Abstract :
[en] The B-CGMS project, started in 1998, is the adaptation to Belgian Conditions of the European Crop Growth Monitoring System (CGMS). This project involved 3 Belgian scientific institutes: the Walloon agricultural research Centre (CRA-W), the Flemish Institute for Technological Research (VITO) and the University of Liège (ULg). The main difference with the European system is that more detailed inputs (meteorological, soil and NUTS inputs) are used.
Crop yields predictions are realised on a monthly basis during the growing season (from April to September) for 6 crops (winter wheat, winter barley, maize, Potato, sugar beet, winter rapeseed). Yields predictions as well as analyses of meteorological situation of the month and RS information on the state of the crops are published in agrometeorological bulletins sent by e-mail since 2002. The information is also available on the Internet website of the project (http://.b-cgms.cra.wallonie.be). Crop yields predictions are produced through a combination of linear regression models which may include different categories of yield indicators (trend, meteo, RS and agrometeorological model outputs). Crop yields predictions procedure is currently semi-automated by the use of a statistical calibration toolbox (StatCaT).
The evaluation of the project after 5 years of monitoring and prediction has first shown that final yields predicted B-CGMS as well as the ones predicted by MARS are coherent compared with official yields: no significant differences are observed. As far as the accuracy according to the month for which the prediction is made is concerned, we can notice that at agricultural circumscriptions level and for winter crops a lower precision of B-CGMS is observed before June and that there is no improvement in July (in comparison with June). The same evolution is observed for summer crops before July but in August and September, the prediction accuracy decreases.
Even if calibration models present high adjusted coefficient of determination, the technological trend explains an important part of the variability and it is therefore necessary to consider the effect of a year factor on the quality of prediction in order to clearly the interest of the agrometeorological model. For some crops (as potato), adding agrometeorological yield outputs to models including already the technological trend allow to improve the quality of prediction especially for “extreme” year i.e. years where official yields move away significantly from the technological trend.
For others crops as winter wheat, this improvement of the quality of prediction is not observed. However, fortunately, adding other yield indicators as meteo indicators can improve in general the quality of prediction and once again especially for “extreme” years.
