• Atlas, R., Wolfson N. , and Terry J. , 1993: The effect of SST and soil moisture anomalies on GLA model simulation of the 1988 U.S. summer drought. J. Climate, 6 , 20342048.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Delworth, T., and Manabe S. , 1989: The influence of soil wetness on near-surface atmospheric variability. J. Climate, 2 , 14471462.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., 2000: Using a global soil wetness dataset to improve seasonal climate simulation. J. Climate, 13 , 29002922.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., 2001: Climate drift in a coupled land–atmosphere model. J. Hydrometeor., 2 , 89100.

  • Dirmeyer, P. A., 2003: The role of the land surface background state in climate predictability. J. Hydrometeor., 4 , 599610.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., and Zeng F. J. , 1999a: An update to the distribution and treatment of vegetation and soil properties in SSiB. COLA Tech. Rep. 78, 25 pp. [Available from the Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].

  • Dirmeyer, P. A., and Zeng F. J. , 1999b: Precipitation infiltration in the Simplified SiB land surface scheme. J. Meteor. Soc. Japan, 77 , 291303.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., and Tan L. , 2001: A multi-decadal global land-surface data set of state variables and fluxes. COLA Tech. Rep. 102, 43 pp. [Available from the Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].

  • Dirmeyer, P. A., and Zhao M. , 2004: Flux replacement as a method to diagnose coupled land–atmosphere model feedback. J. Hydrometeor., 5 , 10341048.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., Dolman A. J. , and Sato N. , 1999: The Global Soil Wetness Project: A pilot project for global land surface modeling and validation. Bull. Amer. Meteor. Soc., 80 , 851878.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douville, H., 2003: Assessing the influence of soil moisture on seasonal climate variability with AGCMs. J. Hydrometeor., 4 , 10441066.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douville, H., 2004: Relevance of soil moisture for seasonal atmospheric predictions: Is it an initial value problem? Climate Dyn., 22 , 429446.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Durre, I., Wallace J. M. , and Lettenmaier D. P. , 2000: Dependence of extreme daily maximum temperatures on antecedent soil moisture in the contiguous United States during summer. J. Climate, 13 , 26412651.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fennessy, M. J., and Shukla J. , 1999: Impact of initial soil wetness on seasonal atmospheric prediction. J. Climate, 12 , 31673180.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and Hunt B. G. , 1987: Interannual variability of the simulated hydrology in a climatic model—Implications for drought. Climate Dyn., 1 , 113130.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gruber, A., Su X. , Kanamitsu M. , and Schemm J. , 2000: The comparison of two merged rain gauge–satellite precipitation datasets. Bull. Amer. Meteor. Soc., 81 , 26312644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kinter, J. L., and Coauthors, 1997: Formulation. Vol. 1, The COLA atmosphere–biosphere general circulation model. COLA Tech. Rep. 51, 46 pp. [Available from the Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].

  • Koster, R. D., and Suarez M. J. , 2001: Soil moisture memory in climate models. J. Hydrometeor., 2 , 558570.

  • Koster, R. D., Suarez M. J. , and Heiser M. , 2000: Variance and predictability of precipitation at seasonal-to-interannual timescales. J. Hydrometeor., 1 , 2646.

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

  • Oglesby, R. J., 1991: Springtime soil moisture, natural climate variability, and North American drought as simulated by the NCAR Community Climate Model I. J. Climate, 4 , 890897.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reale, O., and Dirmeyer P. A. , 2002: Modeling the effect of land surface evaporation variability on precipitation variability. Part I: General response. J. Hydrometeor., 3 , 433450.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reale, O., Dirmeyer P. A. , and Schlosser C. A. , 2002: Modeling the effect of land surface evaporation variability on precipitation variability. Part II: Time- and space-scale structure. J. Hydrometeor., 3 , 451466.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and Smith T. M. , 1994: Improved global sea surface temperature analyses using optimal interpolation. J. Climate, 7 , 929948.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ropelewski, C. F., Janowiak J. E. , and Halpert M. F. , 1985: The analysis and display of real time surface climate data. Mon. Wea. Rev., 113 , 11011107.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schlosser, C. A., and Milly P. C. D. , 2002: A model-based investigation of soil moisture predictability and associated climate predictability. J. Hydrometeor., 3 , 483501.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schneider, E. K., 2002: The causes of differences between equatorial Pacific SST simulations of two coupled ocean–atmosphere general circulation models. J. Climate, 15 , 449469.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sellers, P. J., Mintz Y. , Sud Y. C. , and Dalcher A. , 1986: A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci., 43 , 505531.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shukla, J., and Mintz Y. , 1982: Influence of land-surface evapotranspiration on the earth’s climate. Science, 215 , 14981501.

  • Shukla, J., and Coauthors, 2000: Dynamical seasonal prediction. Bull. Amer. Meteor. Soc., 81 , 25932606.

  • Walker, J., and Rowntree P. R. , 1977: The effect of soil moisture on circulation and rainfall in a tropical model. Quart. J. Roy. Meteor. Soc., 103 , 2946.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xue, Y., Sellers P. J. , Kinter J. L. , and Shukla J. , 1991: A simplified biosphere model for global climate studies. J. Climate, 4 , 345364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xue, Y., Zeng F. J. , and Schlosser C. A. , 1996: SSiB and its sensitivity to soil properties—A case study using HAPEX-Mobilhy data. Global Planet. Change, 13 , 183194.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yeh, T-C., Wetherald R. T. , and Manabe S. , 1984: The effect of soil moisture on the short-term climate and hydrology change—A numerical experiment. Mon. Wea. Rev., 112 , 474490.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery
    Fig. 1.

    (top) Evolution of the climatological root zone soil wetness in GOLD from 1 Jun to Sep and (bottom) the difference between the mean of all GCM simulations and GOLD for the same times of year. Units are percentage of saturation.

  • View in gallery
    Fig. 2.

    Schematic of the evolution of soil wetness from two types of initial conditions (climatological vs realistic) to two different equilibria depending on whether downward surface fluxes are specified or free to interact with the land surface.

  • View in gallery
    Fig. 3.

