GISS climate modeling focuses on decadal time scales. But testing of our model on seasonal predictions has potential research and educational benefits. Just as simulations of paleoclimate are invaluable for checking our model and our understanding, seasonal forecasting provides an equally useful, different, test which may help us develop an improved model.

The GISS Institute on Climate and Planets involves New York City students and educators in GISS research projects. The Pinatubo project is studying the causes of climate variations in the period 1979­present, including the roles of unforced or chaotic variability and climate forcings.

The GISS SI95 Climate Model

The Pinatubo group "froze" a version (SI95) of the GISS atmospheric GCM in 1995 and made a series of model runs for the period 1979­1995 in which four measured radiative forcings (stratospheric aerosols, ozone, other greenhouse gases, and solar irradiance) were added one­by­one. The runs using all of these forcings were repeated with four different ways of treating the ocean: (A) observed SSTs, (B) Q­flux ocean, (C) "GISS" dynamical ocean (Russell et al. 1995), and (D) "GFDL" dynamical ocean (Bryan and Cox 1972). For each case an ensemble of five runs was carried out. The model and experiments are described in a paper being written during the 1996 GISS Summer Institute (Hansen et al. 1996b), which will be submitted to the Journal of Geophysical Research.

The SI95 version of the GISS atmospheric GCM is improved considerably over GISS model II (Hansen et al. 1983). In addition to higher spatial resolution (4 x 5 degrees) and higher accuracy numerical methods [quadratic upstream scheme, equivalent to that of Prather (1986), for heat and moisture and fourth order differencing for the momentum and mass equations (Abram-opoulos 1991)], the greatest changes are in the PBL (Hartke and Rind 1996), clouds (Del Genio et al. 1996), moist convection (Del Genio and Yao 1993) and ground hydrology (Rosenzweig and Abramopoulos 1996). The model including all the major changes has been documented in several papers and AMIP intercomparisons, as delineated by Hansen et al. (1996b). Perhaps the most relevant intercomparison for U. S. forecasts is that of Boyle (1996), which shows that the current GISS model is among the more realistic models.

Predictability and Expected Forecast Skill

The simulations for 1979­1995 with observed time­varying SSTs, with and without the four observed radiative forcings, were compared with observations and with simulations using climatological SSTs, as discussed in our paper in preparation (Hansen et al. 1996b). The results suggest that the interannual variability of seasonal mean temperature and precipitation on this 17 year time scale is principally unforced (chaotic) variability. However there is a significant dependence on the boundary forcings, and thus potential predictability, which is much higher at low latitudes than high latitudes, higher for temperature than precipitation, higher in the summer than in the winter, and higher in some years than in others. For example, SSTs seemed to have had a relatively large influence on climate in the United States during the summer of 1988, when the SSTs were characterized by unusual, large, anomalies, while the SSTs of the early 1990s seemed to provide less predictability. In addition, sensitivity tests with arbitrary changes to springtime soil moisture indicated substantial impact on simulated climate of the following summer.

A measure of the model's skill is provided by global maps of the correlation of the simulation ensemble­mean with observations for the period 1979­1995. Such a map is shown by Hansen et al. (1996b) for tropospheric temperature, based on MSU (Microwave Sounding Unit) channel 2 observations. The correlation for the Jun-Jul-Aug seasonal mean exceeds 0.7 in much of the tropics and varies from about 0.3 at the U.S.­Canada border to 0.6 at the southern border of the U.S. Correlations with surface air temperature are lower (Fig. 1), about 0.2 over the U.S. and 0.6 in the tropics. The very low value over the U.S. is partly a fluke of this particular 5­run ensemble, as 5­run ensembles with other radiative forcings yield correlations of 0.3­0.4 there. Correlation skill for precipitation (not shown) is lower on an overall global basis than that for temperature.

This measure of predictability is optimistic, because it is based on use of simultaneous observed monthly SSTs, while the forecast must predict or extrapolate SSTs for three months. On the other hand, it may be possible to exceed this level of predictability by making use of added information such as initial conditions of soil moisture. Soil moisture should be a particularly relevant variable in summer, and its random variability in the 17­year runs may be a principal reason for the low predictability of surface temperature.

Seasonal forecast experiments;

Forecasts for Jun-Jul-Aug 1996

No tuning of the model was made for the seasonal forecasts, as the original calibrations were carried out with respect to the decadal time scale. We simply extended for three months the model simulations that used observed SSTs. The results that we show are anomalies calculated relative to the model's 12­year base period 1979­1990. We show results for two 5­member ensembles. Ensemble A specifies SST as climatology plus the anomalies observed in the last week of May 1996, with no attempt to predict changes of SST anomalies during Jun-Jul-Aug. Ensemble B adds an estimate of soil moisture anomalies on June 1, as described below. Both ensembles include radiative forcings as extrapolations of recent observations, but these have very little impact on seasonal climate variability (Hansen 1996a,b).

