• Bonan, G. B., K. W. Oleson, M. Vertenstein, S. Levis, X. Zeng, Y. Dai, R. E. Dickinson, and Z-L. Yang, 2002: The land surface climatology of the community land model coupled to the NCAR community climate model. J. Climate, 15 , 31233149.

    • Search Google Scholar
    • Export Citation
  • Cocke, S., 1998: Case study of Erin using the FSU nested regional spectral model. Mon. Wea. Rev., 126 , 13371346.

  • Cocke, S., and T. E. LaRow, 2000: Seasonal predictions using a regional spectral model embedded within a coupled ocean–atmosphere model. Mon. Wea. Rev., 128 , 689708.

    • Search Google Scholar
    • Export Citation
  • Dai, Y., and Coauthors, 2003: The common land model. Bull. Amer. Meteor. Soc., 84 , 10131023.

  • Dirmeyer, P. A., Z. Guo, and X. Gao, 2004: Comparison, validation, and transferability of eight multiyear global soil wetness products. J. Hydrometeor., 5 , 10111033.

    • Search Google Scholar
    • Export Citation
  • Emanuel, K. A., and M. Zivkovic-Rothman, 1999: Development and evaluation of a convective scheme for use in climate models. J. Atmos. Sci., 56 , 17661782.

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

  • Pan, H-L., and W-S. Wu, 1994: Implementing a mass flux convection parameterization package for the NMC Medium-Range forecast model. Preprints, 10th Conf. on Numerical Weather Prediction, Portland, OR, Amer. Meteor. Soc., 96–98.

  • Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Ritchie, J. T., U. Singh, D. C. Godwin, and W. T. Bowen, 1998: Cereal growth, development and yield. Understating Options for Agricultural Production, G. Y. Tsuji, G. Hoogenboom, and P. K. Thornton, Eds., Kluwer Academic, 79–98.

    • Search Google Scholar
    • Export Citation
  • Rosmond, T. E., 1992: The design and testing of the Navy operational global atmospheric system. Wea. Forecasting, 7 , 262272.

  • Schaefer, J. T., 1990: The critical success index as an indicator of warning skill. Wea. Forecasting, 5 , 570575.

  • Shin, D. W., T. E. LaRow, and S. Cocke, 2003: Convective scheme and resolution impacts on seasonal precipitation forecasts. Geophys. Res. Lett., 30 .2078, doi:10.1029/2003GL018297.

    • Search Google Scholar
    • Export Citation
  • Willmott, C. J., and K. Matsuura, cited. 2002: Terrestrial air temperature and precipitation. [Available online at http://climate.geog.udel.edu/climate.].

  • Zeng, X., M. Shaikh, Y. Dai, R. E. Dickinson, and R. Myneni, 2002: Coupling of the common land model to the NCAR community climate model. J. Climate, 15 , 18321854.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33 , 407446.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    A schematic diagram showing the coupling of the CLM2 to the FSU model.

  • View in gallery

    Surface (2 m) air temperature difference between models (FSUc and FSUCLM) and Willmott and Matsuura (2002) observations for DJF and JJA of climatological run (5-yr mean, 1992–96) using the NCEP convective scheme. Values greater than 2 K are shaded light. Values smaller than −2 K are shaded dark.

  • View in gallery

    Surface (2 m) air temperature bias over different geophysical locations for (a) DJF and (b) JJA. Regional area averaging is performed based on Fig. 2. The regions examined are global land, NH, SH, the global Tropics (45°S–45°N), Alaska and northwestern Canada (50°–70°N, 170°–110°W), northern Europe (55°–70°N, 5°–60°W), western Siberia (50°–70°N, 60°–90°E), eastern Siberia (50°–70°N, 90°–140°E), the western United States (30°–50°N, 130°–110°W), central United States (30°–50°N, 110°–90°W), the eastern United States (30°–50°N, 90°–70°W), central Europe (40°–55°N, 10°W–40°E), Central America (10°–25°N, 110°–80°W), the Amazon (10°S–0°, 70°–50°W), the Congo (10°S–5°N, 10°–30°E), India (10°–30°N, 70°–90°E), the Sahara and Arabia (10°–30°N, 20°W–50°E), southern South America (60°–25°S, 80°–50°W), South Africa (35°–10°S, 10°–40°E), and Australia (40°–10°S, 110°–160°E).

  • View in gallery

    As in Fig. 3, but for precipitation.

