1. Introduction
The World Meteorological Organization’s Global Energy and Water Cycle Experiment Continental-Scale International Project (GCIP) is under way. The overall objectives of the multiagency GCIP program, which is being led by the National Oceanic and Atmospheric Administration, are to increase scientific understanding of the hydrologic and energy cycles involved in the complex land–atmosphere–ocean interactions in the Mississippi River basin. Macroscale hydrological models, high-resolution atmospheric models, and coupled hydrological–atmospheric models will be improved as a result of this project. Information retrieval schemes incorporating existing and future satellite observations coupled with enhanced ground-based observations will be developed. This new observing and modeling capability will provide new capability for predicting future variability of water resources on timescales of hours to seasons.
What do we know about energy and water cycles of the Mississippi River basin? A number of previous studies have really only discussed the water cycle using a variety of observations. Rasmusson (1967, 1968) used radiosondes and streamflow measurements. Roads et al. (1994) used the National Centers for Environmental Prediction’s (NCEP, formerly the National Meteorological Center) final pressure analysis, the National Climatic Data Center’s (NCDC) precipitation observations, and the U. S. Geological Survey’s streamflow observations. Mo and Higgins (1996) compared NCEP’s and the National Aeronautical and Space Administration’s reanalysis atmospheric water products with satellite and station observations. From this multivariate combination of observations and model analyses, we have begun to quantitatively understand climatological characteristics of various components of the U.S. water cycle.
Our dependence upon a wide variety of data sources is being simplified by having the complete budgets available from global forecast model analysis output. This kind of information was previously only available from general circulation models (GCMs), but now analysis and forecast models have evolved to the stage of providing much better output, in many respects, than any GCM. In fact, analysis and reanalysis datasets are really beginning to provide the only validation available for certain parameters. Analysis climatologies are, of course, not perfect and need to be constantly monitored and compared to diverse observations, just like GCMs (See Trenberth and Olson 1988; Trenberth and Guillemot 1995; Wang and Paegle 1996).
Various significant improvements have been implemented over the years in the analysis system (see e.g., Caplan and White 1989; Kanamitsu et al. 1991; Roads et al. 1995). Unfortunately, this constant improvement in analyses can be detrimental for studying long-term climate variations, especially if analyses changes are substantial (see Trenberth and Olson 1988; Kalnay et al. 1996). To develop consistency, various reanalysis efforts are under way (see e.g., Kalnay and Jenne 1991; Schubert et al. 1993; Kalnay et al. 1996). Although data inputs will vary, each of these projects will at least use a consistent model to reanalyze long periods in the recent past. In particular, NCEP is using a lower-resolution T62L28 “frozen” (as of 11 January 1995) version of the present operational, T126L28 Global Spectral Model (GSM). The GSM is used for the operational global data assimilation system (GDAS) and medium-range forecast (MRF) predictions. NCEP’s first reanalysis will subsequently be continued as part of a climate data assimilation system (CDAS) (see Kistler et al. 1993). However, since the current operational GDAS is similar in many respects to the current NCEP reanalysis effort, except for having twice the resolution, GDAS products provide an acceptable substitute for GDAS-based reanalysis products.
Besides the approximations in the parameterized physics and dynamics of the forecast models and the model spinup discussed later, global analyses are also affected by the resolution of the global models. To overcome some aspects of the resolution problems, higher-resolution regional models and analysis systems are being developed. In fact, a major analysis system for the GCIP project will be the Eta Model data analysis system (EDAS) (see Berbery et al. 1996). However, GDAS does provide the initial and boundary conditions for the Eta and regional spectral model (RSM); many of the same physical parameterizations in the GSM used for the GDAS are also present in the Eta and other regional GCIP models (the exact same parameterizations are used in the RSM); the resolution of the T126 analysis (1°) is not that much different from the Eta and other regional model analyses (0.5°); and the same basic observations are used in the GDAS, EDAS, and other regional analyses.
Many aspects of the GCIP water and energy budgets can probably already be developed from the GDAS analysis. In any event, it is important to understand what we can currently do with global operational products so we will better understand the advantages of working with new reanalysis and operational GCIP analysis products. Therefore, we have developed, and present here, GDAS thermodynamic energy budgets for the United States with a special focus on the GCIP Mississippi River basin for the period from 11 April 1991 to 5 January 1995.
Before we describe some characteristics of the GDAS thermodynamic energy budget, we first describe the fundamental thermodynamic energy equations that govern the atmospheric and surface thermodynamic energy cycle (section 2), and then we discuss some basic characteristics of the GDAS products (section 3). We then present, in section 4, the surface and atmospheric thermodynamic energy budgets that can be deduced from these products. The anomalous July 1993 event is then examined in section 5. Conclusions are provided in section 6.
2. Basic thermodynamic equations
a. Conventions
In the budget equations below, fluxes are positive upward, eastward, and northward. Thus, in the vertical, the net flux convergence into the atmospheric column is the net flux at the bottom of the atmosphere minus the net flux at the top of the atmosphere. In the horizontal, the net flux convergence is the net flux at the southward and westward boundaries minus the net flux at the northward and eastward boundaries.
The atmospheric sigma coordinate system used in GDAS [see Sela (1980, 1982) and NMC Development Division (1988) for technical descriptions of NCEP’s GSM], σ = p/ps, increases toward increasing pressure, from 0 to 1, with unity reached when pressure is equal to the surface pressure, ps. Therefore, vertical sigma integrals are from the top of the atmosphere down or from 0 to 1. That is, { . . . } = ∫ . . . dσ. Surface variables represent either variables defined at this surface–atmosphere interface or as a vertical integral of near-but-below-surface quantities.