Ensemble A yields a precipitation deficit and high temperatures in the U.S. Southwest in June (Fig. 2). Qualitatively similar results were obtained in May and June in an ensemble of runs initiated May 1 (not illustrated). We have not determined the formal significance of the calculated climate anomalies for the summer of 1996. The June temperature anomaly in the Southwest­southcentral U.S. may be significant, as it has the same sign in four of the five runs and was comparably strong in the ensemble initiated May 1. But in July and August of these ensembles there is a similar frequency of above normal and below normal precipitation and temperature, at face value suggesting that there is a good chance of the drought breaking during the summer.

Among the first qualifications of such a prediction is the question of whether the model includes a realistic soil moisture anomaly at the beginning of the summer, given the sensitivity of seasonal simulations to initial soil moisture inferred by other investigators (e.g., Betts et al. 1996) and also found in sensitivity studies with our model. Thus we carried out a second ensemble of experiments (Ensemble B, Fig. 3) in which we specified a soil moisture anomaly for June 1 based on an approximate Palmer Drought Index (PDI). PDI was calculated from monthly means of precipitation and temperature in preceding months, based on a program provided by Tom Karl and Richard Knight of NCDC. The scale factor required to convert a PDI anomaly to a soil moisture anomaly was set by requiring that the mean magnitude of the soil moisture anomalies equal the mean magnitude of soil moisture anomalies in the model's climatology. In regions without monthly mean data (principally the Amazon basin and the Sahara) the soil moisture anomaly on June 1 is that which was generated by the 17 year model run.

Ensemble B yields a much different forecast in the U.S., with a stronger heat and drought anomaly persisting through June, July and August. This testifies again to the sensitivity of regional climate to soil moisture anomalies in our model. In the 5­run means there are obvious geographical correlations of simulated climate anomalies with soil moisture initial conditions. This positive feedback is at variance with conclusions of Giorgi et al. (1996) based on a mesoscale model, but consistent with the analysis of Betts et al. (1996) and with previous sensitivity studies with our model (Hansen et al. 1996b) and other models.

Interpretation and Discussion

We emphasize that our model has not been designed for regional forecasting. These "forecasts" are not intended for operational purposes, but rather have objectives related to research and education. Our aim is to help understand the significance of different mechanisms in regional climate fluctuations and to improve the realism of our climate model.

The simulations suggest that SST anomalies, possibly the proximate large warm region off the west coast of the U.S., may have been an immediate cause of the current drought in the southwest and south-central U.S. It will be of interest to examine the meteorology associated with the anomalies, quantify their significance, and perhaps do experiments to geographically isolate the effective forcing.

The simulations with only SST forcing would suggest that there is a good chance of the drought in the southwest U.S. breaking in July­August. But inclusion of an initial soil moisture anomaly inferred from a drought index tilts the prediction toward continuation of the drought. Indeed, these simulations suggest that the drought area will expand during the course of the summer.

Although the sense of the soil moisture impact on the forecast seems reasonable, its magnitude suggests that both the procedure for obtaining the soil moisture anomaly and the ground hydrology scheme in the GISS GCM should be carefully scrutinized. The 6­layer hydrology scheme (Rosenzweig and Abramopoulos 1996) includes shallow, intermediate, and deep layers which encompass a range of time constants and a large moisture field capacity.

Perhaps the main issue raised by these forecast experiments concerns the role and time constant of soil moisture anomalies. Either our estimation of the initial anomalies is excessive, or our model unrealistically magnifies the impact of soil moisture, or soil moisture has a greater impact on climate forecasts than is generally believed. If the soil moisture impact is indeed large, is it also possible that the deep soil layers have an effect on interannual time scales?

Both forecast experiments produce some strong anomalies at low latitudes, including below normal temperatures and heavy precipitation in Northeast Brazil and the Sahel. The model has high skill in Northeast Brazil in the period 1979­95 (correlation 0.6­0.7 for temperature, 0.5­0.6 for precipitation), but low skill in the Sahel (about 0.2 for temperature, 0.0 for precipita-tion). The excess precipitation in Northeast Brazil and the low temperatures in both places are consistent with expectations for a La Nina period (Ropelewski and Halpert 1987; Halpert and Ropelewski 1992), although climate anomalies in these regions are also dependent on SST anomalies in the Atlantic (Ward and Folland 1991, Hastenrath and Greischar 1993, Lamb and Peppler 1991).

The present simulations are intended only to study the impact of SSTs and soil moisture on predictability. We do not attempt to initialize atmospheric conditions in our model: the June 1 atmospheric conditions were not used; the model picked up the atmospheric conditions already in progress from the 1979 start time. We also do not use a very high resolution, as would be required for optimal short­term forecasting. Even the two boundary conditions that we examine are handled crudely. One potential improvement would be to replace the fixed SST anomalies with predicted SSTs. This is probably essential for forecasts of more than a few months, and may be an important factor even in single season forecasts. Our simulations also point out the need to determine soil moisture initial conditions more accurately. Although observations of global soil moisture do not appear to be practical at present, it should be possible to use the time history of precipitation and temperature to generate improved soil moisture initial conditions as a function of depth. Empirical use of such calculated soil moisture for long-range temperature prediction has been discussed in this Bulletin (June 1995) and in Huang et al. (1996).