  • View in gallery

    Precipitation difference between models (FSUc and FSUCLM) and Willmott and Matsuura (2002) observations for JJA of climatological run (5-yr mean, 1992–96) using four different convective schemes.

  • View in gallery

    Seasonal surface (2 m) air temperature difference between models (FSUc, FSUCLM, and FSUCLMa) and Willmott and Matsuura (2002) observations over the global land from 1987 to 1996.

  • View in gallery

    Seasonal surface (2 m) air temperature difference between FSUCLMa and FSUCLM for (a) DJF and (b) JJA 1996. Values greater than 2 K are shaded light. Values smaller than −2 K are shaded dark.

  • View in gallery

    ETS (45°S–45°N over land) for seasonal precipitation from 1987 to 1996. Threshold values are (a) > 0.25 and (b) > 2.5 mm day−1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 42 40 4
PDF Downloads 10 9 1

Seasonal Surface Air Temperature and Precipitation in the FSU Climate Model Coupled to the CLM2

View More View Less
  • 1 Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, Florida
© Get Permissions
Full access

Abstract

The current Florida State University (FSU) climate model is upgraded by coupling the National Center for Atmospheric Research (NCAR) Community Land Model Version 2 (CLM2) as its land component in order to make a better simulation of surface air temperature and precipitation on the seasonal time scale, which is important for crop model application. Climatological and seasonal simulations with the FSU climate model coupled to the CLM2 (hereafter FSUCLM) are compared to those of the control (the FSU model with the original simple land surface treatment). The current version of the FSU model is known to have a cold bias in the temperature field and a wet bias in precipitation. The implementation of FSUCLM has reduced or eliminated this bias due to reduced latent heat flux and increased sensible heat flux. The role of the land model in seasonal simulations is shown to be more important during summertime than wintertime. An additional experiment that assimilates atmospheric forcings produces improved land-model initial conditions, which in turn reduces the biases further. The impact of various deep convective parameterizations is examined as well to further assess model performance. The land scheme plays a more important role than the convective scheme in simulations of surface air temperature. However, each convective scheme shows its own advantage over different geophysical locations in precipitation simulations.

Corresponding author address: Dr. D. W. Shin, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840. Email: shin@coaps.fsu.edu

Abstract

The current Florida State University (FSU) climate model is upgraded by coupling the National Center for Atmospheric Research (NCAR) Community Land Model Version 2 (CLM2) as its land component in order to make a better simulation of surface air temperature and precipitation on the seasonal time scale, which is important for crop model application. Climatological and seasonal simulations with the FSU climate model coupled to the CLM2 (hereafter FSUCLM) are compared to those of the control (the FSU model with the original simple land surface treatment). The current version of the FSU model is known to have a cold bias in the temperature field and a wet bias in precipitation. The implementation of FSUCLM has reduced or eliminated this bias due to reduced latent heat flux and increased sensible heat flux. The role of the land model in seasonal simulations is shown to be more important during summertime than wintertime. An additional experiment that assimilates atmospheric forcings produces improved land-model initial conditions, which in turn reduces the biases further. The impact of various deep convective parameterizations is examined as well to further assess model performance. The land scheme plays a more important role than the convective scheme in simulations of surface air temperature. However, each convective scheme shows its own advantage over different geophysical locations in precipitation simulations.

Corresponding author address: Dr. D. W. Shin, Center for Ocean–Atmospheric Prediction Studies, The Florida State University, Tallahassee, FL 32306-2840. Email: shin@coaps.fsu.edu

1. Introduction

Many meteorological institutes around the world possess their own numerical models to predict and study the weather and climate system. To make a better forecast and understand the chaotic nature of the atmospheric system, they have been continuously developing their models mainly by incorporating cutting edge knowledge in physics, that is, advanced physical parameterizations, but still within the Newtonian paradigm in dynamics. To catch up with this trend, the Florida State University (FSU) science team has also introduced several improved physical processes to the FSU numerical model for weather and climate studies during the last few decades.