    The number of boreal summer seasons out of 18 for which the LIC and CTL cases have statistically significant differences in (left) root zone soil wetness, (middle) latent heat flux, and (right) sensible heat flux. Jun is the first month of the integrations, and Sep is the last.

  • View in gallery
    Fig. 4.

    Signal-to-noise ratio for root zone soil wetness for (top) CTL and (bottom) LIC during Jun and Sep.

  • View in gallery
    Fig. 5.

    Signal and noise components, expressed in terms of standard deviation, for root zone soil wetness for (top) CTL and (bottom) LIC during Jul.

  • View in gallery
    Fig. 6.

    The change in spatial anomaly correlation coefficient (land points north of 60°S) for precipitation by month and for the 4-month mean (right shaded area) from CTL case values for each of five cases (see Table 1 for a description of the cases).

  • View in gallery
    Fig. 7.

    As in Fig. 6 but for change in root-mean-square error (units are mm day−1).

  • View in gallery
    Fig. 8.

    As in Fig. 6 but for near-surface air temperature.

  • View in gallery
    Fig. 9.

    As in Fig. 7 but for near-surface air temperature (units are K).

  • View in gallery
    Fig. 10.

    Four regions over which monthly and seasonal precipitation skill are compared.

  • View in gallery
    Fig. 11.

    Seasonal SACC for precipitation in cases LIC (solid) and P (hatched) for each year, and in the last column (shaded) the mean of the 18 yr. Asterisks denote statistical significance (see text for details).

  • View in gallery
    Fig. 12.

    As in Fig. 11 but for Jun of the CTL (solid) and LIC (hatched) cases over North America.

  • View in gallery
    Fig. 13.

    Jun 1988 precipitation anomalies vs the 1982–99 mean for (top) observations and (bottom) six model cases. Spatial correlations of each case with observations are shown in the model panels. Units for anomalies are mm day−1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 85 29 11
PDF Downloads 47 13 6

The Land Surface Contribution to the Potential Predictability of Boreal Summer Season Climate

Paul A. DirmeyerCenter for Ocean–Land–Atmosphere Studies, Calverton, Maryland

Search for other papers by Paul A. Dirmeyer in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The role of the land surface in contributing to the potential predictability of the boreal summer climate is investigated with a coupled land–atmosphere climate model. Ensemble simulations for 1982–99 have been conducted with specified observed sea surface temperatures (SSTs). Several treatments of the land surface are investigated: climatological land surface initialization, realistic initialization of soil wetness, and a series of experiments where downward surface fluxes over land are replaced with observed proxies of precipitation, shortwave, and longwave radiation. Without flux replacement the model exhibits strong drift in soil wetness and both systematic errors and poor simulation of interannual variations of precipitation and near-surface temperature. With flux replacement there are large improvements in simulation of both spatial patterns and interannual variability of precipitation and surface temperature. The land surface apparently does contribute, through positive feedback with the atmosphere, to regional climate anomalies. However, because of the sizeable noise component in precipitation, the strong land–atmosphere feedback may not translate into reliable enhancements in predictability, particularly in years of weak anomalies in the land surface initial conditions at the start of boreal summer.

Corresponding author address: Dr. Paul A. Dirmeyer, Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705-3106. Email: dirmeyer@cola.iges.org

Abstract

The role of the land surface in contributing to the potential predictability of the boreal summer climate is investigated with a coupled land–atmosphere climate model. Ensemble simulations for 1982–99 have been conducted with specified observed sea surface temperatures (SSTs). Several treatments of the land surface are investigated: climatological land surface initialization, realistic initialization of soil wetness, and a series of experiments where downward surface fluxes over land are replaced with observed proxies of precipitation, shortwave, and longwave radiation. Without flux replacement the model exhibits strong drift in soil wetness and both systematic errors and poor simulation of interannual variations of precipitation and near-surface temperature. With flux replacement there are large improvements in simulation of both spatial patterns and interannual variability of precipitation and surface temperature. The land surface apparently does contribute, through positive feedback with the atmosphere, to regional climate anomalies. However, because of the sizeable noise component in precipitation, the strong land–atmosphere feedback may not translate into reliable enhancements in predictability, particularly in years of weak anomalies in the land surface initial conditions at the start of boreal summer.

Corresponding author address: Dr. Paul A. Dirmeyer, Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705-3106. Email: dirmeyer@cola.iges.org

1. Introduction

Within the past three decades, considerable attention has been placed on soil-moisture memory and its role in coupled land–climate variations (e.g., Walker and Rowntree 1977; Yeh et al. 1984; Gordon and Hunt 1987; Oglesby 1991; Atlas et al. 1993; Fennessy and Shukla 1999; Durre et al. 2000, to name a few). The landmark model-based investigation of Delworth and Manabe (1989) introduced a linear, Markovian framework to describe the controls of soil-moisture memory for a simple bucket land model within the Geophysical Fluid Dynamics Laboratory (GFDL) atmospheric general circulation model (AGCM). With this approach, soil-moisture “memory” (i.e., the persistence of soil-moisture autocorrelation) is viewed as an exponentially decaying process with an e-folding time scale determined by the ratio of the soil column’s water-holding capacity to potential evaporation. Recently, some of the simplifying assumptions of the Delworth and Manabe (1989) analysis of soil-moisture memory were addressed by Koster and Suarez (2001) and applied to a more complex land model [based on the Simple Biosphere (SiB) model of Sellers et al. (1986)] coupled to an AGCM. From their analysis, a more general expression of the controls of soil-moisture autocorrelation, ρ(w), was developed, which qualitatively states
i1525-7541-6-5-618-eq1

The expression above implies that because of soil moisture’s covariability with evaporation and runoff, there is an “inherent” memory of soil moisture (in that soil moisture’s influence on surface water fluxes, in turn, regulates soil-moisture memory). However, also important are the controls from the atmosphere and the feedbacks between antecedent soil-moisture conditions and subsequent atmospheric variability (in particular, precipitation variations). Recently, Schlosser and Milly (2002) developed analytic relationships (based on a linear, Markov framework) between soil-moisture memory, soil-moisture predictability, and (locally) associated atmospheric predictability, and these relationships were applied to ensemble climate simulations to quantify the support that soil-moisture memory provides to atmospheric predictability. In addition, Koster et al. (2000) provide numerical evidence to suggest that intermediate soil-moisture regimes (i.e., not excessively dry or wet on average) would likely support the strongest feedbacks between soil moisture and precipitation, found to be true in many AGCMs (Koster et al. 2004). Thus, soil-moisture memory not only breeds soil-moisture predictability but likely contributes to atmospheric predictability—and, potentially, skill in climate prediction.

