Report on an experiment in five-day weather forecasting

Abstract

The following report is presented as a statement of progress made at the Massachusetts Institute of Technology (M.I.T.) in the investigation into the possibility of extending the range of reliable weather forecasts. This project has been supported at M.I.T. and other private institutions by Bankhead-Jones appropriations since September, 1937. This report is concerned only with the work completed or in progress at M.I.T. The complementary program now in progress at the Weather Bureau in Washington is referred to only in so far as it has contributed directly to these investigations. Furthermore, the following report refers only to the last two years of the M.I.T. project. The first year of the three-year project was given over principally to the study of the results obtained by long range forecast methods already in use, and to the establishment of a northern hemisphere synoptic weather map procedure as a necessary precedent to the preparation of weekly forecasts on a synoptic basis. The results of the M.I.T. study of certain long range forecast methods already in practice are included in a general survey of such methods already published. The synoptic charts prepared at M.I.T. during that first year of the investigation are listed in an appendix to this report, together with those of the last two years. The preparation of weekly forecasts carried on during a part of that first year was so experimental in nature, and the procedure was so much changed the following year, that the results obtained were considered neither suffciently significant nor comparable enough with the later forecast results to merit any discussion. The present report is divided into three principal sections. Section I presents in condensed form our present conception of the essential nature of the general circulation, and discusses briefly the background of one or two of Professor Rossby's theoretical considerations concerning the general circulation which have found statistical and synoptic application in this investigation. Section II contains in brief form the results of synoptic and statistical checks of a large number of hypothetical relationships which might be assumed to hold in the earth's atmosphere. These include possible relationships in the large scale features of the general circulation, relationships between the general circulation and its different branches or centers of action, between the different branches or centers of action of the general circulation, between characteristics of the general circulation or its branches and anomalies of the meteorological elements in certain regions, between anomalies of the meteorological elements in one region and those in another region, and even between solar activity (sunspots) and characteristics of the general circulation or anomalies of the meteorological elements. The aim was to investigate possible interrelationships of all kinds, either with or without lag, in order to detect as many interaction principles or points as possible in the earth's atmosphere, whether they had direct or only the most indirect bearing on the forecast problem. The relationships investigated applied to daily, weekly, monthly, seasonal, or annual mean conditions. They were selected for investigation either from theoretical or practical considerations of the nature of the general circulation as outlined in Section I, or on the basis of popular beliefs which have long been current among meteorologists, or on the basis of direct observation of data which looked promising. The majority of these hypothetical relationships are found to be quite weak when subjected to rigid statistical checks, but all such results, whether positive or negative, are summarized in this report. Section III outlines the five-day forecast routine practice which has been carried on at M.I.T. during the greater part of the past two years on a weekly basis. It includes a statistical analysis of the verification results. In the conclusion are summarized the results of the investigation which thus far appear significant enough to justify their consideration in five-day or longer range forecasts. Suggestions are offered as to further steps which might profitably be taken if the investigation is to be continued. Finally there is an appendix in which are listed all the daily synoptic maps and mean charts and diagrams of surface and upper air data which have been plotted and analyzed at M.I.T. in connection with this project during the past three years. The importance of such a list is apparent when it is realized that inevitably in an investigation of this kind much the greater part of the time and effort expended is consumed in the routine or semiroutine duties involved in the preparation of such charts.