The FSU model has recently been updated by introducing new physical parameterizations as well as ocean models (Cocke 1998; Cocke and LaRow 2000). In spite of overall substantial improvements, introducing new physics resulted in a strong wet bias and cold surface temperature bias. It was initially believed that these biases were most likely related to the convective scheme and the land surface treatment in the FSU model. In this connection, several state-of-the-art cumulus parameterizations were introduced and tested in seasonal precipitation forecasts (Shin et al. 2003). In their study, it was found that, although there are benefits from each convective scheme, the wet bias problem was neither reduced much nor eliminated. The wet bias (including cold bias) was mainly due to strong latent heat and weak sensible heat fluxes from the currently employed simple land scheme. Rather than tuning physics to the old land scheme, the authors decided that an advanced sophisticated land model should be incorporated in order to obtain better fluxes from land to atmosphere. Among many available advanced land surface models, the authors chose the National Center for Atmospheric Research (NCAR) Community Land Model Version 2 (CLM2) as our replacement since this model is known to warm surface air temperature and to decrease precipitation amount by reducing latent heat flux and increasing sensible heat flux (Bonan et al. 2002; Zeng et al. 2002; Dai et al. 2003). However, it should be noted that these biases may be model dependent. Therefore, there is no guarantee that the inclusion of this land model will result in the anticipated improvements in seasonal air temperature and precipitation.

This paper describes the inclusion of the CLM2 in the FSU model system and focuses its role in seasonal simulations of surface air temperature and precipitation. This coupling work is actually motivated by our currently ongoing multidisciplinary project, involving downscaling for crop model application, between the Center for Ocean–Atmospheric Prediction Studies at FSU and the Agriculture and Biological Engineering Department at the University of Florida. Crop yields in a crop model depend largely on atmospheric input data, such as surface air temperature, precipitation, and shortwave radiation at the surface. Hence, good seasonal predictions of these variables are crucial to the successful simulation of crop yields. Surface solar radiation shows reasonable agreement with observation in our previous experiments. The analysis of surface solar radiation is not shown in this paper since the impact of the new land model on surface solar radiation is moderately small in the seasonal average sense. The authors are expecting that the FSU climate model coupled to the CLM2 (FSUCLM) in this study will help to provide more realistic surface air temperature and precipitation fields for crop model application.

The paper is organized as follows. Section 2 briefly describes land surface models coupled to the FSU climate model and coupling strategy. Experimental design is discussed in section 3, and results are shown in section 4. Conclusions and future work follow in section 5.

2. Land surface models and coupling strategy

The original FSU simplified land surface scheme includes a three-layer soil temperature model based on the force–restore method. Surface characteristics are determined from the U.S. Geological Survey (USGS) 24-category land use and land cover survey. Seasonally varying climatological values for soil moisture, land albedo, and surface roughness are prescribed based on the USGS data. The original land parameterization in the FSU model estimates surface fluxes of momentum, heat, and moisture via similarity theory. Surface temperature is obtained by the surface energy balance coupled to the similarity theory.

A brief description of the CLM2 is provided here. The surface is represented by five main land types (glacier, lake, wetland, urban, and vegetated) in each grid cell. The vegetated fraction of the grid cell is further divided into patches of up to 4 of 16 possible plant functional types, each with its own leaf and stem area index and canopy height. The soil texture dataset allows vertical profiles of sand and clay. There are 10 layers for soil temperature and soil water with explicit treatment of liquid water and ice (see Bonan et al. 2002 for details).

A schematic diagram showing the coupling of the CLM2 to the FSU model is given in Fig. 1. The FSU atmospheric model provides 1-h-averaged total incident solar radiation (direct beam and diffuse for visible and near-infrared wave bands), precipitation, lowest model-level temperature (T), lowest horizontal wind components (u and υ), specific humidity (q), pressure (p), and height above surface (h) to the CLM2. The land model is sequentially run at every hour to calculate surface temperature (Tg) and to provide latent and sensible heat fluxes, surface wind stress, upward longwave radiation, and surface albedo (direct and diffuse for visible and near-infrared wave bands). In these preliminary experiments, the CLM2 is coupled to the FSU climate model at 1-h intervals. The atmospheric forcings are averaged over the 1-h interval prior to being passed to the land model. The resultant fluxes produced by the land model are then held fixed until the subsequent call to the land model. The choice of the 1-h interval was based partly on computational efficiency and partly to use a time step that is not too short compared to what the CLM2 has been tested within other models. The authors did find that coupling the CLM2 at every time step of the model resulted in what appeared to be an unstable computational mode. This mode appears to be a leapfrog-type instability, perhaps due to a synchronization problem between the time filtered atmospheric input variables and the time tendencies generated by the land model, though this needs to be investigated further. Since the authors intend to use the CLM2 in the FSU model at very high resolutions in future experiments, some coupling interval will need to be decided based on computational efficiency and veracity of the simulations.