The basic variables in the atmospheric thermodynamic energy equation are temperature, T, virtual temperature, Tv, and surface temperature, Ts. The thermodynamic atmospheric energy is obtained by multiplying the atmospheric temperature by the atmospheric heat capacity at constant pressure, Cp; the thermodynamic surface energy is obtained by multiplying the surface temperature by the volumetric surface heat capacity, Cv. The primary variables are horizontal winds v; vertical sigma velocity,
The heating by the net condensation associated with precipitation, LP, is an important component of the atmospheric thermodynamic budget. Cooling of the surface temperatures by the latent heat of evaporation, LE, is also an important component of the surface thermodynamic budget. Another important surface component is sensible heating, SH, which is important, both for cooling the surface, and for heating the atmosphere.
Heating and cooling due to the radiative flux convergence are also important contributors. The vertical integral of the radiative flux convergence considered here is simply the difference between the net shortwave, NSW(1) and net longwave fluxes NLW(1) at the bottom of the atmosphere, and the net shortwave NSW(0), and net longwave fluxes NLW(0), at the top of the atmosphere.
The analysis also contains a ground heat flux that affects the surface temperature by transferring heat from subsurface storage areas to the surface area. Over land, ground heat effects are especially noticeable for the diurnal cycle; the relative contribution to the seasonal cycle is smaller. However, over the ocean, the ground heat flux is associated with ocean advection of heat, which is a very important component, especially in the western boundary Gulf Stream and Kuroshio currents. Both land and ocean ground heat fluxes are determined as part of the surface residual discussed later.
b. Atmospheric thermodynamic energy equation
The tendency, (∂{πCpT})/(∂t), is affected by the large-scale convergence of sensible heat, −∇·{π vCpT}; the adiabatic conversion, {ωα}, associated with rising and sinking air; and the turbulent transfer of sensible heat, SH, from the surface. The latent heat released when water vapor is converted into liquid and solid water (and vice versa when it evaporates before reaching the surface) can be written as a product of latent heat and precipitation, LP, although the actual latent heat released should really depend upon whether water vapor is converted to rain or snow (and thus whether latent heat of condensation or fusion is used). The radiative flux convergence of solar and terrestrial (or shortwave and longwave) radiation is expressed here as a difference between the net flux at the surface minus the net flux at the top of the atmosphere,
In this paper, we use kelvin per day as our thermodynamic energy unit, which is obtained by dividing the GDAS mks (meter, kilogram, second, or m, kg, s) archived fluxes by the atmospheric mass, {π}, and the heat capacity, Cp, and then multiplying by 8.64 × 104 s. The atmospheric mass is approximately 104 kg m−2 over the ocean and less over mountainous regions. Here, Cp is a constant 1.0046 × 103 W m−2.
c. Surface thermodynamic equation
The tendency term ∂{CvTs}/∂t is decreased by sensible heat, SH, as well as by the latent heat of evaporation, LE, that occurs as surface water is transformed into atmospheric water vapor. Again, the latent heat should be dependent upon whether evaporation is occurring over snow and frozen ground and water. The net radiation impinging upon the surface, modified by the surface longwave cooling, [−NSW(1) + NLW(1)] provides the major input into the surface temperature. The ground heat flux [G(−H)] acts to modify the tendency and modulates the temperature variations that would result from a strict thermodynamic energy balance.
Again, we divide and multiply these archived surface thermodynamic energy quantities by the same values used for the atmospheric thermodynamic energy variables in order to express the balances in kelvins per day and in order to have direct comparisons with atmospheric thermodynamic energy processes.
3. GDAS output
Figure 1 shows the U.S. GCIP area and grid points from the T126 GDAS (1°) 384 × 190 grid as well as the grid points for four large-scale domains (LSAs) within the GCIP area. The four LSAs are being emphasized sequentially within the GCIP project. These areas and their focus include (the region marked by multiple As) warm season studies within the LSA southwest (SW); (region B) cold-season studies within the LSA northcentral (NC); (region C) moist climate studies over the Ohio River valley, Appalachians, and Tennessee valley LSA east (region E); (region D) orographic and landcover studies in the upper Missouri LSA northwest (NW). Within the GCIP domain, there are 399 GDAS grid points. There are 129 grid points within the LSA SW (region a); 57 within the LSA NC (region b); 77 within the LSA E (region c); and 136 within the LSA NW (region D). GDAS’ vertical grid was variable: from 6 March 1991 to 11 August 1993, the GDAS had 18 sigma levels; after 11 August the number of levels increased to 28; currently the analysis, as well as NCEP’s reanalysis, has 28 sigma levels.
GDAS upper-air variables, which are archived in what are referred to as sigma files, include spectral coefficients of vorticity, divergence, virtual temperature, and specific humidity at all sigma levels. Also available are spectral coefficients of the natural logarithm of surface pressure. In addition to the traditional variables analyzed for the previously mentioned pressure analyses, the new analyses also contain several relevant hydrologic and thermodynamic energy variables in associated flux files. Precipitation, evaporation, surface sensible heating, ground heat flux, and surface and top-of-the-atmosphere radiation are available. We can now examine precipitation and evaporation as well as just deriving moisture and thermodynamic energy fluxes.
Although incoming solar radiation is not archived in the GDAS flux files, this was fairly straightforward to derive. The GDAS solar constant of 1367.4 W m−2 was used with the integrated the solar zenith angle (see Iacobellis and Somerville 1991), which is dependent upon longitude, latitude, and season, over the daylight hours (the zenith angle was derived using 1-h time steps and averaging for a period centered around the analysis time). Also dependent upon season is the distance of the earth from the sun, which varies (see Fleagle and Businger 1963) by 5 × 106 km between 3 January (perihelion of 147 × 106 km) and 5 July (aphelion of 152 × 106 km). The GDAS solar constant is multiplied by the ratio of the square of the mean distance to the variation in the distance resulting from the elliptical orbit.
