• Alpert, J. C., M. Kanamitsu, P. M. Caplan, J. G. Sela, G. H. White, and E. Kalnay, 1998: Mountain induced gravity wave drag parameterization in the NMC medium-range model. Preprints, Eighth Conf. on Numerical Weather Prediction, Baltimore, MD, Amer. Meteor. Soc., 726–733.

    • Search Google Scholar
    • Export Citation
  • Atlas, R., N. Woldson, and J. Terry, 1993: The effect of SST and soil moisture anomalies on GLA model simulations of the 1998 U.S. summer drought. J. Climate, 6 , 20342048.

    • Search Google Scholar
    • Export Citation
  • Beljaars, A. C. M., P. Viterbo, M. J. Miller, and A. K. Betts, 1996: The anomalous rainfall over the United States during July 1993: Sensitivity to land surface parameterization and soil moisture anomalies. Mon. Wea. Rev., 124 , 362383.

    • Search Google Scholar
    • Export Citation
  • Campana, K. A., Y-T. Hou, K. E. Mitchell, S-K. Yang, and R. Cullather, 1994: Improved diagnostic cloud parameterization in NMC's global model. Preprints, 10th Conf. on Numerical Weather Prediction, Portland, OR, Amer. Meteor. Soc., 324–325.

    • Search Google Scholar
    • Export Citation
  • Chou, M-D., 1992: A solar radiation model for use in climate studies. J. Atmos. Sci., 49 , 762772.

  • Chou, M-D., and K-T. Lee, 1996: Parameterizations for the absorption of solar radiation by water vapor and ozone. J. Atmos. Sci., 53 , 12031208.

    • Search Google Scholar
    • Export Citation
  • Cook, K. H., and A. Gnanadesikan, 1991: Effects of saturated and dry land surfaces on the tropical circulation and precipitation in a general circulation model. J. Climate, 4 , 873889.

    • Search Google Scholar
    • Export Citation
  • de Martonne, E., 1927: Regions of interior basin drainage. Geogr. Rev., 17 , 397414.

  • Dorman, J. L., and P. J. Sellers, 1989: A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SiB). J. Appl. Meteor., 28 , 833855.

    • Search Google Scholar
    • Export Citation
  • Fels, S. B., and M. D. Schwarzkopf, 1975: The simplified exchange approximation: A new method for radiative transfer calculations. J. Atmos. Sci., 32 , 14751488.

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

  • Fennessy, M. J., P. A. Dirmeyer, J. L. Kinter III, A. Schlosser, J. Shukla, J. Huang, and L. Sun, 2000: Impact of initial soil moisture anomalies on seasonal atmospheric predictions over the U.S. during March–April–May 2000. Proc. Climate Diagnostics Workshop, Palisades, NY, IRI, 239–241.

    • Search Google Scholar
    • Export Citation
  • Hong, S-Y., and H-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124 , 23222339.

    • Search Google Scholar
    • Export Citation
  • Hong, S-Y., and E. Kalnay, 2000: Role of sea surface temperature and soil-moisture feedback in the 1998 Oklahoma–Texas drought. Nature, 408 , 842844.

    • Search Google Scholar
    • Export Citation
  • Huang, J., H. M. van den Dool, and K. P. Georgakakos, 1996: Analysis of model-calculated soil moisture over the United States (1931–1993) and applications to long-range temperature forecasts. J. Climate, 9 , 13501362.

    • Search Google Scholar
    • Export Citation
  • Irannejad, P., A. Henderson-Sellers, T. J. Phillips, and K. McGuffie, 2000: Analysis of AMIP II Models' simulations of land surface climates. GEWEX News, Vol. 10, No. 3, 5–8.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kanamitsu, M., and Coauthors. 2002a: NCEP Dynamical Seasonal Forecast System 2000. Bull. Amer. Meteor. Soc., 83 , 10191037.

  • Kanamitsu, M., W. Ebisuzaki, J. Woollen, S-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002b: NCEP/DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 16311643.

    • Search Google Scholar
    • Export Citation
  • Kinter III, J. L., J. Shukla, L. Marx, and E. K. Schneider, 1988: A simulation of the winter and summer circulations with the NMC global spectral model. J. Atmos. Sci., 45 , 24862522.

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

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

  • Mintz, Y., 1984: The sensitivity of numerically simulated climates to land surface boundary conditions. The Global Climate, J. T. Houghton, Ed., Cambridge University Press, 79–105.

    • Search Google Scholar
    • Export Citation
  • Mitchell, K., and Coauthors. 2000: The collaborative GCIP land data assimilation (LDAS) project and supportive NCEP uncoupled land-surface modeling initiatives. Preprints, 15th Conf. on Hydrology, Long Beach, CA, Amer. Meteor. Soc., 1–4.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120 , 9781002.

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

    • Search Google Scholar
    • Export Citation
  • Pan, H. L., and W-S. Wu, 1995: Implementing a mass flux convection parameterization package for the MNC Medium-Range Forecast Model. NMC Office Note 409, 40 pp. [Available from NCEP, 5200 Auth Road, Washington, DC 20233.].

    • Search Google Scholar
    • Export Citation
  • Rind, 1982: The influence of ground moisture conditions in North America in summer climate as modeled in the GISS GCM. Mon. Wea. Rev., 110 , 14871949.

    • Search Google Scholar
    • Export Citation
  • Roads, J. O., S-C. Chen, M. Kanamitsu, and H. Juang, 1999: Surface water characteristics in NECP's Reanalysis and Global Spectral Model. J. Geophys. Res., 104 , 1930719327.

    • Search Google Scholar
    • Export Citation
  • Rowntree, P. R., and J. R. Bolton, 1983: Simulation of the atmospheric response to soil moisture anomalies over Europe. Quart. J. Roy. Meteor. Soc., 109 , 501526.

    • Search Google Scholar
    • Export Citation
  • Sato, N., P. J. Sellers, D. A. Randall, E. K. Schneider, J. Shukla, J. L. Kinter III, Y-T. Hou, and E. Albertazzi, 1989: Effects of implementing the Simple Biosphere Model in a general circulation model. J. Atmos. Sci., 46 , 27572782.

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

    • Search Google Scholar
    • Export Citation
  • Sellers, P. J., W. J. Shuttleworth, and J. Dorman, 1989: Calibrating the Simple Biosphere Model for Amazonian tropical forest using field and remote sensing data. Part I: Average calibration with field data. J. Appl. Meteor., 28 , 727759.