3. Experiment design

Three experimental setups are devised as follows to study the role of the new land model in seasonal simulations of surface air temperature and precipitation. All simulations use observed weekly sea surface temperatures (Reynolds et al. 2002) for the experiment period.

a. Climatological integrations

Climatological simulations of 10-yr length (1987–96) are first performed with each land model and four convective schemes coupled to the FSU climate model at a resolution of T63 (∼1.86°) with 17 vertical levels. Two land models that we use are the original FSU land surface scheme (control, hereafter FSUc) and the CLM2. The four convective schemes are 1) the National Centers for Environmental Prediction (NCEP) Simplified Arakawa–Schubert (SAS) (Pan and Wu 1994): moisture flux, only one cloud type; 2) the NCAR Zhang–McFarlane (1995; ZM): a plume ensemble approach similar to the original Arakawa–Schubert (AS) but with three significant assumptions; 3) the Naval Research Laboratory (NRL) Relaxed Arakawa–Schubert (RAS) (Rosmond 1992): different handling of detrainment; and 4) the Massachusetts Institute of Technology (MIT) (Emanuel and Zivkovic-Rothman 1999): buoyancy-sorting hypothesis, mixing hypothesis, and a stochastic coalescence model. For brevity, the details of these cumulus convection schemes are not given in this paper. The integrations commence on 1 January 1987. The initial land model condition is the same as those of Bonan et al. (2002). Only the last 5 years of the simulations (i.e., for the period 1992–96) are analyzed to allow a 5-yr spinup of soil water and temperature for the FSUCLM run.

b. Seasonal integrations

Seasonal integrations of summer (May–August) and wintertime (November–February) are carried out next with each land model for the 10-yr period (1987–96). The NCEP convective scheme was employed in these simulations since this scheme is currently the default in the FSU climate model, and has been tested more thoroughly. The last three months of each integration are analyzed. Land-model initial conditions for the FSUCLM are obtained from the above final year of 10-yr spinup climatological simulations (1 May 1996 for summer and 1 November 1996 for winter). Each seasonal integration differs from each other in the atmospheric initial conditions and sea surface temperatures.

c. Assimilation experiment

Assimilation of land surface data remains a challenging problem, not the least of which is the lack of a global in situ observational network for soil moisture and other subsurface variables (Fennessy and Shukla 1999). Furthermore, while a number of land data assimilation systems are under development, retrospective land surface reanalyses are not likely to be available soon (Dirmeyer et al. 2004). Even if such land reanalyses were available, it would still present potential problems to adapt them to another land model such as the CLM2, including spinup effects in the land model itself [although some solutions have been proposed; see Dirmeyer et al. (2004)]. Thus there are essentially two practical options for assimilating data into the land surface model: 1) force the land model with an atmospheric model or 2) force the land model with meteorological reanalyses for the upper boundary condition. The latter approach presents at least a couple difficulties. First, the spunup land model using reanalyses may not be in proper balance with the coupled model, yielding further spinup problems when running climate simulations. Second, the reanalyses are generally of coarse temporal resolution, at least 6 h or longer, and thus may not fully capture the diurnal cycle.

The authors thus chose to initialize the land model using option 1) via an indirect assimilation procedure. The coupled land–atmospheric model underwent an approximately 10-yr sequence of 1-day integrations (from 1 June 1986 to 1 December 1996) where the atmospheric model was initialized each day with observed European Centre for Medium-Range Weather Forecasts (ECMWF) analyses. The land model was initialized daily with the 1-day integration from the previous run. The assumption here is that the atmospheric model does not drift much from observation during the 1-day integration, and thus provides reasonable near-surface fields and proper diurnal cycle variation. Prior to the start of this assimilation procedure, a 10-yr climatology spinup run was performed. Using these land model initial conditions for a particular model integration starting date, seasonal simulations are carried out for the same period of 10 yr. This experiment is hereafter called the FSUCLMa.