The filtering and the advective form of the equations (as well as double precision computations) seemed to decrease the overall noise in preliminary diagnostics. Again, as shown by Chen et al. (1996), precipitation and evaporation accumulations cannot be easily compared with these instantaneous moisture and thermodynamic energy fluxes. For example, fluxes are calculated from the initial state, whereas precipitation is really a 6-h accumulation. Comparisons should really have fluxes bracketing the precipitation accumulation period, or better yet the fluxes should be accumulated just like precipitation is. This incompatibility problem is avoided here, superficially, by only considering long-term averages. In that regard, trying to deduce the diurnal budgets from the present GDAS products should be cautiously approached. To really understand the diurnal cycle, we recommend that future analyses should include accumulated fluxes eight times a day and that these accumulated fluxes should be accumulated over the same time period, 3 h, as the precipitation is accumulated.
There are other problems. A few relevant processes are not archived: horizontal smoothing and incoming solar radiation at the top of the atmosphere. Probably the most important problem, previously discussed by Kanamitsu and Saha (1996), is the spinup and spindown inherent in analyses (see also Schubert et al. 1993). In GCM simulations and nature, atmospheric and upper-surface thermodynamic tendencies vanish in budget equations averaged over sufficiently long times. However, if the analysis model is not perfect, it will tend to move away from the observations toward a more compatible model state that is somewhat different from observations. This movement toward the model climatology is systematic and the systematic tendency term or residual is important to the perceived budget. The movement could also be transient if the initial state does not properly represent all aspects of the three-dimensional state; for example, a perfect model with an overly filtered initial state should also show a spinup or spindown.
Many of these (and more) problems are lumped into residual atmospheric and surface terms, which are presented here as a negative tendency term. We might think of these residuals as representing a missing piece, but they are also due to slight inaccuracies in some parameterization or numerical approximation in the model or initial state. We will present the residuals below. In fact, we regard the residuals as one simple measure of the goodness of analysis (or derived analysis) budgets. Of course, having a small residual is not an assurance of goodness. GCMs have infinitesimal residuals but erroneous climatologies. Goodness is really a multivariable and multiprocess measure.
Spectral characteristics of the annual mean budgets were examined (not shown). At the lowest wavenumbers, the precipitation is balanced by the evaporation in the moisture equation and by the radiation and sensible heating in the thermodynamic equation. At moderate wavenumbers, the balance is between the precipitation and moisture convergence in the moisture equation and between precipitation and thermodynamic energy divergence in the thermodynamic equation. At the highest wavenumbers, especially for the thermodynamic energy convergence, the balance occurs between the convergence and residual terms. Especially noticeable, however, was the strong increase at the highest wavenumbers, especially for the thermodynamic energy convergence terms. Because of this spectral increase, we employed a fourth-order diffusion (since a fourth-order diffusion is used in the GSM) with a coefficient that provided a complete damping at n = 126. However, because the residual was still dominant, we subsequently decided to truncate all quantities to T61 (zero out wavenumbers n = 62 and above). As shown by Chen et al. (1996), accumulated fluxes will substantially reduce a posteriori smoothing requirements.
What we cannot really analyze from the present GDAS archive is the vertical structure of the diabatic heating field. Although the vertical structure of the fluxes were initially derived, it soon became apparent that cumulus convection parameterizations were providing dominant vertical transports, especially over the Mississippi River basin during the summer (see Roads et al. 1996). Again, we suggest that the global analysis products could be improved by also providing the vertical structure. In that regard, we do note (see Kalnay et al. 1996) that the reanalysis does at least contain the vertical distribution of the heating terms.
4. Climatology
a. Summary
Table 1 provides the overall summary of the GCIP atmospheric and surface thermodynamic energy budgets for annual, December–January–February (DJF), and June–July–August (JJA) means. The first part of the table shows the terms in the surface thermodynamic energy balance, then the surface temperature and atmospheric temperature, and then the terms in the atmospheric thermodynamic energy balance. Again, residuals are presented with the sign appropriate to a missing term. For example, in the atmospheric temperature equations, if the sign of the residual is positive, something is needed for heating; without this residual the atmospheric temperature will decrease and long-term simulations may have a cold bias.
It should first of all be noted that the residual terms are small but nonnegligible. Imbalances were discussed by Kalnay et al. 1996 (see also White 1994), who compared thermodynamic energy fluxes to climatological estimates by Morel (1994) and Ramanathan (1989). Ramanathan et al. (1996) attribute potential imbalances to deficient atmospheric absorption of solar radiation, especially solar radiation in cloudy regions. Besides errors in radiation parameters, another source of error is the cumulus parameterization. There could also be too much large-scale transport of heat into the region during the summer by the large-scale motions. Wintertime appears to provide the best thermodynamic energy balance in the GCIP region. Anyway, without the residual term, the GDAS GCIP atmospheric and surface temperature would increase, especially during the summer; during the winter, the temperatures would decrease. Essentially, something is needed to cool down the analysis over the GCIP region.
It should next be noted that radiation is the dominant component in the surface and atmosphere. Cooling by outgoing surface heat fluxes is slightly smaller, especially during the winter when the dominant surface balance is between incoming solar radiation and outgoing longwave radiation. However, during the summer the sensible and latent heat fluxes are just as important as the planetary radiation and help to balance the solar radiation. In the atmosphere the radiative cooling is the dominant term, modified slightly by the solar radiation. During the winter the thermodynamic energy convergence along with the latent heating from precipitation help to balance the radiative cooling, which is augmented by net sensible cooling of the atmosphere. During the summer, sensible heating is just as important as precipitation and the net thermodynamic energy convergence is negligible (and may even be negative).
