar
    • Export Citation
  • Reed, S., , Koren V. , , Smith M. , , Zhang Z. , , Moreda F. , , and Seo D.-J. , 2004: Overall distributed model intercomparison project results. J. Hydrol., 298, 2760.

    • Search Google Scholar
    • Export Citation
  • Rosero, E., , Gulden L. E. , , Yang Z.-L. , , De Goncalves L. G. , , Niu G.-Y. , , and Kaheil Y. H. , 2011: Ensemble evaluation of hydrologically enhanced Noah-LSM: Partitioning of the water balance in high-resolution simulations over the Little Washita River experimental watershed. J. Hydrometeor., 12, 4564.

    • Search Google Scholar
    • Export Citation
  • Schaake, J. C., , Koren V. I. , , Duan Q.-Y. , , Mitchell K. E. , , and Chen F. , 1996: Simple water balance model for estimating runoff at different spatial and temporal scales. J. Geophys. Res., 101, 74617475.

    • Search Google Scholar
    • Export Citation
  • Schaake, J. C., , Cong S. , , and Duan Q. , 2006: The US MOPEX data set. Large Sample Basin Experiments for Hydrological Model Parameterization: Results of the Model Parameter Experiment—MOPEX, V. Andrēasean et al., Eds., IAHS, 9–28.

    • Search Google Scholar
    • Export Citation
  • Seneviratne, S. I., , Corti T. , , Davin E. L. , , Hirschi M. , , Jaeger E. B. , , Lehner I. , , Orlowsky B. , , and Teuling A. J. , 2010: Investigating soil moisture–climate interactions in a changing climate: A review. Earth-Sci. Rev., 99 (3–4), 125161.

    • Search Google Scholar
    • Export Citation
  • Smith, M. B., , Laurine D. P. , , Koren V. I. , , Reed S. M. , , and Zhang Z. , 2003: Hydrologic model calibration in the National Weather Service. Calibration of Watershed Models, Q. Duan et al., Eds., Water Science and Application Series, Vol. 6, Amer. Geophys. Union, 133–152.

    • Search Google Scholar
    • Export Citation
  • Smith, M. B., , Seo D. , , Koren V. I. , , Reed S. M. , , Zhang Z. , , Duan Q. , , Moreda F. , , and Cong S. , 2004: The distributed model intercomparison project (DMIP): Motivation and experiment design. J. Hydrol., 298, 426.

    • Search Google Scholar
    • Export Citation
  • Sridhar, V., , Elliot R. L. , , Chen F. , , and Botzge J. A. , 2002: Validation of the NOAH-OSU land surface model using surface flux measurements in Oklahoma. J. Geophys. Res., 107, 4418, doi:10.1029/2001JD001306.

    • Search Google Scholar
    • Export Citation
  • Tang, Y., , Reed P. , , Wagner T. , , and van Werkhoven K. , 2007: Comparing sensitivity of analysis methods to advance lumped watershed model identification and evaluation. Hydrol. Earth. Syst. Sci., 11, 793817.

    • Search Google Scholar
    • Export Citation
  • Tateishi, R., , and Ahn C. H. , 1996: Mapping evapotranspiration and water balance for global land surfaces. J. Photogramm. Remote Sens., 51, 209215.

    • Search Google Scholar
    • Export Citation
  • Taylor, C. M., , and Ellis R. J. , 2006: Satellite detection of soil moisture impacts on convection at the mesoscale. Geophys. Res. Lett., 33, L03404, doi:10.1029/2005GL025252.

    • Search Google Scholar
    • Export Citation
  • Thiemann, M., , Trosset M. , , Gupta H. , , and Sorooshian S. , 2001: Bayesian recursive parameter estimation for hydrologic models. Water Resour. Res., 37, 25212535.

    • Search Google Scholar
    • Export Citation
  • Turnipseed, A. A., , Blanken P. D. , , Anderson D. E. , , and Monson R. K. , 2002: Energy budget above a high-elevation subalpine forest in complex topography. Agric. For. Meteor., 110, 177201.

    • Search Google Scholar
    • Export Citation
  • van Werkhoven, K., , Wagener T. , , Reed P. , , and Tang Y. , 2008: Characterization of watershed model behavior across a hydroclimatic gradient. Water Resour. Res., 44, W01429, doi:10.1029/2007WR006271.

    • Search Google Scholar
    • Export Citation
  • Verseghy, D., 1996: Preliminary considerations for runoff modelling in GCMS. Nord. Hydrol., 27, 117128.

  • Vrugt, J. A., , Gupta H. V. , , Dekker S. C. , , Sorooshian S. , , Wagener T. , , and Bouten W. , 2006: Application of stochastic parameter optimization to the Sacramento Soil Moisture Accounting model. J. Hydrol., 325 (1–4), 288307.

    • Search Google Scholar
    • Export Citation
  • Wang, W., , and Kumar A. , 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103, 28 63728 646.

    • Search Google Scholar
    • Export Citation
  • Wetzel, P. J., , Argentini S. , , and Boone A. , 1996: Role of land surface in controlling daytime cloud amount: Two case studies in the GCIP-SW area. J. Geophys. Res., 101 (D3), 73597370.

    • Search Google Scholar
    • Export Citation
  • Wilson, K., and Coauthors, 2002: Energy balance closure at FLUXNET sites. Agric. For. Meteor., 113 (1–4), 223243.

  • Xiu, A., , and Pleim J. E. , 2001: Development of a land surface model. Part I: Application in a mesoscale meteorological model. J. Appl. Meteor., 40, 192209.

    • Search Google Scholar
    • Export Citation
  • Yapo, P. O., , Gupta H. V. , , and Sorooshian S. , 1998: Multi-objective global optimization for hydrologic models. J. Hydrol., 204, 8397.

  • Zhang, L., , Dawes W. R. , , and Walker G. R. , 2001: Response of mean annual evapotranspiration to vegetation changes at catchment scale. Water Resour. Res., 37, 701708.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Schematic of the Noah LSM, including required forcing variables, evaporative components of transpiration ET, canopy evaporation EC, soil evaporation ES, and snow sublimation SS. Precipitation is partitioned into ET, runoff, and infiltration.

  • View in gallery

    Schematic of the Sac model, including required forcing variables and evaporative components. The UZTWC and UZFWC can vary from 0 to a maximum value. Similarly, LZTWC, LZFPC, and LZFSC can also vary from 0 to a maximum value.

  • View in gallery

    Schematic of the ULM, including required forcing variables, moisture, and energy components. Precipitation P and snowmelt SM are partitioned into direct runoff RD, infiltration, and ET. Infiltration becomes either surface runoff RS or interflow I.F. in the upper zone; the remains of which can then infiltrate farther into the lower zone and become baseflow B. The double arrows represent the transfer of model structure, wherein the soil schematic on the left is only considered for soil moisture computations, whereas the schematic on the right is used for all other model computations.

  • View in gallery

    Location of study basins (shaded areas), flux tower sites (black circles), and ICN soil moisture stations (numbered).

  • View in gallery

    Mean monthly precipitation (right axis, bars) and streamflow (left axis, lines) for the six study basins.

  • View in gallery

    Scatterplots of observed energy balances (sensible SH plus latent LE plus ground heat flux G vs net radiation Rnet). Shown on each plot are the slope of the line of best fit (m) and the bias (W m−2), where a slope of m = 1 and bias = 0 would characterize zero energy balance closure error. A single summer is shown for each site: namely, Blodgett Forest (2004), Niwot Ridge (2006), Brookings (2005), and Howland (2001).

  • View in gallery

    (left to right) Mean diurnal fluxes (W m−2) for ULM during summer for (top to bottom) four Ameriflux sites shown at 30-min intervals for the years with greatest energy balance closure at Blodgett Forest (2004), Niwot Ridge (2006), Brookings (2005), and Howland Forest (2001).

  • View in gallery

    Observed (black circles) volumetric soil moisture, compared with ULMA (red circles) and Noah (blue circles) during the warm season at (left) Blodgett Forest (2004) and (right) Niwot Ridge (2006) at (top to bottom) 10, 30, and 50 cm.

  • View in gallery

    ICN soil moisture data in units of millimeters of water equivalent (observations shown in black), where (left), (middle) the mean monthly soil moisture for the four model soil layers of ULM (red) and Noah (blue) are shown; (top right) the entire soil column, including Sac (green); and (bottom right) the change in monthly soil moisture for the entire soil column.

  • View in gallery

    RMSE between simulated and observed streamflow (1960–69) for (top to bottom) the basins listed, based on 15 parameter values varying uniformly within their plausible range (Table 3) plus an additional simulation using the a priori value for that parameter (black circles). Only shown are the most sensitive parameters for each basin based on this method.

  • View in gallery

    Mean monthly streamflows (1960–69) for ULMA, ULMM, Noah, Sac, and observations for the six basins listed.

  • View in gallery

    (top panels) NSE values (multiplied by −1 for consistency with RMSE minimizations) corresponding to simulated vs observed streamflow for variation of individual parameter values following a Monte Carlo approach for the six basins listed. (bottom panels) For each basin, the scatterplots of the two most sensitive parameters are shown, with the best four simulations circled in blue which correspond to the lowest four points on each scatterplot. (top panels) Trends in the data are grouped into 50-simulation bins, with bin minima in blue showing an envelope of parameter sensitivity and bin means in red. The ADIMP parameter (not shown) was found to be highly sensitive for all basins with an approximate minimum value of 0.225.