4. Results

a. Climatological simulations

Figure 2 shows 5-yr-averaged (1992–96) surface (2 m) air temperature differences between the simulations (FSUc and FSUCLM) and the Willmott and Matsuura (2002) observations for winter [December–February (DJF); Figs. 2a,b] and for summer [June–August (JJA); Figs. 2c,d] from the climatological run. While positive values (> 2 K) are shaded light, negative values (< −2 K) are shaded dark. Simulations using the NCEP convective scheme only are shown here. Since the impact of different convective schemes on surface air temperature is moderately small, other simulations using different convective schemes are not shown in this figure. Surface air temperature bias distributions are very similar to each other, but there are slightly detectable differences in bias magnitudes among the convective schemes. For the boreal winter season, the control run (FSUc; Fig. 2a) shows a prominent cold bias in the Tropics and Southern Hemisphere (SH) and slightly warm bias in the northern high latitudes. The cold bias in the Tropics and SH is reduced or eliminated and even becomes a warm bias in some local regions in the FSUCLM (Fig. 2b). However, a strong cold bias appears over the Eurasia continent. This appears to be a land initialization problem since, as we shall see in the next section, the bias is substantially reduced when using a longer spinup time with assimilation. We note that this region contains significant land-ice coverage, which might require a longer spinup time. If the model spinup time is long enough (probably 60 years), the land condition (soil temperature and soil moisture) will be better described over this area. Meanwhile, as commonly expected, the role of the land model is more important during summertime than wintertime. The strong cold bias is dominant over all land in the FSUc run during the boreal summer season (Fig. 2c). This cold bias is dramatically reduced or eliminated in the FSUCLM (Fig. 2d) and, in fact, a warm bias has been introduced, in particular in the United States. There still exists the cold bias over the Eurasia continent similar to the winter season.

The surface temperature biases are more clearly seen in Figs. 3 for winter (Fig. 3a) and summer (Fig. 3b) seasons. Regional area averaging values of temperature biases are shown over different geophysical locations. The selected locations cover the whole global land, following Bonan et al. (2002). During DJF, the FSUCLM temperatures are closer to the observed ones than those of FSUc except over northern Europe and Siberia. The reason for this has already been explained above. During JJA, the FSUCLM produces improved surface temperature distribution over most regions. This result is due to the reduced latent heat flux and increased sensible heat flux in the FSUCLM simulation (see Table 1).

To quantify the above results, skill scores in terms of rmse for surface air temperature over the global Tropics (45°S–45°N) are computed in Table 2. Scores are displayed by the two land models (FSUc and FSUCLM) and four different convective schemes for winter (DJF) and summer (JJA). The lowest rmse for each season is written in bold numerals. The FSUCLM always produces a lower rmse than the FSUc. The best scores are achieved by the combination of the NCEP convective scheme and the FSUCLM for DJF (4.08) and the NCAR and the FSUCLM for JJA (4.27). The smallest impact of inclusion of the new land model is found in the MIT scheme. The MIT scheme provides the lowest rmse in the control run. However, the introduction of the CLM2 to the FSU model with the NCEP scheme (NCAR scheme) performs better for winter (summer) season. Although there is a recognizable (almost 1°C) difference among the convective schemes, the inclusion of the CLM2 has more impact on overall skill scores. It is hard to say which convective scheme is more suited for surface temperature simulations with this statistic. It is however concluded that the land scheme plays a much more important role than the convective scheme, at least in simulations of surface air temperature.

Similar to Fig. 3, 5-yr-averaged (1992–96) precipitation differences between the simulations (FSUc and FSUCLM) and the Willmott and Matsuura (2002) observations for DJF (Fig. 4a) and JJA (Fig. 4b) are shown in Fig. 4. The global average precipitation amount of the FSUCLM is larger than that of the FSUc during the winter season in this case. This is due to too much precipitation occurring over the Congo. Except for this region, the FSUCLM precipitation amount over different regions is closer to the observation. For summertime, the FSUCLM produced better precipitation amount over most regions. This is due to a better land model specification during summertime.

It is commonly believed that the simulation of precipitation field is much more sensitive to the convective scheme than the land model. However, it should be recognized that the land model plays a basic role in producing latent heat fluxes (evaporation), which eventually control the precipitation amount. Figure 5 illustrates the summer season (JJA) precipitation difference distributions with four different convective schemes. Each convective scheme has its own advantage over different locales. A strong wet bias is noticeable over the Eurasian continent in the FSUc. This problem is noticeably reduced by the new land model (FSUCLM), which includes advanced biogeophysical parameterizations in modulating latent heat fluxes.

