For the second year, European FP7 projects ERMES and RICEGUARD collaborated in the forecast of rice blast infection risk, since the results from the preliminary comparison in 2015 were promising. During 2016, results of the models used in the two project were compared in two areas of northern Italy and one in Greece (Figures 1–3). Comparison was conducted between the Daily Infection Warning Hours (DIWH) data (continuous line)—derived from the RICEGUARD model (modified Yoshino model)—and infection risk percentages (orange bars) derived from WARM model, which is used in the ERMES project.
The RICEGUARD project (“In-field wireless sensor network to predict rice blast”) aims at the development of a prediction system that comprises an in-field customised meteorological station and a rice blast prediction model. On the other hand, the WARM model used in ERMES acquires meteorological data from the national network of weather stations and/or from regional or local weather forecasrts.
Despite the significant differences in meteorological data’s acquisition and effective scale, the infection risk prognoses are very close in several cases. For example:
- Figure 1 (Thessaloniki – Greece): The peak between June 30 to July 4, 2016 and between August 24 to September 12, 2016;
- Figure 2 (Villimpenta – Italy): The time period from the 4th of July until the end of September, when the prognoses of the two models are almost identical;
- Figure 3 (Gazzo Veronese – Italy): The peak between July 21 until July 27, 2016. In this case, the WARM model prediction seems more elevated than RICEGUARD’s. However, the trend of the prognosis remains similar, with high risk predicted until the end of September.
Although the WARM model lacks the much more relevant information provided by RICEGUARD’s in-field meteorological stations, the analysis revealed that it can be used operationally for predicting high risk periods. Ultimately, the inter-comparison of the two FP7 projects is of great importance for validating the integrity and estimating the uncertainty of the two models.