  • View in gallery

    Daily streamflows and precipitation over the 1964 water year for the (a) ILLIN and (b) BROOK basins. Observed flow (solid black), ULMM (dashed black line), ULMA (dashed orange line), Noah (dashed pink line), and Sac (dashed green line) are included. These two basins provided the greatest challenge in modeling, with relatively sporadic precipitation and sharp streamflow peaks (for BROOK, end of March snowmelt and soil thaw streamflow spike was poorly captured by the snow model and frozen soil physics).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 28 28 3
PDF Downloads 22 22 3

Development of a Unified Land Model for Prediction of Surface Hydrology and Land–Atmosphere Interactions

View More View Less
  • 1 Department of Civil and Environmental Engineering, University of Washington, Seattle, Washington
  • | 2 Office of Hydrologic Development, NOAA/National Weather Service, Silver Spring, Maryland
  • | 3 Department of Civil and Environmental Engineering, University of Washington, Seattle, Washington
© Get Permissions
Full access

Abstract

A unified land model (ULM) is described that combines the surface flux parameterizations in the Noah land surface model (used in most of NOAA’s coupled weather and climate models) with the Sacramento Soil Moisture Accounting model (Sac; used for hydrologic prediction within the National Weather Service). The motivation was to develop a model that has a history of strong hydrologic performance while having the ability to be run in the coupled land–atmosphere environment. ULM takes the vegetation, snow model, frozen soil, and evapotranspiration schemes from Noah and merges them with the soil moisture accounting scheme from Sac. ULM surface fluxes, soil moisture, and streamflow simulations were evaluated through comparisons with observations from the Ameriflux (surface flux), Illinois Climate Network (soil moisture), and Model Parameter Estimation Experiment (MOPEX; streamflow) datasets. Initially, a priori parameters from Sac and Noah were used, which resulted in ULM surface flux simulations that were comparable to those produced by Noah (Sac does not predict surface energy fluxes). ULM with the a priori parameters had streamflow simulation skill that was generally similar to Sac’s, although it was slightly better (worse) for wetter (more arid) basins. ULM model performance using a set of parameters identified via a Monte Carlo search procedure lead to substantial improvements relative to the a priori parameters. A scheme for transfer of parameters from streamflow simulations to nearby flux and soil moisture measurement points was also evaluated; this approach did not yield conclusive improvements relative to the a priori parameters.

Corresponding author address: Ben Livneh, Department of Civil and Environmental Engineering, University of Washington, Box 352700, Seattle, WA 98195. E-mail: blivneh@hydro.washington.edu

Abstract

A unified land model (ULM) is described that combines the surface flux parameterizations in the Noah land surface model (used in most of NOAA’s coupled weather and climate models) with the Sacramento Soil Moisture Accounting model (Sac; used for hydrologic prediction within the National Weather Service). The motivation was to develop a model that has a history of strong hydrologic performance while having the ability to be run in the coupled land–atmosphere environment. ULM takes the vegetation, snow model, frozen soil, and evapotranspiration schemes from Noah and merges them with the soil moisture accounting scheme from Sac. ULM surface fluxes, soil moisture, and streamflow simulations were evaluated through comparisons with observations from the Ameriflux (surface flux), Illinois Climate Network (soil moisture), and Model Parameter Estimation Experiment (MOPEX; streamflow) datasets. Initially, a priori parameters from Sac and Noah were used, which resulted in ULM surface flux simulations that were comparable to those produced by Noah (Sac does not predict surface energy fluxes). ULM with the a priori parameters had streamflow simulation skill that was generally similar to Sac’s, although it was slightly better (worse) for wetter (more arid) basins. ULM model performance using a set of parameters identified via a Monte Carlo search procedure lead to substantial improvements relative to the a priori parameters. A scheme for transfer of parameters from streamflow simulations to nearby flux and soil moisture measurement points was also evaluated; this approach did not yield conclusive improvements relative to the a priori parameters.

Corresponding author address: Ben Livneh, Department of Civil and Environmental Engineering, University of Washington, Box 352700, Seattle, WA 98195. E-mail: blivneh@hydro.washington.edu

1. Introduction

The principal role of land schemes in numerical weather and climate prediction models is to partition net radiation into turbulent surface and ground heat fluxes, which are required to characterize the atmospheric model’s lower boundary. Although land surface models (LSMs) perform full land surface hydrologic calculations, they generally focus more on representation of land–atmosphere fluxes than on the processes, such as soil moisture dynamics, that control runoff generation (Koster et al. 2000; Bastidas et al. 2006). As a case in point, the Noah LSM (Ek el al. 2003), which serves as the land surface scheme in the numerical weather and climate prediction models of the National Centers for Environmental Prediction (NCEP) of the National Weather Service (NWS), has been shown to be less skillful in streamflow prediction compared with more hydrologically based models (Bohn et al. 2010). Nonetheless, hydrologic factors, especially soil moisture, play an important role in modulating climate (Wang and Kumar 1998; Mahmood and Hubbard 2003; Koster et al. 2004; Seneviratne et al. 2010). Within an atmospheric model, surface latent heat fluxes are largely controlled by the interaction of soil moisture with evapotranspiration (ET; Xiu and Pleim 2001). Other meteorological processes like cloud formation are also sensitive to soil moisture (Wetzel et al. 1996) and may result in either positive or negative feedbacks (Ek and Holtslag 2004; Taylor and Ellis 2006). Therefore, if an LSM’s representation of these processes is poor, it will produce unrealistic evaporation rates regardless of the quality of the evaporation formulation (Koster et al. 2000). Another consideration is that the runoff that results from an LSM’s soil moisture computation ultimately becomes an input to the oceans (from major river basins) constituting an important boundary condition for the modeling of oceanic circulation and climate. The impact of streamflow on salinity at the continental boundaries can affect both ocean convection and thermohaline circulation and therefore influences sea surface temperature and sea ice, which both exert a strong influence on climate (Verseghy 1996; Arora 2001).

Hydrologic models focus on accurately simulating components of the surface water budget, especially streamflow. The Sacramento Soil Moisture Accounting model (Sac; Burnash et al. 1973), which is the primary model used for river forecasting by the NWS River Forecast Centers (RFCs) across the United States, has been found to perform well in streamflow prediction compared with other models and observations (Reed et al. 2004). A number of recent studies have focused on techniques for Sac parameter estimation based on numerical optimization methods (Duan et al. 1994; Yapo et al. 1998; Gupta et al. 1998; Thiemann et al. 2001; Smith et al. 2003; Vrugt et al. 2006; Gan and Burges 2006; Tang et al. 2007; van Werkhoven et al. 2008). Sac parameters are usually obtained via a calibration process because most model parameters are not directly measureable. An alternative approach is to estimate model parameters from measurable soil characteristics such as percentages of sand and clay and soil field capacity (Koren et al. 2000, 2003; Anderson et al. 2006). Such approaches are attractive because they provide a basis for parameter estimation in ungauged basins, as well as the ability to provide physical constraints on calibration in gauged basins.

Two major obstacles prevent the Sac model from being coupled with atmospheric models. The first is the absence of a surface energy budget, which (in the case of LSMs) includes surface heat fluxes and radiative partitioning. Surface heat fluxes define the near-surface air temperature, ground temperature (sensible heat fluxes), and humidity (latent heat flux). Their estimation is indirectly important for surface hydrology, because feedbacks between soil moisture and precipitation affect the models’ runoff production (McCumber and Pielke 1981; Betts et al. 1996).

The second major shortcoming of Sac is the absence of an explicit representation of vegetation. Vegetation can have a profound influence on climate through surface exchanges of heat, moisture, and momentum (Bonan et al. 1992; Pan and Mahrt 1987; Pielke et al. 1998). The presence of vegetation also alters the rate of moisture movement to and from the soil, via canopy interception and root-zone water uptake for transpiration. Subcanopy soils are frequently moister than intercanopy patches suggesting the possible existence of a positive feedback between vegetation and soil water content (D’Odorico et al. 2007). Additionally, evapotranspiration (ET) rates have been shown to vary according to vegetation type, such as forest versus grassland (Zhang et al. 2001); hence, there is a possible link between vegetation and streamflow production.

Another consideration that is central to nearly all aspects of LSM performance is the estimation of ET. On a global average, between 60% and 80% of precipitation reaching the land surface is returned to the atmosphere through ET, which is the largest component of the terrestrial hydrological cycle (Tateishi and Ahn 1996). In both the Noah and Sac models, ET is a function of potential evapotranspiration (PET), and PET therefore strongly influences ET predictions. PET is a representation of the environmental demand for ET; it is controlled both by the energy available to evaporate water and the ability of the atmosphere to transport the water vapor from the ground into the lower atmosphere. ET is said to equal PET when moisture is freely available at the surface. Both the Noah and Sac models compute actual ET as a fraction of PET that depends on resistances of each ET component (bare soil for both models; canopy evaporation and transpiration for Noah only). The main point of interest is that Noah computes PET dynamically, following a Penman–Monteith approach (Mahrt and Ek 1984), whereas Sac cannot do so (lacking incorporation of radiation and surface roughness). In most cases, Sac requires PET as an input, and it is often prescribed as a fixed value (but seasonally varying). This approach does not account for interannual variability and perhaps more importantly invokes an implicit assumption of stationarity, which arguably no longer is defensible (Milly et al. 2008) because of anthropogenic changes in land cover and changes in the earth’s climate.

For cases where hydrologic model parameters are not readily observable (e.g., Sac), they can be estimated via calibration, in which a set (or sets) of model parameters are obtained that result in differences between observed and simulated states or fluxes (e.g., streamflow) being minimized. Given the complexity of the hydrologic system, parameter estimation generally requires automatic (versus manual trial and error) optimization procedures of multiobjective functions. For those parameters that most directly affect model predictions of surface fluxes, flux tower measurements can be used (Betts et al. 1996; Chen et al. 1997; Gupta et al. 1999, Sridhar et al. 2002; Rosero et al. 2011).

In this paper, we describe a unified land model (ULM), which is a merger of the Noah and Sac models. The motivation for this merger is to incorporate a hydrologically realistic structure within a model construct that can be used in coupled land–atmosphere applications. Because Noah is used operationally at NCEP for offline hydrologic simulations (e.g., for drought characterization) and is coupled with a suite of atmospheric models, the implications of improving its soil moisture–runoff generation scheme would be widespread. Conversely, the Sac model is used operationally for flood forecasting at over 3000 forecast points across the United States, and it would benefit from Noah’s more physically based vegetation and ET algorithms. We follow with a brief description of the heritage and components of each model, the nature of the approach we used to merge key parameterizations from each, and an assessment of ULM performance.