A final point shown in Table 1, somewhat extraneous to the purpose of this paper, is that the precipitation is always larger than the evaporation because of moisture convergence. The strongest convergence occurs during the winter when the evaporation is smallest. During the summer, the evaporation increases as the moisture convergence shuts down. This suggests, but does not prove, that precipitation anomalies may be more strongly affected by anomalous moisture convergence during the winter and more strongly affected by anomalous evaporation during the summer. Previous work (Roads et al. 1994) with evaporation determined as a residual suggested that moisture convergence was always more important. Further efforts to resolve this issue with reanalyses and GCIP regional model analyses are being pursued.
b. Geographic variations
Seasonal geographic characteristics are shown in Figs. 2–5 for the U.S. region. As shown in Fig. 2, during the summer the net solar radiation at the surface is especially strong over the U.S. southwest, with minor variations that must be due to cloud cover, especially over the ocean off southern California. Longwave cooling partially balances this solar heating, especially over the Great Basin. Interestingly, the cooling is weak over the coastal Pacific Ocean. This east–west contrast also shows up in the sensible cooling, which is strongest over the southwestern deserts, and the latent cooling, which is strongest over the eastern U.S. and Gulf of Mexico regions. There is also some residual heating and cooling present over all regions; this residual forcing is small scale and probably indicates potential problems with using these derived fields during the summer.
As shown in Fig. 3, the atmospheric solar heating is relatively weak during the summer, in comparison to the longwave cooling. Note that both fields are geographically uniform, especially in comparison to the precipitation heating fields. Note also that the precipitation is strongly related to the evaporation, with maximum amounts occurring over the eastern portion of the country as well as the Gulf of Mexico. The precipitation is especially strong over the Sierra Madres and the Central American isthmus. There is a clear contrast between the U.S. east and west. In the U.S. west, the atmosphere is heated by sensible heating, whereas in the U.S. east the atmosphere is heated by net condensation, which is equal to the total precipitation at the surface. Heating by net condensation is especially strong over Mexico, which is in the middle of the Central American monsoon circulation during this season. The thermodynamic energy convergence is basically positive over the ocean regions and basically negative over the land regions. Thermodynamic energy divergence is strongly negative over the Sierra Madres, where the precipitation is strongest. Basically, thermodynamic energy convergence appears to act to compensate the latent heat released due to condensation. The residual is not small, but the small-scale nature of the field suggests that this is a random field and is partially due to our inability to accurately calculate thermodynamic energy convergence from current snapshot archives.
As shown in Fig. 4, the atmospheric heating by solar and terrestrial radiation during the winter is again quite uniform. Precipitation is also important in the coastal regions, especially the U.S. northwest coast, and sensible heating is also strong over the east coast. The dominant balance in the interior of the continent is between the thermodynamic energy convergence and the longwave cooling. Unfortunately, this balance is obscured by the large and variable residual over the continent, which is probably related to errors in the derived convergence field. It would probably not be unreasonable to simply combine these two fields. However, part of the residual may be contributed by precipitation deficiencies as well as deficiencies in other parameterizations and it really is not clear how to partition this residual among the convergence as well as the other fields.
As shown in Fig. 5, the wintertime solar radiation at the surface sharply decreases with respect to latitude, whereas the infrared radiative cooling is more uniform over the entire United States. The strongest variations occur off the East Coast where the evaporation and sensible cooling are especially strong. This cooling is largely compensated for with ocean processes in the Gulf Stream as well as the Gulf of Mexico. Overall, the DJF surface balance over the land regions is between the radiation fields. This is in marked contrast to the summertime balances that involve surface thermodynamic energy transfers of latent and sensible heat.
c. Seasonal variations
Seasonal variations over the GCIP region are summarized in Figs. 6–9. As shown in Fig. 6, the dominant increase in solar radiation at the surface during the summer is balanced both by sensible cooling and evaporative cooling. Infrared cooling shows little seasonal variation, but because of the decreased surface fluxes during the winter becomes the dominant term balancing the solar radiation. The residual shows smaller variations although it helps to heat the surface during the winter and cool it during the summer. The dominant balance appears to be one of radiative equilibrium during the winter and one involving surface interactions as well as radiation during the summer.
Surface variations in the four LSAs are shown in Fig. 7. The large-scale domain LSA NW (region D) has the largest variations in the solar radiation. The minimum wintertime amounts in LSA NC (region B) and NW (region D) are due to the low incident angles of the sun in northern regions. The maximum amounts in the LSA NW (region D) and SW (region A) are presumably due to minimum cloud cover during the summer. Longwave cooling is a maximum in the LSA NW (region D) during the summertime and a minimum in the eastern region for almost all seasons. Sensible cooling reaches the largest negative values during the summer in the LSA SW (region A) and LSA NW (region D) regions and reaches the largest positive values in the winter in the LSA NW (region D). Latent cooling by evaporation, by contrast, is a maximum in the LSA E (region C) and a minimum in the LSA NW (region D) in all seasons. The highest temperatures are reached in the LSA SW (region A), and the lowest temperatures are reached in the LSA NC (region B) and LSA NW (region D) regions during the winter. LSA residuals have somewhat different seasonal behavior. LSAs NC (region B) and E (region C) are positive throughout the year, whereas the LSAs SW (region A) and NW (region D) have strong seasonal variations ranging from positive values during the winter and negative values during the summer. A seasonal variation is not inconsistent with ground heat storage and some of this residual is simulating that mechanism as well as indicating systematic tendencies by an analysis system.
GCIP atmospheric variations are shown in Fig. 8. The seasonal radiation variations weakly compensate each other with the maximum solar heating occurring during the summer when the cooling by the longwave radiation is strongest. Sensible heating shows a much stronger seasonal variation reaching amounts comparable to the solar heating during the summer and noticeably negative amounts during the winter when heat is transferred from the atmosphere to the surface. Part of this atmospheric heating comes from the large-scale thermodynamic energy convergence that strongly heats the GCIP atmosphere during the winter and weakly cools it during the summer. Precipitation shows a weak seasonal variation, with the maximum amounts reached during the spring. Residuals are small during the winter but become larger during the summer indicating potential errors in the convergence field. Again, however, the residuals reflect all the errors and could be caused by erroneous precipitation or boundary layer sensible heating variations as well as large-scale thermodynamic energy convergence.