    • Search Google Scholar
    • Export Citation
  • Serafini, Y. V., 1990: The time scale of land surface hydrology in response to initial soil moisture anomalies: A case study. Tellus, 42A , 390400.

    • Search Google Scholar
    • Export Citation
  • Shukla, J., and Y. Mintz, 1982: Influence of land surface evapotranspiration on the earth's climate. Science, 15 , 14981501.

  • Sperber, K. R., and T. N. Palmer, 1996: Interannual tropical rainfall variability in general circulation model simulations associated with the Atmospheric Model Intercomparison Project. J. Climate, 9 , 27272750.

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

    • Search Google Scholar
    • Export Citation
  • Wang, W., and A. Kumar, 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103 , 2863728646.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840858.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. Arkin, 1997: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • Yamazaki, K., 1989: A study of the impact of soil moisture and surface albedo changes on global climate using the MRI-GCM-I. J. Meteor. Soc. Japan, 67 , 123146.

    • Search Google Scholar
    • Export Citation
  • Yang, R., M. J. Fennessy, and J. Shukla, 1994: The influence of initial soil moisture on medium-range surface weather forecasts. Mon. Wea. Rev., 122 , 471485.

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

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Time series of water content (mm) in 2-m soil layer from observation (dashed line) and the reanalysis (solid line: NN-R, black; R-2, gray) over Illinois for the 1982–98 period

  • View in gallery

    Persistence decay curve of deep-layer (10–200 cm) soil moisture (mm) from observations (dashed line) and R-2 analysis (solid line). Persistence is estimated from the anomaly correlation between Apr and later months

  • View in gallery

    Deep-layer (10–200 cm) soil wetness prediction skill decay curve for the 15 regions (see Fig. 4 for the locations). Anomaly pattern correlation in abscissa and lead time (month) in ordinate. The first month score is not plotted. Persistency forecast, forecast started from soil wetness climatology, and forecast started from R-2 soil wetness are marked by black dashed line, black solid line, and light solid line, respectively

  • View in gallery

    Areas selected for the verification of deep-layer (10–200 cm) soil moisture, overlaid on the global field of soil moisture to evaporation ratio in units of days. Each area is approximately 20° × 20°. Only the land regions are used in the computations. The dashed-line box was used for specific regional study, discussed in sections 5 and 6

  • View in gallery

    Comparison of the skill score of (top) deep-layer (10–200 cm) soil moisture, (middle) latent heat flux, and (bottom) near-surface temperature. The skill score was computed from the 10-member ensemble mean of 3-month avg (May–Jul). Initial analysis soil wetness runs are in gray, and initial climatological soil wetness experiments are in black

  • View in gallery

    Concurrent anomaly correlation of deep-layer (10–200 cm) soil moisture anomaly and 2-m air temperature anomaly for the 10-member ensemble mean of Jul forecast. Note that the sign of the temperature anomaly is reversed to make the correlation positive. Initial analysis soil moisture runs are in gray, and initial climatological soil moisture is in black

  • View in gallery

    Comparison of 2-m air temperature anomaly correlation scores of ISM and CTR experiments verified against independent 344 climate division data. The scores for ISM are shaded light and those of CTR are shaded dark. The skill score was computed from the 10-member ensemble average of 3-month mean for Jun–Jul–Aug period

  • View in gallery

    Agreement of deep-layer (10–200 cm) soil moisture anomaly between the initial soil moisture analysis run and the initial climatological soil moisture run after 1-month integration (May). Cold equatorial Pacific SST events (outlined bar), warm episodes (dark shaded bar) and normal years (light shaded bar) are shown. These events were based on NCEP Climate Prediction Center definitions

  • View in gallery

    Scatter diagram of correlations. Ordinate is the correlation of deep-layer (10–20 cm) soil moisture forecast after 1-month integration (May) between initial soil moisture analysis runs and initial climatological soil moisture runs, and the abscissa is the forecast skill of 2-m air temperature verified against R-2 after 3 months (Jul). Cold episodes, warm episodes, and normal years are marked by open square, solid square, and open circle, respectively

  • View in gallery

    Comparison of anomaly correlation skill score of Jun–Jul mean precipitation over the continental United States (dashed-line area in Fig. 4). The forecasts started (top) from analysis soil wetness, and (bottom) from climatological soil moisture. Each bar is the score of the 10-member ensemble mean. Verification was done against CMAP precipitation analyses (Xie and Arkin 1996)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 177 173 6
PDF Downloads 56 55 6

The Predictability of Soil Moisture and Near-Surface Temperature in Hindcasts of the NCEP Seasonal Forecast Model

View More View Less
  • 1 Climate Research Division, Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California
  • | 2 RS Information Systems, Inc., McLean, Virginia
  • | 3 NOAA/NWS/National Centers for Environmental Prediction/Climate Prediction Center, Washington, D.C
© Get Permissions
Full access

Abstract

Using the NCEP–DOE reanalysis (R-2) soil wetness and the NCEP Seasonal Forecast System, seasonal predictability of the soil moisture and near-surface temperature, and the role of land surface initial conditions are examined. Two sets of forecasts were made, one starting from climatological soil moisture as initial condition and the other from R-2 soil moisture analysis. Each set consisted of 10-member ensemble runs of 7-month duration. Initial conditions were taken from the first 5 days of April, 12 h apart, for the 1979–96 period.

The predictive skill of soil moisture was found to be high over arid/semiarid regions. The model prediction surpassed the persisted anomaly forecast, and the soil moisture initial condition was essential for skillful predictions over these areas. Over temperate zones with more precipitation, and over tropical monsoon regions, the predictive skill of the soil moisture declined steeply in the first 3–4 months. This is due to the difficulties in predicting precipitation accurately. In contrast, the situation was very different over tropical South America where tropical SST forcing controlled the precipitation and where the model simulated the precipitation well. The forecast starting from climatological soil moisture approached the forecast skill of initial soil moisture in 3–4 months; after that the effect of initial soil moisture information tended to disappear.

The near-surface temperature anomaly forecast was closely related to the soil moisture anomaly forecast, but the skill was lower. The verification of temperature made against the U.S. 344 climate division data indicated that the improvement in the forecast skill was not an artifact of the R-2 soil moisture analysis.

It was suggested that the equatorial Pacific SST anomaly had an impact on the soil moisture anomaly over the continental United States during the first month of integration, and then it contributed positively toward the prediction of near-surface temperature during the following months.