Douville (2003) concluded that the structure of the experiment can affect the apparent sensitivity of a climate model to land–atmosphere interactions. Typical approaches involve replacing or nudging land surface state variables (e.g., Dirmeyer 2000; Douville 2003) or controlling the fluxes from land to atmosphere (e.g., Koster et al. 2000; Reale and Dirmeyer 2002; Reale et al. 2002). Specification or relaxation of soil wetness assumes that the land surface scheme (LSS) can process anomalies in the fluxes of precipitation, radiation, heat, etc. into a certain kind of soil-moisture anomaly. Here we move upstream in the land–atmosphere feedback loop and ask if the LSS can communicate atmospheric climate anomalies back to the atmosphere. Put another way, what is the role of the land surface in the feedback loop, and is the LSS a hindrance to realizing climate predictability or do the problems reside in the AGCM?

This paper elaborates on the study of land surface impacts on dynamical seasonal prediction during boreal summer (Dirmeyer and Zhao 2004). In Dirmeyer and Zhao, large systematic errors in the climate model are found, and seasonal prediction skill is rather insensitive to the specification of realistic initial soil wetness. It is hypothesized that the errors in downward radiation and precipitation at the surface contribute to the lack of sensitivity. Replacement of the erroneous downward fluxes reduces systematic errors and improves the simulation of climate anomalies, suggesting that the land–atmosphere feedback contributes to the climate drift, but is not necessarily the cause of it. Improvement in the simulation of year-to-year variations in climate is even more evident. The paper concludes that the land surface can communicate climate anomalies back to the atmosphere, given nominal meteorological forcing. This paper focuses specifically on the land surface component of climate predictability, using the same set of climate model integrations as Dirmeyer and Zhao (2004), which are reviewed in section 2. Section 3 examines the character of the simulated soil wetness. Near-surface temperature and precipitation skill are analyzed in section 4. Section 5 examines the model performance in some specific regions for seasonal climate anomalies during the period of study. Conclusions are given in section 6. Throughout the paper we will refer to figures from Dirmeyer and Zhao (2004) to clarify certain points.

2. Models and setup

a. Models and datasets

The climate model and experiments are the same as described in Dirmeyer and Zhao (2004) and are only briefly described here. The atmospheric model is version 2.2 of the COLA AGCM at a spectral resolution of T63 (1.875° longitude and latitude) and 18 levels (Kinter et al.1997; Schneider 2002). It is coupled to an updated version of the Simplified Simple Biosphere Model (SSiB; Xue et al. 1991, 1996) as described by Dirmeyer and Zeng (1999a). Sea surface temperatures (SSTs) are specified from the weekly analysis of Reynolds and Smith (1994).

Ensembles of 10 seasonal hindcasts were generated for each of the boreal summer seasons of 1982–99 following the method of the Dynamical Seasonal Prediction (DSP) Project (Shukla et al. 2000). In the cases with climatological land initial conditions (ICs), initial soil wetness for all integrations is derived from a 2-yr mean SSiB climatology from the Global Soil Wetness Project (GSWP; Dirmeyer et al. 1999) as described by Dirmeyer and Zeng (1999b). For all other cases, the initial soil wetness state is taken from a 21-yr offline integration of SSiB, referred to as the Global Offline Land surface Dataset (GOLD; Dirmeyer and Tan 2001). Table 1 repeats the list of experiments from Dirmeyer and Zhao (2004), with emphasis on the four cases at the top of the table. They comprise a two-by-two matrix of experiments with two-dimensional parameters. One is the choice of soil wetness initial conditions, as described above. The other is the replacement of downward surface fluxes of radiation and precipitation from atmosphere to land at each time step of every integration. The fluxes used to replace the AGCM predicted fluxes (precipitation, downward shortwave and longwave radiation over land points only) are the same forcing fields used to drive SSiB in producing the GOLD dataset. Dirmeyer and Zhao (2004) give a thorough description of the derivation of these flux data. Flux replacement interrupts the feedback loop between land and atmosphere. Note that in flux replacement, the AGCM continues to predict downward radiation and precipitation, even when these fluxes are not coupled to the LSS. The precipitation values that are analyzed and validated throughout this paper are those produced by the AGCM parameterizations.

Dirmeyer and Zhao (2004) provide a description of the model errors in precipitation and temperature (their Figs. 1 and 2), the skill of this model in simulating interannual variations (their Figs. 3 and 4), and biases in downward radiation fluxes over land (their Fig. 6). Discovery of these biases and errors helped to motivate the flux replacement approach. Comparisons to observations use the datasets of Ropelewski et al. (1985) for near-surface air temperature, and Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) data of Gruber et al. (2000).

b. Variability calculations

We examine the impact of soil wetness initialization and flux replacement (which alters the feedback between land and atmosphere) by inspecting the variance in the seasonal runs. Since we have monthly mean values from 18 yr of seasonal simulations with ensembles of 10 members each, we can calculate the grand variance
i1525-7541-6-5-618-e1
where the sums of the means for a given month are over j ensemble members and n years of simulations, w is the root zone soil wetness (dimensionless, where 1 = saturation and 0 = complete desiccation), and is the grand mean over all j and n. The total variance includes a noise component; the mean of the intraensemble variances from each year,
i1525-7541-6-5-618-e2
as well as a signal component; and the interannual variance of the ensemble means [w]n,
i1525-7541-6-5-618-e3
This approached is used in section 3a(2).

