FAIRshare Dataset

Inagro, ILVO2019

A prediction model for the presence of damage caused by wireworms was developed based on the harmful wireworm species collected and their numbers over the last four years, in combination with the soil parameters and the cultivation history of the plots monitored. The 'random forest' machine learning model proved to be the most reliable and transparent method. Based on the data of 235 plots, ten factors seem to determine the presence of harmful wireworm species. The important parcel parameters are soil type, acidity (pH) and organic matter content. For the plot history, the main crops and the number of years of cereals/grass/maize as the main crop over the past five years seem important. The presence of a green cover in the winter prior to cultivation and the type of tillage carried out one to four years ago may in certain circumstances also have an impact on the occurrence of wireworms. Most of the plot parameters and the history of each plot are available through the VLM database (in Flanders) and have been converted into usable map layers for the platform. Any missing model parameters for the intended plot are requested from the user. Linking the predicted wireworm pressure to the main crop of the new season allows the underlying model to calculate the risk of crop damage. The web platform displays a colour code per field, which makes the risk of wireworm damage clear to the grower. An advice is linked to this colour code about the possible measures the grower can take. Among other things, the grower is advised to carry out a sampling on the risk plot. For this purpose, reference is made to the protocol for placing traps.