Corresponding author address: Dr. Masao Kanamitsu, Scripps Institution of Oceanography, MC 0224, CRD/UCSD, La Jolla, CA 92093-0224. Email: kana@ucsd.edu

Abstract

Using the NCEP–DOE reanalysis (R-2) soil wetness and the NCEP Seasonal Forecast System, seasonal predictability of the soil moisture and near-surface temperature, and the role of land surface initial conditions are examined. Two sets of forecasts were made, one starting from climatological soil moisture as initial condition and the other from R-2 soil moisture analysis. Each set consisted of 10-member ensemble runs of 7-month duration. Initial conditions were taken from the first 5 days of April, 12 h apart, for the 1979–96 period.

The predictive skill of soil moisture was found to be high over arid/semiarid regions. The model prediction surpassed the persisted anomaly forecast, and the soil moisture initial condition was essential for skillful predictions over these areas. Over temperate zones with more precipitation, and over tropical monsoon regions, the predictive skill of the soil moisture declined steeply in the first 3–4 months. This is due to the difficulties in predicting precipitation accurately. In contrast, the situation was very different over tropical South America where tropical SST forcing controlled the precipitation and where the model simulated the precipitation well. The forecast starting from climatological soil moisture approached the forecast skill of initial soil moisture in 3–4 months; after that the effect of initial soil moisture information tended to disappear.

The near-surface temperature anomaly forecast was closely related to the soil moisture anomaly forecast, but the skill was lower. The verification of temperature made against the U.S. 344 climate division data indicated that the improvement in the forecast skill was not an artifact of the R-2 soil moisture analysis.

It was suggested that the equatorial Pacific SST anomaly had an impact on the soil moisture anomaly over the continental United States during the first month of integration, and then it contributed positively toward the prediction of near-surface temperature during the following months.

Corresponding author address: Dr. Masao Kanamitsu, Scripps Institution of Oceanography, MC 0224, CRD/UCSD, La Jolla, CA 92093-0224. Email: kana@ucsd.edu

1. Introduction

Many observational and numerical studies have shown that soil moisture significantly impacts model-simulated climate and atmospheric variability. Soil moisture anomalies affect the surface heat balance by altering the partitioning of latent and sensible heat fluxes, thus modifying near-surface fields, such as surface air temperature, humidity, and consequently, low-level circulation, and precipitation.

As reviewed in Mintz (1984) and more recently in Koster et al. (2000), studies based on numerical models have shown that soil moisture has a clear impact on precipitation and surface temperature (Rind 1982; Shukla and Mintz 1982; Rowntree and Bolton 1983; Mintz 1984; Yeh et al. 1984; Yamazaki 1989; Atlas et al. 1993; Yang et al. 1994; Serafini 1990; Koster et al. 2000; Hong and Kalnay 2000). In particular, Atlas et al. (1993) and Beljaars et al. (1996) examined the 1988 U.S. summer drought and found that soil moisture anomalies, once formed, contribute significantly to the maintenance of local and remote hydrologic conditions. This positive feedback was also documented in earlier modeling studies over Africa and Europe (Walker and Rowntree 1977; Rowntree and Bolton 1983), which indicated that the atmosphere responded to soil moisture anomalies in such a way as to perpetuate the initial soil moisture anomaly. In other modeling work, Yeh et al. (1984) showed there was a latitudinal dependence of the soil moisture anomaly persistence.

Almost all the above modeling studies relied on extreme or idealized land surface conditions. For instance, Rind (1982) reduced U.S. soil moisture to one-quarter of its value in the control experiment. Shukla and Mintz (1982) conducted numerical experiments with zero or maximum evapotranspiration over all land surfaces. More recently, Cook and Gnanadesikan (1991) and Beljaars et al. (1996) used similar extreme contrast in the initial soil moisture. These idealized experiments have contributed significantly toward a better understanding of soil moisture's impact on climate and atmospheric variability. However, the existence of such extreme states in nature has not been addressed properly, leaving rooms for further research on the applicability of these studies to real situations.

From a prediction point of view, a considerable effort has been devoted to evaluating the impact of initial soil moisture on atmospheric predictability, using multiyear Atmospheric Model Intercomparison Project (AMIP) type model simulations of atmosphere and soil moisture. Wang and Kumar (1998) assessed seasonal predictability over the United States and suggested that soil moisture anomalies affect near-surface climate variability but have little impact on atmospheric dynamic variability. In addition, they addressed the importance of correctly observed or assimilated initial soil moisture anomalies on atmospheric predictability of summertime precipitation and surface air temperature. Similar results were reported on by Fennessy and Shukla (1999), in which several regions over the globe were studied using the European Centre for Medium-Range Forecasts (ECMWF) soil moisture analysis. Koster et al. (2000) investigated variability and predictability of precipitation at seasonal to interannual timescales using model simulations and concluded that a prior knowledge of the land surface moisture state contributed significantly to predictability in some tropical and midlatitude regions.

In contrast to using model-simulated soil moisture, the use of “realistic” initialization of GCM soil moisture fields has been recognized as a challenging problem (Sellers et al. 1986, 1989; Sato et al. 1989). The difficulty arises due to several factors, including the lack of global-scale long-term records for soil moisture data, considerable uncertainties in initializing and defining soil moisture in the model grid cell, and inaccuracy in the hydrology model. Recently, an effort to use realistic soil moisture in near-real-time seasonal prediction was attempted by Fennessy et al. (2000). They used “NCEP analyzed” (Huang et al. 1996) soil moisture covering only the continental United States for the drought forecast study of the spring of 2000. They demonstrated that initial (realistic) soil moisture was critical for predicting drought over the southern United States. However, the study also noted that the “NCEP soil moisture analyses,” based on a somewhat simplified land model, were incompatible with the Center for Ocean–Land–Atmosphere Studies (COLA) prediction model, which used the more sophisticated Simple Biosphere Model (Kinter et al. 1988). Hence several empirical adjustments were necessary, which introduced uncertainties in the accuracy of the initial soil moisture. This common problem occurs when independent soil moisture analyses are used as the initial condition for model simulations. The problem is that models use different soil layers, soil types, and values for field capacity and wilting point. Consequently models will have different values for “dry” and “wet” conditions. Even observed values can vary as observations, which may have been valid for different depths and different methods, will measure varying amounts of the (unavailable) water.