3. Results

a. Soil wetness properties

1) Mean fields

Figure 1 shows the evolution of the 18-yr mean of root zone soil wetness from the 1 June initial conditions to September from GOLD, and the difference from GOLD of the control (CTL) ensemble mean. The two pairs of plots represent the SSiB simulated soil wetness when driven by combined observations/analyses in an uncoupled mode, and when fully coupled to an AGCM. The initial conditions for the CTL case are standard for Center for Ocean–Land–Atmosphere Studies (COLA) climate model integrations, and are not equal to the GOLD mean. The GOLD rendition of soil wetness clearly shows the global monsoon impact, with soils becoming much wetter during summer over northern South America and Central America, the Sahel and Ethiopia, south and Southeast Asia. Some drying is seen in the Southern Hemisphere and in the middle latitudes of the Northern Hemisphere (e.g., the contiguous United States away from the coast of the Gulf of Mexico). The climatological initial soil wetness in the CTL case at the beginning of June shows a great deal of regional deviation from the 18-yr GOLD climatology, but no appreciable global bias. Only a small fraction of the land surface deviates by more than ±0.10. However, by September, extreme drying in the CTL case is quite evident over most regions. Regions that are already arid are the only large areas that do not dry out more strongly than the GOLD data suggest, as well as some scattered regions where the AGCM shows a strong positive systematic error in rainfall (Dirmeyer 2003).

When one considers the 2 × 2 matrix of experiments at the top of Table 1, we have in essence two different starting points for the land surface (the choice of initial soil wetness) and two different end points (determined by the climate felt by the land surface; either AGCM or specified GOLD fluxes). Table 2 suggests the progression each case takes by showing the monthly rms differences of the 18-yr mean root zone soil wetness between each combination of cases calculated over all land points between 60°S and 80°N. Cases that have the same atmospheric forcing but different land initial conditions (CTL-LIC and cPLS-PLS) converge toward smaller rms differences as the season progresses. The pairings with identical initial conditions and different atmospheric forcings (CTL-cPLS and LIC-PLS) start identically but steadily diverge.

The behavior of the cross cases is particularly interesting. CTL versus PLS shows a marked initial convergence and reduction of rms from their different initial states, followed by divergence to states as different as the initial states by September. The other pair (LIC versus cPLS) maintains a larger rms difference early in the season and shows less change with time.

A similar situation is found when the mean differences are examined (Table 3). Here it appears that the paths of soil wetness of CTL and PLS are crossing over one another. This may be visualized as in Fig. 2, which is an idealization of the evolution of model soil wetness among many grid points. In this schematic, the x axis represents the climate space for soil wetness, and the y axis is time. Initially, there are two possible states, one for the GSWP climatological soil wetness and one in that range represented by GOLD. There are also two asymptotic end points that are not fully approached by September. One is the climate attained by the full coupling of SSiB to the AGCM, which includes a strong drift toward an unrealistic climate due to systematic errors in the AGCM as described by Dirmeyer (2001, 2003) and Dirmeyer and Zhao (2004). The other is the state where SSiB is largely driven by specified conditions of radiation and precipitation. This is expected to be a more realistic climate, although the AGCM can still influence SSiB through its effect on near-surface temperature and humidity, and thus the gradients involved in calculating latent and sensible heat fluxes to the atmosphere. The midseason perigee between CTL and PLS is indicated as something of a crossing of their paths. The GSWP climatological soil wetness appears to be positioned in climate space so as to compensate somewhat for the drift during this season.

The speed of the drift in soil wetness and its affect on the atmosphere can be quantified by comparing CTL and LIC, which have different initialization of soil wetness. Figure 3 shows month by month for three variables (root zone soil wetness, surface latent, and sensible heat flux) what fraction of the 18 yr of seasonal simulations show a statistically significant difference between CTL and LIC. The significance level chosen is 95%. The color shading shows the number of years that meet the criterion for significant difference, using a Student’s t test based on the ratio of interexperiment difference versus intraensemble variance. Root zone soil wetness shows significant differences globally for the first month, because of differences of initialization, that retreat mainly to the semiarid and arid regions by August. Differences in latent heat flux appear to be more persistent than soil wetness differences in boreal high latitudes, but somewhat less persistent in the arid regions. Soil wetness initialization conveys almost no memory to sensible heat flux in arid regions; only in the Tropics and Arctic margins are there significant differences persisting into September.

In summary, there is both a strong drift and little memory of initial soil wetness conditions over most regions. These two characteristics may be related, as suggested by Dirmeyer and Zhao (2004).

2) Variability

Figure 4 shows the spatial distribution for the signal-to-noise ratios of root zone soil wetness for the CTL and LIC simulations during June and September. The pattern in the CTL simulation is rather invariant with time. The regions showing a moderately strong signal-to-noise ratio are mainly regions in the Tropics that have a strong maritime influence, where SST anomalies can affect the climate. The LIC experiment reflects the strong impact of realistic soil-moisture initialization during June. This signal fades in all but the most arid regions, so that by September the signal-to-noise ratio over most of the globe is as weak as in the CTL case.

It is helpful to view the signal and noise components of the variance separately. Figure 5 shows these components, expressed as standard deviations of root zone soil wetness, for the representative month of July. The structure and magnitude of the noise is very similar between the two cases. The CTL case begins the integration with about 30% more noise than LIC, but the difference decreases quickly so that even by July CTL has only slightly more noise. There are substantial differences, however, in the signal. Both cases show the strongest signal at low latitudes, but the LIC case has much more signal over the subtropics and higher latitudes.

Table 4 summarizes the ratio of the global means (land points north of 60°S) of signal versus noise for each month in each experiment. Several features are evident from the table. First, when climatological soil wetness is used to initialize an experiment (CTL, cPLS, cPL) the signal-to-noise ratio starts relatively low. Signal increases with time in these runs, but in the CTL case noise increases even faster. On the other hand, when GOLD initial conditions are specified, the signal starts high and drops with time as the memory of the initial conditions is lost. Second, precipitation flux replacement has a profound impact on both terms in the ratio, suppressing noise while enhancing the signal. This is why the ratio increases in cases cPLS and cPL but not in CTL.