As shown in Fig. 9, spatial variations over the domain are fairly weak in the radiation field. They are much stronger in the sensible heating, which was discussed previously for the surface field and for the precipitation. Clearly the LSA E (C) region is the wettest region and the LSA NW (D) is the driest region. These regions also appear to have displaced seasonal maxima with the LSA E (B) region having a strong summer maximum and the other LSA regions having a spring maximum. The precipitation field is fairly variable as is thermodynamic energy convergence. The maximum energy convergence is reached in the LSA NW (D) regions during the fall and winter, and the maximum divergence is reached in the LSA NC (B) region during the summer. The residual is also significant. The largest residuals occur during the summer in the LSA E (C) as well as the LSA NW (D), which may indicate that some of the residual is caused by precipitation and evaporation deficiencies as well as our derived thermodynamic energy convergence.
5. Anomalies
During the summer of 1993, there was a substantial increase in precipitation centered near the four-state area of Nebraska, Iowa, Missouri, and Kansas. This anomalous precipitation resulted in record flooding (Kunkel et al. 1994) all along the banks of the Mississippi River, as well as record discharges into the Gulf of Mexico. Figure 10 shows some of the observed features for the precipitation as well as the associated temperature depicted in a cleaned and gridded cooperative observer station dataset (COOP) provided by NCDC. Associated with this record-breaking precipitation over the central United States were anomalously dry regions in the eastern and southern United States temperature differences are also striking for these two years. Note that in the dry regions the maximum temperature is anomalously high and in the wet regions the minimum temperature is anomalously high.
As shown in Fig. 10, characteristics of the observed surface temperature variation are also present in the analysis, with some minor quantitative differences between the observed and analysis values. Given the similarity of the temperature patterns, we can thus use the analysis to examine the associated surface thermodynamic energy processes that may affect the surface skin and near-surface atmospheric temperature. As shown in Fig. 11, the latent cooling of the surface is considerably less during July 1993 than during July 1992 in the LSA E (C). By contrast, the sensible cooling of the surface increases in the LSA E during the summer of 1993 and acts to counterbalance the decreased evaporation cooling there. Radiation variations are much weaker but perhaps consistent in that there is a decrease in solar radiation over most of the GCIP domain as well as an increase in net longwave cooling in the LSA SW (A) region and a decrease in the cloudy LSA NW (D) and NC (B) regions. The surface residual also shows only a minor contribution, fortunately. This dominant surface evaporation and surface temperature relationship is consistent with previous studies (see e.g., Huang et al. 1996; Georgakakos et al. 1995) that showed that land surface temperature could be predicted by previous precipitation, evaporation, or soil moisture anomalies.
As shown in Fig 10, observed precipitation variations are also present, qualitatively, in the analysis, although there are same distinct differences between the analyzed and observed precipitation variations. In particular, the maximum precipitation differences are too far to the north and even outside of the GCIP area. Still, the large-scale moist variations in the U.S. northwest and the dry variations in the eastern third of the United States are present. As shown in Fig. 12, these dry regions are associated with strong sensible heat convergence into the area by the large-scale and turbulent-scale transport of thermodynamic energy. That is, there is a usually reasonably good inverse correspondence in most places between the large-scale divergence of heat and precipitation and large-scale convergence of heat and lack of precipitation. Unfortunately, the atmospheric residual term is especially large. Although it is tempting to ascribe the residual to defects in the large-scale transport terms, it should also be noted that the GDAS divergence term appears to be doing a better job than GDAS precipitation in describing variations in the observed precipitation field shown in Fig. 10. This anomalous convergence is not balanced by the precipitation and instead is balanced by a residual term. Although this residual term could certainly be related to errors in how we calculate the thermodynamic energy convergence, it also indicates that there may be potential errors in other parameterizations. For example, precipitation spinup and spindown may be especially affected by intense precipitation events. Another possibility discussed by Paegle et al. (1996) and Beljaars et al. (1996) is that the GDAS surface evaporation is deficient.
6. Conclusions
The purpose of this paper is to describe the GCIP thermodynamic energy budgets in NCEP’s GDAS. Basically, solar radiation warms the atmosphere and surface and planetary radiation cools the atmosphere and surface. The surface is further cooled by sensible and latent heat transfer to the atmosphere. The surface transfers are especially strong during the summer. The GCIP atmosphere is also warmed during the winter by the transfer of sensible heat from the surrounding land and ocean; this transfer is negligible and may even be negative during the summer. The GCIP atmosphere is also warmed by the precipitation.
There are significant east–west and north–south gradients within the Mississippi basin. The maximum solar radiation at the surface occurs in the LSAs SW and NW during the summer and the minimum amounts occur in the LSA NC during the winter. The maximum longwave surface cooling occurs in the LSA NW during the summer. The maximum sensible heat transfer to the atmosphere occurs during the summer in the LSA SW and the maximum latent heat transfer occurs in the LSA E during the summer. The latent heat released by precipitation is also a maximum in the LSA E, although the seasonal variations are not as strong as the evaporation variations since increased moisture transports into the GCIP basin occur during the winter. Transport of thermodynamic energy into the Mississippi basin is also influential, especially during the winter. During the summer, in all regions except the LSA NW, the thermodynamic energy transports are out of the basin and thus aid the longwave radiation in moderating the summer warming of the GCIP atmosphere from increased solar, sensible, and latent heating.