The objective of this study was to investigate the seasonal predictability of soil moisture and near-surface temperature using long-term records of soil moisture analyses that are more realistic and consistent with the forecast model. This was accomplished using the National Centers for Environmental Prediction–Department of Energy (NCEP–DOE) reanalysis (R-2; Kanamitsu et al. 2002b) for the period 1979 to 1996 and an improved version of the general circulation model used in R-2. The R-2 analysis uses observed precipitation in its land data assimilation, which likely makes the soil moisture analysis more realistic than the NCEP–National Center for Atmospheric Research (NCAR) reanalysis (NN-R; Kalnay et al. 1996). Use of these long-term analyses makes it possible to study the predictability of soil moisture over regions of different climatological characteristics with greater confidence. It should be noted that although R-2 land surface analysis used observed precipitation, it is still affected by the atmospheric and hydrological models used in the data assimilation system. Since the model used in R-2 and the forecast model used in this study share the same origin, the model results here will be biased. Attempts to overcome this bias have been made by comparing R-2 soil moisture with limited observations, and validating near-surface temperature with the independent observations from the 344 U.S. climate divisions.

The brief outline of this paper is as follows. The soil moisture assimilation used in R-2 is briefly reviewed in section 2, where a comparison between the analysis and observed soil moisture over a U.S. site is presented. The model and the experimental design are presented in section 3. Section 4 discusses predictability of soil moisture over various climatological zones over the globe. In section 5, prediction skill of the near-surface temperature is presented. The possibility of land processes transferring sea surface temperature (SST) anomaly signal to summertime continental temperature anomaly is discussed in section 6. The conclusions and discussions are presented in section 7.

2. Land data assimilation in the NCEP–DOE reanalysis

A significant improvement of R-2 analysis upon NN-R was the analysis of land surface conditions (Kanamitsu et al. 2002b). The soil moisture evolution in the NCEP–NCAR reanalysis data assimilation cycle is determined by the model precipitation, radiation forcing computed from atmospheric analysis, and near-surface atmospheric analyses. Furthermore, an artificial nudging toward climatology was added in NN-R to minimize deviation of the soil moisture analysis from climatology. Unfortunately, this nudging dominated the soil moisture evolution (Roads et al. 1999), and the resulting soil moisture analysis had an excessively large amplitude annual cycle with too little interannual variability. Therefore, the utility of NN-R soil moisture analysis has been limited.

In order to correct this undesirable feature, R-2 replaced model precipitation with an observed 5-day mean precipitation constructed from rain gauge and satellite observations (Xie and Arkin 1997). In this procedure, the difference between the 5-day mean model and observed precipitation was computed and the excess or deficit of precipitation was applied to adjust the soil moisture of the top soil layer (the details on this procedure are described in Kanamitsu et al. 2002b). This procedure is similar to the Land Data Assimilation System (LDAS) proposed by Mitchell et al. (2000) except that they are using observed radiation forcing and surface conditions instead of objectively analyzed and calculated forcings.

Figure 1 shows a comparison of Illinois soil moisture from R-2, NN-R, and field observations. Here, soil moisture was defined as water contained in the top 2 m of soil expressed in millimeters. The model (that predicts volumetric soil moisture content) and observed (that measures water content at various layers) soil moisture were converted to this common unit before comparison. In addition, when comparing the soil water content, one wants to remove the unavailable soil moisture as models and observations often treat it differently. For the observed unavailable water, we used the minimum observed soil moisture in the 1982–98 record. This is an overestimate that is compensated for by estimating the model's unavailable water by using the minimum Illinois R-2 and NN-R analyses during 1982–98 periods. We combined the R-2 and NN-R analyses as they used the same soil model and we wanted to illustrate the systematic differences between the reanalyses. Clearly R-2 exhibits soil moisture variability closer to observations than NN-R, which is quite deficient. Except for the spinup period (1979–80, not shown), the R-2 soil moisture analyses are in good agreement with observations.

Another important aspect of soil moisture is the persistence or memory of soil moisture anomalies. It is recognized that the model-simulated soil moisture in AMIP-type runs tends to have more memory than the observation (e.g., Wang and Kumar 1998). Since better memory tends to inflate the predictability, it is important to compare the persistence of R-2 soil moisture anomalies and observations. In Fig. 2, we show the correlation between the 10–200-cm April soil moisture anomaly and that of the following months for both observations and R-2. This figure shows the persistence of R-2 April soil moisture anomalies and indicates that R-2 has better short-term memory and comparable long-term memory (4 or more months) to observations. The comparable long-term memory encouraged us to use R-2 analysis for the soil moisture predictability study, although the applicability of these results from Illinois to other areas needs further study. More detailed comparisons of R-2 soil moisture with observations over various parts of the globe are in progress and will be presented in C.-H. Lu and M. Kanamitsu (2002, unpublished manuscript).

3. Model and experiment design

The model used for this study was a slightly outdated version of the global spectral model used in the operational seasonal forecast of NCEP Seasonal Forecast System (SFM), which was described in detail in Kanamitsu et al. (2002a). The model was run at T42 horizontal resolution (about 300 km) using 28 vertical levels. Model parameterizations include the relaxed Arakawa–Schubert convection scheme (Moorthi and Suarez 1992), longwave radiation (Fels and Schwarzkopf 1975), shortwave radiation (Chou 1992; Chou and Lee 1996), cloud–radiation interaction (Campana et al. 1994), nonlocal vertical diffusion (Hong and Pan 1996), gravity wave drag (Alpert et al. 1988), Oregon State University land surface model (Mahrt and Pan 1984; Pan and Mahrt 1987), and mean orography.

The model had an upper 10-cm soil layer and a lower 190-cm layer for total of 2-m soil depth for all land surfaces. The sources of ground water include precipitation, snowmelt, and condensation of atmospheric water vapor onto the ground. Sinks included surface evaporation and runoff. The runoff is crudely modeled by removing the amount of water in excess of the soil's maximum retaining capacity. The vegetation type, cover, and soil type vary geographically and are from Simple Biosphere model (SiB) climatology (Dorman and Sellers 1989). R-2, however, used an older model, which fixed the vegetation cover to 0.7 and the soil type to sandy clay loam for all land surfaces.

