Optimization of Fire blight scouting with a decision support system based on infection risk

Tsvi Kuflik, Ilaria Pertot, Robert Moskovitch, Rosaly Zasso, Elisabetta Pellegrini, Cesare Gessler

Research output: Contribution to journalArticlepeer-review


This paper presents FireFight, which is a model-driven decision support system designed to help optimization of scouting efforts for Fire blight disease symptoms in an area where the disease is not yet established. Fire blight is the most destructive bacterial disease of apple and pear. The most effective method to prevent disease spreading inside a new area is the accurate monitoring of orchards to identify symptoms and destroy the infected material. Infections occur only if the pathogen is present during favorable environmental and plant conditions. Scouting for the disease is expensive and time-consuming, due to the vast areas to be monitored in a very short period of time (immediately after the end of the disease incubation period). In some areas, as in the Trentino region (northern Italy), exhaustive monitoring is almost impossible due to the large number of trees, and therefore only a random sample is examined after the infection risk period. This approach is problematic, since it may fail to identify infected plants. Fire blight warning systems are widely used to forecast disease outbreak and support treatment planning. However, these systems are known to suit the specific regional conditions for which they were developed. In order to overcome the uncertainty due to the local nature of disease prediction models, several algorithms were applied in parallel, so that a wide range of different environmental conditions could be covered. FireFight is based on three of the most widely used Fire blight forecasting models (MARYBLIGHT, BIS95, and Fire blight Control Advisory). Disease incubation period was also included to advice scouting efforts when visible infection symptoms are timed to appear. Historical weather data analysis of the 57 subareas of Trentino revealed that there is relatively little risk (average of 1.13 risky days per subarea per year) during bloom. In 2003, 2004, and 2005 there were risky conditions in 37%, 51%, and 67% of the subareas, respectively. By comparing the current sample-based scouting (where no symptoms were found even if they were present in the orchards) and the positive FireFight risk prediction in the areas where disease symptoms were actually found, we can conclude that FireFight can assist in the optimization of scouting efforts. It seems that the FireFight DSS approach can be applied in general to others plant diseases where accurate monitoring is needed in areas where disease is not yet present.

Original languageEnglish
Pages (from-to)118-127
Number of pages10
JournalComputers and Electronics in Agriculture
Issue number2
StatePublished - Jul 2008

Bibliographical note

Funding Information:
The research was supported by SafeCrop Centre funded by Fondo per la ricerca, Autonomous Province of Trento. We acknowledge Dani Shtienberg, Eduard Hollinger and Stefano Corradini for the information and help in the development of the FireFight DSS.


  • Erwinia amylovora
  • Fire blight risk prediction
  • Model-based agricultural decision support system
  • Optimization of resource allocation

ASJC Scopus subject areas

  • Forestry
  • Agronomy and Crop Science
  • Computer Science Applications
  • Horticulture


Dive into the research topics of 'Optimization of Fire blight scouting with a decision support system based on infection risk'. Together they form a unique fingerprint.

Cite this