Optimizing Variable Rate Granular Fertilizer Spreader Performance for Single-Tree Prescription Zones

Abstract

Variable rate application (VRA) granular fertilizer spreaders are increasingly important for improved nutrient management of Florida’s citrus groves. Performance characteristics of selected VRA spreaders tested in Florida showed that they were not properly equipped or configured for applying accurate fertilizer rates to single tree prescription zones. The major limitations identified were slow VRA controller and hydraulic valve combinations, resulting in excessive response and transition times (2-5 s) during rate changes. The objective of this study was to modify a split-chain, spinner-type, hydraulically driven VRA fertilizer spreader for optimized performance in single-tree prescription zones. The existing rate control of the fertilizer spreader was set through two hydraulic servo valves actuated with small DC motors. Faster electro-proportional flow control valves and a new matching VRA controller plus rate sensors (system 2) were installed for comparison with the existing valves and controller (system 1). The performance of these two VRA systems was tested as a complete closed loop control system on the spreader by simulating a sequence of representative fertilizer rate changes in response to typical speeds and travel distances encountered in variable citrus groves. Performance data were collected with a digital flow meter installed between the hydraulic valve and hydraulic motor. Sigmoidal regression and time series plots showed significant errors and bias in the actual fertilizer rates obtained with system 1, where in general the low rates were too high and the high rates too low due to a slow system response that could not keep up with the rate change commands. Overall accuracy estimated by the RMSE of the actual versus target rate regression (R2=0.85) of system 1 was 62.5 kg/ha, and response times were up to 3.7 s, making it unsuitable for use in single-tree fertilization. System 2 was significantly faster (maximum 0.46 s response time) and more accurate (R2=0.99; RMSE=14.9 kg/ha), with negligible bias. The merits of a ‘look-ahead’ feature for further fine tuning of these two systems was investigated. The inclusion of 'look-ahead' compensation in system 1 had no effect, while in system 2, the RMSE was further reduced to 6.77 kg/ha, making the modified VRA spreader highly suitable for fertilization of citrus on a single tree basis.