The model integrations were performed in the following manner. Two sets of 180 forecasts of 7-month duration were performed starting from April initial conditions and using either R-2 or climatological soil moisture. The initial conditions were taken from the 0000 and 1200 UTC analyses for 1–5 April from 1979–96 for 10 forecasts per year over a span of 18 yr. This large sample was adequate to provide statistically meaningful model climatology and model forecast skill statistics. For the control experiment (CTR), climatological soil moisture (computed as the 18-yr average of R-2 soil moisture), climatological snow depth (computed as 1979–96 average of R-2 snow depth), and R-2 land surface and deep-layer temperatures analyses were used as surface initial conditions. In the initial soil moisture experiment (ISM), R-2 soil moisture analysis was used for initial conditions, while other conditions were identical to the CTR. For both experiments, observed sea surface temperature was used during the integration and atmospheric initial conditions were taken from the R-2 analyses.

For each year, the 10-member ensemble mean was computed where the anomaly was defined as a departure from the model climatology, which was computed as the average of the ensemble means over the 18 yr. In this manner, the model systematic error was removed. These forecast anomalies were compared to the analysis anomalies computed from the same 18 yr of R-2 analyses.

4. Predictability of soil moisture

The persistence of soil moisture depends on the latitude (Yeh et al. 1984). It also depends on climatological characteristics such as the seasonal amount and nature of precipitation. We selected 15 land zones over the globe covering temperate climate zones, monsoon areas, desert regions, and tropical forest areas. These areas were chosen based on the Martonne aridity index (de Martonne 1927) computed by Irannejad et al. (2000) from Atmospheric Model Intercomparison Project II (AMIP-II) analysis. Figure 4 presents the zones selected. Zones 2, 6, 7, 11, and 12 covered temperate climate areas, zones 8 and 9 covered monsoon regions, zones 3, 4, and 14 covered tropical monsoon or forests, and zones 1, 5, 10, 13, and 15 were desert or semiarid areas. Each selected zone covers an area of 20° × 20° in latitudinal and longitudinal directions.

Figure 3 shows the correlation between the R-2 deep-layer soil moisture analyses and the ISM (gray), CTR (dark solid), and persistence (dashed) forecasts for the selected zones. Figure 3 can be summarized as follows. The model forecast and anomaly persistence have a high predictive skill over zones 5, 7, 10, 11, 12, 13, and 15 throughout the 7-month period. These zones correspond to temperate and desert/semiarid areas. Characterized by little precipitation during the summertime, the soil moisture evolution is mainly controlled by evaporation. It is particularly interesting to note that over zones 5, 10, 11, 12, 13, and 15, the model skill exceeds that of the anomaly persistence. This is likely due to the dominant role of evaporation in the soil moisture evolution that is well represented by the model physics, but could also be due to the similarity of the models used in R-2 and in this study. Note that the zones 5, 10, and 15 are located in the Southern Hemisphere; therefore, part of the high predictive skill may be due to the season changing into more-predictable winter. Over these temperate and desert/semiarid areas, the control runs have much lower scores for all lead times. This indicates that factors such as specified SST and initial atmospheric state were not strong enough to produce the analyzed soil moisture anomaly.

Over zones 1, 2 (both located over the continental U.S.), and 6, the skill of the soil moisture forecast degraded with lead time for model prediction as well as for anomaly persistence. This is typical, where the “noisy” behavior of the precipitation controls the soil moisture evolution. The model seems to be competitive with anomaly persistence over zone 1 (where precipitation is less) but worse over zones 2 and 6 where precipitation is stronger and more likely poorly simulated.

Over zones 8, 9, and 14, prediction skill of the soil moisture decreased quickly and reached the same level as the control run in 4–5 months. These zones, particularly 8 and 9, are monsoon areas and the soil moisture evolution is mostly determined by precipitation. The model was not very skillful in reproducing precipitation over these areas (less so over zone 14). This difficulty in predicting and simulating precipitation over monsoon regions is consistent with the results reported in Sperber and Palmer (1996).

Figure 4 shows the ratio between soil moisture and evaporation, which is a timescale for soil moisture anomalies. It is clearly seen that this timescale is very short over zones 8 and 9 in agreement with the poor forecast skill there. Zones 2 and 6 also have relatively short timescales and had poor soil moisture forecasts.

Zones 3 and 4 have different behavior in prediction skill than the other areas. In these zones, the skill of the CTR forecast increased with lead time. This behavior was particularly apparent in zone 4. We hypothesize that this increased skill was caused by using observed SSTs, which produced better precipitation forecasts that influenced the soil moisture. After the memory of the initial soil moisture was lost, both the ISM and CTR forecasts had a similar positive skill. Zone 4 covers Nordeste, Brazil, which is one of the key tropical regions where precipitation anomaly is predictable when a tropical SST anomaly over the Pacific or Atlantic is specified (Sperber and Palmer 1996). The poor skill of the persistence forecast is consistent with this observation. In addition, zone 3 stretches from Panama to the Amazon basin, which includes regions that show an ENSO response.

In summary, the initial condition of soil moisture was important for summertime forecasts except over certain areas in the Tropics where precipitation was controlled by the SST anomaly. The prediction skill was high over areas receiving little precipitation. In these areas, the model evaporation was reasonably accurate. The skill declined in midlatitude temperate zones, and more rapidly in the tropical monsoon region, where precipitation was the main forcing. The only tropical regions showing good long-lead skill were in tropical South America, the Nordeste area in particular, where tropical SST anomalies strongly control the anomalous precipitation.

5. Predictability of near-surface temperature

The skillful prediction of soil moisture has significant implications to agricultural applications by itself. Furthermore, since strong anticorrelation between soil moisture and near-surface temperature is anticipated, due to the compensation between latent heat flux and sensible heat flux, the skillful prediction of soil moisture may be translated into a skillful prediction of near-surface temperature. The prediction of near-surface temperature has a much wider application and in fact, it is one of the two major forecast parameters in the long-range operational forecast of NCEP, together with precipitation. In this section, the predictability of the near-surface temperature from the same experiments is presented.

Figure 5 compares forecast skill of soil moisture, latent heat flux, and 2-m air temperature over the continental United States. Dark-and light-shaded bars show the anomaly correlations for two experiments, CTR and ISM, respectively. The verifying latent heat flux and 2-m temperature were obtained from the R-2 archive. The R-2 latent heat flux was the average of 6-hr model integration during the data assimilation. It can be seen that the forecast skill decreased for latent heat flux, and even more so for near-surface temperature, although the skills are still high enough to be useful. The skill for near-surface temperature was lower because it was determined not only by soil moisture but also by other factors that were less predictable than soil moisture.