Acknowledgments: The paper in preparation (Hansen et al. 1996b), describing the SI95 model and experiments, will include a large number of collaborators at other NASA centers and universities. Tom Karl and Dick Knight of NOAA NCDC kindly provided a program for calculating a drought index from monthly precipitation and temperature.

Abramopoulos, F., 1991: A new fourth­order enstrophy and energy conserving scheme, Mon. Wea. Rev., 119, 128­133.

Betts, A.K., J.H. Hall, A.C.M. Beljaars, M.J. Miller and P.A. Viterbo, 1996: The land­surface- atmosphere interaction, J. Geophys. Res., 101, 7209­7225.

Boyle, J.S., 1996: Seasonal characteristics of precipitation over the United States in AMIP simulations. PCMDI Report No. 31, Lawrence Livermore Laboratory, Livermore, California, 30 pp.

Bryan, K. and M.D. Cox, 1972: An approximate equation of state for numerical model of ocean circulation, J. Phys. Oceanogr., 15, 1255­1273.

Del Genio, A.D. and M.S. Yao, 1993: Efficient cumulus parameterization for long­term climate studies: GISS scheme, AMS Mono., 46, 181­184.

Del Genio, A.D., M.S. Yao, W. Kovari, and K.K. Lo, 1996: A prognostic cloud water parameterization for global climate models, J. Climate, 9, in press.

Giorgi, F., L. Mearns, C. Shields and L. Mayer, 1996: A regional model study of the importance of local versus remote controls of the 1988 drought and the 1993 flood over the central United States, J. Climate., 9, 1150­1162.

Halpert, M.S. and C.F. Ropelewski, 1992: Surface temperature patterns associated with the Southern Oscillation, J. Climate, 5, 577-593.

Hansen, J., G. Russell, D. Rind, P. Stone, A. Lacis, S. Lebedeff, R. Ruedy, and L. Travis, 1983: Efficient three­dimensional global models for climate studies: models I and II, Mon. Wea. Rev., 111, 609­662.

Hansen, J., M. Sato, R. Ruedy, A. Lacis, K. Asamoah, S. Borenstein, E. Brown, B. Cairns, G. Caliri, M. Campbell, B. Curran, S. de Castro, L. Druyan, M. Fox, C. Johnson, J. Lerner, M.P. McCormick, R. Miller, P. Minnis, A. Morrison, L. Pandolfo, I. Ramberran, F. Zaucker, M. Robinson, P. Russell, K. Shah, P. Stone, I. Tegen, L. Thomason, J. Wilder, and H. Wilson, 1996a: A Pinatubo climate modeling investigation, in NATO ASI Series Volume, Subseries I Global Environment Change, Eds. G. Fiocco, D. Fua and G. Viscount, Springer­Verlag.

Hansen, J. and APinatubo Team@, 1996b: Forcings and chaos in interannual to decadal climate change, J. Geophys Res, to be submitted.

Hartke, G.J. and D. Rind: An improved boundary layer model for the GISS GCM, 1996: J. Climate, 9, in press.

Hastenrath, S. and L. Greischar, 1993: Further work on the prediction of northeast Brazil rainfall anomalies. J. Climate, 6, 743-758.

Huang, J., H.M. van den Dool and K.P. Georgakakos, 1996: Analysis of model-calculated soil moisture over the United States (1931-1993) and applications to long-range temperature forecasts. J. Climate, 9, in press (to be in July or August issue).

Lamb, P. and R. Peppler, 1991: West Africa. Chapter 5. In Teleconnections: Linkages between ENSO, Worldwide Climate Anomalies, and Societal Impacts. M.H. Glantz. R.W. Katz, N. Nicholls, Eds., Cambridge University Press, 121-189.

Prather, M.J., 1986: Numerical advection by conservation of second order moments. J. Geophys. Res., 91, 6671­6680.

Ropelewski, C.F. and M.S. Halpert, 1987: Global and regional scale precipitation patterns associated with ENSO, Mon. Wea. Rev., 115, 1606­1626.

Rosenzweig, C. and F. Abramopoulos, 1996: Land surface model development for the GISS GCM, J. Climate, in press.

Russell, G.L., J. Miller and D. Rind, 1995: A coupled atmosphere­ocean model for transient climate change studies, Atmos. Ocean, 33, 687­730.

Ward, M.N., and C.K. Folland, 1991: Prediction of seasonal rainfall in the North Nordeste of Brazil using eigenvectors of sea surface temperature. Int. J. Climatol., 11, 711-743.

Figures

*Subset of "Pinatubo" group preparing summer forecast, prior to arrival of most high school participants:

J. Hansen, K. Beckford, S. Borenstein, E. Brown, B. Cairns, S. DeCastro, L. Druyan, M. Kelly, A. Luckett, R. Miller, R. Ruedy, and J. Wilder.