Prediction does not always help in control but does enable farmers to reduce their losses in other ways. For instance, there is no practicable short-time control measure available for wheat leaf rust, but forecasts issued early in the season allow farmers to plow up their wheat and plant some other crop or to pasture their fields if a serious outbreak is indicated. Thus they can recover at least part of their season's investment.

Predictions can never be entirely right. This is just as true of plant disease forecasting as of weather or political polls or anything else. The average correctness for the limited number of plant diseases that are being forecast so far is more than 80 percent. The possible accuracy depends on how complicated the critical periods are and on how far in advance they operate. For instance, if winter temperature by itself determines the amount of the disease for the following season, a prediction is a simple matter of calculation and can be made with practically complete certainty that it will be right. In contrast, if the pathogen can cause disease at any time that its temperature or moisture demands or both are met, both disease and weather must be watched throughout the season, and frequently revised short-term forecasts are necessary.

Forecasting plant disease occurrence on the basis of known weather requirements is well established. In this country apple scab spray warning services have been routine for almost 30 years. Some other diseases for which forecasting is regularly a part of the control program in one region or another include bacterial wilt of sweet corn, wheat leaf rust, late blight,tobacco blue mold, cucurbit downy mildew (Pseudoperonospora cuhensis), and lima bean downy mildew (Phytophthora phaseoli). The weather relations that make it possible to predict for some of these diseases have already been indicated.

The 1946 tomato late blight outbreak proved beyond doubt that diseases that spread so rapidly, and attack so destructively, require more than local attention to cope with them adequately. As a result, the Crop Plant Disease Forecasting Project of the Plant Disease Survey, Bureau of Plant Industry, Soils, and Agricultural Engineering, was organized. The project is a concrete expression of the importance of the weather-disease relation. It was set up especially to keep watch over circumstances that might lead to outbreaks of plant diseases, to use all available information about conditions favoring diseases as an immediate basis for making forecasts, and to study the requirements for development and spread so that predictions can be improved in accuracy and duration.

The chart on the next page shows how the warning service of the project operates. A pathologist in each State works with it and performs the same task for his own State that the project does for regions. Because information is obtained from such a wide area, ample time is given for local warning and preparation if a disease outbreak seems likely. An indispensable feature of the project, noted in the chart, is the cooperation of fungicide and equipment manufacturers, which assures availability of chemicals and control equipment wherever and whenever they are needed. The role of the Weather Bureau, in a program that has weather as one of its components, is obvious.

Late blight, tobacco blue mold, and cucurbit downy mildew, up to now, have been the chief diseases considered by the warning service. Regional forecasting for potato late blight has been tried in the North Central States, with good results. The forecasts are based on temperature and humidity recorded by hygrothermographs placed in potato fields in various parts of the region, on the weather forecasts, and on observations on the prevalence of the disease in selected locations. This regional forecasting on the basis of instrumental recording, much the same as for weather, is a new trend and shows promise of further development.

We can make useful predictions without knowing all the reasons for the observed reaction of disease to weather. Neither do we need to know at the time of predicting exactly what the weather will be. We can say that on the basis of what we know now, if the weather is this, then the disease will be that, and even the conditional forecast will be helpful. Obviously, however, the more we know about the operation and timing of the factors favoring or inhibiting the development and spread of disease, and the longer ahead of time we can be sure Of the weather, the more accurate our predictions will be and the earlier we can make them. Improvement and extension of the long-range weather forecasts is the answer to the latter need. The research program of the forecasting project is designed to add to our knowledge on the connection between environment and disease. Detailed and continuous observation of weather and microclimate is as much a part of its investigation as are the purely disease factors. By means of weather-recording instruments, the weather can be correlated directly with the development of the disease in experimental fields. The regional late blight forecasts in the Central States are a result of this sort of study.

PAUL. R. MILLER, a pathologist in charge of the Plant Disease Survey, Bureau of Plant Industry, Soils, and Agricultural Engineering, has spent 16 years developing survey techniques and conducting field studies of diseases that occur on major economic crops, including peanuts, tobacco, cotton and truck crops. Dr. Miller directs a plant disease forecasting service in cooperation with the agricultural experiment station pathologists of the 48 States, the United States Weather Bureau, and the National Fungicide and Farm Equipment Association.