The analysis above shows how initialization and flux replacement affect the model simulation of soil wetness variability. They do not indicate whether the climate model simulates a realistic degree of variability. Unfortunately, we have no way of gauging the noise component in the real world. However, we can compare the model’s grand variance directly to the interannual variance of the proxy observed (GOLD) soil wetness, which presumably contains both signal and noise components of the actual climate. Table 5 presents the ratio of global mean model grand standard deviation to GOLD standard deviation. A ratio near unity suggests the climate model is maintaining a good degree of total variance, although there is no way of knowing from this statistic whether the partitioning between signal and noise is correct. Looking only at June, it is clear that initialization with climatological soil wetness results in too little variance, while use of the GOLD initial conditions gives about the right variance. Otherwise, there is a slight overall decrease of grand variance with time, but CTL and LIC show a reasonable level of overall variance in the later months. That is, once away from the influence of the initial conditions, the free-running model has the proper level of grand variance compared to observations. Specification of downward precipitation fluxes artificially limits the grand variance, mainly by confining the noise level within the ensembles. It is particularly interesting to note that among the flux replacement cases, L and S have the highest grand variance. Perhaps that is because the specification of some but not all fluxes makes for inconsistent combinations of surface fluxes, contributing to unrealistic extremes (note that L and S also have low signal-to-noise ratios in Table 4, particularly in latter months).

Overall, we see that initialization of soil wetness provides a great deal of signal that is rapidly lost in the climate model. However, the grand variance of the model is about right. Table 4 does show a slow decline of grand variance relative to observations over the course of the 4 months. This drop probably reflects to some degree the drift of the model soil wetness to its extreme limits. Yet it may also be a symptom of the premature loss of the initialization signal component of the grand variance. Finally, flux replacement greatly limits variance, namely the noise. This does not concern us, as we are using flux substitution as a tool for understanding climate processes, and not as a component of a viable climate simulation or forecast model.

b. Other climate variables

The skill of seasonal near-surface air temperature and precipitation forecasts, measured in terms of local temporal correlations across the 18-yr period, was detailed by Dirmeyer and Zhao (2004). In summary, precipitation skill was low, but showed signs of improvement with flux replacement. Skill of temperature forecasts was fairly good, and improved markedly with flux replacement. This section details the impact of land surface initialization and flux replacement on the simulation of spatial anomaly correlation coefficients (SACC) and root-mean-square error (rmse) of the monthly and seasonal anomalies of precipitation and temperature.

Table 6 gives the average of the global terrestrial SACC and rmse for precipitation and temperature for each month [June–July–August–September (JJAS)] of the boreal summer simulations, and for the 4-month mean fields from the CTL case. Correlations are generally quite low for both fields, with the exception of temperature during the first month of simulation, when atmospheric initial conditions still have a prominent effect on the monthly mean. Monthly rmse for the anomaly fields are about 1.4 mm day−1 for precipitation, and 1.6 K for temperature, with rmse for the season about one-third smaller in magnitude.

Figures 6 –9 show the impact of realistic land surface initialization and flux replacement on these statistics. The 18-yr mean change for each case relative to CTL is shown by the central black bars for each month. Each bar encloses a range of ±one standard deviation, and the whiskers denote the full range of changes simulated. The first element for each month shows the impact of initial conditions alone (LIC). The others show the impact of initial conditions plus the replacement of the indicated flux (refer to Table 1). Not all cases are shown in these figures—for clarity we have limited the figures to five cases including the one multiflux replacement that gives the greatest improvement.

Dirmeyer and Zhao (2004) showed that large systematic errors in the climate model induce a strong drift that suppresses the model’s sensitivity to land surface initialization in simulating interannual variations. Figures 6 –9 show that the same is true for the simulation of spatial patterns of climate anomalies. Figure 6 shows the impacts on the SACC of precipitation. Replacement of model precipitation with observed precipitation at the land–atmosphere interface induces the greatest improvement in the simulation of precipitation. Replacement of downward longwave radiation also has a positive, but weaker, effect on rainfall simulation. A combination of precipitation and longwave flux replacement induces the greatest improvement precipitation patterns. Shortwave radiation replacement by itself actually eliminates what little SACC skill there is in the model.

Figure 7 shows the impact on the rmse of precipitation. Here all three fluxes reduce error when applied individually or in combination. The greatest improvement for a single flux is again induced by replacement of precipitation, and the best overall improvement is found for a combination of precipitation and longwave radiation replacement. There is, however, a wide range of impact in the later months, especially September, suggesting that improvement is not guaranteed at longer lead times.

It may seem obvious that replacement of the precipitation fluxes should lead to the greatest improvement in model precipitation. After all, this is the one flux that is a branch of the hydrologic cycle. However, outside of arid regions, where the availability of energy has an important role in determining evapotranspiration, changes in longwave and shortwave radiation might be expected to be important as well. In this model, drifts are so large that in many places the soil wetness spends too much time at extreme values where the sensitivity of upward fluxes to AGCM anomalies is muted. Specification of precipitation is the best way to keep soil wetness in a realistic range in this model.

Figures 8 and 9 repeat the analyses of Figs. 6 and 7 for near-surface air temperature. For this variable, replacement of all three fluxes induces the greatest model improvement. Here, longwave radiation is the key flux for improving the simulations. Shortwave flux replacement alone is detrimental to the model’s rmse, but does improve anomaly correlations slightly during July and August. Outstanding improvements are found in the PLS case, where global SACC is generally over 0.4, and as high as 0.67 for September 1992. PLS also has the lowest rmse, averaging less than 1.00 K for the seasonal mean.

These results suggest that the land surface, when given the opportunity to supply realistic spatial and temporal variations in heat and moisture fluxes to the atmosphere, can induce an improved atmospheric response. This appears to be true even when the atmospheric model has a strong tendency to drift toward a poor simulation of mean climate.