The relationship between soil moisture anomaly and near-surface temperature anomaly was examined by looking at the pattern correlation between July soil moisture anomaly and the negative of the 2-m temperature anomaly (Fig. 6). There are high anticorrelations between the soil moisture and temperature, particularly when the forecast was started from R-2 analysis soil moisture. This may be attributed to a tendency for weaker soil moisture anomalies in the ISM forecasts, which would allow other factors to affect the 2-m temperature.

The forecast verifications presented so far have been against the R-2 analyses, which can artificially increase the forecast skill because the model used here is similar to that used by R-2. The main difference between the two models is the relaxed Arakawa–Schubert (RAS) convective parameterization used in this study versus the simplified Arakawa–Schubert scheme (Pan and Wu 1995) used in R-2. In order to remove this model dependency in the verification, the 2-m temperature was also verified against climate division data from 344 stations (CD344). These temperature observations were not used by R-2 and provide a ground truth. Figure 7 illustrates the comparison of anomaly correlation of CTR and ISM experiments over U.S. land area for the period 1979–96. Each bar is the verification of the 10-member ensemble mean. For ISM, only 2 yr had negative correlation and the score exceeds 0.4 in many years. By comparison, the anomaly correlation for the CTR experiment was much worse. Thus, the improvement demonstrated in our result was not an artifact of the model dependency but was a real improvement.

6. A possibility of delayed SST forcing

Finally, we examined occasional cases when the near-surface temperature forecasts were reasonably skillful even when the forecasts were initialized from the climatological soil moisture condition. We explored the possibility of a two-step process in which the equatorial Pacific sea surface temperature anomaly produced a soil moisture anomaly during the early part of the forecast, and then this soil moisture anomaly lead to a near-surface temperature anomaly in the following months. Since these steps involve many physical processes, it is difficult to clearly show the linkage of events. Furthermore, there were not so many clear-cut cases in the 18-yr period examined, and hence, the results shown here are not extremely statistically significant.

Figure 8 shows the agreement between CTR and ISM predicted soil moisture after a 1-month integration, using pattern correlation as a measure. The years were stratified with the tropical Pacific SST anomaly events. Warm events are shown with dark shading, normal years with light shading, and cold events are unshaded [events as defined by the NCEP Climate Prediction Center (CPC)]. During warm events, the correlation between ISM and CTR soil moisture varied between 0.25 and 0.57. This suggests that the warm tropical SST forcing was sufficient to generate a significant soil moisture anomaly over the United States during the first month of integration. For neutral and cold events the average correlation was positive but not as significant as for the warm events. This positive correlation in warm events suggests that the tropical SST anomaly information was transferred to the soil moisture anomaly over the United States. This soil moisture anomaly of the first month then persisted for several months and caused a near-surface temperature anomaly. To show this, we present a scatter diagram of the correlation of soil moisture between CTR and ISM after 1-month integration versus correlation of 2-m temperature forecast by the CTR model after 3 months (Fig. 9). This figure shows that when the CTR model soil forecast was similar to that of the ISM after 1 month, the 2-m temperature forecast after 3 months tended to have higher skill. Although the scatter is very large, we see an indication that greater similarity in the soil moisture forecast between the ISM and CTR after 1 month (in May) resulted in higher surface temperature forecast skill after 3 months (in July). This chain of events is probably pronounced when the forecasts are made from early to midspring when the impact of the equatorial SST on the weather over the continental United States is especially significant.

7. Conclusions

Using the NCEP–DOE reanalysis soil moisture analysis and the NCEP Seasonal Forecast System, the seasonal predictability of soil moisture, and near-surface temperature over the globe were examined. The NCEP–DOE reanalysis soil moisture evolution agrees well with Illinois measurements including the soil moisture having a comparable long-term memory. Unfortunately, the quality of the soil moisture analysis over other land areas was still uncertain, and the objective evaluation of the NCEP–DOE soil moisture analysis is still left for future investigation.

Two sets of forecasts were made, one starting from climatological soil moisture as an initial condition and the other starting from NCEP–DOE reanalysis soil moisture. Each set was made for 180 7-month integrations made using April 1979–96 initial conditions.

The predictability of the soil moisture was examined over varied regions around the globe. The predictive skill of the initial soil moisture conditions was very high over regions that had little precipitation. Over these areas, the evaporation process was the primary forcing for evolution for soil moisture, and the model did an excellent job in predicting this process. In many of these relatively dry or low-precipitation areas, model prediction surpassed the anomaly persistence forecast. Over temperate zones with more precipitation, and over tropical monsoon regions, the prediction skill of the soil moisture dropped quickly in the first 3–4 months. The timescale of the soil moisture memory computed as a ratio between soil moisture and evaporation showed that the timescale is particularly short over the monsoon regions and some temperate zones. For the regions noted above, the forecast starting from climatological soil moisture was quite poor, indicating that the soil moisture anomalies cannot be generated by model precipitation or by model evaporation. From a practical point of view, a correct soil moisture initial condition is essential in making seasonal prediction over these areas. In contrast, the characteristics were very different over tropical South America. The forecast starting from climatological soil moisture quickly caught up the forecast skill of initial soil moisture as well as the anomaly persistence. Particularly in the Nordeste region of Brazil, the anomaly persistence became the worst forecast after 4 months. Over these areas, SST forcing determined the precipitation anomaly over land, and subsequently forced the soil moisture evolution. The effect of initial soil moisture information tended to disappear after 3–4 months over these areas.

The near-surface temperature anomaly forecast was closely related to the soil moisture anomaly forecast. The correlation, however, was not perfect. Thus the near-surface temperature forecast skill was lower than the soil moisture forecast skill. The skill of predicting latent heat flux was lower than that of the soil moisture, and the skill in predicting the surface temperature was even lower. Additional physical processes, such as cloudiness, advection, and vertical mixing influence the near-surface air temperature thus reducing the predictive value of the soil moisture. The abundance of near-surface temperature observations allowed us to further verify these hindcasts with independent data that were not used in the NCEP–DOE reanalysis. The verification of temperature made against the 344 climate division data clearly indicated that the use of the NECP–DOE reanalysis soil moisture initial condition truly improved the forecasts.

Finally, we looked at cases when the near-surface temperature in summer was well predicted even though initial soil moisture was taken from climatology. It is suggested that the equatorial Pacific SST anomaly information created the soil moisture anomaly over the continental United States during the first month of integration, and persistence of the soil moisture anomaly would produce some skill in the near-surface temperature in the following months.

