Field Demonstration of a Low-Cost NDVI-Guided Stepwise Variable-Rate Nitrogen Topdressing Applicator for Rice

Low-cost variable-rate applicators are needed to translate crop-sensing nitrogen (N) recommendations into field-scale fertilizer delivery. This study presents a field demonstration of a normalized difference vegetation index (NDVI)-guided stepwise variable-rate granular urea topdressing applicator for rice in a single-site, single-season comparison with a uniform split-management treatment. The tractor-mounted system used an active optical canopy sensor, a deterministic NDVI-to-rate rule, pulse-width-modulation actuator control, and fluted-roller metering. Fertilizer rates are reported as urea product, with elemental N equivalents calculated from 46% N content. The uniform split-management treatment prescribed three 75.00 kg urea ha−1 splits, totaling 225.00 kg urea ha−1 seasonally. The NDVI-guided split-management treatment prescribed two fixed manual splits of 56.25 kg urea ha−1 each and used a sensor-guided third split at 45 days after transplanting. Because the first two splits differed between treatments, the comparison represents two complete split-management strategies rather than the isolated effect of the sensor-guided third application. The aggregate field-collection estimate for the sensor-guided third split was 100.92 kg urea ha−1, which was 25.92 kg urea ha−1 greater than the uniform third split; the estimated seasonal total was 213.42 kg urea ha−1, or 5.1% less than the prescribed uniform seasonal total because of the two lower early manual splits. Harvest measurements adjusted to 14% grain moisture showed mean grain yields of 7.35 and 8.02 t ha−1 for the uniform and NDVI-guided split-management treatments, respectively. Treatment-level partial factor productivity, calculated from treatment-mean yield and estimated seasonal N input, was 71.0 and 81.7 kg grain kg−1 N, respectively. These results show the effectiveness of the applicator within the tested split-management comparison; however, the retained records do not verify rate-class frequency, spatial as-applied delivery, or complete strip coverage, and the yield and partial-factor-productivity differences should not be attributed solely to NDVI sensing or variable-rate actuation.