Un système de prévision de gelées de printemps à échelle fine dans le vignoble champenois

EnglishA System of local forecasting of minimum screen temperatures and actinothermal indices at 10 cm height in four sites of the Champagne vineyard, two located dovvn slope and two located at mid-slope, has been developed. This System combines the use of output data of the generai numerical weather forecasting (cloud cover, wind speed and direction), of a simplified simulation model of night air cooling on flat terrain in the case of a clear sky anticyclonic situation, and a procedure for local adaptation for the calculation of air temperatures in the vineyard. This third step is based on a topoclimatic study done previously, using 9-year temperature data series, which enabled us to calculate thermal differences between a site within the wine region and the reference synoptical station (in Reims), as a function of the wind direction. This forecasting System, activated at 15 h UTC for the next morning, has been tested during spring 1989 (50 nights altogether). The results of the tests are the following : for clear sky nights (18 nights out of 50) the mean absolute deviation between observed and forecasted values of minimum screen temperature in Reims is 1.7°C. The absolute deviation (calculated over these 18 nights) between observed values and forecasted values of the actinothermal index is less than 1.5°C in 5 0% of the cases with some differences between sites. The diagnosis of frost risk, defined as the occurrence of an actinothermal index (i.e. surface temperature) less than -2.5°C, has been successf ully forecasted for the most frost-exposed site. francaisCet article presente un systeme de prevision des temperatures minimales sous abri et des indices actinothermiques en quatre points du vignoble champenois : Avize, Bouzy, Chambrecy, Mailly. Les differentes etapes de realisation de la prevision sont decrites : utilisation des donnees de la prevision generale, utilisation eventuelle d'un logiciel de simulation de refroidissement nocturne par temps anticyclonique, et adaptation locale. Cette derniere necessite une etude topoclimatique prealable des series de mesures thermometriques observees aux quatre sites. Le systeme a ete teste au cours des printemps 1988 et 1989 par la station meteorologique de Reims, la prevision etant realisee a 15h UTC pour la nuit suivante. Les resultats du test de l'annee 1989 sont les suivants: pour les nuits de ciel clair l'ecart absolu moyen entre valeurs observees et prevues de la temperature minimale sous abri a Reims est de 1,7°C tandis que l'ecart absolu moyen entre valeurs observees et prevues de l'indice actinothermique est inferieur a 1,5°C dans 6 0% des cas a Avize et Mailly. Pour Chambrecy, le poste le plus gelif, les quatre nuits avec risque de gel ont ete correctement prevues, une fausse alerte ayant ete declenchee une seule fois.