For better prediction of soil moisture and near-surface temperature during summer, it is important that the soil moisture initial conditions be accurately specified and also that the model produces a better precipitation forecast. Unfortunately, these two requirements are very difficult to satisfy in practice, and further advances in observational and modeling studies are needed. Independent measures of the soil moisture are also extremely important to eliminate any model-dependent result. In this regard, satellite observations that provide soil moisture measurement are becoming more widely available, and the development of such data is strongly encouraged.

Figure 10 shows impact of soil moisture initial conditions on the precipitation forecast over the continental United States for the June–July average verified against CPC Merged Analysis of Precipitation (CMAP) observation (Xie and Arkin 1996). Overall, the forecasts were not skillful and furthermore, we did not find systematic positive impact of initial soil moisture on precipitation forecast as verified against observations. The major part of this difficulty is an inability of the numerical model to predict observed precipitation. The studies based on model simulations seems to artificially increase the model predictability of precipitation since the verification is made against the precipitation produced by the same model. The improvement of the precipitation physics is another essential requirement for better summertime seasonal prediction.

Acknowledgments

A portion of this research was funded by a cooperative agreement from NOAA-NA17RJ1231. The views expressed herein are those of the authors and do not necessarily reflect the views of NOAA. Thanks are due to Dr. Ken Mitchell who provided invaluable comments during the course of the study. We also thank Dr. John Roads who provided useful comments. Ms. Diane Boomer helped to improve the text.

REFERENCES

  • Alpert, J. C., M. Kanamitsu, P. M. Caplan, J. G. Sela, G. H. White, and E. Kalnay, 1998: Mountain induced gravity wave drag parameterization in the NMC medium-range model. Preprints, Eighth Conf. on Numerical Weather Prediction, Baltimore, MD, Amer. Meteor. Soc., 726–733.

    • Search Google Scholar
    • Export Citation
  • Atlas, R., N. Woldson, and J. Terry, 1993: The effect of SST and soil moisture anomalies on GLA model simulations of the 1998 U.S. summer drought. J. Climate, 6 , 20342048.

    • Search Google Scholar
    • Export Citation
  • Beljaars, A. C. M., P. Viterbo, M. J. Miller, and A. K. Betts, 1996: The anomalous rainfall over the United States during July 1993: Sensitivity to land surface parameterization and soil moisture anomalies. Mon. Wea. Rev., 124 , 362383.

    • Search Google Scholar
    • Export Citation
  • Campana, K. A., Y-T. Hou, K. E. Mitchell, S-K. Yang, and R. Cullather, 1994: Improved diagnostic cloud parameterization in NMC's global model. Preprints, 10th Conf. on Numerical Weather Prediction, Portland, OR, Amer. Meteor. Soc., 324–325.

    • Search Google Scholar
    • Export Citation
  • Chou, M-D., 1992: A solar radiation model for use in climate studies. J. Atmos. Sci., 49 , 762772.

  • Chou, M-D., and K-T. Lee, 1996: Parameterizations for the absorption of solar radiation by water vapor and ozone. J. Atmos. Sci., 53 , 12031208.

    • Search Google Scholar
    • Export Citation
  • Cook, K. H., and A. Gnanadesikan, 1991: Effects of saturated and dry land surfaces on the tropical circulation and precipitation in a general circulation model. J. Climate, 4 , 873889.

    • Search Google Scholar
    • Export Citation
  • de Martonne, E., 1927: Regions of interior basin drainage. Geogr. Rev., 17 , 397414.

  • Dorman, J. L., and P. J. Sellers, 1989: A global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SiB). J. Appl. Meteor., 28 , 833855.

    • Search Google Scholar
    • Export Citation
  • Fels, S. B., and M. D. Schwarzkopf, 1975: The simplified exchange approximation: A new method for radiative transfer calculations. J. Atmos. Sci., 32 , 14751488.

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

  • Fennessy, M. J., P. A. Dirmeyer, J. L. Kinter III, A. Schlosser, J. Shukla, J. Huang, and L. Sun, 2000: Impact of initial soil moisture anomalies on seasonal atmospheric predictions over the U.S. during March–April–May 2000. Proc. Climate Diagnostics Workshop, Palisades, NY, IRI, 239–241.

    • Search Google Scholar
    • Export Citation
  • Hong, S-Y., and H-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev., 124 , 23222339.

    • Search Google Scholar
    • Export Citation
  • Hong, S-Y., and E. Kalnay, 2000: Role of sea surface temperature and soil-moisture feedback in the 1998 Oklahoma–Texas drought. Nature, 408 , 842844.

    • Search Google Scholar
    • Export Citation
  • Huang, J., H. M. van den Dool, and K. P. Georgakakos, 1996: Analysis of model-calculated soil moisture over the United States (1931–1993) and applications to long-range temperature forecasts. J. Climate, 9 , 13501362.

    • Search Google Scholar
    • Export Citation
  • Irannejad, P., A. Henderson-Sellers, T. J. Phillips, and K. McGuffie, 2000: Analysis of AMIP II Models' simulations of land surface climates. GEWEX News, Vol. 10, No. 3, 5–8.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors. 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kanamitsu, M., and Coauthors. 2002a: NCEP Dynamical Seasonal Forecast System 2000. Bull. Amer. Meteor. Soc., 83 , 10191037.

  • Kanamitsu, M., W. Ebisuzaki, J. Woollen, S-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002b: NCEP/DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 16311643.

    • Search Google Scholar
    • Export Citation
  • Kinter III, J. L., J. Shukla, L. Marx, and E. K. Schneider, 1988: A simulation of the winter and summer circulations with the NMC global spectral model. J. Atmos. Sci., 45 , 24862522.

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

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

  • Mintz, Y., 1984: The sensitivity of numerically simulated climates to land surface boundary conditions. The Global Climate, J. T. Houghton, Ed., Cambridge University Press, 79–105.

    • Search Google Scholar
    • Export Citation
  • Mitchell, K., and Coauthors. 2000: The collaborative GCIP land data assimilation (LDAS) project and supportive NCEP uncoupled land-surface modeling initiatives. Preprints, 15th Conf. on Hydrology, Long Beach, CA, Amer. Meteor. Soc., 1–4.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120 , 9781002.

