Uniform phosphorous fertilisation has economical, ecological and agronomical shortcomings. This study was undertaken to optimise the variable rate (VR) elemental P application using a previously developed on-the-go visible (VIS) and near infrared (NIR) soil sensor. This VIS–NIR sensor consists of a chisel unit, to which the optical unit to detect soil extractable phosphorous (P-ext) was attached.Amobile, fibre-typeVIS–NIR spectrophotometer (ZeissCorona 45 visnir, Germany) with ameasurement range between 305 and 1711 nmwas used to measure soil spectra in reflectancemode. On-the-go measurement of soil spectrawas carried out in two fields (A and B) situated near Leuven in Belgium. From the spectra, P-ext was calculated in soil and subsequently the required elemental P was determined. Different averaging windows (AW) of the predicted P-ext from successive spectra (2–22) and five recommendation classification intervals (RCI) of elemental P of 20, 10, 5, 2 and 1 kg ha_1 were assigned and tested. The VR of elemental P was compared with uniform rate (UR) application.Results showed that among the fiveRCIs, theminimumelemental P application ratewas for interval of 5 kg ha_1, with small differences of among the different RCIs. In the fields under study, the amount of elemental P fertiliser according to theVR approach was higher than theURapplicationwith an extra elemental P of 4 and 2.38 kg ha_1 for fields A and B, respectively. However, this higher elemental P fertiliser recommendation of VR is only valid when an equal number of samples (1200 in fieldAand 660 in fieldB) is considered for bothVRandURmethods. Largeramounts of elemental P fertiliser were needed forplots and/or fields having higher variation in measured P-ext. The results also showed that in both fields the application rate decreased with larger AWs. Averaging of less than five P-ext successive values was not a proper choice with any RCIs due to the large deviation between the target and classified elemental P in