2. Model structure

a. Noah

The heritage of Noah dates to the early 1990s, when NCEP adopted the Oregon State University (OSU) LSM (Mahrt and Pan 1984; Pan and Mahrt 1987) for use in its numerical weather prediction model. Subsequently, the OSU model became NOAH, with many upgrades described by Ek el al. (2003). NOAH originally stood for the collaborators in the project which adopted the OSU model [NCEP, OSU, the Air Force—both the Air Force Weather Agency (AFWA) and the Air Force Research Laboratory (AFRL)—and the Hydrologic Research Laboratory at the NWS); however, the model acronym has since been dropped and it is referred to simply as Noah.

Noah has been run at spatial resolutions ranging from several kilometers to hundreds of kilometers. Sridhar et al. (2002) found that Noah’s surface heat fluxes compared favorably with observations, however other studies have shown Noah is less skillful than other land surface models in streamflow simulation (Reed et al. 2004; Bohn et al. 2010). Figure 1 shows the main elements of Noah. The model uses a bulk surface layer with a single (dominant) vegetation class and snowpack, overlying a (dominant) soil texture divided into four layers. The vegetation canopy is assumed to cover a fraction of the land surface that varies spatially and temporally by an input greenness fraction Gvf (Gutman and Ignatov 1998), derived from the photosynthetically active portion of leaf area index (LAI), based on a monthly 5-yr climatology of Advanced Very High Resolution Radiometer (AVHRR) satellite data. The remainder of the grid cell is bare soil. Water can be intercepted by the vegetation canopy up to a prescribed maximum threshold.

Fig. 1.
Fig. 1.

Schematic of the Noah LSM, including required forcing variables, evaporative components of transpiration ET, canopy evaporation EC, soil evaporation ES, and snow sublimation SS. Precipitation is partitioned into ET, runoff, and infiltration.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

A Richards equation approach is used to solve for the movement of moisture through the four soil layers. The soil temperature profile is determined using nonlinear functions for the thermal conductivity of each soil layer (Johansen 1977). Both of these computations require parameters such as porosity, wilting point, dry density, and quartz content that relate to soil texture. The model does not explicitly form a water table and capillary rise does not occur in the strict sense but rather as the result of vertical dispersion via the solution to the Richards equation. Infiltration into the soil follows Schaake et al. (1996) as a nonlinear function of soil saturation, bounded above by precipitation and below by soil hydraulic conductivity. Sensible heat flux and ground heat flux are computed by the thermal diffusion equation (Chen et al. 1996), as differences between skin and air temperatures and soil and skin temperatures, respectively, whereas latent heat flux is a function of the actual ET. In the absence of snow, ET occurs either by canopy evaporation, bare soil evaporation, or transpiration through the root zones, which is described in greater detail in section 2c. The Noah snow model prescribes a seasonally varying snow albedo decay function and provides for liquid water retention within the snowpack and partial snow coverage. Frozen soil physics follow Koren et al. (1999), which provides for a reduction in moisture movement in response to increased soil ice content. Further details of the Noah snow model can be found in Livneh et al. (2010).

b. Sac

Sac is the operational flood forecasting model of the U.S. National Weather Service. It is also used for seasonal ensemble forecast applications at most of the 13 RFCs throughout the United States (Anderson et al. 2006). The model was developed by Burnash et al. (1973) with the initial charge of being “a generalized streamflow simulation system” that could be used to aid in water management decision making. The model was designed to be computationally efficient (run at a daily time step), run over an entire basin using a single set of model parameters (i.e., spatially lumped). Although there are many exceptions, the model has most often been used to generate river forecasts for rivers with response times of greater than 12 h and drainage areas ranging from 300 to 5000 km2 (Finnerty et al. 1997). Recent work by Koren et al. (2003) generalized the model for use in a spatially distributed context. Other recent enhancements include implementation of frozen soil physics representations, which also have resulted in an ability to map the model’s moisture storage contents (in five zones) to physical layers (Koren 2006).

Absent an explicit representation of vegetation, the model’s ET representation utilizes monthly ET factors that adjust a prescribed (monthly varying but otherwise constant) PET. Snow processes are represented in a separate, temperature-index-based snow model, SNOW-17 (Anderson 1973).

Figure 2 shows the model conceptually. The five storage zones represent “free” and “tension” water reservoirs in an upper and a lower zone. Free water is a representation of the quantity of water in excess of the soil’s field capacity, for which gravity governs the moisture movement through the soil. The tension water zones represent the quantity of water between the soil’s field capacity and soil’s wilting point that is bound more closely to the soil and hence must be satisfied before any moisture can be extracted from the free water zones. Movement between upper and lower zones is controlled by a nonlinear percolation function, whereas subsurface flow is computed based on parameters derived from the hydraulic conductivity of each zone and other factors. Surface infiltration is a linear function of upper-zone tension water saturation and direct runoff is controlled by an impervious fraction, which increases up to a prescribed threshold depending on the degree of the upper-zone saturation.

Fig. 2.
Fig. 2.

Schematic of the Sac model, including required forcing variables and evaporative components. The UZTWC and UZFWC can vary from 0 to a maximum value. Similarly, LZTWC, LZFPC, and LZFSC can also vary from 0 to a maximum value.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

c. ULM

The Noah and Sac models have been widely used in operational settings to simulate soil moisture (both models), energy fluxes (Noah only), and streamflow (primarily Sac). Figure 3 illustrates the components that are preserved from each of the parent models in ULM. In general, we retained the land surface components from Noah (e.g., vegetation and ET), as well as the Noah snow model, and Noah’s algorithms for computing surface heat and radiative fluxes. We also retained Noah’s frozen soil algorithm. The soil moisture and runoff generation algorithms (including infiltration) were taken from Sac. A key element of the merger is conversion from Sac’s conceptual soil moisture storage zones to physical layers, which is achieved through an adaptation of the Sac heat transfer (SAC-HT) mechanism, in which tension and free water storages are transferred to physical layers as moisture that exceeds the soil wilting point (Koren 2006).

Fig. 3.
Fig. 3.

Schematic of the ULM, including required forcing variables, moisture, and energy components. Precipitation P and snowmelt SM are partitioned into direct runoff RD, infiltration, and ET. Infiltration becomes either surface runoff RS or interflow I.F. in the upper zone; the remains of which can then infiltrate farther into the lower zone and become baseflow B. The double arrows represent the transfer of model structure, wherein the soil schematic on the left is only considered for soil moisture computations, whereas the schematic on the right is used for all other model computations.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

ET is an essential flux in hydrological models that defines the soil moisture balance and hence storm runoff production, as well as the cycling of moisture to the atmosphere. During snow-free periods, ET in Noah (and hence ULM) is based on a relationship with PET taken from Mahrt and Ek (1984),
e1
where Ro is the radiation flux density; Δ is the slope of the saturation vapor pressure curve; A is a function of the specific humidity of the air with respect to saturation; Rr is a function of surface air temperature and surface pressure; Ch is the surface exchange coefficient for heat and moisture; ρ is the air density; cp is the specific heat capacity; U is wind speed; and Lυ is the latent heat of vaporization.
Very generally, each ET component (soil, canopy, and transpiration) is a fraction (≤1) of PET, scaled by its resistance to moisture transfer. Soil evaporation ES only occurs over the nongreen fraction of the grid cell, over which PET is scaled accordingly. This demand is then applied to the Sac upper and lower zones as follows:
e2a
e2b
e2c
UZTWC is the upper zone tension water contents, UZTWM is the upper zone tension water maximum storage, UZFWC is the upper zone free water contents, UZFWM is the upper zone free water maximum storage, and similarly for lower zone. The above logic keeps the Sac soil ET extraction scheme intact with the exceptions that (i) ET from riparian vegetation as represented by Sac (usually small) is neglected in favor of the more complete and explicit vegetation scheme from Noah and (ii) soil evaporation from the Sac lower zone ESOIL-LOWER, which was intended to represent deeper soil moisture extraction via transpiration (e.g., by trees), is replaced by root water uptake from Noah’s explicit vegetation scheme. Soil evaporation is a function of soil saturation, which is indexed to the relative storage in each zone. Hence, the zone capacities influence model moisture movement through the soil, and the quasi-equilibrium moisture state of ULM will differ from Sac because of the moisture demands from canopy evaporation and transpiration.
Canopy evaporation EC in ULM is a Noah analog, equal to PET over the “green” area reduced by a nonlinear canopy factor n, which is applied to the degree of canopy saturation such that
e3
Here, Wi is the current canopy moisture content, always less than or equal to the vegetation class–defined maximum Wmax.
The transpiration computation in ULM is similar to Noah’s. It uses a Jarvis-type canopy resistance scheme (Jarvis 1976), which is described in detail by Niyogi et al. (2008). Essentially transpiration ET is a function of PET reduced by the canopy resistance and scaled by saturation between zero and one, during wet and dry canopy conditions, respectively, such that canopy evaporation dominates in the former case. Therefore,
e4
where FC is the canopy resistance, which is derived empirically from four noninteracting environmental stress functions, where each of which represents a statistical relationship between canopy resistance and incident solar radiation, humidity, air temperature, and leaf water potential. The removal of transpiration water from the physical soil layer structure is weighted based on soil-class-defined root-zone distributions.

3. Model testing and evaluation strategy

ULM was tested with respect both to its hydrologic prediction capabilities and its ability to predict land–atmosphere moisture and energy fluxes. The evaluation strategy included comparisons of flux tower measurements of surface energy and moisture, point observations of soil moisture, and predicted and observed streamflows over catchments of varying size and hydroclimatic characteristics. To the extent possible, the study catchments and flux towers were collocated. The evaluation criteria included the ability to reproduce the observed diurnal cycle of turbulent heat and radiative fluxes, seasonal patterns of soil moisture, and timing and magnitude of streamflow variations.