Perhaps a bit surprising was the strong influence of the surface sensible heating on the thermodynamic energy budget over the GCIP region, especially in the western regions. In the past, latent heating has been perceived to be globally dominant. Clearly, sensible heating is important over arid land regions like the U.S. southwest in general and the LSA SW in particular. Another surprising feature was the importance that horizontal thermodynamic energy convergence plays in balancing the local thermodynamic budgets. This term has usually been ignored in the past, presumably due to the difficulty in calculating it, but now that analysis products are coming on-line, we can begin to better understand its contribution, which is significant.
Surface differences between the summers of 1993 and 1992 appeared to be qualitatively portrayed by the GDAS analysis. Presumably the positive differences (1993 − 1992) between maximum temperatures in the eastern portion (LSA E) of the GCIP region are related to the drying of the soil and the decrease in surface latent cooling in this region. Sensible cooling increased in this region during this period indicating an obvious association but not a cause of higher surface temperatures.
In the atmosphere, the observed increased precipitation over the four-state area of Nebraska, Iowa, Missouri, and Kansas was qualitatively related to increased analysis precipitation, although the maximum analysis precipitation was too far to the north. Interestingly, the thermodynamic energy convergence was strongly negative over this region but was balanced instead by the residual term rather than the precipitation indicating potential deficiencies in the precipitation parameterization. Thermodynamic energy divergence and precipitation appeared to be related more strongly over the LSA E.
The residuals turned out to be more important than we had first realized. Over the ocean, they appear to realistically represent ocean advection and upwelling; similar to the atmosphere, convergence and divergence of heat by oceanic currents are probably just as important as latent, sensible, and radiative heating processes. Over land, the residuals may represent some seasonal heat storage, however, they may also represent some potential analysis model drift (Kanamitsu and Saha 1996). Atmospheric residuals are also probably mainly due to model drift as well as inaccurate computations of nonlinear terms from instantaneous samples four times a day. Climatologically, over land, and in the atmosphere, the residuals are slightly smaller than the other terms and we can ignore them when talking about the climatological balances, which is the major aspect of this paper. Unfortunately, when we consider monthly anomalies, atmospheric residuals become one of the dominant atmospheric processes. This makes it difficult to attribute cause and effect to other atmospheric processes. For example, we noted that the increased divergence over the Mississippi during July 1993 was balanced by the residual rather than by the precipitation.
We are eager to learn whether high-resolution GCIP analyses really do diminish the residuals. If the residuals are larger, then all we can really hope to get at the moment is a large-scale climatological understanding of the atmospheric climatological thermodynamic energy budget. On the other hand, if the residuals in the GCIP regional analysis products become smaller and moreover the differences between the primary variables and observed variables also become smaller, then we will know that GCIP has accomplished what it set out to do, which was to develop comprehensive and useful high-resolution thermodynamic energy and water budgets for a continental region.
Understanding characteristics of the analysis residuals is one critical element of analysis diagnostics. Properly calculated from complete accumulated budgets, residuals can provide an indication of model deficiencies and biases (Klinker and Sardeshmukh 1992; Schubert and Chang 1996). Unfortunately, we feel that some of the residual shown here may arise from an improper representation of thermodynamic energy convergence calculated from instantaneous samples four times daily. We are therefore developing short-range global and regional forecasts (Juang and Kanamitsu 1994) from the present operational analysis and reanalysis as well as long-term GCM simulations. The forecasts and GCM simulations are accumulating all the fluxes and they will therefore have a decreased residual. The GCM simulations will provide an especially consistent assessment of low-frequency surface atmosphere feedbacks; unfortunately we will gain this increased understanding at the expense of dealing with a less realistic climate. Nonetheless, we feel that all of these kinds of efforts, in addition to just simply comparing GCIP regional analysis products with GCIP observations, are required in order to truly understand GCIP’s and other regional hydrologic experiments’ water and energy budgets.
Acknowledgments
This study was supported by NOAA Grants NA37GP0372 and NA36GP0377. E. Kalnay and P. Caplan provided us the detailed history of GDAS changes. S. Iacobellis gave us a routine to calculate the hourly solar zenith angle. Dr. Chi-Fan Shih of NCAR and E. Bainto of Scripps helped decode the NCEP GRIB0 and GRIB1 GDAS products archived at NCAR. COOP station observations were obtained from NCDC. Reviewers’ comments improved the presentation of the work.
REFERENCES
Beljaars, A. C. M., P. Viterbo, and M. J. Miller, 1996: The anomalous rainfall over the United States during July 1993: Sensitivity to land surface parameterization and soil moisture anomalies. Mon. Wea. Rev.,124, 362–383.
Berbery, E. H., E. M. Rasmusson, and K. E. Mitchell, 1996: Studies of North American continental-scale hydrology using eta model forecast products. J. Geophys. Res.,101, 7305–7319.
Caplan, P. M., and G. H. White, 1989: Performance of the National Meteorological Center’s medium-range model. Wea. Forecasting,4, 391–400.
Chen, S. C., C. L. Norris, and J. O. Roads, 1996: Balancing the atmospheric hydrologic budget. J. Geophys. Res.,101, 7341–7358.
Fleagle, R. G., and J. A. Businger, 1963: An Introduction to Atmospheric Physics. Academic Press, 346 pp.
Georgakakos, K. P., D.-H. Bae, and D. R. Cayan, 1995: Hydroclimatology of continental watersheds, temporal analyses. Water Resour. Res.,31 (3), 655–675.
Huang, J., J. M. van den Dool, and K. Georgakakos, 1996: Analysis of model-calculated soil moisture over the United States (1931–93) and applications to long-range temperature forecasts. J. Climate,9, 1350–1362.
Iacobellis, S., and R. Somerville, 1991: Diagnostic modeling of the Indian monsoon onset. Part I: Model description and validation. J. Atmos. Sci.,48, 1949–1959.
