Weed on open fields is harming farmers by suppressing growth of the planted crops. Applying agronomic practices along with chemical herbicides can limit the damage from weeds. Global field trial experiments are carried out to gain insights about effectiveness and good combinations of agronomic practices while taking regional differences into account. The main goal of this project was to gain a comprehensive evaluation of costs and effectiveness for the surveyed weed management practices and to derive recommended actions from it. A global business intelligence platform with the information of regional trials displaying as dynamic dashboards was built to reach the goal of the project.
The previous recorded assessment data is highly heterogenous across different localities. To create the central platform, a global homogenous data format needed to be introduced to adapt the regional scheme. The dashboards needed to be dynamically adaptable by the user, so that specific relevant reports can be created and exported.
A broadly adaptable template for data acquisition was designed for the heterogenous data sources. The existing data was transferred to the templates and future assessments can be documented in the same format. R, along with its wrangling package dplyr was involved in the data transformation process. The global, central platform was built on Tibco Spotfire. It combines various regional platforms and evaluates the cost-effect across regions. The created dashboards and reports are accessible via web browser.
The assessment data for several countries and the corresponding experiments was successfully imported to the central platform for weed management analysis. Results of more countries, as well as future experiments will be added continuously. All existing graphs and filters will expand automatically as new data is added.
Dynamic dashboards provide producers a fast solution on economic evaluation of individual weed management practices as well as usage of particular products. Through the standardized data acquisition and data wrangling, the data can be further analyzed by using machine learning methods. Also the weed management data can be merged with assessment data of other experiments, weather data or more detailed products information. This opens up completely new analytics possibilities.