Our main objective in evaluation of ULM was to determine the ability of the model to produce plausible results with a set of a priori parameters from its parent models. The Noah parameters were obtained from the North American Land Data Assimilation System (NLDAS) which uses existing high-resolution vegetation and soil coverages derived from satellite and other remote sensing sources. The Sac parameters were derived by the method of Koren et al. (2003), which relates model parameters to soil texture characteristics. A secondary objective of this research was to develop strategies for geographic transfer of ULM model parameters that account for differences in hydroclimatic conditions and avoid the necessity for computationally intensive site-specific model calibration.

In an attempt to reduce the relatively large number of Sac parameters (13) that need to be estimated, we examined the quality of their a priori values within ULM (ULMA) via individual parameter sensitivity tests at each study basin. These tests involve uniformly sampling each parameter over its plausible range of values while holding all other parameters at their a priori value over the respective catchment. Comparing the resulting root-mean-square error (RMSE) from these simulations with observed streamflow allows for a preliminary assessment of the quality of the a priori value and the sensitivity of the parameter to streamflow. The amplitude of RMSE variability associated with each parameter describes its sensitivity, whereas the quality of the a priori value itself is described by the proximity of its ensuing RMSE to the minimum RMSE over the sampled parameter space.

To further understand sensitivities and higher-order parameter interactions, we employed a Monte Carlo search procedure with the ultimate aim of defining an improved set of parameters (ULMM). Performance using these parameters was then compared with performance: (i) using strictly a priori parameters, (ii) by preserving adjusted values from a subset of only the three most sensitive model parameters while keeping the remaining parameters at their a priori value (ULM3), and (iii) using the parent models (Noah and Sac). A Monte Carlo procedure was selected because other methods for parameter sampling, such as iterated fractional factorial design (IFFD) or Sobol’s method (for a complete discussion, see Tang et al. 2007), require a prohibitively large number of samples (simulations) to adequately account for the effects of second-order parameter interactions given the number of parameters (13) (e.g., >103 simulations for IFFD, >8 × 103 simulations for Sobol’s method). The Monte Carlo procedure varies all parameters simultaneously and randomly, thus having the potential to reveal higher-order interactions with fewer simulations (in this case 250). We acknowledge that this approach is more approximate and less exhaustive than the systems mentioned above, but nevertheless it should capture the essence of each parameter’s sensitivity given the random component. Finally, it should be noted that the emphasis here is not on extensive calibrations but rather to show that ULM can produce plausible results with limited parameter tuning.

4. Study areas and data

Streamflow, soil moisture, and flux tower sites were chosen to represent a range of elevation, climatic, soil, and vegetation conditions, subject to the availability of observation sites and the quality of data. To the extent possible, stream gauges and flux towers in close proximity to one another were selected to facilitate comparison of the streamflow and flux predictions. The selected study locations are shown in Fig. 4.

Fig. 4.
Fig. 4.

Location of study basins (shaded areas), flux tower sites (black circles), and ICN soil moisture stations (numbered).

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

a. Study basins

In addition to hydroclimatic and geographic diversity, selected stream gauge locations were either free of significant anthropogenic effects upstream (reservoir storage and/or diversions) or had naturalized flow data available (flows that would occur in the absence of upstream water management effects). In addition, two periods of at least 10 yr were required for which quality assured streamflow data were available. The selected river basins are listed in Table 1, along with relevant information on the local environment and climate for each basin. Figure 5 summarizes mean monthly precipitation and streamflow for each basin.

Table 1.

Description of river basins used in this study.

Table 1.
Fig. 5.
Fig. 5.

Mean monthly precipitation (right axis, bars) and streamflow (left axis, lines) for the six study basins.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

For the three largest river basins, model forcing data were obtained from the Maurer et al. (2002) dataset, which is a ⅛° latitude–longitude grid of the required model inputs derived from climatological station data (precipitation and daily maximum and minimum temperatures). For smaller basins, such as CARNA, ILLIN, and SANDY (<3000 km2), station data were gridded to spatial resolution using the Maurer et al. (2002) approach to derive model forcings. For each basin, the model was run at the same resolution as the forcing data.

For one of the stations (OROVI), Bohn et al. (2010) conducted a rigorous statistical analysis of the naturalized streamflows (obtained from the California Department of Water Resources) and meteorological data and found them to be in good agreement with respect to major storm and drought events. The other five basins are part of the Model Parameter Estimation Experiment (MOPEX; Schaake et al. 2006) for which naturalized (or minimally regulated) streamflow, mean-areal precipitation, and daily maximum and minimum temperatures had already been assembled. A major effort of MOPEX was to assemble a large number of high-quality historical hydrometeorological and river basin characteristics datasets for a wide range of river basins (500–10 000 km2) for model development and understanding. The model forcings for the selected MOPEX basins were obtained as described above; however, we performed a postprocessing adjustment to make the monthly average of the gridded temperature and precipitation values averaged over the basins match the mean-areal values produced by MOPEX. To assure consistency of daily and monthly values, the daily gridded values were adjusted by the ratio (precipitation) or difference (temperature) between the monthly means of the gridded data and the MOPEX mean-areal values. Finally, to account for topographic effects, each model grid cell was subdivided into up to five elevation bands, depending on the elevation range within the grid cell. Within each band, the air temperature was lapsed to the band’s average elevation using a lapse rate of 6.5°C km−1 and precipitation was redistributed to reflect the nonlinearity in precipitation with respect to topographic effects, as obtained from the Parameter-Elevation Regressions on Independent Slopes Method (PRISM) dataset (Daly et al. 1994).

b. Surface fluxes and soil moisture observations

Flux data were taken from the Ameriflux network, which consists of flux towers at approximately 50 sites (in the continental United States) that represent a range of hydroclimatic and ecological conditions. A central issue to flux measurement is energy balance closure. By construct, latent (λE) and sensible (H) heat fluxes must be balanced by net radiation (Rnet), ground heat flux (G), heat storage change between the soil and the height of the flux measurement system (S), and other advected (source and sink) fluxes (A),
e5
However, each of the terms is measured independently, so closure is not assured. Typically, the advection term is small and can be neglected, leaving five main terms to be considered for energy balance closure. Latent and sensible heat fluxes are measured at the flux towers using the eddy covariance method, a direct, micrometeorological approach that relies on a simplification of the conservation equation (Baldocchi 2003). Within the Ameriflux network, these fluxes are calculated at half-hour intervals, as the time average covariance between the (essentially) instantaneous vertical wind speed fluctuations w′ and the instantaneous scalar quantity c′ (temperature and water vapor). Other quantities of interest measured at Ameriflux stations include precipitation, net radiation and/or its components (downward and upward solar and longwave radiation), soil heat flux, soil moisture, and other micrometeorological variables. Specific details of Ameriflux measurement standards can be found online (at http://public.ornl.gov/ameriflux/sop.shtml).

Studies such as Baldocchi et al. (2000), Wilson et al. (2002), and Loescher et al. (2006) have examined potential error sources in eddy covariance measurements including those used in the Ameriflux network. These studies have reported energy imbalances on the order of 20% at stations across a wide range of vegetation and climate types.

The two criteria used in selecting flux tower sites were (i) to use sites with a high degree of energy balance closure and (ii) to select sites (where possible) in proximity to study basins described in section 4a, that span a range of hydroclimatic conditions. Only sites with the maximum Ameriflux quality rating, L4, were selected. These data include quality-control flags and specify any gap filling algorithms that were applied to the final, processed dataset (for details, see, e.g., Falge et al. 2001). Table 2 summarizes the selected flux tower sites, including their respective principal investigators. An inspection of energy balance closure across these sites indicates <20% closure imbalance during the warm season at half-hourly intervals (Fig. 6), with the Blodgett generally having greatest closure and Howland having the least. Local observations of precipitation, air temperature, wind speed, and downward solar radiation were used to force model simulations at each flux tower site. Additional forcing variables not directly measured but required by the models were derived following the techniques described in section 4a.

Table 2.

Summary of characteristics of Ameriflux sites.

Table 2.
Fig. 6.
Fig. 6.

Scatterplots of observed energy balances (sensible SH plus latent LE plus ground heat flux G vs net radiation Rnet). Shown on each plot are the slope of the line of best fit (m) and the bias (W m−2), where a slope of m = 1 and bias = 0 would characterize zero energy balance closure error. A single summer is shown for each site: namely, Blodgett Forest (2004), Niwot Ridge (2006), Brookings (2005), and Howland (2001).

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

Only two of the flux towers produced usable soil moisture data. For this reason, soil moisture simulations were compared with 10 stations from the Illinois Climate Network (ICN; Hollinger and Isard 1994) that are summarized in Table 3. Previous studies have found these data to be of good quality for purposes of model evaluations, given the range of measurement depths and completeness of data record (e.g., Mishra et al. 2010; Maurer et al. 2002).

Table 3.

Summary of characteristics of ICN stations.

Table 3.

5. Model validation and discussion

In this section, we compare ULM simulations with observations and with Noah and Sac simulations. The analysis begins with a comparison of the models’ surface fluxes, followed by an evaluation of their soil moisture predictions. Finally, we compare the models’ streamflow predictions, along with an analysis of parameter sensitivities.

a. Surface fluxes

Given Noah’s history as the land scheme in coupled land–atmosphere models, the ability for ULM to, at minimum, match Noah skill in simulating surface energy fluxes is of key importance. Figure 7 summarizes the observed diurnal cycles of fluxes at the four flux towers, Blodgett, Niwot, Brookings, and Howland, that are in the vicinity of OROVI, MAYBE, BROOK, and SANDY basins, respectively. The net radiation cycle encompasses variations in surface albedo (fraction of reflected shortwave radiation), surface temperature (quantity of emitted longwave radiation), and emissivity. Sensible and latent heat fluxes describe how much of the available energy goes into heating the near-surface air and how much goes into changing the phase of near-surface water; both are also affected by the moisture state of the soil through its effect on soil temperature (sensible heat flux) and saturation, which drives evaporative efficiency (latent heat flux). Ground heat flux is affected by soil moisture, through its effect on soil heat capacity and thermal conductivity. Recognizing the nature and potential magnitude of flux measurement errors, we focus on major overall features, such as the timing and relative magnitude of the diurnal cycles, and less on accumulated values.

