PART IV — DYNAMICS OF THE ATMOSPHERE-OCEAN SYSTEM 



ing such critical problems as whether 

 precipitation will be in the form of 

 rain or snow. These problems, in- 

 volving interactions of wind, tem- 

 perature, and moisture fields, are of a 

 different order of difficulty. Short- 

 range predictions also suffer some- 

 what from data deficiencies, particu- 

 larly in oceanic and adjoining regions. 

 Satellite observations have, however, 

 alleviated the data deficiencies to a 

 considerable degree in recent years. 



There are several avenues for ad- 

 vancing the science of short-range 

 prediction of synoptic-scale phenom- 

 ena; all of them are being actively 

 pursued and deserve encouragement. 

 First, fine-grid scale models are being 

 developed which accept data at grid 

 intervals of half or less the current 

 standard mesh length of about 400 

 kilometers. Use of a finer grid per- 

 mits better resolution and more accu- 

 rate depiction of the synoptic patterns 

 and improves the accuracy of the 

 computational procedures. Unfortu- 

 nately, the presently available obser- 

 vations are not ideally suited for 

 fine-grid computation. Though a net- 

 work of surface observations exists 

 which makes it possible to represent 

 surface weather features more pre- 

 cisely than is presently done in nu- 

 merical prediction, no corresponding 

 closely spaced upper air observations 

 are available. High-resolution, scan- 

 ning radiometric sounders aboard 

 satellites offer a promising means of 

 overcoming this gap, and every effort 

 should be made to speed their devel- 

 opment and application. Data from 

 more advanced satellites can also be 

 expected to improve further the qual- 

 ity of ocean analysis, and thereby 

 contribute to better short-range fore- 

 casts over ocean areas and adjacent 

 coastal regions. 



Another important avenue for ad- 

 vancing short-range prediction is 

 through continued efforts at improv- 

 ing the physical basis of the predic- 

 tion models. Such efforts can be 

 carried out in part bv theoretical 

 means, using presently available 

 knowledge of the physical processes. 



But they will also almost certainly 

 require the acquisition of special data 

 sets of the sort planned under GARP 

 and other large observational pro- 

 grams. Better modeling of the physi- 

 cal processes will not only widen the 

 scope of the phenomena that can be 

 forecast successfully by objective 

 means but will result in greater ac- 

 curacy of the forecast as a whole. 



A final important new direction in 

 short-range prediction is in modeling 

 of the near surface layer. This is the 

 layer that affects man most directly. 

 Accurate predictions of its structure 

 will contribute to successful predic- 

 tions of the dispersal of pollutants in 

 the atmosphere, and of fog and other 

 visibility- and ceiling-reducing fac- 

 tors that hamper aircraft operations. 

 Modeling of this layer is a difficult 

 undertaking, since its characteristics 

 and behavior are controlled in large 

 measure by turbulent processes. Both 

 theoretical work and field observa- 

 tional programs will be required to 

 advance this effort. We are still a 

 long way from being able to make 

 surface-layer prediction a part of the 

 routine prognosis. 



Medium-Range Prediction 



During the past dozen years, the 

 greatest gains in forecast skill have 

 probably occurred at medium range 

 (1-5 days). These gains are the direct 

 outcome of the development and ap- 

 plication of numerical prediction 

 models capable of forecasting the for- 

 mation and movement of synoptic- 

 scale weather systems. The method 

 differs in no way from that described 

 in connection with short-range pre- 

 diction; it is simply extended for a 

 longer period. 



Surface weather predictions are 

 now quite satisfactory for periods of 

 about 48 hours. Upper-level prog- 

 noses show some degree of skill for 

 periods as long as three to five days. 

 Again, pressure and wind patterns 

 are better forecast than such elements 

 as precipitation. At medium ranges 



it is still possible to infer likely areas 

 of convective activity — thunder- 

 storms and the like — but prediction 

 of individual small-scale disturbances 

 is completely beyond the realm of 

 possibility. 



Numerical experiments conducted 

 as part of GARP suggest that it is 

 possible, in principle, to forecast day- 

 to-day weather changes for periods 

 as long as two to three weeks in ad- 

 vance — though some critics feel this 

 is an excessive figure in terms of 

 what the public would judge to be 

 successful forecasting. In any event, 

 there is good reason to believe that 

 useful forecasts can be made by nu- 

 merical methods for periods well in 

 excess of the present three-to-five day 

 limit. 



The main obstacles in the way of 

 increasing the time range of forecasts 

 (and thereby also their accuracy, even 

 at shorter ranges) are the lack of an 

 adequate observational network on a 

 worldwide basis and deficiencies in 

 the physical formulation of the pre- 

 diction models. A principal aim of 

 GARP is to overcome these observa- 

 tional and physical shortcomings. 



A number of areas or industries 

 have been identified in which more 

 accurate predictions in the five- to 

 twenty-day range would result in 

 great economic benefit. Among these 

 are agriculture, transportation, public 

 utilities, and the construction and 

 fishing industries. 



Long-Range Prediction 



Long-range prediction is a contro- 

 versial subject. Its proponents make 

 a variety of claims, ranging from the 

 ability to forecast a given day's 

 weather weeks or months in advance 

 to the ability to forecast, with some 

 small degree of skill, departures of 

 temperature or precipitation from 

 their monthly or seasonal means. 

 Skeptics contend that the whole busi- 

 ness is a waste of time, either that we 

 do not know how to make long-range 

 predictions or that long-range pre- 



96 