The tendency for degradation of the simulations when shortwave fluxes are replaced requires further explanation. Whenever a partial set of fluxes are replaced, inconsistencies are introduced at the land surface. Case S usually has the largest inconsistency among the downward fluxes at the surface. Imagine a situation, for example, where the GCM is producing heavy precipitation over a grid box, yet the specified shortwave flux is large, consistent with a clear-sky situation. Excessive evaporation would result. In case S the inconsistencies are large enough and frequent enough that the mean climate is compromised. Problems of inconsistency are also suggested by the fact that the simulations improve soundly when all three fluxes are replaced. It should also be recognized that the LSS was originally tuned to compensate to some extent for the errors in the surface fluxes, which are quite large for shortwave radiation (see Dirmeyer and Zhao 2004, their Fig. 6). Thus, this coupled land–atmosphere model may not always deal correctly with nominal surface fluxes, particularly when only some of the fluxes are replaced. In effect, some but not all compensating errors were removed, creating new problems. Both of these factors likely play a role in the poor performance of case S.

c. Regional cases

Global statistics do not reveal how well the model performs in simulating observed regional anomalies. We have seen that the climate model does rather well in simulating temperature anomalies, particularly when flux replacement is invoked. Here we focus on the precipitation response. We have chosen four large regions of particular agricultural interest (see Fig. 10), each measuring 36° latitude by 71° longitude, over which we compute the SACC between model and observed precipitation for each ensemble mean. Calculations are confined to the land regions only in each box.

For the seasonal anomalies, there is an overall tendency for substitution of downward longwave fluxes to improve SACC slightly, and for shortwave fluxes to degrade the simulations when applied by themselves. Substitution of precipitation fluxes often leads to a marked improvement of SACC over all of the areas, as shown in Fig. 11. The first bar in each pair represents the SACC for the LIC case, and the second is for the P case. Note that for the Sahel, we define the season as July–September because these are the three wettest months over the region. Otherwise, June–August is used. One, two, or three asterisks above the bars indicate improvements that are significant at the 90%, 95%, or 99% level, respectively (conservatively assuming 25 degrees of freedom over each area for precipitation at the climate model’s resolution). The final pair of bars shows the average SACC for all years and the significance for this larger sample. The SACC improves over each area for nearly every year. In some cases the improvements are quite dramatic, particularly over North America and Europe. Over the monsoon regions of Africa and south Asia, most years show improvements, but in only one case (south Asia in 1984) is the change significant at even the 90% level. However, the average improvement over 18 yr is significant at the 99% level over south Asia, and at the 90% level for the Sahel region. Substitution of all fluxes (case PLS) generally leads to improvements that are slightly poorer than in case P.

Initial land surface conditions have a less profound impact on precipitation simulation. Only over North America is there a sign of systematic improvement of SACC during June (Fig. 12), and even here there are no years with significantly large improvements. The 18-yr average change, though nearly a doubling in SACC over the mean CTL value, is only significant at the 69% level.

For the same 18 Junes, there are seven cases where precipitation substitution leads to a statistically significant improvement at 90% or higher. Figure 13 shows an example from one particular case—the 1988 drought over North America. The observed precipitation anomaly (relative to CMAP) and the ensemble mean model simulation from six of the cases are shown, along with the SACC over the domain (land only) for each case. The observed anomaly covers most of the eastern two-thirds of the contiguous United States and extends into southern and eastern Canada. Positive anomalies exist over the Gulf of Mexico, Atlantic Ocean, and western Canada. The model persistently places a negative anomaly over the northern Gulf of Mexico, Florida, and coastal Atlantic waters, regardless of the case. This feature appears to be unrelated to surface fluxes over land, and may be part of the GCM response to the specified SST (errors over ocean are very robust in this model). The CTL simulation has a slightly negative pattern correlation with observations, placing a wet anomaly over the southeastern United States and dry anomalies over the Northwest (note that the magnitudes of the model anomalies are necessarily weakened by the averaging across 10 ensemble members—the SACC only measures the correlation between the patterns of anomalies). The anomalies over Canada are better represented. Realistic land initial conditions improve the SACC to 0.14, strengthening and expanding the dry anomaly over the center of the continent. Specification of observed precipitation fluxes to the LSS recovers much of large-scale pattern, giving a SACC of 0.54 over the domain. Specification of observed downward radiation has a similar but weaker effect on the simulation. Shortwave flux replacement alone generates an erroneous wet anomaly over the eastern third of the United States. The PL case is the best at capturing the strong core of the drought, but has a slightly lower SACC than the P case for the entire region.

Referring back to Fig. 11, there is a tendency for the response to flux replacement to be stronger in the years when anomalies were large. This supposition is born out when we calculate the root-mean-square precipitation anomaly for the 3-month period over each region (as a measure of strength of the year’s anomaly pattern) and correlate that with the change in SACC realized as a result of flux replacement. Over North America and south Asia, there is a positive correlation between these two 18-yr time series that is significant at the 99% level. Over the African region the significance level is more than 80%. This positive relationship is intuitive under the assumption that specification of observed fluxes leads to a more accurate time series of land surface state variables, which should lead to improved upward surface heat fluxes and improved atmospheric states and circulation in a regime of land–atmosphere feedback. Seasons with strong anomalies in observed rainfall will have stronger resulting anomalies of model soil wetness under flux replacement, which ought to drive the AGCM in turn to enhance anomalies in the vein of Shukla and Mintz (1982). Of course, the strength of such feedbacks are highly model dependent (Koster et al. 2004) and the strength of land–atmosphere coupling in the real world is still poorly understood. Nevertheless, if the land–atmosphere feedbacks are realistically represented in the climate model, then the anomalies in land surface state variables and upward fluxes should be better represented with specified observed downward surface fluxes than in the absence of flux replacement. We see strong evidence for that in these experiments. Therefore, these results provide further evidence that large anomalies in rainfall may indeed be the product of land–atmosphere feedbacks on local and regional scales.

4. Conclusions

A global climate model has been integrated in ensemble mode for 18 successive boreal summer seasons with climatological land surface initialization, realistic initialization from GOLD, and a series of experiments with replacement of downward surface fluxes over land of all combinations of precipitation, shortwave, and longwave radiation with observed proxies. The role of the land surface in potential climate predictability is examined in this modeling framework.