Fig. 7.
Fig. 7.

(left to right) Mean diurnal fluxes (W m−2) for ULM during summer for (top to bottom) four Ameriflux sites shown at 30-min intervals for the years with greatest energy balance closure at Blodgett Forest (2004), Niwot Ridge (2006), Brookings (2005), and Howland Forest (2001).

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

At all flux towers, both Noah and ULM capture the net radiation cycle well, with the main exception being Brookings and Howland, where both models slightly overpredict the peak magnitude. Latent and sensible heat fluxes were slightly more variable between the models. The timing of peak latent heat flux was generally predicted correctly; however, its magnitude was underpredicted at Blodgett and Howland by both models, whereas it was slightly overpredicted by Noah and underpredicted by ULM at Brookings; the opposite was true at Niwot. Sensible heat flux timing was better predicted at the lower elevation sites (Brookings and Howland), with a noticeable disparity in sensible heat magnitude for both models throughout the diurnal cycle at Niwot, which was by far the windiest site. ULM matched sensible heat magnitudes equally or slightly better than Noah at each site. The ground heat flux simulations had the poorest match between simulations and observations across the four sites. This is the result of two features. (i) Ground heat flux measurements are notoriously prone to errors. Particularly at Blodgett, Niwot, and Howland, the magnitude of observed ground heat flux was extremely small compared to other fluxes. (ii) Regardless of measurement errors, peak timing of observed ground heat flux was lagged compared with other flux peaks, likely because of heat storage. Altogether, the model ground heat flux predictions were too large (small) during day (night) compared with observations.

Our overall assessment of these comparisons is that ULM satisfies the minimum stated objective of performing comparably to Noah at the flux tower sites, although a few notable disparities persisted between simulated and observed values for both models. Taken over all the flux towers, ULM had slightly smaller average bias than Noah for each individual flux. For both models, the net radiation cycle was better captured at the alpine versus nonalpine sites; the timing of peaks in latent and sensible heat fluxes was generally better at the nonalpine sites; and ground heat flux had the poorest match with observations, although it is not clear whether the reason is model performance, observation errors, or both.

b. Soil moisture

At two of the four flux towers, continuous soil moisture measurements overlapped the flux measurement periods which allowed comparisons of modeled and observed soil moisture. Figure 8 shows daily time series of soil moisture for Blodgett and Niwot over a 7-month period, illustrating the evolution of soil moisture throughout the warm season. The Blodgett site, which experiences very dry summers, exhibits two successive periods of more or less monotonic decline in soil moisture beginning near the middle of June, without a significant increase until the last half of September. An important difference between Noah and ULM at this site is that ULM captures the two separate periods of decreasing soil moisture, particularly in the near-surface layers, whereas Noah predicts a longer single period of declining soil moisture throughout the summer. ULM’s soil moisture evolution is dominated by direct soil evaporation in the early part of summer, which proceeds until the upper-zone soil moisture becomes stressed, after which transpiration is dominant. During this second period, the rate at which soil moisture is extracted is reduced because the soil is under greater stress (i.e., near wilting), and thus soil moisture remains nearly constant. On the other hand, Noah combines direct soil evaporation and transpiration (from all layers) into its solution for the Richards equation and both of these soil moisture processes proceed together. At greater depth, ULM soil moisture is greater than observed, likely because of an excessively large soil moisture reservoir and perhaps inadequate representation of the root density, which affects soil moisture removal via transpiration at this depth.

Fig. 8.
Fig. 8.

Observed (black circles) volumetric soil moisture, compared with ULMA (red circles) and Noah (blue circles) during the warm season at (left) Blodgett Forest (2004) and (right) Niwot Ridge (2006) at (top to bottom) 10, 30, and 50 cm.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

The Niwot site has a less uniform progression of soil moisture as compared with the Blodgett warm season, because frequent precipitation events and even lingering snowmelt supply the soil column with new moisture during spring and summer. The late May spike in soil moisture from a snowmelt event was missed in the snow model that is common to Noah and ULM. On the other hand, both models capture a large precipitation event in early July relatively well, in addition to several smaller events in August and September. However, the magnitudes of both Noah and ULM near-surface soil moisture are too high. This may be due in part to problems with the soil texture data, such as porosity and inferred wilting point. Furthermore, the sensor appears to be in error during most of April, during which anomalously low and nearly constant soil moisture readings were produced.

The mean soil moisture from 10 ICN stations is plotted in Fig. 9. Because these data were available at approximately weekly intervals, a monthly analysis was performed, consistent with other studies that have used these data (see section 4b). Overall, ULM has smaller bias than Noah across the 10 stations. Evaluation of the total column (~2.0 m) soil moisture and its monthly change allows for a comparison with the Sac’s conceptual storages. The change in monthly soil moisture provides a representation of how much water is available for plants and runoff versus the amount going to storage. With respect to the observed seasonal cycle, Noah (ULM) slightly overpredicts (underpredicts) the total range of variability, whereas Sac significantly lacks seasonal variability. Overall, ULM performs best over the ICN stations with roughly equal performance to Noah in terms of the models’ representation of seasonal variations. ULM, however, had the least overall bias.

Fig. 9.
Fig. 9.

ICN soil moisture data in units of millimeters of water equivalent (observations shown in black), where (left), (middle) the mean monthly soil moisture for the four model soil layers of ULM (red) and Noah (blue) are shown; (top right) the entire soil column, including Sac (green); and (bottom right) the change in monthly soil moisture for the entire soil column.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

c. Streamflow and parameter sensitivities

Streamflow represents an integrated basin response to precipitation and evaporative demand. Given the model forcings, variability in simulated streamflow can result from a multidimensional space of soil parameter combinations. To address this problem, Sac, which is an operational model, typically is calibrated in practice. Although ULM uses Sac soil parameters, the ULM model structure calls into question the extent to which Sac parameters are transferrable. Here, we attempt to reduce the relatively large number of Sac soil parameters (13; defined in Table 4) to a smaller set that control the bulk of simulated streamflow performance. Figure 10 shows the most sensitive soil parameters in each basin in terms of local RMSE, in the context of the corresponding a priori parameter values. Common among all basins but CARNA was sensitivity to the parameter UZTWM, which plays a major role in controlling runoff and ET, because it represents the non-gravity-driven water content within the upper soil zone. In both ULM and Sac, UZTWM capacity must be filled before runoff and infiltration can occur, whereas its relative saturation linearly controls the rate of evaporation from the upper soil zone. Notable sensitivities were also exhibited to the parameters that control subsurface flow among the upper and lower zones (UZK, LZPK, and LZSK). The results of the comparisons suggest that the a priori parameters for OROVI, SANDY, ILLIN, and BROOK are close to the values associated with minimum RMSE, whereas the a priori values for CARNA and MAYBE are farther from minimum RMSE.

Table 4.

List of Sac soil parameters and their plausible ranges.

Table 4.
Fig. 10.
Fig. 10.

RMSE between simulated and observed streamflow (1960–69) for (top to bottom) the basins listed, based on 15 parameter values varying uniformly within their plausible range (Table 3) plus an additional simulation using the a priori value for that parameter (black circles). Only shown are the most sensitive parameters for each basin based on this method.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

Despite the model performance inferred from individual RMSE minimizations, the hydrographs produced by ULMA (Fig. 11) do not adequately match observations, with the exceptions of OROVI and SANDY. This warrants further analysis of parameter sensitivities to identify the effects of higher-order parameter interactions on simulated streamflow. Following a Monte Carlo approach, wherein all 13 soil parameters were allowed to vary simultaneously within their plausible ranges over 250 simulations per basin, a clearer maximum envelope of model efficiency evolved. Model performance as measured by Nash–Sutcliffe efficiency (NSE; Nash and Sutcliffe 1970), which effectively is the MSE normalized by the variance, was selected for this extended analysis because it facilitates interbasin performance comparisons. Figure 11 shows that an improved set of ULM simulations results from use of the parameter set obtained by the Monte Carlo search procedure (ULMM), which shows notable improvements over ULMA. Next, the 3 parameters to which NSE was most sensitive were identified for each basin (Fig. 12); these parameters (ULM3) are responsible for most of the increase in NSE for ULMM relative to ULMA.

Fig. 11.
Fig. 11.

Mean monthly streamflows (1960–69) for ULMA, ULMM, Noah, Sac, and observations for the six basins listed.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

Fig. 12.
Fig. 12.

(top panels) NSE values (multiplied by −1 for consistency with RMSE minimizations) corresponding to simulated vs observed streamflow for variation of individual parameter values following a Monte Carlo approach for the six basins listed. (bottom panels) For each basin, the scatterplots of the two most sensitive parameters are shown, with the best four simulations circled in blue which correspond to the lowest four points on each scatterplot. (top panels) Trends in the data are grouped into 50-simulation bins, with bin minima in blue showing an envelope of parameter sensitivity and bin means in red. The ADIMP parameter (not shown) was found to be highly sensitive for all basins with an approximate minimum value of 0.225.