Juang, H., and M. Kanamitsu, 1994: The NMC nested regional spectral model. Mon. Wea. Rev.,122, 3–26.
Kalnay, E., and R. Jenne, 1991: Summary of the NMC/NCAR Reanalysis Workshop of April 1991. Bull. Amer. Meteor. Soc.,72, 1897–1904.
——, M. Kanamitsu, and W. E. Baker, 1990: Global numerical weather prediction at the National Meteorological Center. Bull. Amer. Meteor. Soc.,71, 1410–1428.
——, and Coauthors, 1996: The NMC/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc.,77, 437–471.
Kanamitsu, M., and S. Saha, 1995: Spectral budget analysis of the short-range forecast error of the NMC MRF model. Mon. Wea. Rev.,123, 1834–1850.
——, and ——, 1996: Systematic tendency error in budget calculations. Mon. Wea. Rev.,124, 1145–1160.
——, J. C. Alpert, K. A. Campana, P. M. Caplan, D. G. Deaven, M. Iredell, B. Katz, H. L. Pan, J. Sela, and G. H. White, 1991: Recent changes implemented into the global forecast system at NMC. Wea. Forecasting,6, 425–435.
Kistler, R., E. Kalnay, and M. Kanamitsu, 1993: The NMC/NCAR CDAS reanalysis project. Proc. 18th Annual Climate Diagnostics Workshop, Boulder, CO, NOAA, 326–329.
Klinker, E., and P. D. Sardeshmukh, 1992: The diagnosis of mechanical dissipation in the atmosphere from large-scale balance requirements. J. Atmos. Sci.,49, 608–627.
Kunkel, K. E., S. A. Changnon, and J. R. Angel, 1994: Climate aspects of the 1993 upper Mississippi River basin flood. Bull. Amer. Meteor. Soc.,75, 811–822.
Mo, K. C., and R. W. Higgins, 1996: Large-scale atmospheric moisture transport as evaluated in the NMC/NCAR and NASA/DAO reanalysis. J. Climate,9, 1531–1545.
Morel, P., 1994: Scientific issues underlying the global energy and water cycle. Extended Abstracts, European Conf. on the Global Energy and Water Cycle, London, United Kingdom, The Royal Meteorological Society, 23 pp.
Newell, R. E., J W. Kidson, D. G. Vincent, G. J. Boer, J. R. Holton, J. M. Wallace, T. G. Dopplick, and A. C. Kyle, 1974: The General Circulation of the Tropical Atmosphere and Interactions with Extratropical Latitudes. Vol. 2. The MIT Press, 371 pp.
NOAA NMC Development Division, 1988: Documentation of the NMC Global Model. 244 pp. [Available from NOAA/NCEP Environmental Modeling Center, 5200 Auth Rd., WWB, Washington, DC 20233.].
Paegle, J., K. C. Mo, and J. Nogues-Paegle, 1996: Dependence of simulated precipitation on surface evaporation during the 1993 United States summer floods. Mon. Wea. Rev.,124, 345–361.
Ramanathan, V., B. R. Bartstrom, and E. F. Harrison, 1989: Climate and Earth’s radiation budget. Phys. Today,42, 22–32.
Rasmusson, E. M., 1967: Atmospheric water vapor transport and the water balance of North America. Part I. Characteristics of the water vapor flux field. Mon. Wea. Rev.,95, 403–426.
——, 1968: Atmospheric water vapor transport and the water balance of North America. II. Large-scale water balance investigations. Mon. Wea. Rev.,96, 720–734.
Roads, J. O., S.-C. Chen, J. Kao, D. Langley, and G. Glatzmaier, 1992: Global aspects of the Los Alamos general circulation model hydrologic cycle. J. Geophys. Res.,97, 10 051–10 068.
——, ——, A. Guetter, and K. Georgakakos, 1994: Large-scale aspects of the United States hydrologic cycle. Bull. Amer. Meteor. Soc.,75, 1589–1610.
——, ——, and K. Ueyoshi, 1995: Comparison of NMC’s global pressure analysis to NCDC’s U.S. observations. J. Climate,8, 1410–1428.
Schubert, S., and Y. Chang, 1996: An objective method for inferring sources of model error. Mon. Wea. Rev.,124, 325–340.
——, R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science applications. Bull. Amer. Meteor. Soc.,74, 2331–2342.
Sela, J. G., 1980: Spectral modeling at the National Meteorological Center. Mon. Wea. Rev.,108, 1279–1292.
——, 1982: The NMC Spectral Model. NOAA Tech. Rep. NWS 30, 36 pp. [NTIS PB81 247256.].
Trenberth, K., 1991: Climate diagnostics from global analyses: Conservation of mass in ECMWF analyses. J. Climate,4, 707–722.
——, and J. G. Olson, 1988: An evaluation and inter-comparison of global analyses from the National Meteorological Center and the European Centre for Medium Range Weather Forecasts. Bull. Amer. Meteor. Soc.,69, 1047–1057.
——, and C. J. Guillemot, 1995: Evaluation of the global atmospheric moisture budgets as seen from analysis. J. Climate,8, 2255–2272.
Wang, M., and J. Paegle, 1996: Impact of analysis uncertainty upon regional atmospheric moisture flux. J. Geophys. Res.,101, 7291–7303.
White, G. H., 1994: Diagnostics of atmospheric forcing from reanalysis. Proc. 19th Climate Diagnostics Workshop, College Park, MD, U.S. Dept. of Commerce, 246–249.