On the seasonal time scale, the climate model exhibits significant drift in the land surface state relative to offline analyses generated using the same land surface scheme driven by observed meteorological and radiation forcing. This drift is sufficient after the first month to overcome the signal in soil wetness introduced by the use of “realistic” (interannually varying) initial conditions, rather than climatological initial conditions. This drift is driven by the errors in downward surface fluxes over land, motivating the experiments with flux replacement. Those test cases are also motivated by a desire to understand how climate signals and model errors propagate around the hydrologic cycle.

The grand variance of the model is comparable to observations, but slowly declines through the season. This may be an indication of the loss of signal due to drift, or the loss of the initial condition signal. Comparison of model signal to noise shows that only in arid regions are anomalies in initial soil wetness preserved without flux replacement. Specification of observed fluxes suppresses noise while enhancing the signal.

Replacement of a single flux yields uneven results, and often degrades skill, as the LSS experiences an inconsistent set of downward fluxes. A combination of precipitation and longwave flux replacement induces the greatest improvement in precipitation patterns. Replacement of all three fluxes leads to the highest skill in surface temperature simulations. These results suggest that the land surface can stimulate an improved atmospheric response through coupled feedback. This appears to be true even when the atmospheric model has strong systematic errors.

Investigation of the model’s ability to simulate regional patterns of precipitation anomalies shows that flux replacement leads to dramatic improvements, particularly over North America and Europe, but also over monsoon regions of Asia and Africa. This is evidence that large anomalies in rainfall may indeed be the product of positive land–atmosphere feedbacks on local and regional scales. However, these are also regions of high intraensemble precipitation variability, or noise, so this evidence of strong land–atmosphere feedback in important agricultural regions may not fully translate into enhanced predictability. In the cases of extreme anomalies, there is evidence of above-average improvement in the flux substitution cases. So perhaps these extreme years will be the most likely to benefit from improvements in model simulations. The improvements may be model parameterization developments or statistical corrections to model simulations, but they must reduce the drift and errors in downward fluxes at the lower boundary of the AGCM, so as to allow anomalies in land surface initial conditions to have an undiminished effect on the seasonal climate forecast.

As with all model experiments where observational data are inserted into a retrospective forecast, we are not quantifying realizable predictability but rather potential predictability. In a true forecast we will not know what the radiative fluxes, precipitation, soil wetness, or even the SST will be a season in advance. Forecasting is an initial value problem, and indeed studies of the impact of initial soil wetness on predictions tend to show weaker impacts (e.g., case LIC in this study; Douville 2004). Such experiments can place an upper bound on prediction skill, but more importantly help us understand how each model component and branch of coupling between land and atmosphere affect predictability.

Finally, it should be noted that the specific details of the results of this investigation are almost certainly dependent on the choice of model. Yet the methodology of flux replacement provides a tool that is useful for all coupled model systems, adding to existing approaches such as fixing or nudging of land surface state variables, or manipulating the upward fluxes from land to atmosphere. Each of these approaches amounts to an intercession in a segment of the water and energy cycle for the purpose of understanding and improving model climate simulations.

Acknowledgments

The author wishes to thank reviewers H. Douville, N. Diffenbaugh, and an anonymous reviewer for their thoughtful and detailed comments that have helped to improve the clarity of this paper. This work was conducted as part of omnibus research at the Center for Ocean–Land–Atmosphere Studies, supported by NSF Grant ATM 9814265, NOAA Grant NA96GP0056, and NASA Grant NAG5–8202.

REFERENCES

  • Atlas, R., Wolfson N. , and Terry J. , 1993: The effect of SST and soil moisture anomalies on GLA model simulation of the 1988 U.S. summer drought. J. Climate, 6 , 20342048.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Delworth, T., and Manabe S. , 1989: The influence of soil wetness on near-surface atmospheric variability. J. Climate, 2 , 14471462.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., 2000: Using a global soil wetness dataset to improve seasonal climate simulation. J. Climate, 13 , 29002922.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., 2001: Climate drift in a coupled land–atmosphere model. J. Hydrometeor., 2 , 89100.

  • Dirmeyer, P. A., 2003: The role of the land surface background state in climate predictability. J. Hydrometeor., 4 , 599610.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., and Zeng F. J. , 1999a: An update to the distribution and treatment of vegetation and soil properties in SSiB. COLA Tech. Rep. 78, 25 pp. [Available from the Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].

  • Dirmeyer, P. A., and Zeng F. J. , 1999b: Precipitation infiltration in the Simplified SiB land surface scheme. J. Meteor. Soc. Japan, 77 , 291303.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., and Tan L. , 2001: A multi-decadal global land-surface data set of state variables and fluxes. COLA Tech. Rep. 102, 43 pp. [Available from the Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].

  • Dirmeyer, P. A., and Zhao M. , 2004: Flux replacement as a method to diagnose coupled land–atmosphere model feedback. J. Hydrometeor., 5 , 10341048.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Dirmeyer, P. A., Dolman A. J. , and Sato N. , 1999: The Global Soil Wetness Project: A pilot project for global land surface modeling and validation. Bull. Amer. Meteor. Soc., 80 , 851878.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douville, H., 2003: Assessing the influence of soil moisture on seasonal climate variability with AGCMs. J. Hydrometeor., 4 , 10441066.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Douville, H., 2004: Relevance of soil moisture for seasonal atmospheric predictions: Is it an initial value problem? Climate Dyn., 22 , 429446.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Durre, I., Wallace J. M. , and Lettenmaier D. P. , 2000: Dependence of extreme daily maximum temperatures on antecedent soil moisture in the contiguous United States during summer. J. Climate, 13 , 26412651.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fennessy, M. J., and Shukla J. , 1999: Impact of initial soil wetness on seasonal atmospheric prediction. J. Climate, 12 , 31673180.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and Hunt B. G. , 1987: Interannual variability of the simulated hydrology in a climatic model—Implications for drought. Climate Dyn., 1 , 113130.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Gruber, A., Su X. , Kanamitsu M. , and Schemm J. , 2000: The comparison of two merged rain gauge–satellite precipitation datasets. Bull. Amer. Meteor. Soc., 81 , 26312644.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kinter, J. L., and Coauthors, 1997: Formulation. Vol. 1, The COLA atmosphere–biosphere general circulation model. COLA Tech. Rep. 51, 46 pp. [Available from the Center for Ocean–Land–Atmosphere Studies, 4041 Powder Mill Road, Suite 302, Calverton, MD 20705.].