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

The mean monthly hydrographs presented thus far illustrate the general relationship between modeled and observed flows for the two basins with particularly sporadic, peaked streamflow (ILLIN and BROOK); an examination of daily streamflows (Fig. 13) over a single water year provides additional insights. ILLIN was part of the Distributed Model Intercomparison Project (DMIP; Smith et al. 2004; Reed et al. 2004) wherein a number of well-known models (including Noah and Sac) had relatively poor average NSE of approximately 40% and required calibration to achieve average efficiencies of approximately 65%. This was due in part to the challenging hydroclimatology, from a hydrologic modeling perspective, which consists of extended dry periods followed by high intensity rainfall. These peaked conditions are apparent in the daily hydrograph (Fig. 13a). Noah underestimates high flows with overly broad peaks and overestimates low flows, whereas Sac and ULMA simulate peaks better but generally underestimate low flows. ULMM shows modest improvements in both peak and low-flow response relative to ULMA. A major part of the improvement in ULMM performance resulted from an increase in UZTWM, which dampens the rapid storm runoff response and increases low flows slightly as well.

Fig. 13.
Fig. 13.

Daily streamflows and precipitation over the 1964 water year for the (a) ILLIN and (b) BROOK basins. Observed flow (solid black), ULMM (dashed black line), ULMA (dashed orange line), Noah (dashed pink line), and Sac (dashed green line) are included. These two basins provided the greatest challenge in modeling, with relatively sporadic precipitation and sharp streamflow peaks (for BROOK, end of March snowmelt and soil thaw streamflow spike was poorly captured by the snow model and frozen soil physics).

Citation: Journal of Hydrometeorology 12, 6; 10.1175/2011JHM1361.1

BROOK is also characterized by low streamflow for most of the year, followed by large surges in late spring and early summer, complicated by the effects of snowmelt and frozen soil thawing, which are not present (or negligible) at ILLIN. It has the lowest annual precipitation of all study basins, and, in part for this reason, all models struggle to capture the peak flow response (Fig. 13b) that is somewhat masked in the multiyear average, because year-to-year differences in timing appear as a broad peak (see Fig. 11). Most notably, a substantial decrease in UZTWM is needed to increase model peak response as confirmed in both the parametric scatterplot in Fig. 12 and the daily hydrograph (Fig. 13b).

The impact of snowmelt is more visible in the hydrographs of the other four basins. MAYBE has the lowest annual precipitation of these four, and thus the soil response to snowmelt is of great importance (because much of the winter precipitation is released in spring as snowmelt). Noah and ULMA have flashy response to the large spring snowmelt, whereas an increase in the upper-zone storage in ULMM buffers the runoff rate to more closely match observations. Sac, which uses a different snow model than Noah and ULM, has similar streamflow timing to both ULMM and observations; however, it has higher ET rates (not shown), which ultimately result in an overall reduced hydrograph. Both the OROVI and CARNA basins receive most of their precipitation between November and March, of which a majority is stored as snow that ultimately runs off in the warm season. The large surplus of water during melt season for these basins makes the larger lower-zone storages of increased consequence for accurately simulating baseflow, runoff, and ET rates into the warm season. The wettest basin (CARNA) shows a marked improvement in streamflow simulations when the lower-zone hydraulic conductivity LZSK is reduced to attenuate baseflow. Similarly, an increase in the lower-zone storage LZFPM for the OROVI basin provides for more realistic spring and early summer streamflow.

Precipitation over the SANDY basin is distributed nearly evenly throughout the year with a slight peak in autumn. However, the storage and release of winter precipitation as snowmelt in the spring is the dominant seasonal streamflow feature. It follows that a reduction in LZSK attenuates the snowmelt response to yield a modest improvement in streamflow in ULMM. Sac gets the peak timing correct but with flows that are generally too low, whereas the Noah peaks are too broad and low flows are overestimated.

One way to organize the importance of model parameters by basin is through classification by an aridity index (AI), a metric first proposed by Budyko (1974),
e6
where Rnet-ann is the annual average net radiation; L is the latent heat of vaporization; and Pann is the mean annual precipitation, such that LPann is the amount of energy needed to evaporate the available precipitation Pann. As AI exceeds 1, a basin becomes increasingly arid (or water limited), whereas for AI less than 1 radiation becomes limiting (increasingly moist); the potential evaporation is theoretically reached at AI = 1. Comparing the respective AI values for each basin and its most sensitive parameters in Table 5 highlights the importance of upper-zone parameters for more arid basins and lower-zone parameters for wetter basins. For water-limited basins, UZTWM plays a critical role in partitioning infiltration and runoff, as well as controlling direct soil evaporation, whereas, for all basins, an increase of the allowable area that becomes impermeable during soil saturation (ADIMP) was beneficial in capturing streamflow timing related to large moisture input events, such as storms and snowmelt. Wetter basins benefited from an adjustment in the rate of baseflow response from the larger lower-zone reservoir, where LZSK was lowered from its a priori value and LZPK was increased.
Table 5.

AI for each study basin and the key parameters for improved streamflow, where a decreasing parameter is given in parentheses. Supplemental to the parameters listed, the ADIMP parameter was increased to 0.225 for all basins. Seasonal AI are for December–February (DJF), March–May (MAM), JJA, and September–October (SON).

Table 5.

Table 6 lists the NSE for ULMA, ULMM, ULM3, Noah, and Sac, computed for the six study basins for both training and evaluation periods (i.e., 12 cases). Beginning with the simulations using only a priori parameters (ULMA, Noah, and Sac), ULMA scored the highest four times, the median six times, and the lowest twice. The latter cases were for the same basin (BROOK, the most arid basin) for which none of the a priori simulations performed distinctively better than climatology, so this was somewhat of an outlier. Considering the entire set of simulations in Table 6, ULMM was the best model nine times, ULM3 was best twice, and Sac was best once. ULM3 was capable of realizing much of the performance gains from ULMM for most basins, including three exceptions where it outperformed ULMM. We attribute these exceptions to the random component in the ULMM parameter set (inherent in the Monte Carlo procedure), whereas ULM3 used a priori values for less-sensitive parameters that performed better in these cases. Sac outperformed all models during the validation period in the snowmelt-dominated MAYBE basin, for which Sac’s snow model was particularly well suited.

Table 6.

Summary statistics: Nash–Sutcliffe model efficiencies (%) for training (1960–69) and validation periods (1990–99) for the Noah, Sac, ULMA, ULMM, and ULM3 based on sensitivity and climate. Daily statistics are without parentheses and monthly statistics are in parentheses. Cases where ULM3 scored higher than ULMM for the respective period are bolded, and any model scoring higher than ULMM is italicized.

Table 6.

In summary, ULMA performed comparably or slightly better than Noah and Sac for the wetter basins, whereas it was generally on par or slightly poorer for the drier basins. With several or all parameters estimated (ULM3 and ULMM, respectively), significant improvements were realized for all basins with the largest occurring for the drier basins. The parameter adjustments of ULMM and ULM3 are underscored because, unlike Noah and Sac, ULM does not yet have an established set of default (or a priori) parameters. Such a parameter set depends in large part on the dynamic equilibrium of model soil moisture, which itself is dependent on the ET scheme, which comes from Sac and Noah, respectively. Therefore, the potential advantage of the model structure change was offset by the lack of customized ULM-specific parameters for the ULMA comparisons.

An experiment was conducted in which the parameter adjustments from ULM3 and ULMM were transferred to corresponding flux tower parameter sets, to attempt to improve model results for both surface fluxes and soil moisture (where applicable). However, the resulting simulations were not conclusively improved by this experiment, calling into question the merit of the parameter transfer strategy. In addition, comparing the seasonal [June–August (JJA)] and annual AI values from basin to point, there were notable discrepancies (Blodgett/OROVI: AIJJA = 65.7/64.8 and AIannual = 0.8/1.4; Niwot/MAYBE: AIJJA = 9.1/21.4 and AIannual = 1.5/1.6; Howland/SANDY: AIJJA = 4.7/5.5 and AIannual = 0.7/0.9; Brookings/BROOK: AIJJA = 8.7/13.9 and AIannual = 1.5/3.0). The apparent lack of transferability for Sac parameters is consistent with the findings of Gan and Burges (2006) and also follows the findings for other models (e.g., Heuvelmans et al. 2004; Abdulla and Lettenmaier 1997). Alternate approaches likely will be required for effective parameter transferability and regionalization of ULM.

6. Conclusions

The objective of this work was to assess the potential of ULM to improve upon the weaknesses of its parent models, Noah and Sac, while taking advantage of their respective strengths. Model performance was examined in ways relevant to the typical uses of these models, including prediction of land–atmosphere moisture and energy fluxes and streamflow. We established a default set of soil parameters for ULM, which we expected would perform differently from the Noah and Sac a priori values given an alternate equilibrium state within the new model structure. Finally, we evaluated a simple approach for transferring parameters estimated on the basis of basin streamflow prediction performance for energy and moisture flux predictions.

Our key findings are as follows:

  1. On the basis of observed flux data from four Ameriflux sites, the alternate soil moisture states of ULM result in diurnal variations of surface fluxes that are on par or modestly superior to Noah (Sac does not predict moisture or energy fluxes other than evapotranspiration and runoff).
  2. The nonlinearity of observed soil drying was better captured by ULM (as compared with Noah), whereas soil moisture response during wetter periods was more comparable between models. ULM had the smallest bias in a 10-site, multiyear soil moisture comparison. ULM and Noah comparably matched the observed seasonal cycle, with Sac’s seasonality notably too small.
  3. ULMA generally outperformed the Noah and Sac models over wetter basins; however, performance was more variable for drier basins, for which parameter adjustments were required to achieve competitive performance.
  4. Much of the streamflow accuracy improvements achievable by calibration (ULMM) can be realized by adjusting only three parameters (ULM3) while leaving the remaining parameters at the a priori values of the parent models (Noah and Sac).
  5. For drier basins, simulated streamflow was generally most sensitive to upper soil zone parameters, whereas the opposite was generally true for wetter basins.
Overall, ULM performance was encouraging. Parameter adjustments were ultimately needed to improve simulations comprehensively relative to the parent models. However, this has to be tempered by the fact that a suitable set of default (a priori) parameters has not yet been established for ULM, and the a priori parameters of its parent models may not be fully representative of the alternative quasi-equilibrium state within the new model structure. Work remains to find a reliable set of a priori parameters for ULM to further improve simulations of both moisture fluxes (streamflow and soil moisture) and energy fluxes without compromising one for the other. In its present form, ULM could be applied for any simulations for which Noah and Sac are currently used, with essentially the same capabilities for data assimilation and coupling as Noah. Additional testing will be required to evaluate the impact of ULM on atmospheric feedbacks and lateral moisture redistribution. Finally, further examination of parameter transfer strategies across gradients of vegetation, topography, soils, and climate is in order that will make ULM more applicable over large areas without the need for site-specific parameter estimation.