APPENDIX A
Analysis Characteristics
NCEP’s GDAS has undergone and continues to undergo change. References to some of the early analysis characteristics and changes are provided in Trenberth and Olson (1986), Caplan and White (1989), Kanamitsu et al. (1991), and Roads et al. (1995), which basically cover the period through the end of 1993. Below we describe the basic features pertinent to the GDAS flux and sigma products, which began to be archived at the National Center for Atmospheric Research starting in 1990 but were only really examined here starting in 1991 once the T126L18 model had been implemented (pentad 21). Of special note are the changes made on 10 January 1995, since NCEP’s frozen reanalysis started with this model on 11 January 1995 (see Kalnay et al. 1996). It should be mentioned that the NCEP GSM was originally designed by Sela (1982) but has since been enhanced by a large number of NCEP scientists and visitors (including Campana, Yu, Pan, Iredell, Katz, Derber, Ballish, Wu, and Hong; Caplan is the best initial contact for questions about specific parameterizations).
1991: T126L18 model implemented, mean orography replaced silhouette orography, marine stratus parameterization, mass conservation invoked, new horizontal diffusion parameterization, new sea surface temperature analysis, new spectral statistical interpolation (SSI) analysis, incorporation of land surface pressure observations, expansion of quality control to more observations.
January 1992: Radiation every 3 h (vs 12).
1993: SSM/I wind speeds over water Arakawa–Schubert convection and 28 sigma layers Interactive retrievals in NH (31 Mar 1993), (11 Aug 1993), (11 Aug 1993), (Dec 1993).
1994: No significant changes.
10 January 1995: Hydrology, soil physics, sfc layer; clouds, radiation; reanalysis; SSM/I precipitable water over sea; Track checking for aircraft obs.
25 October 1996: Direct use of satellite radiances; near-surface winds from ERS-1; PBL, diffusion based on bulk gradients; convection, adjustment of limiting CAPE.
APPENDIX B
Approximations to the Thermodynamic Equations
For historical reasons, the virtual temperature is the primary prognostic variable in the GDAS. Although virtual temperature simplifies the hydrostatic law and hence geopotential calculations, there are some associated complications; the virtual temperature equation must then include forcing from moist physical processes such as precipitation and evaporation that affect the moisture dynamics implicit in the virtual temperature equation (see Kanamitsu and Saha 1995). Presumably similar forcings could be included in the atmospheric mass equation discussed below. Anyway, we avoid these effects here in our diagnostic calculations by first (after calculating geopotential heights) transforming the virtual temperature to temperature before calculating associated fluxes and tendencies. It should also be noted here that the heat capacity is similarly affected by moisture; that is, Cp = (1 + .9q)Cpd, where Cpd is the dry air heat capacity, but variations in the heat capacity due to water vapor are not considered in the GDAS or here. Similarly, the surface volume heat capacity should be affected by the amount of surface water.
Grid points in the GDAS T126 Gaussian grid pertinent to the Mississippi River basin outlined by the thick solid line. The grid points over the four LSAs are labeled as (region A) LSA SW, (region B) LSA NC, (region C) LSA E, and (region D) LSA NW. Also shown is the U.S. orography, on a higher-resolution 0.25° grid, contoured at 600-m increments, starting at 200 m.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
JJA mean surface energy values for the U.S. region. (a)–(e) Positive values indicate the surface is being warmed and negative values indicate the surface is being cooled. Counter increments are 0.5 K day−1. (f) Units are kelvin and contour increments are 10 K: (a) Shortwave or solar radiative heating, (b) longwave or planetary radiative cooling, (c) sensible cooling (and heating), (d) latent evaporative cooling, (e) residual heating (and cooling) needed for balance, and (f) surface temperature.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
JJA mean atmospheric energy values for the U.S. region. Units are kelvin per day and contour increments are 0.5 K day−1. Positive values indicate the atmosphere is being warmed and negative values indicate it is being cooled: (a) Shortwave or solar radiative heating, (b) longwave or planetary radiative cooling, (c) sensible heating (and cooling), (d) latent heat of precipitation, (e) residual heating (and cooling) needed for balance, and (f) thermodynamic energy convergence (and divergence).
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 3 except shown here are the DJF mean atmospheric values for the U.S. region.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 2 except shown here are the DJF mean surface values for the U.S. region.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 2 except shown here are seasonal variations in the surface thermodynamic budget, averaged over the GCIP region. The abscissa indicates the month. The ordinate provides the amplitude of the heating or cooling in K day−1 [except for (f) which is in units of K].
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 6 except shown here are variations for each of the LSAs: (a) southwest, (b) northcentral, (c) east, and (d) northwest.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 3 except shown here are seasonal variations in the atmospheric thermodynamic budget, averaged over the GCIP region. The abscissa indicates the month. The ordinate provides the amplitude of the heating or cooling (in K day−1).
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 8 except shown here are variations for each of the LSAs: (a) southwest, (b) northcentral, (c) east, and (d) northwest.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
July 1993 minus July 1992 differences in (a) GDAS precipitation (mm day−1); (b) COOP precipitation (mm day−1), (c) GDAS skin temperature (K), (d) COOP daily average temperature (K) (min T + max T)/2.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 2 except shown here is the July 1993 minus July 1992 differences for the surface thermodynamic energy terms. Positive values for the thermodynamic energy rates indicate that these rates will increase the surface temperature for July 1993 and reduce it for July 1992.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
Same as Fig. 3 except shown here are the July 1993 minus July 1992 differences for the atmospheric thermodynamic energy terms. Positive values for the thermodynamic energy rates indicate that these rates will increase the atmospheric temperature for July 1993 and reduce it for July 1992.
Citation: Journal of the Atmospheric Sciences 54, 13; 10.1175/1520-0469(1997)054<1776:GSGEB>2.0.CO;2
GCIP-averaged atmospheric and surface thermodynamic energy terms for the annual DJF and JJA means. The first part of the table shows the terms in the surface thermodynamic energy balance in units of K day−1, followed by the surface temperature and atmospheric temperature, and then followed by the terms in the atmospheric thermodynamic energy balance in units of K day−1.