To determine which convective scheme performs more skillfully for precipitation distribution and intensity, the so-called equitable threat score (ETS) is employed. The definition of ETS is given in Schaefer (1990). The ETS scores the forecast purely on its ability to predict rain existence above a specified threshold value. The higher the value, the more skillful the model is for a particular threshold value. The score can vary from a small negative number to 1.0, where 1.0 represents a perfect forecast. In Table 3, ETSs for 5-yr-averaged (1992–96) precipitation over the global Tropics are shown in terms of two threshold values (> 0.25 and > 2.5 mm day−1), four convective schemes, and two land models for DJF and JJA. The highest scores are written in bold numerals for each category. All of the FSUCLM simulations give higher scores than the FSUc simulations. This is because the new land model (CLM2) plays a crucial role in modulating evaporation from land to atmosphere, that is, reducing latent heat flux (see Table 1). As a threshold value increases from 0.25 to 2.5 mm day−1, higher scores are seen. However, the scores are decreased for very high threshold values (extreme events, e.g., 10 mm day−1, not shown here). During DJF, the NCEP convective scheme produces the highest score for the lower threshold value (> 0.25 mm day−1) and the NCAR scheme for the higher threshold value (> 2.5 mm day−1). Meanwhile, during JJA, the NRL scheme gives the best scores for both thresholds. Overall, the NRL scheme has a higher skill for precipitation than other schemes, at least in the current simulations. Selecting the best convective scheme in the FSU climate model requires more experiments and ensemble forecasting techniques. The NCEP convective scheme is currently employed as the default in the FSU climate model system. Although the scores are slightly lower or equivalent compared to others, this scheme will be employed in the following seasonal simulations.

b. Seasonal simulations

As evident from the climatological simulations, the FSU climate model with CLM2 improves the simulations of both surface air temperature and precipitation compared to the control (FSUc). The FSUCLM reduced much of the cold bias noted in the FSUc run. The wet bias in the FSUc was reduced as well, especially over Eurasia during the JJA. In this subsection, the authors will examine the predictability of the FSU climate system in the seasonal simulations of surface air temperature and precipitation.

The pattern of temperature bias distribution is very similar to the climatological run (Fig. 2), even in the seasonal simulations. It is therefore not shown here. Figure 6 shows seasonal surface (2 m) air temperature difference between the models (FSUc, FSUCLM, and FSUCLMa) and Willmott and Matsuura (2002) observations over the global land from 1987 to 1996. Here the FSUCLMa is the simulation with the ECMWF daily assimilation mentioned in section 3. The winter bias is smaller than the summer bias in the FSUc run. The FSU model coupled to the CLM2 has less bias than the FSUc in both seasons except the 1993/94 DJF. The assimilation impact is clearly shown in every summer, but not in every winter simulation. The FSUCLMa bias is slightly smaller than the FSUCLM bias during the summer season. This is due to the improved land-model initial condition for the FSUCLMa. As the assimilation period increases, temperature biases are more reduced in the FSUCLMa than in the FSUCLM even during the winter season. If a longer assimilation is introduced, the bias might be further decreased.

In Fig. 7, temperature differences between the FSUCLMa and the FSUCLM for DJF (Fig. 7a) and JJA (Fig. 7b) 1996 are shown to demonstrate the impact of the assimilation. While positive values (> 2 K) are shaded light, negative values (< −2 K) are shaded dark. During 1996 DJF, several differences are detected in several locations all over the Northern Hemisphere landmass. There is no difference over the Southern Hemisphere outside of ±2°C. The cold temperature biases in northern Europe, Southeast Asia, and the southeast United States shown in the FSUCLM simulation (see Fig. 2b) are reduced in the FSUCLMa. Although the FSUCLMa introduces a colder bias over northern Africa and southern Europe, the FSUCLMa is about 0.2° closer to the observation than the FSUCLM over the global land. Meanwhile, during 1996 JJA, the FSUCLMa is warmer than the FSUCLM in the Siberian region. That is, the cold bias over this region (see Fig. 2d) is reduced due to the assimilation and partly due to a longer time spinup. This reduces the FSUCLM cold bias over the global land by approximately 0.7°.

Similar to the climatological run (Figs. 5a,b), the FSUc shows a strong wet bias in the precipitation field and less bias is achieved in the FSUCLM in this seasonal simulation. Slightly different precipitation bias distribution is shown by the FSUCLMa (not shown here). It is not easy to say which precipitation simulation is better directly from these kinds of distribution maps. Hence, ETSs for precipitation (45°S–45°N over land) are again computed for every single season from 1987 to 1996 in Fig. 8. Two threshold values are employed as before. For a 0.25 mm day−1 threshold value, the FSUc shows lower scores during summer season and relatively higher scores during the winter season. The scores of both the FSUCLM and the FSUCLMa are higher than those of the FSUc. The FSUCLMa shows slightly higher scores for most seasons as the assimilation time increases. However, it is difficult to conclude that the FSUCLMa is more skillful than the FSUCLM. Not much impact of the assimilation is found in the precipitation field compared to the surface temperature. Overall scores for the 2.5 mm day−1 threshold value (Fig. 8b) are higher than those for the 0.25 mm day−1 threshold value. The impact of the land model is still higher during summer than winter. This result is similar to the climatological run. The FSUc simulation is sometimes better than the FSU model coupled to CLM2 for some seasons in this threshold value (three seasons for this study). There are almost no differences between the FSUCLM and the FSUCLMa. Another physical process, such as the convective scheme, might be more important than the land scheme for higher threshold precipitation value simulations.