Acknowledgments

The authors are grateful to the numerous ICN and Ameriflux investigators who made data available for this research from the across the state of Illinois, Howland Forest, Brookings, Niwot Ridge, and Blodgett Forest Ameriflux sites. Additionally, the descriptions and assistance by Victor Koren were greatly appreciated. The work on which this paper is based was supported by NOAA Grant NA070AR4310210 to the University of Washington.

REFERENCES

  • Abdulla, F. A., , and Lettenmaier D. P. , 1997: Development of regional parameter estimation equations for a macroscale hydrologic model. J. Hydrol., 197, 230257.

    • Search Google Scholar
    • Export Citation
  • Anderson, E. A., 1973: National Weather Service River Forecast System—Snow accumulation and ablation model. NOAA Tech. Memo. NWS HYDRO-17, 217 pp.

    • Search Google Scholar
    • Export Citation
  • Anderson, R. M., , Koren V. I. , , and Reed S. M. , 2006: Using SSURGO data to improve Sacramento model a priori parameter estimates. J. Hydrol., 320, 103116.

    • Search Google Scholar
    • Export Citation
  • Arora, V. K., 2001: Streamflow simulations for continental-scale river basins in a global atmospheric general circulation model. Adv. Water Resour., 24, 775791.

    • Search Google Scholar
    • Export Citation
  • Baldocchi, D. D., 2003: Assessing the eddy-covariance technique for evaluating the carbon dioxide exchange rates of ecosystems: Past, present and future. Global Change Biol., 9, 479492.

    • Search Google Scholar
    • Export Citation
  • Baldocchi, D. D., , Finnigan J. , , Wilson K. , , Paw U K. T. , , and Falge E. , 2000: On measuring net ecosystem carbon exchange over tall vegetation on complex terrain. Bound.-Layer Meteor., 96, 257291.

    • Search Google Scholar
    • Export Citation
  • Bastidas, L. A., , Hogue T. S. , , Sorooshian S. , , Gupta H. V. , , and Shuttleworth W. J. , 2006: Parameter sensitivity analysis for different complexity land surface models using multicriteria methods. J. Geophys. Res., 111, D20101, doi:10.1029/2005JD006377.

    • Search Google Scholar
    • Export Citation
  • Betts, A. K., , Ball J. H. , , Beljaars A. C. M. , , Miller M. J. , , and Viterbo P. A. , 1996: The land surface-atmosphere interaction: A review based on observational and global modeling perspectives. J. Geophys. Res., 101 (D3), 72097225.

    • Search Google Scholar
    • Export Citation
  • Bohn, T. J., , Sonessa M. Y. , , and Lettenmaier D. P. , 2010: Seasonal hydrologic forecasting: Is there a role for multimodel ensemble methods? J. Hydrometeor., 11, 13581372.

    • Search Google Scholar
    • Export Citation
  • Bonan, G. B., , Pollard D. , , and Thompson S. L. , 1992: Effects of boreal forest vegetation on global climate. Nature, 359, 716718.

  • Budyko, M. I., 1974: Climate and Life. Academic Press, 508 pp.

  • Burnash, R. J. C., , Ferral R. L. , , and McGuire R. A. , 1973: A generalized streamflow simulation system—Conceptual modeling for digital computers. U.S. National Weather Service and California Department of Water Resources Joint Federal and State River Forecast Center Tech. Rep., 204 pp.

    • Search Google Scholar
    • Export Citation
  • Chen, F., and Coauthors, 1996: Modeling of land-surface evaporation by four schemes and comparison with FIFE observations. J. Geophys. Res., 101, 72517268.

    • Search Google Scholar
    • Export Citation
  • Chen, F., , Janjic Z. , , and Mitchell K. E. , 1997: Impact of atmospheric surface-layer parameterizations in the new land-surface scheme of the NCEP mesoscale Eta model. Bound.-Layer Meteor., 85, 391421.

    • Search Google Scholar
    • Export Citation
  • Daly, C., , Neilson R. P. , , and Phillips D. L. , 1994: A statistical topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33, 140158.

    • Search Google Scholar
    • Export Citation
  • D’Odorico, P., , Caylor K. , , Okin G. S. , , and Scanlon T. M. , 2007: On soil moisture–vegetation feedbacks and their possible effects on the dynamics of dryland ecosystems. J. Geophys. Res., 112, G04010, doi:10.1029/2006JG000379.

    • Search Google Scholar
    • Export Citation
  • Duan, Q., , Sorooshian S. , , and Gupta V. K. , 1994: Optimal use of the SCE-UA global optimization method for calibrating watershed models. J. Hydrol., 158 (3–4), 265284.

    • Search Google Scholar
    • Export Citation
  • Ek, M. B., , and Holtslag A. A. M. , 2004: Influence of soil moisture on boundary layer cloud development. J. Hydrometeor., 5, 8699.

  • Ek, M. B., , Mitchell K. E. , , Lin Y. , , Rogers E. , , Grunmann P. , , Koren V. , , Gayno G. , , and Tarpley J. D. , 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model. J. Geophys. Res., 22, 8851, doi:10.1029/2002JD003296.

    • Search Google Scholar
    • Export Citation
  • Falge, E., and Coauthors, 2001: Gap filling strategies for long term energy flux data sets. Agric. For. Meteor., 107, 7177.

  • Finnerty, B. D., , Smith M. B. , , Seo D. J. , , Koren V. I. , , and Moglen G. , 1997: Sensitivity of the Sacramento Soil Moisture Accounting model to space–time scale precipitation inputs from NEXRAD. J. Hydrol., 203 (1–4), 2138.

    • Search Google Scholar
    • Export Citation
  • Gan, T. Y., , and Burges S. J. , 2006: Assessment of soil-based and calibrated parameters of the Sacramento model and parameter transferability. J. Hydrol., 320 (1–2), 117131.

    • Search Google Scholar
    • Export Citation
  • Goldstein, A. H., and Coauthors, 2000: Effects of climate variability on the carbon dioxide, water, and sensible heat fluxes above a ponderosa pine plantation in the Sierra Nevada (CA). Agric. For. Meteor., 101 (2–3), 113129.

    • Search Google Scholar
    • Export Citation
  • Gupta, H. V., , Sorooshian S. , , and Yapo P. O. , 1998: Toward improved calibration of hydrologic models: Multiple and noncommensurable measures of information. Water Resour. Res., 34, 751763.

    • Search Google Scholar
    • Export Citation
  • Gupta, H. V., , Bastidas L. A. , , Sorooshian S. , , Shuttleworth W. J. , , and Yang Z. L. , 1999: Parameter estimation of a land surface scheme using multicriteria methods. J. Geophys. Res., 104, 491503.

    • Search Google Scholar
    • Export Citation
  • Gutman, G., , and Ignatov A. , 1998: The derivation of the green vegetation fraction from NOAA/AVHRR data for use in numerical weather prediction models. Int. J. Remote Sens., 19, 15331543.

    • Search Google Scholar
    • Export Citation
  • Heuvelmans, G., , Muys B. , , and Feyen J. , 2004: Evaluation of hydrological model parameter transferability for simulating the impact of land use on catchment hydrology. Phys. Chem. Earth, 29 (11–12), 739747.

    • Search Google Scholar
    • Export Citation
  • Hollinger, D. Y., , Goltz S. M. , , Davidson E. A. , , Lee J. T. , , Tu K. , , and Valentine H. T. , 1999: Seasonal patterns and environmental control of carbon dioxide and water vapour exchange in an ecotonal boreal forest. Global Change Biol., 5, 891902.

    • Search Google Scholar
    • Export Citation
  • Hollinger, S. E., , and Isard S. A. , 1994: A soil moisture climatology of Illinois. J. Climate, 7, 822833.

  • Jarvis, P. G., 1976: The interpretation of leaf water potential and stomatal conductance found in canopies in the field. Philos. Trans. Roy. Soc. London, B273, 593610.

    • Search Google Scholar
    • Export Citation
  • Johansen, O., 1977: Thermal conductivity of soils (in Norwegian). Ph.D. thesis, University of Trondheim, 236 pp.

  • Koren, V. I., 2006: Parameterization of frozen ground effects: Sensitivity to soil properties. Predictions in Ungauged Basins: Promises and Progress, M. Sivapalan, Ed., IAHS, 125–133.

    • Search Google Scholar
    • Export Citation
  • Koren, V. I., , Schaake J. , , Mitchell K. , , Duan Q.-Y. , , Chen F. , , and Baker J. M. , 1999: A parameterization of snowpack and frozen ground intended for NCEP weather and climate models. J. Geophys. Res., 104 (D16), 19 56919 585.

    • Search Google Scholar
    • Export Citation
  • Koren, V. I., , Smith M. B. , , Wang D. , , and Zhang Z. , 2000: Use of soil property data in the derivation of conceptual rainfall-runoff model parameters. Preprints, 15th Conf. on Hydrology, Long Beach, CA, Amer. Meteor. Soc., 2.16. [Available online at http://ams.confex.com/ams/annual2000/techprogram/paper_6074.htm.]

