1ADAS Pwllpeiran, St Agnes Road, Cardiff CF4 4YH, UK; 2Central Science Laboratory, Sand Hutton, York Y04 1 LZ, UK

Background and objectives
The Smith Period [1] has been the mainstay of potato blight forecasting for growers in the UK since 1975. Together with local knowledge of crop development it has provided an invaluable aid to blight risk assessment. The relatively straightforward criteria are calculated from the data collected by the UK network of synoptic stations. These stations are frequently sited at airfields, not necessarily close to the major potato-growing areas. The conditions experienced in many potato fields may be quite different from those at the meteorological stations, leading to false alarms, or more seriously, missed conditions. Since the 1950s, more complex schemes have been developed which attempt to give more accurate forecasts provided their weather data requirements can be met. This paper describes the results of an evaluation of selected forecasting schemes under UK conditions using 'in-field' weather data.

Materials and methods
Five models were compared over four years - Smith [1], Sparks [2], Schrodter and Ullrich [3], Blitecast [4] and NEGFRY [5]. Data were collected from in-crop meteorological stations via cellphone analogue networks at six sites chosen to provide a range of disease pressures and crop development relative to infection date. The blight-susceptible cv. King Edward was used. The number of sprays applied under each system and the number of days between the actual date when blight was found in the crop and the time to initiate the spray programme as dictated by the various forecasting schemes were recorded.

Results and conclusions
In 1994, all schemes except Smith failed to predict the first occurrence of blight at Trawsgoed. At High Mowthorpe, Sparks and NEGFRY predicted the outbreak too close to be practical and Blitecast failed to predict blight by a day. At Arthur Rickwood, all schemes gave a warning before blight appeared but only the NEGFRY and Schrodter & Ullrich systems were close to an optimum period of 14 days before blight appeared. At Starcross, traditionally a high-risk blight location, all schemes except Blitecast warned well in advance of the actual outbreak. In 1995, all systems, except Smith and Sparks, failed at Starcross, but even these were closer than the 14-day optimum. At Trawsgoed, all schemes warned too far in advance, whilst at High Mowthorpe, Smith was too premature and Blitecast within the 14-day optimum. In 1996, the schemes showed a great degree of inconsistency and at High Mowthorpe only Smith and Sparks warned well before the outbreak. NEGFRY and Schrodter & UlIrich were too close to the outbreak day, and Blitecast failed altogether. At Trawsgoed, all schemes except Blitecast, which was very close to the ideal, warned too far in advance. At Starcross, the systems gave warnings ranging from 35 days (NEGFRY) to 53 days (Smith and Sparks) in advance of the blight outbreak. 1997 was a major blight year and all schemes predicted blight at all sites. Blight forecasting schemes must be 100% successful. Those tested in this study were at their least reliable in low-blight years when spray programmes were initiated too far in advance of infection, even where blight did not occur. These were the years when the potential to reduce unnecessary fungicide applications was at its greatest. In high-blight years the schemes were more reliable but less likely to recommend fewer fungicide sprays than routine programmes. Of the schemes evaluated, only the Smith Period did not fail to warn of forthcoming blight infection. It is concluded that more generalised but robust systems are likely to be of more practical value to potato growers, but further refinements are considered necessary in order to increase precision.

1. Smith LP, 1956. Plant Pathology 5, 83-87.
2. Sparks WR, 1984. Agricultural Memorandum 1020, British Meteorological Office, Bracknell.
3. Schrodter H, Ullrich J, 1967. Agricultural Meteorology 4,119-135.
4. Krause RA, Massie LB, Hyre RA, 1975. Plant Disease Reporter 59, 95-98.
5. Hanson JG, Andersson B, Hermansen A, 1995. Phytophthora infestans. Dublin, Ireland: EAPR, pp. 220-225.