  • Koster, R. D., and Suarez M. J. , 2001: Soil moisture memory in climate models. J. Hydrometeor., 2 , 558570.

  • Koster, R. D., Suarez M. J. , and Heiser M. , 2000: Variance and predictability of precipitation at seasonal-to-interannual timescales. J. Hydrometeor., 1 , 2646.

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

  • Oglesby, R. J., 1991: Springtime soil moisture, natural climate variability, and North American drought as simulated by the NCAR Community Climate Model I. J. Climate, 4 , 890897.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reale, O., and Dirmeyer P. A. , 2002: Modeling the effect of land surface evaporation variability on precipitation variability. Part I: General response. J. Hydrometeor., 3 , 433450.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reale, O., Dirmeyer P. A. , and Schlosser C. A. , 2002: Modeling the effect of land surface evaporation variability on precipitation variability. Part II: Time- and space-scale structure. J. Hydrometeor., 3 , 451466.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., and Smith T. M. , 1994: Improved global sea surface temperature analyses using optimal interpolation. J. Climate, 7 , 929948.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ropelewski, C. F., Janowiak J. E. , and Halpert M. F. , 1985: The analysis and display of real time surface climate data. Mon. Wea. Rev., 113 , 11011107.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schlosser, C. A., and Milly P. C. D. , 2002: A model-based investigation of soil moisture predictability and associated climate predictability. J. Hydrometeor., 3 , 483501.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Schneider, E. K., 2002: The causes of differences between equatorial Pacific SST simulations of two coupled ocean–atmosphere general circulation models. J. Climate, 15 , 449469.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sellers, P. J., Mintz Y. , Sud Y. C. , and Dalcher A. , 1986: A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci., 43 , 505531.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Shukla, J., and Mintz Y. , 1982: Influence of land-surface evapotranspiration on the earth’s climate. Science, 215 , 14981501.

  • Shukla, J., and Coauthors, 2000: Dynamical seasonal prediction. Bull. Amer. Meteor. Soc., 81 , 25932606.

  • Walker, J., and Rowntree P. R. , 1977: The effect of soil moisture on circulation and rainfall in a tropical model. Quart. J. Roy. Meteor. Soc., 103 , 2946.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xue, Y., Sellers P. J. , Kinter J. L. , and Shukla J. , 1991: A simplified biosphere model for global climate studies. J. Climate, 4 , 345364.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Xue, Y., Zeng F. J. , and Schlosser C. A. , 1996: SSiB and its sensitivity to soil properties—A case study using HAPEX-Mobilhy data. Global Planet. Change, 13 , 183194.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Yeh, T-C., Wetherald R. T. , and Manabe S. , 1984: The effect of soil moisture on the short-term climate and hydrology change—A numerical experiment. Mon. Wea. Rev., 112 , 474490.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

(top) Evolution of the climatological root zone soil wetness in GOLD from 1 Jun to Sep and (bottom) the difference between the mean of all GCM simulations and GOLD for the same times of year. Units are percentage of saturation.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 2.
Fig. 2.

Schematic of the evolution of soil wetness from two types of initial conditions (climatological vs realistic) to two different equilibria depending on whether downward surface fluxes are specified or free to interact with the land surface.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 3.
Fig. 3.

The number of boreal summer seasons out of 18 for which the LIC and CTL cases have statistically significant differences in (left) root zone soil wetness, (middle) latent heat flux, and (right) sensible heat flux. Jun is the first month of the integrations, and Sep is the last.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 4.
Fig. 4.

Signal-to-noise ratio for root zone soil wetness for (top) CTL and (bottom) LIC during Jun and Sep.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 5.
Fig. 5.

Signal and noise components, expressed in terms of standard deviation, for root zone soil wetness for (top) CTL and (bottom) LIC during Jul.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 6.
Fig. 6.

The change in spatial anomaly correlation coefficient (land points north of 60°S) for precipitation by month and for the 4-month mean (right shaded area) from CTL case values for each of five cases (see Table 1 for a description of the cases).

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 7.
Fig. 7.

As in Fig. 6 but for change in root-mean-square error (units are mm day−1).

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 8.
Fig. 8.

As in Fig. 6 but for near-surface air temperature.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 9.
Fig. 9.

As in Fig. 7 but for near-surface air temperature (units are K).

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 10.
Fig. 10.

Four regions over which monthly and seasonal precipitation skill are compared.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 11.
Fig. 11.

Seasonal SACC for precipitation in cases LIC (solid) and P (hatched) for each year, and in the last column (shaded) the mean of the 18 yr. Asterisks denote statistical significance (see text for details).

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 12.
Fig. 12.

As in Fig. 11 but for Jun of the CTL (solid) and LIC (hatched) cases over North America.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Fig. 13.
Fig. 13.

Jun 1988 precipitation anomalies vs the 1982–99 mean for (top) observations and (bottom) six model cases. Spatial correlations of each case with observations are shown in the model panels. Units for anomalies are mm day−1.

Citation: Journal of Hydrometeorology 6, 5; 10.1175/JHM444.1

Table 1.

List of experiments and their attributes.

Table 1.
Table 2.

Rms differences in root zone soil wetness between pairs of integrations.

Table 2.
Table 3.

As in Table 2 but for mean differences.

Table 3.
Table 4.

Global mean (land only) signal-to-noise ratio for root zone soil wetness (see text for details).

Table 4.
Table 5.

As in Table 4, but for (first row) GOLD standard deviation and (other rows) ratio of GCM grand standard deviation to GOLD standard deviation.

Table 5.
Table 6.

The 18-yr mean of global (land only 60°S–90°N) spatial anomaly correlation coefficients (corr) and rmse for precipitation and temperature. Rmse units are mm day−1 for precipitation, and K for temperature.

Table 6.
Save