    • Search Google Scholar
    • Export Citation
  • Koren, V. I., , Smith M. , , and Duan Q. , 2003: Use of a priori parameter estimates in the derivation of spatially consistent parameter sets of rainfall-runoff models. Calibration of Watershed Models, Q. Duan et al., Eds., Vol. 6, Water Science and Applications, Amer. Geophys. Union, 239–254.

    • Search Google Scholar
    • Export Citation
  • Koster, R. D., , Suarez M. J. , , Ducharne A. , , Stieglitz M. , , and Kumar P. , 2000: A catchment-based approach to modeling land surface processes in a general circulation model 1. Model structure. J. Geophys. Res., 105 (D20), 24 80924 822.

    • Search Google Scholar
    • Export Citation
  • Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305, 11381140.

  • Livneh, B., , Xia Y. , , Mitchell K. E. , , Ek M. B. , , and Lettenmaier D. P. , 2010: Noah LSM snow model diagnostics and enhancements. J. Hydrometeor., 11, 721738.

    • Search Google Scholar
    • Export Citation
  • Loescher, H. W., , Law B. E. , , Mahrt L. , , Hollinger D. , , Campbell J. L. , , and Wofsy S. E. , 2006: Uncertainties in, and interpretation of, carbon flux estimates using the eddy covariance technique. J. Geophys. Res., 111, D21S90, doi:10.1029/2005JD006932.

    • Search Google Scholar
    • Export Citation
  • Mahmood, R., , and Hubbard K. G. , 2003: Simulating sensitivity of soil moisture and evapotranspiration under heterogeneous soils and land uses. J. Hydrol., 280, 7290.

    • Search Google Scholar
    • Export Citation
  • Mahrt, L., , and Ek M. , 1984: The influence of atmospheric stability on potential evaporation. J. Climate Appl. Meteor., 23, 222234.

  • Mahrt, L., , and Pan H. L. , 1984: A two-layer model of soil hydrology. Bound.-Layer Meteor., 29, 120.

  • Maurer, E. P., , Wood A. W. , , Adam J. C. , , Lettenmaier D. P. , , and Nijssen B. , 2002: A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States. J. Climate, 15, 32373251.

    • Search Google Scholar
    • Export Citation
  • McCumber, M., , and Pielke R. A. , 1981: Simulation of effects of surface fluxes of heat and moisture in a mesoscale numerical model. Part 1: Soil layer. J. Geophys. Res., 86, 99299938.

    • Search Google Scholar
    • Export Citation
  • Milly, P. C. D., , Betancourt J. , , Falkenmark M. , , Hirsch R. M. , , Kundzewicz Z. W. , , Lettenmaier D. P. , , and Stouffer R. J. , 2008: Stationarity is dead: Whither water management? Science, 319, 573574.

    • Search Google Scholar
    • Export Citation
  • Mishra, V., , Cherkauer K. A. , , and Shukla S. , 2010: Assessment of drought due to historic climate variability and projected future climate change in the midwestern United States. J. Hydrometeor., 11, 4668.

    • Search Google Scholar
    • Export Citation
  • Nash, J. E., , and Sutcliffe J. V. , 1970: River flow forecasting through conceptual models part I — A discussion of principles. J. Hydrol., 10, 282290.

    • Search Google Scholar
    • Export Citation
  • Niyogi, D., , Alfieri J. G. , , Blanken P. D. , , Chen F. , , LeMone M. A. , , Mitchell K. E. , , and Ek M. B. , 2008: Estimation of the minimum canopy resistance for croplands and grasslands using data from the 2002 International H20 Project. Mon. Wea. Rev., 136, 44524469.

    • Search Google Scholar
    • Export Citation
  • Pan, H. L., , and Mahrt L. , 1987: Interaction between soil hydrology and boundary layer development. Bound.-Layer Meteor., 38, 185202.

  • Pielke, R. A., , Avissar R. , , Raupach M. , , Dolman A. J. , , Zeng X. , , and Denning A. S. , 1998: Interactions between the atmosphere and terrestrial ecosystems: Influence on weather and climate. Global Change Biol., 4, 461475.

    • Search Google Scholar
    • Export Citation
  • Reed, S., , Koren V. , , Smith M. , , Zhang Z. , , Moreda F. , , and Seo D.-J. , 2004: Overall distributed model intercomparison project results. J. Hydrol., 298, 2760.

    • Search Google Scholar
    • Export Citation
  • Rosero, E., , Gulden L. E. , , Yang Z.-L. , , De Goncalves L. G. , , Niu G.-Y. , , and Kaheil Y. H. , 2011: Ensemble evaluation of hydrologically enhanced Noah-LSM: Partitioning of the water balance in high-resolution simulations over the Little Washita River experimental watershed. J. Hydrometeor., 12, 4564.

    • Search Google Scholar
    • Export Citation
  • Schaake, J. C., , Koren V. I. , , Duan Q.-Y. , , Mitchell K. E. , , and Chen F. , 1996: Simple water balance model for estimating runoff at different spatial and temporal scales. J. Geophys. Res., 101, 74617475.

    • Search Google Scholar
    • Export Citation
  • Schaake, J. C., , Cong S. , , and Duan Q. , 2006: The US MOPEX data set. Large Sample Basin Experiments for Hydrological Model Parameterization: Results of the Model Parameter Experiment—MOPEX, V. Andrēasean et al., Eds., IAHS, 9–28.

    • Search Google Scholar
    • Export Citation
  • Seneviratne, S. I., , Corti T. , , Davin E. L. , , Hirschi M. , , Jaeger E. B. , , Lehner I. , , Orlowsky B. , , and Teuling A. J. , 2010: Investigating soil moisture–climate interactions in a changing climate: A review. Earth-Sci. Rev., 99 (3–4), 125161.

    • Search Google Scholar
    • Export Citation
  • Smith, M. B., , Laurine D. P. , , Koren V. I. , , Reed S. M. , , and Zhang Z. , 2003: Hydrologic model calibration in the National Weather Service. Calibration of Watershed Models, Q. Duan et al., Eds., Water Science and Application Series, Vol. 6, Amer. Geophys. Union, 133–152.

    • Search Google Scholar
    • Export Citation
  • Smith, M. B., , Seo D. , , Koren V. I. , , Reed S. M. , , Zhang Z. , , Duan Q. , , Moreda F. , , and Cong S. , 2004: The distributed model intercomparison project (DMIP): Motivation and experiment design. J. Hydrol., 298, 426.

    • Search Google Scholar
    • Export Citation
  • Sridhar, V., , Elliot R. L. , , Chen F. , , and Botzge J. A. , 2002: Validation of the NOAH-OSU land surface model using surface flux measurements in Oklahoma. J. Geophys. Res., 107, 4418, doi:10.1029/2001JD001306.

    • Search Google Scholar
    • Export Citation
  • Tang, Y., , Reed P. , , Wagner T. , , and van Werkhoven K. , 2007: Comparing sensitivity of analysis methods to advance lumped watershed model identification and evaluation. Hydrol. Earth. Syst. Sci., 11, 793817.

    • Search Google Scholar
    • Export Citation
  • Tateishi, R., , and Ahn C. H. , 1996: Mapping evapotranspiration and water balance for global land surfaces. J. Photogramm. Remote Sens., 51, 209215.

    • Search Google Scholar
    • Export Citation
  • Taylor, C. M., , and Ellis R. J. , 2006: Satellite detection of soil moisture impacts on convection at the mesoscale. Geophys. Res. Lett., 33, L03404, doi:10.1029/2005GL025252.

    • Search Google Scholar
    • Export Citation
  • Thiemann, M., , Trosset M. , , Gupta H. , , and Sorooshian S. , 2001: Bayesian recursive parameter estimation for hydrologic models. Water Resour. Res., 37, 25212535.

    • Search Google Scholar
    • Export Citation
  • Turnipseed, A. A., , Blanken P. D. , , Anderson D. E. , , and Monson R. K. , 2002: Energy budget above a high-elevation subalpine forest in complex topography. Agric. For. Meteor., 110, 177201.

    • Search Google Scholar
    • Export Citation
  • van Werkhoven, K., , Wagener T. , , Reed P. , , and Tang Y. , 2008: Characterization of watershed model behavior across a hydroclimatic gradient. Water Resour. Res., 44, W01429, doi:10.1029/2007WR006271.

    • Search Google Scholar
    • Export Citation
  • Verseghy, D., 1996: Preliminary considerations for runoff modelling in GCMS. Nord. Hydrol., 27, 117128.

  • Vrugt, J. A., , Gupta H. V. , , Dekker S. C. , , Sorooshian S. , , Wagener T. , , and Bouten W. , 2006: Application of stochastic parameter optimization to the Sacramento Soil Moisture Accounting model. J. Hydrol., 325 (1–4), 288307.

    • Search Google Scholar
    • Export Citation
  • Wang, W., , and Kumar A. , 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103, 28 63728 646.

    • Search Google Scholar
    • Export Citation
  • Wetzel, P. J., , Argentini S. , , and Boone A. , 1996: Role of land surface in controlling daytime cloud amount: Two case studies in the GCIP-SW area. J. Geophys. Res., 101 (D3), 73597370.

    • Search Google Scholar
    • Export Citation
  • Wilson, K., and Coauthors, 2002: Energy balance closure at FLUXNET sites. Agric. For. Meteor., 113 (1–4), 223243.

  • Xiu, A., , and Pleim J. E. , 2001: Development of a land surface model. Part I: Application in a mesoscale meteorological model. J. Appl. Meteor., 40, 192209.

    • Search Google Scholar
    • Export Citation
  • Yapo, P. O., , Gupta H. V. , , and Sorooshian S. , 1998: Multi-objective global optimization for hydrologic models. J. Hydrol., 204, 8397.

  • Zhang, L., , Dawes W. R. , , and Walker G. R. , 2001: Response of mean annual evapotranspiration to vegetation changes at catchment scale. Water Resour. Res., 37, 701708.

    • Search Google Scholar
    • Export Citation
Save