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

    • Search Google Scholar
    • Export Citation
  • Pan, H. L., and W-S. Wu, 1995: Implementing a mass flux convection parameterization package for the MNC Medium-Range Forecast Model. NMC Office Note 409, 40 pp. [Available from NCEP, 5200 Auth Road, Washington, DC 20233.].

    • Search Google Scholar
    • Export Citation
  • Rind, 1982: The influence of ground moisture conditions in North America in summer climate as modeled in the GISS GCM. Mon. Wea. Rev., 110 , 14871949.

    • Search Google Scholar
    • Export Citation
  • Roads, J. O., S-C. Chen, M. Kanamitsu, and H. Juang, 1999: Surface water characteristics in NECP's Reanalysis and Global Spectral Model. J. Geophys. Res., 104 , 1930719327.

    • Search Google Scholar
    • Export Citation
  • Rowntree, P. R., and J. R. Bolton, 1983: Simulation of the atmospheric response to soil moisture anomalies over Europe. Quart. J. Roy. Meteor. Soc., 109 , 501526.

    • Search Google Scholar
    • Export Citation
  • Sato, N., P. J. Sellers, D. A. Randall, E. K. Schneider, J. Shukla, J. L. Kinter III, Y-T. Hou, and E. Albertazzi, 1989: Effects of implementing the Simple Biosphere Model in a general circulation model. J. Atmos. Sci., 46 , 27572782.

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

    • Search Google Scholar
    • Export Citation
  • Sellers, P. J., W. J. Shuttleworth, and J. Dorman, 1989: Calibrating the Simple Biosphere Model for Amazonian tropical forest using field and remote sensing data. Part I: Average calibration with field data. J. Appl. Meteor., 28 , 727759.

    • Search Google Scholar
    • Export Citation
  • Serafini, Y. V., 1990: The time scale of land surface hydrology in response to initial soil moisture anomalies: A case study. Tellus, 42A , 390400.

    • Search Google Scholar
    • Export Citation
  • Shukla, J., and Y. Mintz, 1982: Influence of land surface evapotranspiration on the earth's climate. Science, 15 , 14981501.

  • Sperber, K. R., and T. N. Palmer, 1996: Interannual tropical rainfall variability in general circulation model simulations associated with the Atmospheric Model Intercomparison Project. J. Climate, 9 , 27272750.

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

    • Search Google Scholar
    • Export Citation
  • Wang, W., and A. Kumar, 1998: A GCM assessment of atmospheric seasonal predictability associated with soil moisture anomalies over North America. J. Geophys. Res., 103 , 2863728646.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9 , 840858.

    • Search Google Scholar
    • Export Citation
  • Xie, P., and P. Arkin, 1997: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • Yamazaki, K., 1989: A study of the impact of soil moisture and surface albedo changes on global climate using the MRI-GCM-I. J. Meteor. Soc. Japan, 67 , 123146.

    • Search Google Scholar
    • Export Citation
  • Yang, R., M. J. Fennessy, and J. Shukla, 1994: The influence of initial soil moisture on medium-range surface weather forecasts. Mon. Wea. Rev., 122 , 471485.

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

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Time series of water content (mm) in 2-m soil layer from observation (dashed line) and the reanalysis (solid line: NN-R, black; R-2, gray) over Illinois for the 1982–98 period

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 2.
Fig. 2.

Persistence decay curve of deep-layer (10–200 cm) soil moisture (mm) from observations (dashed line) and R-2 analysis (solid line). Persistence is estimated from the anomaly correlation between Apr and later months

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 3.
Fig. 3.

Deep-layer (10–200 cm) soil wetness prediction skill decay curve for the 15 regions (see Fig. 4 for the locations). Anomaly pattern correlation in abscissa and lead time (month) in ordinate. The first month score is not plotted. Persistency forecast, forecast started from soil wetness climatology, and forecast started from R-2 soil wetness are marked by black dashed line, black solid line, and light solid line, respectively

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 4.
Fig. 4.

Areas selected for the verification of deep-layer (10–200 cm) soil moisture, overlaid on the global field of soil moisture to evaporation ratio in units of days. Each area is approximately 20° × 20°. Only the land regions are used in the computations. The dashed-line box was used for specific regional study, discussed in sections 5 and 6

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 5.
Fig. 5.

Comparison of the skill score of (top) deep-layer (10–200 cm) soil moisture, (middle) latent heat flux, and (bottom) near-surface temperature. The skill score was computed from the 10-member ensemble mean of 3-month avg (May–Jul). Initial analysis soil wetness runs are in gray, and initial climatological soil wetness experiments are in black

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 6.
Fig. 6.

Concurrent anomaly correlation of deep-layer (10–200 cm) soil moisture anomaly and 2-m air temperature anomaly for the 10-member ensemble mean of Jul forecast. Note that the sign of the temperature anomaly is reversed to make the correlation positive. Initial analysis soil moisture runs are in gray, and initial climatological soil moisture is in black

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 7.
Fig. 7.

Comparison of 2-m air temperature anomaly correlation scores of ISM and CTR experiments verified against independent 344 climate division data. The scores for ISM are shaded light and those of CTR are shaded dark. The skill score was computed from the 10-member ensemble average of 3-month mean for Jun–Jul–Aug period

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 8.
Fig. 8.

Agreement of deep-layer (10–200 cm) soil moisture anomaly between the initial soil moisture analysis run and the initial climatological soil moisture run after 1-month integration (May). Cold equatorial Pacific SST events (outlined bar), warm episodes (dark shaded bar) and normal years (light shaded bar) are shown. These events were based on NCEP Climate Prediction Center definitions

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 9.
Fig. 9.

Scatter diagram of correlations. Ordinate is the correlation of deep-layer (10–20 cm) soil moisture forecast after 1-month integration (May) between initial soil moisture analysis runs and initial climatological soil moisture runs, and the abscissa is the forecast skill of 2-m air temperature verified against R-2 after 3 months (Jul). Cold episodes, warm episodes, and normal years are marked by open square, solid square, and open circle, respectively

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Fig. 10.
Fig. 10.

Comparison of anomaly correlation skill score of Jun–Jul mean precipitation over the continental United States (dashed-line area in Fig. 4). The forecasts started (top) from analysis soil wetness, and (bottom) from climatological soil moisture. Each bar is the score of the 10-member ensemble mean. Verification was done against CMAP precipitation analyses (Xie and Arkin 1996)

Citation: Journal of Climate 16, 3; 10.1175/1520-0442(2003)016<0510:TPOSMA>2.0.CO;2

Save