5. Conclusions and future work

This paper described seasonal surface air temperature and precipitation fields in simulations by the recently upgraded FSU climate model. The main upgrade made in this paper was the inclusion of the NCAR CLM2 as the land parameterization. Climatological and seasonal simulations were performed with the FSU atmospheric model coupled to the previously used FSU land scheme (FSUc) and the CLM2 (FSUCLM). Four convective schemes were employed as well to further assess their impact on surface air temperature and precipitation.

In the climatological runs, the FSUCLM experiment improved the simulations of both surface air temperature and precipitation compared to the FSUc in spite of the fact that the 5-yr analysis is too short to assess the statistical significance of these changes in temperature and precipitation. The FSUCLM reduced much of the surface temperature cold bias noted in the FSUc run. Although there were recognizable differences among convective schemes, the inclusion of the CLM2 showed more impact in the simulation of surface air temperature. It was difficult to conclude which convective scheme is better for surface temperature simulations. The wet bias in the FSUc was reduced as well, especially over the Eurasia continent during the JJA. The CLM2 played a crucial role in modulating the evaporation (reduced latent heat flux) over that region. Four convective schemes employed showed their own advantage over different geophysical locations. The role of the land model was found to be more important during the summer than winter season.

Similar to the climatological simulations, the FSUc showed a strong cold and wet bias and much less bias was obtained in the FSUCLM in the seasonal simulations. One more experiment was done with the ECMWF daily assimilation (FSUCLMa) in order to produce an improved land-model initial condition. As the assimilation period increases, the temperature biases become smaller and smaller in the FSUCLMa than in the FSUCLM due to the better land-model initial condition. Not much impact of the assimilation was found in the precipitation field compared to the surface temperature, although the FSUCLMa showed slightly higher scores for most seasons as the assimilation time increases.

The coupled model (FSUCLM) will be used in our ongoing project, downscaled for crop models. Since the current simulations were carried out using the FSU global climate model at a very low resolution (∼200 km), downscaling the parameters for a particular station may result in inaccurate results, in other words, unsatisfactory fields for the crop model application. In this connection, the CLM2 has to be coupled to the FSU regional climate model to allow more accurate representation of the station data. The regional model will be placed over the southeast United States and run at 20-km resolution, roughly resolving the county level. To be precise, an attempt will be made to integrate outputs from the FSU regional climate model with agricultural models to forecast maize yield in the southeast United States using the Crop Environment Resource Synthesis (CERES) maize crop model (Ritchie et al. 1998). This work will be presented in a future publication.

Acknowledgments

Computations were performed on the IBM SP4 at the FSU. COAPS receives its base support from the Applied Research Center, funded by the NOAA Office of Global Programs awarded to Dr. James J. O’Brien.

REFERENCES

  • Bonan, G. B., K. W. Oleson, M. Vertenstein, S. Levis, X. Zeng, Y. Dai, R. E. Dickinson, and Z-L. Yang, 2002: The land surface climatology of the community land model coupled to the NCAR community climate model. J. Climate, 15 , 31233149.

    • Search Google Scholar
    • Export Citation
  • Cocke, S., 1998: Case study of Erin using the FSU nested regional spectral model. Mon. Wea. Rev., 126 , 13371346.

  • Cocke, S., and T. E. LaRow, 2000: Seasonal predictions using a regional spectral model embedded within a coupled ocean–atmosphere model. Mon. Wea. Rev., 128 , 689708.

    • Search Google Scholar
    • Export Citation
  • Dai, Y., and Coauthors, 2003: The common land model. Bull. Amer. Meteor. Soc., 84 , 10131023.

  • Dirmeyer, P. A., Z. Guo, and X. Gao, 2004: Comparison, validation, and transferability of eight multiyear global soil wetness products. J. Hydrometeor., 5 , 10111033.

    • Search Google Scholar
    • Export Citation
  • Emanuel, K. A., and M. Zivkovic-Rothman, 1999: Development and evaluation of a convective scheme for use in climate models. J. Atmos. Sci., 56 , 17661782.

