in the way of transmission of intelligence during the night.

Except in the large towns, the majority of telegraphic stations are closed till 8 a.m., so that while the evening forecasts do not reach them till past eight o'clock in the morning, the information for that morning arrives about three hours later. Thus we see the practical difficulties in the way of forecasting. There is no doubt that in time some of them will be successfully overcome.

WHEN MOST SUCCESSFUL.

A few remarks on the circumstances under which the most successful forecasts can be issued will also much help a general apprehension of the subject. We will confine our observations to Great Britain only. It is very obvious that the more striking the weather-changes, the more have we something definite to forecast. When we have a well-formed cyclone, which traverses a welldefined path, we have strongly marked sequences of wind and weather, and any error in the forecast will only arise from some slight difference between the expected and the actual track. But when we have what we have seen is the more usual state of things in Great Britain—illdefined depressions which move irregularly, and one or more of which fuse into a fresh cyclone with a new centre -then we have no definite sequence of weather to deal with, but a change which is produced by the weather at each station gradually conforming to the varying shapes of isobars. The best that can be done then is to forecast generally broken weather, and more or less rain

generally; but no attempt can be made to foretell any definite series of wind-shifts, as in a true cyclone.

Experience has shown that in Great Britain no serious gale has ever been experienced, unless there is more than half an inch of difference of barometric pressure between some two stations. Synoptic charts will always detect even much smaller differences; so that, though some uncertainty will always remain as to the direction of the wind, the force will generally be at least approximately forecast correctly, except in the case of a very sudden and unexpected fall of the barometer.

Very different, however, is the case of rain. Secondaries and non-isobaric rains are the forecaster's bugbear; they form so quickly, show so little on a synoptic chart, and move so irregularly, that rain in general terms is all that the forecaster can usually say. In summer, when he sees the characteristic loops in the isobars which constitute secondaries, he can safely predict thunder and rain; but he cannot attempt to localize either of these phenomena.

Sometimes, too, secondaries are so small that they do not show at all on a synoptic chart, which is constructed on reports received from stations often a hundred and fifty or two hundred miles apart. The whole loop of a secondary need not be nearly so large; and then a depression of that class might lie between two stations, and yet be indicated at neither. The weather, however, would be profoundly modified, and the forecasts would probably be

erroneous.

There is also always the important difference between wind and rain, that the former is always in the main

determined by the steepness of the gradients, while the amount of precipitation bears no relation to any known meteorological element.

In many shapes of isobars we know that there will be rainfall, but whether much or little, we cannot tell at present.

From these considerations we need not be surprised to find that in all offices, except in Japan, wind is better forecast than rain.

SOURCES OF FAILURE.

From the conditions of successful forecasts, we can readily turn to those of unsuccessful predictions. Besides the uncertainty of rainfall due to the action of secondaries, there are four principal sources of failure: the sudden formation of an intense cyclone; the sudden dying out of an existing cyclone; the motion of a cyclone in an unexpected path; and, lastly, an error in the judgment of the forecaster.

In the first case, of the sudden formation of a new cyclone, the whole forecast is necessarily totally upset, and the weather which is experienced is worse than had been anticipated.

The converse occurs when an intense cyclone suddenly dies out. Then the weather is much better than was expected; but neither in this case nor in the preceding one can settled weather be expected.

When a cyclone takes an unusual path, the general character of the weather will remain bad, but the direction of the wind and the details in different districts will

be wrongly forecast. We have already given instances of cyclones which move in no well-defined path, and more complicated cases often occur. Sometimes the path will describe a complete circle of no very great diameter; but the commonest case in Western Europe is when the path of a cyclone takes the form of the letter V. For instance, a cyclone comes in from the Atlantic from about due west, and after it has gone as far as England, it moves back again in a north-westerly direction, as it has not been able to pass the area of high pressure which would then be lying over Northern and Central Europe. In another common case, the cyclone comes down from the north-west on to England, and then passes off in a northeasterly direction towards Norway. In all such cases the forecaster is at a great disadvantage.

err.

Lastly, the judgment of the forecaster will sometimes We have shown that no absolute law of cyclonemotion can be laid down, and that, in fact, the tracking of well-defined depressions forms but a small portion of the forecaster's business. On the larger number of days he has to estimate how, or where, cyclones will form in an ill-defined area of low pressure, or how far an area of low pressure will encroach on another region of high barometer. In this, he must rely on his own opinion and experience alone; that must be fallible sometimes, but better results are obtained by trusting to personal skill than by attempting to use any mechanical rules or maxims.

Men differ in their aptitude to forecast weather in the same way as physicians differ as to the accuracy of their diagnosis; but just as the best results are obtained by

selecting the doctor whom experience has shown to be the most successful practitioner, so the best forecasts are got by selecting the meteorologist who has been the most successful in that branch of the subject. In the United States Signal Office at the present time, four men take the duty of forecasting in rotation. They have so far all been ground in the same mill, by passing through a twoyears' course of the same hard training; and it is found in practice that the difference between the best and worst is two per cent. in the number of successful forecasts. For instance, if the best man gets ninety per cent., the worst will attain to eighty-eight per cent. of success.

SOME COUNTRIES EASIER THAN OTHERS.

From all that we have now explained, it will be very evident that forecasting is much easier in some countries than others. In the tropics, the great seasonal changes come on regularly, and the smaller changes from day to day are insignificant. In the two or three days of any year on which a regular cyclone may form, the premonitory symptoms are so obvious that there is no difficulty in framing a forecast.

In temperate regions, those countries will be the best situated which lie to the east of a well-observed land area, because most disturbances in the temperate zone move from the west.

Thus Germany and Norway are much more favourably located for weather-prevision than either England or France.

In the year 1869 twenty-three storms were felt in