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

  • Pan, H-L., and W-S. Wu, 1994: Implementing a mass flux convection parameterization package for the NMC Medium-Range forecast model. Preprints, 10th Conf. on Numerical Weather Prediction, Portland, OR, Amer. Meteor. Soc., 96–98.

  • Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15 , 16091625.

    • Search Google Scholar
    • Export Citation
  • Ritchie, J. T., U. Singh, D. C. Godwin, and W. T. Bowen, 1998: Cereal growth, development and yield. Understating Options for Agricultural Production, G. Y. Tsuji, G. Hoogenboom, and P. K. Thornton, Eds., Kluwer Academic, 79–98.

    • Search Google Scholar
    • Export Citation
  • Rosmond, T. E., 1992: The design and testing of the Navy operational global atmospheric system. Wea. Forecasting, 7 , 262272.

  • Schaefer, J. T., 1990: The critical success index as an indicator of warning skill. Wea. Forecasting, 5 , 570575.

  • Shin, D. W., T. E. LaRow, and S. Cocke, 2003: Convective scheme and resolution impacts on seasonal precipitation forecasts. Geophys. Res. Lett., 30 .2078, doi:10.1029/2003GL018297.

    • Search Google Scholar
    • Export Citation
  • Willmott, C. J., and K. Matsuura, cited. 2002: Terrestrial air temperature and precipitation. [Available online at http://climate.geog.udel.edu/climate.].

  • Zeng, X., M. Shaikh, Y. Dai, R. E. Dickinson, and R. Myneni, 2002: Coupling of the common land model to the NCAR community climate model. J. Climate, 15 , 18321854.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33 , 407446.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

A schematic diagram showing the coupling of the CLM2 to the FSU model.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 2.
Fig. 2.

Surface (2 m) air temperature difference between models (FSUc and FSUCLM) and Willmott and Matsuura (2002) observations for DJF and JJA of climatological run (5-yr mean, 1992–96) using the NCEP convective scheme. Values greater than 2 K are shaded light. Values smaller than −2 K are shaded dark.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 3.
Fig. 3.

Surface (2 m) air temperature bias over different geophysical locations for (a) DJF and (b) JJA. Regional area averaging is performed based on Fig. 2. The regions examined are global land, NH, SH, the global Tropics (45°S–45°N), Alaska and northwestern Canada (50°–70°N, 170°–110°W), northern Europe (55°–70°N, 5°–60°W), western Siberia (50°–70°N, 60°–90°E), eastern Siberia (50°–70°N, 90°–140°E), the western United States (30°–50°N, 130°–110°W), central United States (30°–50°N, 110°–90°W), the eastern United States (30°–50°N, 90°–70°W), central Europe (40°–55°N, 10°W–40°E), Central America (10°–25°N, 110°–80°W), the Amazon (10°S–0°, 70°–50°W), the Congo (10°S–5°N, 10°–30°E), India (10°–30°N, 70°–90°E), the Sahara and Arabia (10°–30°N, 20°W–50°E), southern South America (60°–25°S, 80°–50°W), South Africa (35°–10°S, 10°–40°E), and Australia (40°–10°S, 110°–160°E).

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 4.
Fig. 4.

As in Fig. 3, but for precipitation.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 5.
Fig. 5.

Precipitation difference between models (FSUc and FSUCLM) and Willmott and Matsuura (2002) observations for JJA of climatological run (5-yr mean, 1992–96) using four different convective schemes.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 6.
Fig. 6.

Seasonal surface (2 m) air temperature difference between models (FSUc, FSUCLM, and FSUCLMa) and Willmott and Matsuura (2002) observations over the global land from 1987 to 1996.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 7.
Fig. 7.

Seasonal surface (2 m) air temperature difference between FSUCLMa and FSUCLM for (a) DJF and (b) JJA 1996. Values greater than 2 K are shaded light. Values smaller than −2 K are shaded dark.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Fig. 8.
Fig. 8.

ETS (45°S–45°N over land) for seasonal precipitation from 1987 to 1996. Threshold values are (a) > 0.25 and (b) > 2.5 mm day−1.

Citation: Journal of Climate 18, 16; 10.1175/JCLI3470.1

Table 1.

Five-year-averaged (1992–96) latent and sensible heat fluxes over the global land. Units: W m−2.

Table 1.
Table 2.

Surface (2 m) air temperature rmse (45°N–45°S).

Table 2.
Table 3.

ETS (45°S–45°N over land) for 5-yr-averaged (1992–96) precipitation.

Table 3.
Save